{"categories":{"Correct class":1,"Correct rejection":4,"False rejection":3},"completed_decisions":8,"completed_surveys":6,"decisive_mentions":{"autocorrelation":6,"bands":8,"envelope_2":2,"envelope_3":1,"envelope_4":3,"fft":7,"kurtosis_bank":2,"ridge":6,"spectral_kurtosis":1,"stft_detail":1,"wavelet":1,"welch":2},"failures":[],"registered_decisions":8,"rows":[{"baseline":{"category":"Correct class","dataset":"Jin","name":"Jin-known-T01","phase":"known","predicted":"C01","query_id":"T01"},"category":"Correct class","contract_sha256":"fceff32a6a0d3f326ca8bc674346a2c9df13f5cbf06498bc82a688a0eee732fd","dataset":"Jin","expected":"C01","name":"Jin-known-T01","original_sha256":"4a129e2706146fe1624ef37b784218bb49e991c53dc16fc52eed2ec7df2d8280","phase":"known","predicted":"C01","query_id":"T01","result":{"counterevidence":"Q01 has a strong, persistent approximately 0.94\u20131.0 kHz line family and substantially greater midband filter-bank impulsiveness (maximum 18.77) than R01 (8.73); its dominant ridge median is 940 Hz rather than R01's 16 kHz. Its 8\u201316 kHz modulation also peaks at 137.55 Hz rather than R01's 13.78 Hz. These are outside-reference or conflicting features, so resemblance to C01 is not exact.","decision":{"comparisons":[{"condition_id":"C01","differences":["Q01 adds a strong approximately 942 Hz cluster and has less relative power above 8 kHz than R01.","Q01 filter-bank kurtosis is stronger and extends farther through the midband; spectral-kurtosis maxima shift from R01's 2.14 kHz to Q01's 3.38 kHz.","Q01 high-carrier envelope peaks at 137.55 Hz rather than R01's 13.78 Hz."],"evidence_ids":["R01.fft","T01.fft","R01.bands","T01.bands","R01.envelope_2","T01.envelope_2","R01.envelope_3","T01.envelope_3","R01.kurtosis_bank","T01.kurtosis_bank","R01.spectral_kurtosis","T01.spectral_kurtosis"],"similar":true,"similarities":["Both show periodic approximately 3.5 Hz activity and harmonics across 500\u20134000 Hz carrier envelopes.","Both have broad low/midband spectral content, a weaker persistent approximately 16 kHz feature, and impulsiveness concentrated principally below about 5 kHz.","The mean-centered full-band Welch control is much nearer R01 (3.15 dB) than R02\u2013R04 (7.62\u201310.60 dB)."]},{"condition_id":"C02","differences":["R02 concentrates 59.7% below 100 Hz and 39.2% above 8 kHz, whereas Q01 is midband-heavy from 500\u20134000 Hz.","R02 lacks Q01's strong cross-band approximately 3.5/7 Hz modulation and has high-frequency rather than midband filter-bank impulsiveness.","The Welch control distance is 10.60 dB, substantially larger than for R01."],"evidence_ids":["R02.bands","T01.bands","R02.envelope_2","T01.envelope_2","R02.envelope_3","T01.envelope_3","R02.kurtosis_bank","T01.kurtosis_bank","R02.welch","T01.welch"],"similar":false,"similarities":["Both contain low-frequency content and a persistent approximately 16 kHz line, but these generic structures have different relative weighting."]},{"condition_id":"C03","differences":["R03 is overwhelmingly below 100 Hz (94.1%) with centroid 0.638 kHz, unlike Q01's broad midband distribution and 3.736 kHz centroid.","R03 is near-Gaussian in the filter bank (maximum 0.658), whereas Q01 is strongly impulsive below about 10 kHz (maximum 18.77).","R03 carrier envelopes are led by 81\u2013212 Hz components rather than Q01's strong approximately 3.5/7 Hz family."],"evidence_ids":["R03.bands","T01.bands","R03.welch","T01.welch","R03.kurtosis_bank","T01.kurtosis_bank","R03.envelope_2","T01.envelope_2","R03.envelope_3","T01.envelope_3"],"similar":false,"similarities":["Both include several midband narrow lines, but their total spectral weighting and modulation differ strongly."]},{"condition_id":"C04","differences":["R04 places 61.1% of power above 8 kHz and has a dense dominant approximately 16 kHz family; Q01 places only 17.7% above 8 kHz and is strongest around 0.94 kHz plus lower bands.","R04 filter-bank impulsiveness is low-frequency sub-Gaussian and peaks near 15.7\u201317.0 kHz, opposite Q01's strong 1.3\u20135.3 kHz impulsiveness.","R04 broadband/high-carrier modulation emphasizes approximately 13.2 Hz, unlike Q01's dominant approximately 3.5 Hz broadband family and 137.55 Hz high-carrier maximum."],"evidence_ids":["R04.bands","T01.bands","R04.fft","T01.fft","R04.kurtosis_bank","T01.kurtosis_bank","R04.envelope","T01.envelope","R04.envelope_5","T01.envelope_5"],"similar":false,"similarities":["Both retain an approximately 16 kHz line and show some periodic low-frequency structure, but those are not the dominant Q01 characteristics."]}],"condition_id":"C01","explanation":"Exploratory binary judgment: Q01 is most consistent with known condition C01 (excess_hall_adhesive). The observable basis is the R01-like broad low/midband spectrum, persistent periodic approximately 3.5 Hz modulation across multiple carrier bands, and sub-5 kHz impulsiveness. The strong approximately 0.94 kHz Q01 family and greater impulsiveness prevent an exact-match claim. This does not establish fault identity, affected component, or severity.","outcome":"similar"},"decision_basis":"Visual and reported numeric observations, not numerical controls alone, drive the decision. The most decisive evidence is the joint spectral distribution, cross-carrier approximately 3.5 Hz modulation, and location of filter-bank impulsiveness. The Welch, estimator, reassigned-STFT, modulation-map, cepstral, and kurtosis controls independently rank R01 nearest within their own incompatible scales, but they are treated only as corroborating controls and not as calibrated acceptance evidence.","decisive_diagrams":["bands","fft","envelope_2","envelope_3","kurtosis_bank"],"diagram_review":{"autocorrelation":{"effect":"supports","evidence_ids":["R01.autocorrelation","R02.autocorrelation","R03.autocorrelation","R04.autocorrelation","T01.autocorrelation","Q01.C01.S0001"],"favored_condition":"C01","finding":"Q01's strongest nonzero peak is 68.0 \u00b5s/0.813, matching R01's 68.0 \u00b5s/0.870 rather than R02/R04's 249.4 \u00b5s or R03's 181.4 \u00b5s; long-lag shape is not identical and no speed inference is justified.","priority":"medium"},"bands":{"effect":"supports","evidence_ids":["R01.bands","R02.bands","R03.bands","R04.bands","T01.bands","Q01.C01.S0001"],"favored_condition":"C01","finding":"Q01 is broadly midband weighted: 54.3% from 500\u20134000 Hz and 20.2% below 100 Hz. This is closest in shape to broad R01, while contradicting R03's 94.1% below 100 Hz and R04's 61.1% above 8 kHz; Q01 nevertheless has less high-frequency power than R01.","priority":"high"},"cepstrum":{"effect":"supports","evidence_ids":["R01.cepstrum","R02.cepstrum","R03.cepstrum","R04.cepstrum","T01.cepstrum"],"favored_condition":"C01","finding":"Q01's largest peak is 22.7 \u00b5s/0.266, close to R01's 22.7 \u00b5s/0.271; the nearest-view control also selects R01. Echo/response alternatives prevent interpreting this as a mechanical period.","priority":"medium"},"cyclic":{"effect":"neutral","evidence_ids":["R01.cyclic","R02.cyclic","R03.cyclic","R04.cyclic","T01.cyclic"],"favored_condition":null,"finding":"Q01's maximum squared coherence, 0.716, exceeds every reference (0.131\u20130.309), while its median 0.00286 lies in the common range. With no significance threshold and deterministic tones present, this is outside-reference evidence rather than class support.","priority":"low"},"envelope":{"effect":"supports","evidence_ids":["R01.envelope","R02.envelope","R03.envelope","R04.envelope","T01.envelope","Q01.C01.S0001"],"favored_condition":"C01","finding":"Q01 has a strong 3.535 Hz fundamental with 7.071, 10.606, and 14.141 Hz multiples, most resembling R01's 3.434 Hz-led broadband modulation; R04 instead leads at 13.232 Hz and R03 has weaker low-rate structure.","priority":"high"},"envelope_1":{"effect":"supports","evidence_ids":["R01.envelope_1","R02.envelope_1","R03.envelope_1","R04.envelope_1","T01.envelope_1"],"favored_condition":"C01","finding":"For 100\u2013500 Hz carriers Q01 leads at 3.571 Hz/0.331 with 3.469, 7.041, and 14.082 Hz companions. It is weaker than R01's 3.469 Hz/0.599 but shares its low-rate family better than R02\u2013R04.","priority":"medium"},"envelope_2":{"effect":"supports","evidence_ids":["R01.envelope_2","R02.envelope_2","R03.envelope_2","R04.envelope_2","T01.envelope_2"],"favored_condition":"C01","finding":"For 500\u20131500 Hz carriers Q01 peaks at 3.571 Hz/1.292 with a 7.041 Hz harmonic, closely matching R01's exceptional 3.469 Hz/1.262 and 6.939 Hz/0.979 pattern; R02/R04 lead near 6.5 Hz and R03 at 81.3 Hz.","priority":"high"},"envelope_3":{"effect":"supports","evidence_ids":["R01.envelope_3","R02.envelope_3","R03.envelope_3","R04.envelope_3","T01.envelope_3"],"favored_condition":"C01","finding":"For 1.5\u20134 kHz carriers Q01 has strong 3.571/7.041 Hz modulation (0.946/0.884), matching R01's low-rate family and sharply differing from R02, R03, and R04 maxima at 338, 187, and 172 Hz.","priority":"high"},"envelope_4":{"effect":"supports","evidence_ids":["R01.envelope_4","R02.envelope_4","R03.envelope_4","R04.envelope_4","T01.envelope_4"],"favored_condition":"C01","finding":"For 4\u20138 kHz carriers Q01 remains led by 3.571 and 7.041 Hz, resembling R01's 3.469/6.939 Hz structure, though Q01's maximum 0.323 exceeds R01's 0.205.","priority":"medium"},"envelope_5":{"effect":"contradicts","evidence_ids":["R01.envelope_5","R02.envelope_5","R03.envelope_5","R04.envelope_5","T01.envelope_5"],"favored_condition":"C03","finding":"For 8\u201316 kHz carriers Q01 peaks at 137.55 Hz/0.158 and is weak overall. No reference matches that maximum; amplitude is closest to R03, while R01/R04 lead near 13 Hz and R02 near 6.5 Hz.","priority":"medium"},"estimators":{"effect":"supports","evidence_ids":["R01.estimators","R02.estimators","R03.estimators","R04.estimators","T01.estimators"],"favored_condition":"C01","finding":"All Q01 estimators preserve broad low/midband energy, a strong approximately 0.95 kHz region, and a weaker approximately 16 kHz line. Overall shape is nearest R01 by control distance 3.96 versus 9.98\u201310.32, although the 0.95 kHz feature is Q01-specific.","priority":"medium"},"fft":{"effect":"supports","evidence_ids":["R01.fft","R02.fft","R03.fft","R04.fft","T01.fft","Q01.C01.S0001"],"favored_condition":"C01","finding":"Q01 shares R01's 5.3 Hz line, approximately 169\u2013180 Hz group, broad midband structure, and weaker approximately 16.03 kHz line. Q01 uniquely emphasizes a 934\u2013995 Hz cluster, so the match is partial rather than exact.","priority":"high"},"harmonics":{"effect":"supports","evidence_ids":["R01.harmonics","R02.harmonics","R03.harmonics","R04.harmonics","T01.harmonics"],"favored_condition":"C01","finding":"Q01's largest spacing count is 3 Hz/count 17, closest to R01's 3 Hz/count 14; Q01 also has 7 Hz/count 13. Unrelated tonal clusters can generate these counts, so this is secondary evidence only.","priority":"medium"},"kurtosis_bank":{"effect":"supports","evidence_ids":["R01.kurtosis_bank","R02.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank","T01.kurtosis_bank"],"favored_condition":"C01","finding":"Q01 and R01 both concentrate positive impulsiveness below about 5 kHz, peaking in 1.33\u20132.64 kHz. Q01 is substantially stronger (18.77 versus 8.73) and extends through 5\u201310 kHz; R02/R04 instead peak near 15\u201317 kHz and R03 is near Gaussian.","priority":"high"},"levels":{"effect":"supports","evidence_ids":["R01.levels","R02.levels","R03.levels","R04.levels","T01.levels","Q01.C01.S0001"],"favored_condition":"C01","finding":"Q01 is unclipped, near-zero-DC, peak-normalized, and periodically impulsive; RMS 0.120 and crest factor 17.98 dB are closest to R01's 0.137/16.84 dB. Normalization precludes acoustic-level or severity claims.","priority":"medium"},"modulation_map":{"effect":"supports","evidence_ids":["R01.modulation_map","R02.modulation_map","R03.modulation_map","R04.modulation_map","T01.modulation_map"],"favored_condition":"C01","finding":"Q01's map maximum 0.245 and median 0.00222 are closest to R01's 0.252/0.00325; the carrier-specific views confirm shared low-rate modulation rather than relying on the scalar maximum.","priority":"medium"},"orders":{"effect":"unavailable","evidence_ids":["R01.orders","R02.orders","R03.orders","R04.orders","T01.orders"],"favored_condition":null,"finding":"Unavailable for Q01 and every reference because no independent RPM trace was supplied; no order-domain comparison is possible.","priority":"unavailable"},"persistence":{"effect":"neutral","evidence_ids":["R01.persistence","R02.persistence","R03.persistence","R04.persistence","T01.persistence"],"favored_condition":"C04","finding":"Q01 maximum occupancy is 29.47%, visually retaining periodic low/midband and approximately 16 kHz populations. The raw signature control ranks R04 nearest, conflicting with the spectral-shape evidence; chronology is absent and frames overlap.","priority":"low"},"reassigned":{"effect":"supports","evidence_ids":["R01.reassigned","R02.reassigned","R03.reassigned","R04.reassigned","T01.reassigned"],"favored_condition":"C01","finding":"Q01 localizes periodic low/midband lines plus a weaker approximately 16 kHz ridge, visually closest to R01; the control distance is 3.60 for R01 versus 8.94\u20139.81 for the others. The strong approximately 0.94 kHz family remains Q01-specific.","priority":"medium"},"ridge":{"effect":"contradicts","evidence_ids":["R01.ridge","R02.ridge","R03.ridge","R04.ridge","T01.ridge"],"favored_condition":"C03","finding":"Q01's dominant-bin median is 940 Hz, unlike R01's 16 kHz and R02/R04's 50 Hz; among references the raw control is nearest R03, whose median is 20 Hz, so no close ridge match exists.","priority":"medium"},"spectral_kurtosis":{"effect":"supports","evidence_ids":["R01.spectral_kurtosis","R02.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis","T01.spectral_kurtosis"],"favored_condition":"C01","finding":"Q01 has broad positive intermittency below about 5 kHz with maximum 5.03 at 3.38 kHz, most resembling R01's sub-5 kHz pattern and maximum 5.87 at 2.14 kHz. The frequency shift is material, and finite-sample bias is uncorrected.","priority":"high"},"stft":{"effect":"supports","evidence_ids":["R01.stft","R02.stft","R03.stft","R04.stft","T01.stft"],"favored_condition":"C01","finding":"Q01 shows regular vertical low/midband bursts, persistent approximately 0.94 kHz structure, and a weaker approximately 16 kHz line. Its periodic broadband pattern most resembles R01, though the 0.94 kHz dominance is additional evidence.","priority":"medium"},"stft_detail":{"effect":"supports","evidence_ids":["R01.stft_detail","R02.stft_detail","R03.stft_detail","R04.stft_detail","T01.stft_detail"],"favored_condition":"C01","finding":"At 0.1 s frames and 10 Hz bins, Q01's repeating low/midband structures resemble R01 more than the steadier R02\u2013R04 patterns; control distance is 3.52 for R01 versus 9.30\u20139.81.","priority":"medium"},"synchronous":{"effect":"unavailable","evidence_ids":["R01.synchronous","R02.synchronous","R03.synchronous","R04.synchronous","T01.synchronous"],"favored_condition":null,"finding":"Unavailable for Q01 and every reference because no independent RPM trace was supplied; phase-locked waveform evidence is missing.","priority":"unavailable"},"wavelet":{"effect":"supports","evidence_ids":["R01.wavelet","R02.wavelet","R03.wavelet","R04.wavelet","T01.wavelet"],"favored_condition":"C01","finding":"Below the 4.41 kHz wavelet limit, Q01 has strong regular periodic low/midband patches resembling R01; the control ranks R01 nearest (3.61), followed by R03 (5.44). This view cannot assess approximately 16 kHz evidence.","priority":"medium"},"welch":{"effect":"supports","evidence_ids":["R01.welch","R02.welch","R03.welch","R04.welch","T01.welch","Q01.C01.S0001"],"favored_condition":"C01","finding":"Q01 centroid is 3.736 kHz with broad low/midband power and a weaker approximately 16 kHz hump. It is nearer R01's mixed spectrum than R03's 0.638 kHz or R04's 9.93 kHz extremes; the baseline control ranks R01 nearest at 3.15 dB.","priority":"high"}},"limitations":"Only one peak-normalized 10 s reference represents each condition, so within-condition variability is unknown and no calibrated acceptance threshold exists. Microphone response, placement, reverberation, speed, load, and acquisition regime can explain differences despite gain normalization. The diagram views are correlated transforms, not independent votes; numerical distances use incompatible scales and are only controls. Orders and synchronous averages are unavailable, wavelets exclude the approximately 16 kHz region, cyclic coherence lacks a significance threshold, persistence discards chronology, and kurtosis estimates can reflect unrelated impacts or regime changes. The result supports resemblance to a known condition but does not prove the unknown's actual label, fault identity, component, or severity.","most_informative":["bands","fft","envelope_2","envelope_3","kurtosis_bank"],"priority_changes":"Relative to the frozen survey, envelope_3 rises to decisive because Q01's strong approximately 3.5/7 Hz modulation sharply separates R01 from R02\u2013R04. FFT remains high but is tempered by Q01's outside-reference approximately 0.94 kHz family. Spectral kurtosis remains informative but is not decisive because Q01's maximum shifts to 3.38 kHz. Ridge is downgraded as contradictory/source-selective, and cyclic coherence is downgraded because Q01 exceeds all references without a significance threshold."},"stage":"decision","survey":"survey-aa9448d1719da544","usage":{"completion_tokens":4734,"completion_tokens_details":{"accepted_prediction_tokens":0,"audio_tokens":0,"reasoning_tokens":92,"rejected_prediction_tokens":0},"latency_checkpoint":{"engine_tbt_ms":8,"engine_ttft_ms":667,"engine_ttlt_ms":37899,"pre_inference_ms":1544,"service_tbt_ms":8,"service_ttft_ms":2297,"service_ttlt_ms":39525,"user_visible_ttft_ms":752},"prompt_tokens":51698,"prompt_tokens_details":{"audio_tokens":0,"cached_tokens":0},"total_tokens":56432}},{"baseline":{"category":"Correct rejection","dataset":"Jin","name":"Jin-excluded-T01","phase":"excluded","predicted":"different","query_id":"T01"},"category":"Correct rejection","contract_sha256":"dc25e0ec7ffdb429d09c8a1ae9e8afc97f9e83d442e3c1a85d04fefb4d4a3ea0","dataset":"Jin","expected":"different","name":"Jin-excluded-T01","original_sha256":"5b2cf83dfd0bc3727ce31b5275244d0ec6f4a4297f990f64beaf4d6ef0e053bf","phase":"excluded","predicted":"different","query_id":"T01","result":{"counterevidence":"The strongest counterevidence is partial resemblance to C04/R04 in the persistent approximately 16 kHz component and cyclic features, and to C03/R03 in low-frequency envelope spacing and the wavelet control. The baseline numerical control also ranks R04 nearest (7.62 dB versus R03 7.82 and R02 10.60). These are forced rankings without an acceptance threshold and are outweighed by Q01's repeated broadband transients, 0.5\u20134 kHz power concentration, approximately 3.57 Hz cross-band modulation, and exceptionally high low/mid-band filter-bank kurtosis.","decision":{"comparisons":[{"condition_id":"C02","differences":["Q01 concentrates 54.3% of Welch power from 500\u20134000 Hz, whereas R02 has only about 0.67% there and is dominated by below 100 Hz plus above 8 kHz.","Q01 has strongly periodic broadband impulses and filter-bank excess kurtosis up to 18.77 in 1.33\u20132.64 kHz; R02's impulsiveness peaks at only 1.95 near 14.4\u201315.7 kHz.","Q01's carrier envelopes are dominated across 100\u20138000 Hz by approximately 3.57 Hz and harmonics, unlike R02's carrier-dependent 6.53, 12.96 and 337.76 Hz families."],"evidence_ids":["Q01.C01.S0001","T01.bands","T01.stft_detail","T01.kurtosis_bank","T01.envelope_2","T01.envelope_3","R02.C01.S0001","R02.bands","R02.stft_detail","R02.kurtosis_bank","R02.envelope_2","R02.envelope_3"],"similar":false,"similarities":["Both contain a persistent approximately 16 kHz component and a 5.3 Hz FFT line, but these shared features do not reproduce the query's dominant transient low/mid-frequency regime."]},{"condition_id":"C03","differences":["R03 is overwhelmingly below 100 Hz (94.13%), while Q01 has only 20.19% there and instead concentrates power across 500\u20134000 Hz.","Q01 shows regular broadband transient columns and filter-bank kurtosis up to 18.77; R03 is comparatively persistent and has a maximum filter-bank value of 0.658.","R03's carrier-specific envelope peaks at 81.33, 187.35 and 212.04 Hz in bands 2\u20134, whereas Q01 consistently emphasizes approximately 3.57 Hz and its harmonics."],"evidence_ids":["Q01.C01.S0001","T01.bands","T01.stft","T01.kurtosis_bank","T01.envelope_2","T01.envelope_3","T01.envelope_4","R03.C01.S0001","R03.bands","R03.stft","R03.kurtosis_bank","R03.envelope_2","R03.envelope_3","R03.envelope_4"],"similar":false,"similarities":["Q01 shares R03's approximately 3.57 Hz modulation in the 100\u2013500 Hz band, and the wavelet numerical control is nearest R03, but the broader modulation and transient structures conflict strongly."]},{"condition_id":"C04","differences":["R04 puts 61.07% of power above 8 kHz and only about 3.83% from 500\u20134000 Hz; Q01 puts 17.67% above 8 kHz and 54.27% from 500\u20134000 Hz.","Q01 has repeated broadband transients and very high 1.33\u20135.25 kHz filter-bank kurtosis (6.78\u201318.77), while R04 is comparatively stationary and peaks at 0.745 near 15.7\u201317.0 kHz.","Q01's strongest modulation in carrier bands 2\u20134 is approximately 3.57 Hz, unlike R04's 6.63, 171.73 and 26.43 Hz peaks."],"evidence_ids":["Q01.C01.S0001","T01.bands","T01.stft_detail","T01.kurtosis_bank","T01.envelope_2","T01.envelope_3","T01.envelope_4","R04.C01.S0001","R04.bands","R04.stft_detail","R04.kurtosis_bank","R04.envelope_2","R04.envelope_3","R04.envelope_4"],"similar":false,"similarities":["Both show a persistent approximately 16 kHz line and cyclic-coherence horizontal features; the baseline and persistence numerical controls rank R04 nearest."]}],"condition_id":null,"explanation":"Q01 is substantially different from all three supplied recordings. Its observable signature is a train of broadband impulses at approximately 3.57 Hz, strong 0.5\u20134 kHz energy, cross-band modulation at that spacing and harmonics, and extreme low/mid-band impulsiveness. None of R02, R03 or R04 combines those properties. R04 is the forced nearest baseline reference, but nearest is not acceptance and several high-priority diagnostics directly conflict. This does not identify an actual unknown label, fault component or severity.","outcome":"different"},"decision_basis":"Visual and numeric query measurements\u2014not the nearest-reference controls\u2014drive the decision. The decisive evidence is the query's transient STFT/wavelet pattern, markedly different band-power distribution, coherent approximately 3.57 Hz carrier-envelope family, and filter-bank kurtosis far outside every supplied reference. Numerical controls are mixed: baseline, estimators, persistence and spectral kurtosis favor R04; wavelet and several envelopes favor R03; kurtosis-bank and reassigned distances favor R02. Their disagreement and absence of calibrated thresholds prevent treating them as acceptance evidence.","decisive_diagrams":["bands","stft_detail","envelope_2","kurtosis_bank","wavelet"],"diagram_review":{"autocorrelation":{"effect":"supports","evidence_ids":["T01.autocorrelation","Q01.C01.S0001","R02.autocorrelation","R02.C01.S0001","R03.autocorrelation","R03.C01.S0001","R04.autocorrelation","R04.C01.S0001"],"favored_condition":null,"finding":"Q01's strongest nonzero peak is 0.813 at 0.0680 ms with rapidly damped oscillation, unlike R02/R04 at 0.249 ms and R03 at 0.181 ms; none is a validated shaft period.","priority":"medium"},"bands":{"effect":"supports","evidence_ids":["T01.bands","Q01.C01.S0001","R02.bands","R02.C01.S0001","R03.bands","R03.C01.S0001","R04.bands","R04.C01.S0001"],"favored_condition":null,"finding":"Q01 places 54.27% in 500\u20134000 Hz and only 17.67% above 8 kHz. R02 and R04 are much more high-frequency weighted, while R03 has 94.13% below 100 Hz; no reference has Q01's distribution.","priority":"high"},"cepstrum":{"effect":"supports","evidence_ids":["T01.cepstrum","R02.cepstrum","R03.cepstrum","R04.cepstrum"],"favored_condition":null,"finding":"Q01 peaks at 22.68 \u00b5s (0.266) and 68.03 \u00b5s (0.246), substantially different from R02's 45.35 \u00b5s maximum, R03's 181.41 \u00b5s maximum and R04's weaker 68.03 \u00b5s maximum; shared secondary peaks are ambiguous.","priority":"medium"},"cyclic":{"effect":"contradicts","evidence_ids":["T01.cyclic","R02.cyclic","R03.cyclic","R04.cyclic"],"favored_condition":"C04","finding":"Q01 has strong horizontal cyclic features and maximum squared coherence 0.716, exceeding R04's 0.299 and R02/R03 near 0.13. Its morphology is relatively closest to R04, but magnitude is outside the references and no significance threshold exists.","priority":"medium"},"envelope":{"effect":"supports","evidence_ids":["T01.envelope","Q01.C01.S0001","R02.envelope","R02.C01.S0001","R03.envelope","R03.C01.S0001","R04.envelope","R04.C01.S0001"],"favored_condition":null,"finding":"Q01 forms a clear 3.535 Hz family at 3.54, 7.07, 10.61, 14.14 and 17.68 Hz. References have only partial overlaps: R02 mixed 2.93/6.46/13.03 Hz, R03 slow plus 43\u201344 Hz, and R04 dominant 13.23 Hz.","priority":"high"},"envelope_1":{"effect":"contradicts","evidence_ids":["T01.envelope_1","R02.envelope_1","R03.envelope_1","R04.envelope_1"],"favored_condition":"C03","finding":"In 100\u2013500 Hz, Q01 is led by 3.571 Hz with 7.04 and 14.08 Hz harmonics; the fundamental matches R03's 3.571 Hz more closely than R02/R04 at 3.265 Hz, although Q01 is stronger and cleaner.","priority":"medium"},"envelope_2":{"effect":"supports","evidence_ids":["T01.envelope_2","R02.envelope_2","R03.envelope_2","R04.envelope_2"],"favored_condition":null,"finding":"In 500\u20131500 Hz, Q01 has exceptionally strong 3.571 Hz modulation (ordinary 1.292; squared 1.651) and harmonics. R02/R04 are led near 6.5 Hz and R03 by 81.33 Hz, so all conflict.","priority":"high"},"envelope_3":{"effect":"supports","evidence_ids":["T01.envelope_3","R02.envelope_3","R03.envelope_3","R04.envelope_3"],"favored_condition":null,"finding":"In 1.5\u20134 kHz, Q01 is dominated by 3.571/7.041 Hz with amplitudes near 0.95/0.88, unlike R02 at 337.76 Hz, R03 at 187.35 Hz and R04 at 171.73 Hz.","priority":"high"},"envelope_4":{"effect":"supports","evidence_ids":["T01.envelope_4","R02.envelope_4","R03.envelope_4","R04.envelope_4"],"favored_condition":null,"finding":"In 4\u20138 kHz, Q01 again has a strong 3.571 Hz harmonic family; reference leaders are 12.96 Hz (R02), 212.04 Hz (R03) and 26.43 Hz (R04).","priority":"high"},"envelope_5":{"effect":"contradicts","evidence_ids":["T01.envelope_5","R02.envelope_5","R03.envelope_5","R04.envelope_5"],"favored_condition":"C03","finding":"In 8\u201316 kHz, Q01 is led by 137.55 Hz and has weaker 3.57 Hz content; no reference matches, although its overall low modulation amplitude is relatively closer to R03 than the strong low-frequency modulation of R02/R04.","priority":"medium"},"estimators":{"effect":"supports","evidence_ids":["T01.estimators","R02.estimators","R03.estimators","R04.estimators"],"favored_condition":"C04","finding":"All Q01 estimators show a broad low/mid-frequency structure including approximately 0.94 kHz and a weaker 16 kHz line, unlike R03's dominant near-DC spectrum and R02/R04's stronger 16 kHz regions. The numerical distance narrowly favors R04 but remains large.","priority":"high"},"fft":{"effect":"supports","evidence_ids":["T01.fft","Q01.C01.S0001","R02.fft","R02.C01.S0001","R03.fft","R03.C01.S0001","R04.fft","R04.C01.S0001"],"favored_condition":null,"finding":"Q01 has prominent clusters near 173 Hz and 0.94\u20130.99 kHz plus a 5.3 Hz line. R02 centers its dense cluster near 16 kHz, R03 is dominated by 5.3 Hz with several low-kHz lines, and R04 by 50 Hz plus 16 kHz.","priority":"high"},"harmonics":{"effect":"supports","evidence_ids":["T01.harmonics","R02.harmonics","R03.harmonics","R04.harmonics"],"favored_condition":null,"finding":"Q01's largest spacing count is 3 Hz (17), followed by 7 Hz (13), unlike R02's 7/17 Hz, R03's 1 Hz and R04's 1/13 Hz maxima; spacing counts can mix unrelated tones.","priority":"medium"},"kurtosis_bank":{"effect":"supports","evidence_ids":["T01.kurtosis_bank","R02.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank"],"favored_condition":null,"finding":"Q01 reaches excess kurtosis 18.77 at 1.33\u20132.64 kHz and remains very high through 5.25 kHz. Reference maxima are only 1.95 (R02), 0.658 (R03) and 0.745 (R04), located in different bands.","priority":"high"},"levels":{"effect":"supports","evidence_ids":["T01.levels","Q01.C01.S0001","R02.levels","R02.C01.S0001","R03.levels","R03.C01.S0001","R04.levels","R04.C01.S0001"],"favored_condition":"C02","finding":"Q01 is unclipped and DC-free like all references, but has the lowest normalized RMS (0.1199) and highest crest factor (17.98 dB), with regular sharp transients. R02 is the nearest crest-factor reference at 15.58 dB; normalization bars acoustic-level or severity claims.","priority":"medium"},"modulation_map":{"effect":"supports","evidence_ids":["T01.modulation_map","R02.modulation_map","R03.modulation_map","R04.modulation_map"],"favored_condition":null,"finding":"Q01's maximum normalized modulation is 0.245 and strong low-frequency columns span several carrier bands, exceeding and differing from R02 0.142, R03 0.0839 and R04 0.122; 155 masked cells per map remain missing.","priority":"medium"},"orders":{"effect":"unavailable","evidence_ids":["T01.orders","R02.orders","R03.orders","R04.orders"],"favored_condition":null,"finding":"Order spectra are unavailable for Q01 and all references because no independent RPM trace was supplied.","priority":"unavailable"},"persistence":{"effect":"contradicts","evidence_ids":["T01.persistence","R02.persistence","R03.persistence","R04.persistence"],"favored_condition":"C04","finding":"Q01 reaches 29.47% persistence and retains a 16 kHz population; the numerical control is nearest R04. However, Q01 also has broad low/mid-frequency populations absent from R04, and chronology is discarded.","priority":"medium"},"reassigned":{"effect":"supports","evidence_ids":["T01.reassigned","R02.reassigned","R03.reassigned","R04.reassigned"],"favored_condition":"C02","finding":"Q01 localizes repeated broadband low/mid-frequency transients plus a persistent 16 kHz ridge. None of the references shows that repeated vertical structure; the numerical summary happens to be nearest R02 but pooled weight is not PSD.","priority":"medium"},"ridge":{"effect":"supports","evidence_ids":["T01.ridge","R02.ridge","R03.ridge","R04.ridge"],"favored_condition":null,"finding":"Q01's dominant-bin median is 940 Hz with switches to approximately 16.05 kHz, unlike medians of 50 Hz for R02/R04 and 20 Hz for R03. The ridge has no continuity or speed validation.","priority":"high"},"spectral_kurtosis":{"effect":"supports","evidence_ids":["T01.spectral_kurtosis","R02.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis"],"favored_condition":"C04","finding":"Q01 has broad positive intermittency below roughly 5 kHz and a maximum 5.03 near 3.38 kHz. R02/R03 have much larger narrow low-frequency maxima above 22, while R04 peaks at 2.52 near 15.66 kHz; numerical distance favors R04 despite different location.","priority":"high"},"stft":{"effect":"supports","evidence_ids":["T01.stft","R02.stft","R03.stft","R04.stft"],"favored_condition":null,"finding":"Q01 displays regular broadband vertical events over 10 s, strongest below about 4 kHz, together with a persistent 16 kHz line. All references are dominated by comparatively stationary horizontal components without this event train.","priority":"high"},"stft_detail":{"effect":"supports","evidence_ids":["T01.stft_detail","R02.stft_detail","R03.stft_detail","R04.stft_detail"],"favored_condition":null,"finding":"At identical 0.1 s frames and 10 Hz bins, Q01 clearly shows periodic vertical low/mid-frequency bursts and a median floor of \u221281.70 dB; R02/R03/R04 show stable ridges and no comparable broadband repetition.","priority":"high"},"synchronous":{"effect":"unavailable","evidence_ids":["T01.synchronous","R02.synchronous","R03.synchronous","R04.synchronous"],"favored_condition":null,"finding":"Synchronous averages are unavailable for Q01 and every reference because no independent RPM trace was supplied.","priority":"unavailable"},"wavelet":{"effect":"supports","evidence_ids":["T01.wavelet","R02.wavelet","R03.wavelet","R04.wavelet"],"favored_condition":"C03","finding":"Within the 0\u20134.41 kHz wavelet range, Q01 has conspicuous periodic vertical events across scales, absent from all references' mostly persistent structures. Its median power and numerical distance are closest to R03, but morphology conflicts.","priority":"high"},"welch":{"effect":"supports","evidence_ids":["T01.welch","Q01.C01.S0001","R02.welch","R02.C01.S0001","R03.welch","R03.C01.S0001","R04.welch","R04.C01.S0001"],"favored_condition":"C04","finding":"Q01's centroid is 3.74 kHz with broad 0.5\u20134 kHz power and a weaker 16 kHz peak. It lies between R03's 0.638 kHz and R02/R04's 6.44/9.93 kHz centroids but matches none in shape; the baseline distance narrowly favors R04.","priority":"high"}},"limitations":"Only one peak-normalized recording represents each known condition, so acquisition response, microphone placement, reverberation, speed and load may explain part of the separation. Correlated DSP views are not independent votes. No calibrated distance or acceptance threshold, bearing geometry, independent RPM, cyclic-coherence significance threshold or estimator confidence interval is available. Orders and synchronous averages are unavailable; modulation-map masks and wavelet edge masks are missing data. The evidence supports only an exploratory outside-reference judgment, not an actual label, fault identity, component or severity.","most_informative":["bands","stft_detail","envelope_2","kurtosis_bank","wavelet"],"priority_changes":"After adding Q01, stft_detail and wavelet rise to high priority because they expose its repeated broadband transients, while envelope_2 remains high for the unusually strong approximately 3.57 Hz modulation. Spectral_kurtosis remains informative but is no longer among the five most decisive because its numerical nearest reference conflicts with peak location. FFT is retained as high but is less decisive than integrated bands and transient morphology. Numerical controls did not set the decision."},"stage":"decision","survey":"survey-4bf405b6bcc03509","usage":{"completion_tokens":4620,"completion_tokens_details":{"accepted_prediction_tokens":0,"audio_tokens":0,"reasoning_tokens":141,"rejected_prediction_tokens":0},"latency_checkpoint":{"engine_tbt_ms":8,"engine_ttft_ms":1140,"engine_ttlt_ms":37293,"pre_inference_ms":1140,"service_tbt_ms":8,"service_ttft_ms":2379,"service_ttlt_ms":38527,"user_visible_ttft_ms":1239},"prompt_tokens":43138,"prompt_tokens_details":{"audio_tokens":0,"cached_tokens":0},"total_tokens":47758}},{"baseline":{"category":"Correct class","dataset":"Ottawa","name":"Ottawa-known-D01","phase":"known","predicted":"C01","query_id":"D01"},"category":"False rejection","contract_sha256":"95e305d589f5ed6c0e815d77b4d7d08df0b6c50f7237aea76ed0942ebf83a148","dataset":"Ottawa","expected":"C01","name":"Ottawa-known-D01","original_sha256":"92a1772623b857bcb5010a4599a014f18549cfb0482719e29ddaa16bcfa9e869","phase":"known","predicted":"different","query_id":"D01","result":{"counterevidence":"The strongest counterevidence to rejection is that several broad-spectrum/acquisition-sensitive views are nearest to C01/R01: similar crest factor, low integrated power, broad background, persistence occupancy, wavelet maximum, and reassigned/STFT floors. The uncalibrated Welch control instead ranks R06 first, and individual modulation controls favor several other references. These conflicts show no coherent known-condition match rather than support for a particular fault identity.","decision":{"comparisons":[{"condition_id":"C01","differences":["Q01 is broader than R01: 39.06% versus 68.27% below 100 Hz, centroid 369.8 versus 212.5 Hz, and dominant ridge median 110 versus 50 Hz.","Q01's strongest FFT lines are 92.0/109.8 Hz rather than R01's 45.9 Hz, and its autocorrelation is much less periodic (0.688 versus 0.905)."],"evidence_ids":["R01.bands","D01.bands","R01.welch","D01.welch","R01.fft","D01.fft","R01.autocorrelation","D01.autocorrelation","R01.ridge","D01.ridge"],"similar":false,"similarities":["Both have broad spectra, high crest factors near 12 dB, low multitaper power relative to most references, and 50.38% maximum persistence."]},{"condition_id":"C02","differences":["R02 is 97.25% below 100 Hz with a dominant 45.1 Hz line and autocorrelation 0.979 at 0.0665 s; Q01 is 39.06% below 100 Hz, peaks at 92.0/109.8 Hz, and has only 0.688 correlation at 0.0640 s.","Carrier-band modulation families differ substantially, including 91.73/183.47 Hz for R02 versus 123.16/61.63 Hz for Q01 in envelope bands 4/5."],"evidence_ids":["R02.bands","D01.bands","R02.fft","D01.fft","R02.autocorrelation","D01.autocorrelation","R02.envelope_4","D01.envelope_4","R02.envelope_5","D01.envelope_5"],"similar":false,"similarities":["Both contain some approximately 30/61 Hz spacing or modulation components, but these are nonspecific and may combine unrelated tones."]},{"condition_id":"C03","differences":["R03 is low-frequency dominated with a 45.0 Hz primary and strong 15 Hz modulation across several carrier bands; Q01 is broadband with 92.0/109.8 Hz primaries and lacks that cross-band 15 Hz family.","R03 has strong high-frequency filter-bank impulsiveness (median 0.506, maximum 1.850), unlike Q01 (median 0.021, maximum 0.183)."],"evidence_ids":["R03.fft","D01.fft","R03.bands","D01.bands","R03.envelope_2","D01.envelope_2","R03.envelope_3","D01.envelope_3","R03.envelope_5","D01.envelope_5","R03.kurtosis_bank","D01.kurtosis_bank"],"similar":false,"similarities":["The uncalibrated baseline Welch distance to R03 is relatively small (4.240), but the observable peak, modulation, and impulsiveness structures conflict."]},{"condition_id":"C04","differences":["R04 has strong 44.9/74.8 Hz lines, 92.46% of power below 100 Hz, and an intermittent 40\u201380 Hz ridge; Q01 has 92.0/109.8/452 Hz lines, only 39.06% below 100 Hz, and ridge excursions to 450 Hz.","R04's envelope bands emphasize about 44.9, 29.9, and 30.4 Hz, unlike Q01's 109.8, 3.27, 166.1, and 123.2 Hz maxima."],"evidence_ids":["R04.fft","D01.fft","R04.bands","D01.bands","R04.ridge","D01.ridge","R04.envelope_1","D01.envelope_1","R04.envelope_2","D01.envelope_2","R04.envelope_4","D01.envelope_4"],"similar":false,"similarities":["Both show detected level changes and relatively modest filter-bank kurtosis, but timing and spectral structure differ."]},{"condition_id":"C05","differences":["R05 is dominated by 43.9 Hz and has 95.86% below 100 Hz; Q01 is dominated by 92.0/109.8 Hz and has more power at 100\u20134000 Hz than below 100 Hz.","R05 has a very strong approximately 30 Hz envelope-band-4 family (0.487), whereas Q01 peaks at 123.16 Hz with only 0.105."],"evidence_ids":["R05.fft","D01.fft","R05.bands","D01.bands","R05.envelope_4","D01.envelope_4","R05.persistence","D01.persistence"],"similar":false,"similarities":["The modulation-map numerical control is nearest R05, but the map summarizes correlated envelope views and its principal bright-cell structure differs."]},{"condition_id":"C06","differences":["Although the Welch control ranks R06 nearest, R06 remains 92.12% below 100 Hz with a 44.9 Hz primary and stable 40\u201350 Hz ridge; Q01 is 39.06% below 100 Hz with 92.0/109.8 Hz primaries and a 110 Hz median ridge reaching 450 Hz.","R06 has strong 15/30 Hz modulation and markedly higher filter-bank impulsiveness (median 0.352, maximum 1.043) than Q01."],"evidence_ids":["R06.welch","D01.welch","R06.bands","D01.bands","R06.fft","D01.fft","R06.ridge","D01.ridge","R06.envelope_1","D01.envelope_1","R06.envelope_2","D01.envelope_2","R06.kurtosis_bank","D01.kurtosis_bank"],"similar":false,"similarities":["R06 is the forced nearest baseline control (4.107 dB RMS), and both show some approximately 30/61 Hz components; neither observation supplies an acceptance threshold."]},{"condition_id":"C07","differences":["R07 has a dominant 45.9 Hz line, 95.13% below 100 Hz, autocorrelation 0.979 at 0.0654 s, and a nearly fixed 40\u201350 Hz ridge; Q01 differs strongly on every item.","R07's envelope bands emphasize 15.31, 62.35, and 187.04 Hz, unlike Q01's 109.80, 123.16, and 61.63 Hz maxima."],"evidence_ids":["R07.fft","D01.fft","R07.bands","D01.bands","R07.autocorrelation","D01.autocorrelation","R07.ridge","D01.ridge","R07.envelope_1","D01.envelope_1","R07.envelope_4","D01.envelope_4","R07.envelope_5","D01.envelope_5"],"similar":false,"similarities":["Both have low or near-zero median filter-bank excess kurtosis, but their frequency and modulation structures are dissimilar."]},{"condition_id":"C08","differences":["R08 has 96.995% below 100 Hz, a 44.6 Hz primary, and autocorrelation 0.980 at 0.0672 s; Q01 has 39.06% below 100 Hz, 92.0/109.8 Hz primaries, and correlation 0.688.","R08's strong 30.31 Hz band-4 modulation and 181.73 Hz band-5 modulation are absent as Q01 maxima."],"evidence_ids":["R08.bands","D01.bands","R08.fft","D01.fft","R08.autocorrelation","D01.autocorrelation","R08.envelope_4","D01.envelope_4","R08.envelope_5","D01.envelope_5"],"similar":false,"similarities":["Both have modest filter-bank kurtosis and some approximately 61 Hz envelope content, but these weak similarities do not overcome the spectral and periodicity differences."]}],"condition_id":null,"explanation":"Q01 is outside the supplied reference set. Its power distribution, principal lines, ridge behavior, and weak/irregular autocorrelation differ substantially from every reference. C01/R01 is the closest visible acquisition regime in several broadband views, while the uncalibrated Welch control ranks C06/R06; neither forms a coherent accepted match. This is an exploratory rejection, not an identification of Q01's actual condition, component, or severity.","outcome":"different"},"decision_basis":"The decision is driven primarily by observable numeric evidence rather than pixel estimates: Q01 has only 39.06% of Welch power below 100 Hz and 48.22% at 100\u2013500 Hz, centroid 369.8 Hz, strongest FFT peaks at 92.0 and 109.8 Hz, autocorrelation coefficient 0.688, and a 110 Hz median ridge with 450 Hz excursions. All references are materially different: even broadest R01 has 68.27% below 100 Hz, centroid 212.5 Hz, a 45.9 Hz primary, and a 50 Hz median ridge; R02\u2013R08 have 92.12\u201397.25% below 100 Hz and strong approximately 45 Hz primaries. Numerical distance controls are conflicting and uncalibrated, so they do not drive acceptance.","decisive_diagrams":["bands","fft","autocorrelation","ridge","welch"],"diagram_review":{"autocorrelation":{"effect":"supports","evidence_ids":["R01.autocorrelation","R02.autocorrelation","R03.autocorrelation","R04.autocorrelation","R05.autocorrelation","R06.autocorrelation","R07.autocorrelation","R08.autocorrelation","D01.autocorrelation"],"favored_condition":"C01","finding":"Q01's strongest nonzero peak is at 0.06405 s but only 0.688 and the curve is irregular. R02\u2013R08 are strongly sinusoidal with 0.971\u20130.983 peaks near 0.0654\u20130.0684 s; R01 is also stronger at 0.905 but is the least dissimilar shape. No RPM inference is justified.","priority":"high"},"bands":{"effect":"supports","evidence_ids":["R01.bands","R02.bands","R03.bands","R04.bands","R05.bands","R06.bands","R07.bands","R08.bands","D01.bands"],"favored_condition":"C01","finding":"Q01 has 39.06% below 100 Hz and 48.22% at 100\u2013500 Hz, outside R01's already broad 68.27%/26.65% and far outside R02\u2013R08's 92.12\u201397.25%/2.53\u20137.11%. Identical Welch bands make this decisive outside-reference evidence.","priority":"high"},"cepstrum":{"effect":"neutral","evidence_ids":["R01.cepstrum","R02.cepstrum","R03.cepstrum","R04.cepstrum","R05.cepstrum","R06.cepstrum","R07.cepstrum","R08.cepstrum","D01.cepstrum"],"favored_condition":"C01","finding":"Q01's first-bin magnitude 0.807 exceeds R01's 0.526 and all others, but all are dominated by floor-sensitive near-zero quefrency structure without a validated nonzero spacing peak. It weakly resembles broad-spectrum R01 but is not decisive.","priority":"low"},"cyclic":{"effect":"neutral","evidence_ids":["R01.cyclic","R02.cyclic","R03.cyclic","R04.cyclic","R05.cyclic","R06.cyclic","R07.cyclic","R08.cyclic","D01.cyclic"],"favored_condition":"C06","finding":"Q01's maximum cyclic coherence is 0.168, below every reference (0.245\u20130.699), while its valid-cell count equals R01's 62,565. Unequal masking and deterministic-tone coherence prevent a reliable class match.","priority":"low"},"envelope":{"effect":"supports","evidence_ids":["R01.envelope","R02.envelope","R03.envelope","R04.envelope","R05.envelope","R06.envelope","R07.envelope","R08.envelope","D01.envelope"],"favored_condition":"C01","finding":"Q01 emphasizes 17.88, 360.10, 109.80, and 61.31 Hz; no reference has this leading family. R01 is relatively closest only through mixed broadband modulation, not matching peaks.","priority":"medium"},"envelope_1":{"effect":"supports","evidence_ids":["R01.envelope_1","R02.envelope_1","R03.envelope_1","R04.envelope_1","R05.envelope_1","R06.envelope_1","R07.envelope_1","R08.envelope_1","D01.envelope_1"],"favored_condition":"C05","finding":"Q01's 100\u2013500 Hz envelope peaks at 109.80 Hz (0.235). R05's 102.45 Hz maximum is the nearest family, but Q01 also has strong 0.10/0.82 Hz components and does not reproduce R05's detailed pattern.","priority":"medium"},"envelope_2":{"effect":"neutral","evidence_ids":["R01.envelope_2","R02.envelope_2","R03.envelope_2","R04.envelope_2","R05.envelope_2","R06.envelope_2","R07.envelope_2","R08.envelope_2","D01.envelope_2"],"favored_condition":"C01","finding":"Q01's 500\u20131500 Hz maximum is 3.27 Hz (0.088), with 30.61 Hz secondary energy. No reference shares the maximum; the numerical control slightly favors R02, illustrating ambiguity.","priority":"low"},"envelope_3":{"effect":"supports","evidence_ids":["R01.envelope_3","R02.envelope_3","R03.envelope_3","R04.envelope_3","R05.envelope_3","R06.envelope_3","R07.envelope_3","R08.envelope_3","D01.envelope_3"],"favored_condition":"C07","finding":"Q01's 1500\u20134000 Hz maximum at 166.12 Hz is absent from reference maxima. Its low amplitude resembles several references, but the modulation family is outside the set.","priority":"low"},"envelope_4":{"effect":"supports","evidence_ids":["R01.envelope_4","R02.envelope_4","R03.envelope_4","R04.envelope_4","R05.envelope_4","R06.envelope_4","R07.envelope_4","R08.envelope_4","D01.envelope_4"],"favored_condition":"C06","finding":"Q01 peaks at 123.16 Hz with secondary 246.33/61.63 Hz and a modest 0.105 maximum. References peak near 30, 61, 62, or 92 Hz; R06 contains a strong 122.24 Hz secondary but peaks at 61.12 Hz.","priority":"medium"},"envelope_5":{"effect":"neutral","evidence_ids":["R01.envelope_5","R02.envelope_5","R03.envelope_5","R04.envelope_5","R05.envelope_5","R06.envelope_5","R07.envelope_5","R08.envelope_5","D01.envelope_5"],"favored_condition":"C01","finding":"Q01's 8000\u201316000 Hz maximum is 61.63 Hz, close to R01's 61.73 Hz, but Q01's normalized amplitude is much smaller (0.057 versus 0.265) and its associated family differs. This is isolated counterevidence to rejection.","priority":"low"},"estimators":{"effect":"supports","evidence_ids":["R01.estimators","R02.estimators","R03.estimators","R04.estimators","R05.estimators","R06.estimators","R07.estimators","R08.estimators","D01.estimators"],"favored_condition":"C01","finding":"Q01's multitaper power 0.055 is nearest R01's 0.0367, and both have elevated broadband floors; however Q01's strongest smoothed region is 90\u2013110 Hz rather than R01's 40\u201370 Hz. Absolute power is acquisition-sensitive.","priority":"medium"},"fft":{"effect":"supports","evidence_ids":["R01.fft","R02.fft","R03.fft","R04.fft","R05.fft","R06.fft","R07.fft","R08.fft","D01.fft"],"favored_condition":"C01","finding":"Q01's strongest lines are 92.0 and 109.8 Hz, followed by 452.0 and 219.6 Hz. Every reference has its strongest line near 44\u201346 Hz; even broad R01 retains a 45.9 Hz primary. This directly contradicts every reference peak hierarchy.","priority":"high"},"harmonics":{"effect":"supports","evidence_ids":["R01.harmonics","R02.harmonics","R03.harmonics","R04.harmonics","R05.harmonics","R06.harmonics","R07.harmonics","R08.harmonics","D01.harmonics"],"favored_condition":"C05","finding":"Q01's largest spacing count is 61 Hz (9), unlike all references whose largest family is near 15 Hz (11\u201324). Peak-pair spacing can mix unrelated tones, so this supports rejection only moderately.","priority":"medium"},"kurtosis_bank":{"effect":"contradicts","evidence_ids":["R01.kurtosis_bank","R02.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank","R05.kurtosis_bank","R06.kurtosis_bank","R07.kurtosis_bank","R08.kurtosis_bank","D01.kurtosis_bank"],"favored_condition":"C08","finding":"Q01 has modest filter-bank impulsiveness (median 0.021, maximum 0.183), closest in level to R08 (0.021/0.258) and unlike R03/R06. This isolated resemblance conflicts with rejection but does not match spectral structure or establish bearing fault identity.","priority":"medium"},"levels":{"effect":"neutral","evidence_ids":["R01.levels","R02.levels","R03.levels","R04.levels","R05.levels","R06.levels","R07.levels","R08.levels","D01.levels"],"favored_condition":"C01","finding":"Q01 is unclipped and nonsilent, has crest factor 11.86 dB and two late level changes; R01 has a similar 11.93 dB crest factor but a different DC offset and mid-record disturbance. Peak normalization prevents acoustic-level or severity conclusions.","priority":"low"},"modulation_map":{"effect":"neutral","evidence_ids":["R01.modulation_map","R02.modulation_map","R03.modulation_map","R04.modulation_map","R05.modulation_map","R06.modulation_map","R07.modulation_map","R08.modulation_map","D01.modulation_map"],"favored_condition":"C07","finding":"Q01's map maximum 0.057 is below every reference and closest to R07's 0.078, but its bright-cell locations differ. Each map has 1,100 valid cells; masked cells are missing, and this view duplicates band-envelope evidence.","priority":"low"},"orders":{"effect":"unavailable","evidence_ids":["R01.orders","R02.orders","R03.orders","R04.orders","R05.orders","R06.orders","R07.orders","R08.orders","D01.orders"],"favored_condition":null,"finding":"Unavailable for Q01 and every reference because no independent RPM trace was supplied; frequency evidence cannot be converted into validated orders.","priority":"unavailable"},"persistence":{"effect":"contradicts","evidence_ids":["R01.persistence","R02.persistence","R03.persistence","R04.persistence","R05.persistence","R06.persistence","R07.persistence","R08.persistence","D01.persistence"],"favored_condition":"C01","finding":"Q01 and R01 share the same reported maximum occupancy, 50.38%, and both show broad level populations. Q01's populations are shifted toward higher frequencies. Chronology is discarded and overlapping frames are dependent.","priority":"medium"},"reassigned":{"effect":"contradicts","evidence_ids":["R01.reassigned","R02.reassigned","R03.reassigned","R04.reassigned","R05.reassigned","R06.reassigned","R07.reassigned","R08.reassigned","D01.reassigned"],"favored_condition":"C01","finding":"Q01's broad reassigned background and retained weight 0.99970 most resemble R01's broad regime, though Q01's median pooled weight is much higher (-71.85 versus -76.01 dB) and its ridge frequencies differ. This is correlated with PSD/STFT evidence.","priority":"low"},"ridge":{"effect":"supports","evidence_ids":["R01.ridge","R02.ridge","R03.ridge","R04.ridge","R05.ridge","R06.ridge","R07.ridge","R08.ridge","D01.ridge"],"favored_condition":"C01","finding":"Q01 has median dominant bin 110 Hz and repeated 450 Hz selections, versus R01 median 50 Hz and maximum 220 Hz and R02\u2013R08 almost entirely 40\u201350 Hz except R04 to 80 Hz. At 10 Hz bins, Q01 remains clearly outside all references.","priority":"high"},"spectral_kurtosis":{"effect":"neutral","evidence_ids":["R01.spectral_kurtosis","R02.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis","R05.spectral_kurtosis","R06.spectral_kurtosis","R07.spectral_kurtosis","R08.spectral_kurtosis","D01.spectral_kurtosis"],"favored_condition":"C06","finding":"Q01's maximum 0.634 at 11.83 kHz is within the broad reference range and nearest R06's 0.600 in magnitude, but peak bands differ. Unequal exclusions, finite-sample bias, and no significance threshold make this nondiagnostic.","priority":"low"},"stft":{"effect":"supports","evidence_ids":["R01.stft","R02.stft","R03.stft","R04.stft","R05.stft","R06.stft","R07.stft","R08.stft","D01.stft"],"favored_condition":"C01","finding":"Q01 is broadly stationary but has a much wider low/mid-frequency background and late amplitude disturbance. R01 is the broadest reference and thus relatively closest, while R02\u2013R08 are dominated by persistent low-frequency lines.","priority":"medium"},"stft_detail":{"effect":"supports","evidence_ids":["R01.stft_detail","R02.stft_detail","R03.stft_detail","R04.stft_detail","R05.stft_detail","R06.stft_detail","R07.stft_detail","R08.stft_detail","D01.stft_detail"],"favored_condition":"C01","finding":"Q01's median displayed PSD is -84.18 dB, above R01's -88.33 dB and R02\u2013R08's -90.35 to -94.37 dB, with broader 100\u2013500 Hz energy. The 10 Hz resolution confirms an outside-reference background rather than a matching line pattern.","priority":"medium"},"synchronous":{"effect":"unavailable","evidence_ids":["R01.synchronous","R02.synchronous","R03.synchronous","R04.synchronous","R05.synchronous","R06.synchronous","R07.synchronous","R08.synchronous","D01.synchronous"],"favored_condition":null,"finding":"Unavailable for Q01 and every reference because no independent RPM trace was supplied; phase-locked waveform evidence cannot be assessed.","priority":"unavailable"},"wavelet":{"effect":"contradicts","evidence_ids":["R01.wavelet","R02.wavelet","R03.wavelet","R04.wavelet","R05.wavelet","R06.wavelet","R07.wavelet","R08.wavelet","D01.wavelet"],"favored_condition":"C01","finding":"Q01's maximum wavelet power (-35.31 dB) nearly equals R01 (-35.31 dB), and both are broader than most references, but Q01's median is higher (-55.42 versus -60.65 dB) with stronger time-varying bands. Edge-masked cells remain unavailable.","priority":"low"},"welch":{"effect":"supports","evidence_ids":["R01.welch","R02.welch","R03.welch","R04.welch","R05.welch","R06.welch","R07.welch","R08.welch","D01.welch"],"favored_condition":"C01","finding":"Q01 has centroid 369.76 Hz and flatness 0.00397, beyond R01's 212.52 Hz/0.00307 and far beyond R02\u2013R08's 51.85\u201365.85 Hz/0.000076\u20130.000225. The 61.5 Hz Welch bandwidth limits line resolution but not this large shape difference.","priority":"high"}},"limitations":"Only one peak-normalized recording represents each condition, so condition effects cannot be separated from microphone response, placement, reverberation, speed, load, or acquisition regime. Correlated views are not independent votes. No calibrated acceptance threshold or distance confidence interval is supplied. Orders and synchronous averages are unavailable; cyclic and modulation maps contain masked missing cells; wavelet edges are masked; spectral-kurtosis exclusions differ. The result does not infer Q01's true label and makes no fault-component or severity claim.","most_informative":["bands","fft","autocorrelation","ridge","welch"],"priority_changes":"Compared with the frozen survey, bands, FFT, autocorrelation, and ridge remain high priority. Welch is promoted into the decisive set because Q01's centroid/flatness and broadband shape lie beyond every reference. Envelope_4 and kurtosis_bank are demoted from the original top five: Q01's envelope maxima do not form a coherent reference match, while modest kurtosis resembles R08 but conflicts with the decisive spectral evidence. Numerical controls do not replace these observations: baseline favors R06, whereas several visual/acquisition-sensitive views favor R01, confirming ambiguity rather than acceptance."},"stage":"decision","survey":"survey-8ca096c2c8dd560f","usage":{"completion_tokens":6118,"completion_tokens_details":{"accepted_prediction_tokens":0,"audio_tokens":0,"reasoning_tokens":164,"rejected_prediction_tokens":0},"latency_checkpoint":{"engine_tbt_ms":8,"engine_ttft_ms":913,"engine_ttlt_ms":49292,"pre_inference_ms":2999,"service_tbt_ms":8,"service_ttft_ms":4127,"service_ttlt_ms":52418,"user_visible_ttft_ms":1128},"prompt_tokens":84843,"prompt_tokens_details":{"audio_tokens":0,"cached_tokens":0},"total_tokens":90961}},{"baseline":{"category":"Correct rejection","dataset":"Ottawa","name":"Ottawa-excluded-D01","phase":"excluded","predicted":"different","query_id":"D01"},"category":"Correct rejection","contract_sha256":"6ee90b47cdb503664853071168f3afd4496d4fd54922ff15ca4a407d4b800c4c","dataset":"Ottawa","expected":"different","name":"Ottawa-excluded-D01","original_sha256":"2cad2bc176c894d9c18ad36493e6df6f23bd0809728f2b95096fc9f97157c54b","phase":"excluded","predicted":"different","query_id":"D01","result":{"counterevidence":"The strongest counterevidence is forced nearest-reference ranking: R06 is nearest on the baseline Welch control (4.1067 dB RMS), cepstrum, cyclic, spectral kurtosis and wavelet, while several modulation controls favor R04/R05/R07. These rankings conflict across views and lack an acceptance threshold. Q01 also retains a roughly 0.064 s autocorrelation peak and some approximately 30/61 Hz spectral or modulation components seen in references, but its correlation is much weaker and its dominant spectrum is fundamentally different [Q01.C01.S0001, D01.autocorrelation, D01.fft; R02.autocorrelation-R08.autocorrelation, R02.fft-R08.fft].","decision":{"comparisons":[{"condition_id":"C02","differences":["Q01 has only 39.06% below 100 Hz versus 97.25% for R02, and its dominant FFT lines are 92.0/109.8/452.0 Hz rather than R02's 45.1 Hz-led family.","Q01 autocorrelation is much weaker (0.6877 versus 0.9786), and its ridge median is 110 Hz with excursions to 450 Hz rather than R02's 40/50 Hz ridge."],"evidence_ids":["Q01.C01.S0001","D01.bands","D01.fft","D01.autocorrelation","D01.ridge","R02.C01.S0001","R02.bands","R02.fft","R02.autocorrelation","R02.ridge"],"similar":false,"similarities":["Both contain approximately 30 and 61 Hz components and are unclipped."]},{"condition_id":"C03","differences":["Q01's band distribution, centroid 369.76 Hz and flatness 0.00397 are far outside R03's low-frequency-dominated spectrum, centroid 54.59 Hz and flatness 0.000192.","R03 has strong 15 Hz modulation and broadly high-frequency kurtosis, whereas Q01's envelope leaders and low filter-bank kurtosis differ."],"evidence_ids":["Q01.C01.S0001","D01.bands","D01.welch","D01.envelope_3","D01.kurtosis_bank","R03.C01.S0001","R03.bands","R03.welch","R03.envelope_3","R03.kurtosis_bank"],"similar":false,"similarities":["The numerical estimator and persistence controls rank R03 nearest, but no calibrated threshold makes that acceptance evidence."]},{"condition_id":"C04","differences":["R04 is dominated by 44.9 and 74.8 Hz and switches among 40/50/80 Hz ridges; Q01 is led by 92.0/109.8/452.0 Hz with median ridge 110 Hz and 450 Hz episodes.","Q01 has 48.22% in 100\u2013500 Hz versus 7.18% for R04 and lacks R04's strongly low-frequency concentration."],"evidence_ids":["Q01.C01.S0001","D01.fft","D01.ridge","D01.bands","R04.C01.S0001","R04.fft","R04.ridge","R04.bands"],"similar":false,"similarities":["R04 is nearest on kurtosis-bank and modulation-map numerical controls, but Q01's visible maps and absolute band distribution remain unlike R04."]},{"condition_id":"C05","differences":["R05's spectrum is led by 43.9 Hz and has 95.86% below 100 Hz; Q01 is led above 90 Hz and has only 39.06% below 100 Hz.","R05 has a dominant approximately 30 Hz 4\u20138 kHz envelope component (0.4869), while Q01 leads at 123.16 Hz with only 0.1051."],"evidence_ids":["Q01.C01.S0001","D01.fft","D01.bands","D01.envelope_4","R05.C01.S0001","R05.fft","R05.bands","R05.envelope_4"],"similar":false,"similarities":["The modulation-map and envelope_5 controls rank R05 closely, but individual carrier-band leaders conflict."]},{"condition_id":"C06","differences":["Although R06 is the forced baseline nearest, Q01's Welch centroid is 369.76 Hz versus 62.04 Hz, integrated power 0.0553 versus 0.1136 FS\u00b2, and flatness 0.00397 versus 0.000225.","Q01's ridge median is 110 Hz with 450 Hz switches; R06 remains at 40/50 Hz. Q01's autocorrelation coefficient 0.6877 is also far below R06's 0.9714."],"evidence_ids":["Q01.C01.S0001","D01.welch","D01.ridge","D01.autocorrelation","R06.C01.S0001","R06.welch","R06.ridge","R06.autocorrelation"],"similar":false,"similarities":["R06 is nearest in the baseline, cepstrum, cyclic, spectral-kurtosis and wavelet controls; both show reported level changes."]},{"condition_id":"C07","differences":["R07 is dominated by 45.9 Hz with a nearly fixed 50 Hz ridge, unlike Q01's 92/109.8/452 Hz spectrum and 110/450 Hz ridge behavior.","Q01 has much more 500\u20134000 Hz relative power and far weaker autocorrelation than R07."],"evidence_ids":["Q01.C01.S0001","D01.fft","D01.ridge","D01.bands","D01.autocorrelation","R07.C01.S0001","R07.fft","R07.ridge","R07.bands","R07.autocorrelation"],"similar":false,"similarities":["Both show a near-61 Hz modulation component, and envelope_5 is numerically closest to R07 despite different leading frequencies."]},{"condition_id":"C08","differences":["R08 has 96.995% below 100 Hz and a 44.6 Hz dominant line; Q01 has 39.06% below 100 Hz and dominant lines at 92.0, 109.8 and 452.0 Hz.","R08's 4\u20138 kHz envelope is strongly led near 30 Hz (0.4357), unlike Q01's weaker 123.16 Hz leader (0.1051)."],"evidence_ids":["Q01.C01.S0001","D01.bands","D01.fft","D01.envelope_4","R08.C01.S0001","R08.bands","R08.fft","R08.envelope_4"],"similar":false,"similarities":["Both have relatively low filter-bank kurtosis compared with R03/R06 and contain approximately 30/60 Hz components."]}],"condition_id":null,"explanation":"Q01 is substantially outside every supplied reference in the decisive observable structure: it shifts power from below 100 Hz into 100\u20134000 Hz, has a 369.76 Hz Welch centroid and 0.00397 flatness, lacks the references' dominant approximately 44\u201346 Hz line, has much weaker periodic autocorrelation, and shows a 110 Hz median ridge with intermittent 450 Hz dominance [Q01.C01.S0001, D01.bands, D01.fft, D01.welch, D01.autocorrelation, D01.ridge]. All references remain approximately 92\u201397% below 100 Hz, have centroids 51.85\u201365.85 Hz, strong approximately 44\u201346 Hz fundamentals, autocorrelation near 0.97\u20130.98 and mainly 40/50 Hz ridges [R02.C01.S0001-R08.C01.S0001, R02.bands-R08.bands, R02.fft-R08.fft, R02.welch-R08.welch, R02.autocorrelation-R08.autocorrelation, R02.ridge-R08.ridge]. Therefore nearest-reference controls do not establish acceptance.","outcome":"different"},"decision_basis":"The decision is driven by query/reference numeric measurements and corresponding visible spectral, band-power, autocorrelation and ridge observations\u2014not by numerical nearest-reference controls. The controls are inconsistent across views and provide forced rankings only.","decisive_diagrams":["fft","bands","welch","autocorrelation","ridge"],"diagram_review":{"autocorrelation":{"effect":"supports","evidence_ids":["D01.autocorrelation","Q01.C01.S0001","R02.autocorrelation","R03.autocorrelation","R04.autocorrelation","R05.autocorrelation","R06.autocorrelation","R07.autocorrelation","R08.autocorrelation"],"favored_condition":null,"finding":"Q01's strongest positive-lag peak is 0.06405 s but only 0.6877, versus 0.9714\u20130.9828 for every reference; periodicity is therefore visibly and numerically much weaker.","priority":"high"},"bands":{"effect":"supports","evidence_ids":["D01.bands","Q01.C01.S0001","R02.bands","R03.bands","R04.bands","R05.bands","R06.bands","R07.bands","R08.bands"],"favored_condition":null,"finding":"Q01 has 39.06% at 0\u2013100 Hz and 48.22% at 100\u2013500 Hz; every reference has 92.12\u201397.25% below 100 Hz and at most 7.18% at 100\u2013500 Hz. This is outside all references under identical bands.","priority":"high"},"cepstrum":{"effect":"supports","evidence_ids":["D01.cepstrum","R02.cepstrum","R03.cepstrum","R04.cepstrum","R05.cepstrum","R06.cepstrum","R07.cepstrum","R08.cepstrum"],"favored_condition":null,"finding":"Q01's first-bin maximum is 0.8065 versus 0.0984\u20130.2056 for references, with additional short-quefrency structure; it does not establish a spacing or RPM.","priority":"medium"},"cyclic":{"effect":"neutral","evidence_ids":["D01.cyclic","R02.cyclic","R03.cyclic","R04.cyclic","R05.cyclic","R06.cyclic","R07.cyclic","R08.cyclic"],"favored_condition":"C06","finding":"Q01 has low maximum squared coherence 0.1685 and 62,565 valid cells; R06/R08 are nearest numerically, but all reference maxima are higher (0.2446\u20130.6995), masked-cell populations differ, and no significance threshold exists.","priority":"low"},"envelope":{"effect":"supports","evidence_ids":["D01.envelope","Q01.C01.S0001","R02.envelope","R03.envelope","R04.envelope","R05.envelope","R06.envelope","R07.envelope","R08.envelope"],"favored_condition":null,"finding":"Q01's broadband leader is 17.88 Hz with other strong 360.10/109.80/61.31 Hz components; references lead near 14.95, 29.9\u201330.6 or 44.65 Hz, so no close broadband-envelope match appears.","priority":"medium"},"envelope_1":{"effect":"neutral","evidence_ids":["D01.envelope_1","R02.envelope_1","R03.envelope_1","R04.envelope_1","R05.envelope_1","R06.envelope_1","R07.envelope_1","R08.envelope_1"],"favored_condition":"C05","finding":"Q01's 100\u2013500 Hz leader is 109.80 Hz/0.2348; it lies between R05's 102.45 Hz and R08's 118.98 Hz patterns, while the numerical control slightly favors R04. Evidence is ambiguous.","priority":"low"},"envelope_2":{"effect":"supports","evidence_ids":["D01.envelope_2","R02.envelope_2","R03.envelope_2","R04.envelope_2","R05.envelope_2","R06.envelope_2","R07.envelope_2","R08.envelope_2"],"favored_condition":"C02","finding":"Q01's 500\u20131500 Hz leader is 3.27 Hz/0.0883; only R02 similarly emphasizes very-low modulation, but R02 leads at 0.51 Hz/0.1229 and the full signatures are not close.","priority":"low"},"envelope_3":{"effect":"supports","evidence_ids":["D01.envelope_3","R02.envelope_3","R03.envelope_3","R04.envelope_3","R05.envelope_3","R06.envelope_3","R07.envelope_3","R08.envelope_3"],"favored_condition":null,"finding":"Q01's 1500\u20134000 Hz leader is 166.12 Hz; all references lead at 0.51\u201358.57 Hz, placing the principal modulation outside the supplied patterns.","priority":"medium"},"envelope_4":{"effect":"supports","evidence_ids":["D01.envelope_4","R02.envelope_4","R03.envelope_4","R04.envelope_4","R05.envelope_4","R06.envelope_4","R07.envelope_4","R08.envelope_4"],"favored_condition":"C06","finding":"Q01 leads at 123.16 Hz with a modest 0.1051 amplitude; references lead near 30, 61\u201362 or 91.73 Hz and generally show sharper peaks. R06 is only a weak relative resemblance through its 122.24 Hz secondary component.","priority":"medium"},"envelope_5":{"effect":"supports","evidence_ids":["D01.envelope_5","R02.envelope_5","R03.envelope_5","R04.envelope_5","R05.envelope_5","R06.envelope_5","R07.envelope_5","R08.envelope_5"],"favored_condition":"C06","finding":"Q01 leads at 61.63 Hz/0.0571; reference leaders are approximately 15, 89.8, 181.7\u2013187.0 or 274 Hz. R06 shares a 59.8 Hz secondary but is dominated by 14.9 Hz.","priority":"medium"},"estimators":{"effect":"supports","evidence_ids":["D01.estimators","R02.estimators","R03.estimators","R04.estimators","R05.estimators","R06.estimators","R07.estimators","R08.estimators"],"favored_condition":"C03","finding":"Q01's DPSS maximum is 90 Hz and integrated power 0.0553, whereas references peak at 40\u201360 Hz with 0.1136\u20130.2161 power. R03 is numerically nearest only under the uncalibrated RMS control.","priority":"medium"},"fft":{"effect":"supports","evidence_ids":["D01.fft","Q01.C01.S0001","R02.fft","R03.fft","R04.fft","R05.fft","R06.fft","R07.fft","R08.fft"],"favored_condition":null,"finding":"At the common 0.1 Hz bins/0.15 Hz Hann bandwidth, Q01 is led by 92.0, 109.8 and 452.0 Hz; every reference is led by approximately 43.9\u201345.9 Hz. This is decisive outside-reference evidence.","priority":"high"},"harmonics":{"effect":"supports","evidence_ids":["D01.harmonics","R02.harmonics","R03.harmonics","R04.harmonics","R05.harmonics","R06.harmonics","R07.harmonics","R08.harmonics"],"favored_condition":null,"finding":"Q01's strongest pair spacing is 61 Hz with count 9, unlike every reference's approximately 15 Hz leader with counts 11\u201324. Peak-pair spacing remains non-diagnostic by itself.","priority":"medium"},"kurtosis_bank":{"effect":"neutral","evidence_ids":["D01.kurtosis_bank","R02.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank","R05.kurtosis_bank","R06.kurtosis_bank","R07.kurtosis_bank","R08.kurtosis_bank"],"favored_condition":"C08","finding":"Q01 has low, fairly uniform impulsiveness (maximum 0.1826, median 0.0208), nearest in level to R08 (0.2579/0.0214) but with a different peak band; this weak resemblance cannot identify a condition.","priority":"medium"},"levels":{"effect":"supports","evidence_ids":["D01.levels","Q01.C01.S0001","R02.levels","R03.levels","R04.levels","R05.levels","R06.levels","R07.levels","R08.levels"],"favored_condition":"C04","finding":"Q01 is unclipped but has RMS 0.2426 FS, crest factor 11.86 dB and two late >6 dB changes; reference RMS is 0.3375\u20130.4680 and only R04/R06 report changes. Peak normalization prevents acoustic-level or severity claims.","priority":"medium"},"modulation_map":{"effect":"neutral","evidence_ids":["D01.modulation_map","R02.modulation_map","R03.modulation_map","R04.modulation_map","R05.modulation_map","R06.modulation_map","R07.modulation_map","R08.modulation_map"],"favored_condition":"C05","finding":"Q01's map is diffuse with maximum 0.0570, below all references (0.0784\u20130.4404). The numerical control favors R05, but Q01 lacks R05's strong 4\u20138 kHz approximately 30 Hz cell.","priority":"medium"},"orders":{"effect":"unavailable","evidence_ids":["D01.orders","R02.orders","R03.orders","R04.orders","R05.orders","R06.orders","R07.orders","R08.orders"],"favored_condition":null,"finding":"Unavailable for Q01 and every reference because no independent RPM trace was supplied.","priority":"unavailable"},"persistence":{"effect":"supports","evidence_ids":["D01.persistence","R02.persistence","R03.persistence","R04.persistence","R05.persistence","R06.persistence","R07.persistence","R08.persistence"],"favored_condition":"C03","finding":"Q01 maximum occupancy is 50.38%, below all references except near R03's 54.66%, and its persistent energy extends more broadly upward in frequency. Chronology is discarded.","priority":"medium"},"reassigned":{"effect":"supports","evidence_ids":["D01.reassigned","R02.reassigned","R03.reassigned","R04.reassigned","R05.reassigned","R06.reassigned","R07.reassigned","R08.reassigned"],"favored_condition":"C03","finding":"Q01 localizes a broad higher-frequency structure with median \u221271.85 dB, versus reference medians around \u221278.4 to \u221282.6 dB and their low-frequency line dominance. This is correlated with FFT/STFT, not an independent vote.","priority":"low"},"ridge":{"effect":"supports","evidence_ids":["D01.ridge","R02.ridge","R03.ridge","R04.ridge","R05.ridge","R06.ridge","R07.ridge","R08.ridge"],"favored_condition":null,"finding":"Q01's dominant ridge spans 30\u2013450 Hz, median 110 Hz, with repeated 450 Hz episodes; references are confined mainly to 40/50 Hz, except R04 reaching 80 Hz. This is decisive separation.","priority":"high"},"spectral_kurtosis":{"effect":"neutral","evidence_ids":["D01.spectral_kurtosis","R02.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis","R05.spectral_kurtosis","R06.spectral_kurtosis","R07.spectral_kurtosis","R08.spectral_kurtosis"],"favored_condition":"C06","finding":"Q01 peaks at 11.83 kHz/0.634 with only two excluded bins; references peak at disparate 0.62\u201320.83 kHz locations. R06 is nearest numerically, but finite-sample bias and differing exclusions limit interpretation.","priority":"low"},"stft":{"effect":"supports","evidence_ids":["D01.stft","R02.stft","R03.stft","R04.stft","R05.stft","R06.stft","R07.stft","R08.stft"],"favored_condition":null,"finding":"Q01 shows broader elevated energy and time-varying low/mid-frequency structure rather than the references' predominantly persistent approximately 45 Hz-led horizontal pattern.","priority":"medium"},"stft_detail":{"effect":"supports","evidence_ids":["D01.stft_detail","R02.stft_detail","R03.stft_detail","R04.stft_detail","R05.stft_detail","R06.stft_detail","R07.stft_detail","R08.stft_detail"],"favored_condition":"C03","finding":"At common 0.1 s/10 Hz resolution, Q01 median PSD is \u221284.18 dB versus \u221290.35 to \u221294.37 dB for references and its energy is distributed much more broadly. R03 is merely the nearest RMS control.","priority":"medium"},"synchronous":{"effect":"unavailable","evidence_ids":["D01.synchronous","R02.synchronous","R03.synchronous","R04.synchronous","R05.synchronous","R06.synchronous","R07.synchronous","R08.synchronous"],"favored_condition":null,"finding":"Unavailable for Q01 and every reference because no independent RPM trace was supplied.","priority":"unavailable"},"wavelet":{"effect":"supports","evidence_ids":["D01.wavelet","R02.wavelet","R03.wavelet","R04.wavelet","R05.wavelet","R06.wavelet","R07.wavelet","R08.wavelet"],"favored_condition":"C06","finding":"Q01 has pronounced time-varying mid-frequency bands and median coefficient power \u221255.42 dB, above the reference range \u221259.58 to \u221265.11 dB. R06 is nearest numerically but not visually accepting.","priority":"low"},"welch":{"effect":"supports","evidence_ids":["D01.welch","Q01.C01.S0001","R02.welch","R03.welch","R04.welch","R05.welch","R06.welch","R07.welch","R08.welch"],"favored_condition":"C06","finding":"With identical 41.02 Hz bins/61.52 Hz bandwidth, Q01 has centroid 369.76 Hz, flatness 0.00397 and power 0.0553 FS\u00b2; references have centroids 51.85\u201365.85 Hz, flatness at most 0.000225 and power 0.1136\u20130.2162. R06 is the nearest baseline control but remains substantially different.","priority":"high"}},"limitations":"Only one reference is supplied per condition, so acquisition regime, microphone response/placement, reverberation, speed and load are inseparable from condition effects. Peak normalization removes acoustic-level and severity interpretation but not frequency-response effects. The 26 views are correlated transforms, not independent votes. Masked cyclic cells are missing data; no cyclic significance threshold or calibrated distance threshold is supplied. Orders and synchronous averages are unavailable without RPM. No fault identity, component or severity is inferred, and the frozen survey did not establish predictive accuracy.","most_informative":["fft","bands","welch","autocorrelation","ridge"],"priority_changes":"Compared with the frozen reference-only priorities, autocorrelation and ridge rise to high because Q01 is dramatically outside the reference ranges; Welch remains high; FFT and bands remain high. Modulation map, kurtosis bank and spectral kurtosis are demoted to medium/low because their nearest-reference indications conflict and do not overcome the broad spectral mismatch."},"stage":"decision","survey":"survey-ab7169c4a978e367","usage":{"completion_tokens":5484,"completion_tokens_details":{"accepted_prediction_tokens":0,"audio_tokens":0,"reasoning_tokens":181,"rejected_prediction_tokens":0},"latency_checkpoint":{"engine_tbt_ms":8,"engine_ttft_ms":784,"engine_ttlt_ms":43631,"pre_inference_ms":1983,"service_tbt_ms":8,"service_ttft_ms":3042,"service_ttlt_ms":45885,"user_visible_ttft_ms":1059},"prompt_tokens":76106,"prompt_tokens_details":{"audio_tokens":0,"cached_tokens":0},"total_tokens":81590}},{"baseline":{"category":"Wrong class","dataset":"Ottawa","name":"Ottawa-known-D03","phase":"known","predicted":"C06","query_id":"D03"},"category":"False rejection","contract_sha256":"91c62ad1883dba17bf1c6f3593e72c48bfcf9e22d367e7e5ad114dba0cc824dc","dataset":"Ottawa","expected":"C02","name":"Ottawa-known-D03","original_sha256":"06f395cb020d94e8393a62f239bd48691153eea149c09a7dc91ef5a55a1f6adb","phase":"known","predicted":"different","query_id":"D03","result":{"counterevidence":"Q01 retains the same approximately 0.067 s autocorrelation period as R02\u2013R08, and several normalized secondary features have reference-level analogues: its 29.9 Hz broadband modulation resembles R04/R05, its low filter-bank kurtosis resembles R04/R05/R08, and numerical controls rank R06 nearest in baseline Welch, harmonics, cyclic coherence and envelope_4. These similarities conflict with an entirely novel-signal interpretation, but they do not overcome the persistent 89.6/90 Hz dominant line and the outside-reference 100\u2013500 Hz power fraction.","decision":{"comparisons":[{"condition_id":"C01","differences":["Unlike R01, Q01 has strong long-lag periodicity at 0.06698 s rather than a reported 0.000381 s peak.","Q01 is dominated by 89.6 Hz with a fixed 90 Hz ridge, whereas R01 is led by 45.9 Hz and has broad 40\u2013220 Hz ridge switching.","Q01 lacks R01's large DC offset and mid-record level disturbance."],"evidence_ids":["R01.autocorrelation","D03.autocorrelation","R01.fft","D03.fft","R01.ridge","D03.ridge","R01.levels","D03.levels"],"similar":false,"similarities":["Both allocate unusually little power below 100 Hz relative to R02\u2013R08 and unusually much power to 100\u2013500 Hz, although Q01 is still more concentrated in 100\u2013500 Hz.","Both are broader than R02\u2013R08 in low-frequency band allocation."]},{"condition_id":"C02","differences":["R02 is dominated by 45.1 Hz and a 40\u201350 Hz ridge; Q01 is dominated by 89.6 Hz with a constant 90 Hz ridge.","Q01 has 27.15% of power at 100\u2013500 Hz versus 2.53% for R02.","Q01's strongest high-carrier modulation families do not match R02's 91.7 and 183.5 Hz maxima consistently."],"evidence_ids":["R02.fft","D03.fft","R02.ridge","D03.ridge","R02.bands","D03.bands","R02.envelope_4","D03.envelope_4","R02.envelope_5","D03.envelope_5"],"similar":false,"similarities":["Both have strong approximately 0.067 s autocorrelation recurrence.","Q01's 119.4 Hz envelope_1 peak is closer to neither R02's 90.2 Hz maximum nor its complete family, though both show narrow modulation lines."]},{"condition_id":"C03","differences":["R03 is led by 45.0 Hz with a strong 30 Hz secondary, while Q01 is led by 89.6 Hz and 59.7 Hz.","R03 has conspicuous 15 Hz modulation across several carrier bands and very high filter-bank kurtosis; Q01 instead peaks at 1.12, 358.3, 120.6 and 89.6 Hz by successive higher bands and has low kurtosis."],"evidence_ids":["R03.fft","D03.fft","R03.envelope_2","D03.envelope_2","R03.envelope_3","D03.envelope_3","R03.envelope_5","D03.envelope_5","R03.kurtosis_bank","D03.kurtosis_bank"],"similar":false,"similarities":["Cepstral numerical control is closest to R03, but both cepstra are dominated by floor-dependent near-zero quefrency bins.","Both show approximately 0.067 s autocorrelation recurrence."]},{"condition_id":"C04","differences":["R04's principal FFT lines are 44.9 and 74.8 Hz, whereas Q01 is dominated by 89.6 Hz.","R04 has intermittent 40/50/80 Hz ridge selections and reported level changes; Q01 has an invariant 90 Hz ridge and no reported level changes.","Q01's 100\u2013500 Hz fraction is 27.15%, far above R04's 7.18%."],"evidence_ids":["R04.fft","D03.fft","R04.ridge","D03.ridge","R04.levels","D03.levels","R04.bands","D03.bands"],"similar":false,"similarities":["Q01's 29.90 Hz broadband-envelope maximum resembles R04's 29.90 Hz maximum.","Their filter-bank kurtosis medians are similarly modest positive values."]},{"condition_id":"C05","differences":["R05 is dominated by 43.9 Hz with a 40\u201350 Hz ridge; Q01 is dominated by 89.6 Hz and remains at 90 Hz.","R05 has only 3.77% power at 100\u2013500 Hz versus Q01's 27.15%.","R05's 4\u20138 kHz envelope is dominated by 29.9 Hz with amplitude 0.487, while Q01 is dominated by 120.6 Hz with amplitude 0.242."],"evidence_ids":["R05.fft","D03.fft","R05.ridge","D03.ridge","R05.bands","D03.bands","R05.envelope_4","D03.envelope_4"],"similar":false,"similarities":["Both broadband envelopes are led by approximately 29.9 Hz.","Their filter-bank kurtosis medians are close and modestly positive."]},{"condition_id":"C06","differences":["Although R06 is the nearest baseline numerical control, its FFT is led by 44.9/29.9 Hz and its ridge stays at 40\u201350 Hz, unlike Q01's 89.6 Hz line and fixed 90 Hz ridge.","R06 has 92.12% below 100 Hz and 7.11% at 100\u2013500 Hz, versus Q01's 71.71% and 27.15%.","R06 has substantially greater high-frequency filter-bank impulsiveness."],"evidence_ids":["R06.fft","D03.fft","R06.ridge","D03.ridge","R06.bands","D03.bands","R06.kurtosis_bank","D03.kurtosis_bank"],"similar":false,"similarities":["The uncalibrated baseline Welch, harmonic-spacing and cyclic numerical controls rank R06 nearest.","Both have approximately 29.9/59.7 Hz spacing families and approximately 0.067 s autocorrelation recurrence."]},{"condition_id":"C07","differences":["R07 is dominated by 45.9 Hz and has a 40\u201350 Hz ridge, unlike Q01's 89.6 Hz dominance and constant 90 Hz ridge.","R07's envelope_1 maximum is 15.31 Hz and envelope_4 maximum is 62.35 Hz; Q01's are 119.39 and 120.61 Hz.","Q01 has much more 100\u2013500 Hz power."],"evidence_ids":["R07.fft","D03.fft","R07.ridge","D03.ridge","R07.envelope_1","D03.envelope_1","R07.envelope_4","D03.envelope_4","R07.bands","D03.bands"],"similar":false,"similarities":["Both have high persistence occupancy of 88.41% and no reported waveform level changes.","Both have modest filter-bank kurtosis rather than the high R03/R06 pattern."]},{"condition_id":"C08","differences":["R08 is dominated by 44.6 Hz with strong 74.4/89.3 Hz secondary lines, while Q01 makes approximately 89.6 Hz the primary line and has no comparable dominant 44\u201345 Hz component.","R08's ridge remains at 40\u201350 Hz, whereas Q01 remains at 90 Hz.","R08 has 2.68% power at 100\u2013500 Hz versus Q01's 27.15%."],"evidence_ids":["R08.fft","D03.fft","R08.ridge","D03.ridge","R08.bands","D03.bands"],"similar":false,"similarities":["R08 contains an 89.3 Hz secondary FFT line near Q01's 89.6 Hz primary line.","Both have low positive filter-bank kurtosis and approximately 0.067 s autocorrelation recurrence."]}],"condition_id":null,"explanation":"Exploratory judgment: Q01 is substantially different from every supplied condition. The decisive observable mismatch is a persistent 89.6 Hz dominant component and invariant 90 Hz STFT ridge, while all references are principally 43.9\u201345.9 Hz and have 40\u201350 Hz dominant ridges except intermittent R04 excursions. Q01 also lies outside the reference range in band allocation: 27.15% at 100\u2013500 Hz versus 2.53\u20137.18% for R02\u2013R08 and 26.65% for R01, while its strong 0.06698 s autocorrelation excludes the otherwise band-similar R01. R06 is only a forced nearest numerical ranking; no calibrated acceptance threshold exists, and visual/numeric physical features rather than the distance controls drive the decision.","outcome":"different"},"decision_basis":"The decision is driven by direct query/reference observations, especially FFT, ridge, bands and autocorrelation\u2014not by numerical controls. Q01's 89.6 Hz primary line, fixed 90 Hz ridge, 71.71% below 100 Hz and 27.15% at 100\u2013500 Hz form a combination absent from all references. R01 approaches the band split but conflicts in autocorrelation, ridge behavior, DC offset and waveform disturbance; R02\u2013R08 share the autocorrelation period but are much more concentrated below 100 Hz and led near 45 Hz.","decisive_diagrams":["fft","ridge","bands","autocorrelation","envelope_4"],"diagram_review":{"autocorrelation":{"effect":"contradicts","evidence_ids":["R01.autocorrelation","R02.autocorrelation","R03.autocorrelation","R04.autocorrelation","R05.autocorrelation","R06.autocorrelation","R07.autocorrelation","R08.autocorrelation","D03.autocorrelation"],"favored_condition":"C06","finding":"Q01's strongest positive peak is 0.06698 s at 0.9767, matching the 0.0654\u20130.0684 s, 0.971\u20130.983 family of R02\u2013R08 and strongly conflicting with R01's 0.000381 s report. This common periodicity counters the final different decision but is not condition-specific.","priority":"high"},"bands":{"effect":"supports","evidence_ids":["R01.bands","R02.bands","R03.bands","R04.bands","R05.bands","R06.bands","R07.bands","R08.bands","D03.bands"],"favored_condition":"C01","finding":"Q01 has 71.71% below 100 Hz and 27.15% at 100\u2013500 Hz. This is close only to R01's 68.27%/26.65% split and is far outside R02\u2013R08's 92.12\u201397.25%/2.53\u20137.11%; because other decisive views reject R01, the unusual band combination supports different.","priority":"high"},"cepstrum":{"effect":"neutral","evidence_ids":["R01.cepstrum","R02.cepstrum","R03.cepstrum","R04.cepstrum","R05.cepstrum","R06.cepstrum","R07.cepstrum","R08.cepstrum","D03.cepstrum"],"favored_condition":"C03","finding":"Q01 and all references are dominated by the first near-zero-quefrency bins; Q01's first-bin value 0.160 is nearest in scale to R03's 0.150. No distinct nonzero spacing peak is established, so this floor-dependent view is neutral.","priority":"low"},"cyclic":{"effect":"neutral","evidence_ids":["R01.cyclic","R02.cyclic","R03.cyclic","R04.cyclic","R05.cyclic","R06.cyclic","R07.cyclic","R08.cyclic","D03.cyclic"],"favored_condition":"C06","finding":"Q01's maximum coherence is 0.432 with only 38,497 valid cells, fewer than every reference; maxima overlap the reference range. Unequal masking and deterministic tones prevent a reliable class match; the numerical control happens to favor R06.","priority":"low"},"envelope":{"effect":"contradicts","evidence_ids":["R01.envelope","R02.envelope","R03.envelope","R04.envelope","R05.envelope","R06.envelope","R07.envelope","R08.envelope","D03.envelope"],"favored_condition":"C04","finding":"Q01's broadband envelope is led by 29.90 Hz with 59.70/89.60/119.39 Hz companions, resembling R04's 29.90 Hz lead and R05's 29.29/58.48 Hz family. This counters complete separation but does not validate a shaft or fault frequency.","priority":"medium"},"envelope_1":{"effect":"neutral","evidence_ids":["R01.envelope_1","R02.envelope_1","R03.envelope_1","R04.envelope_1","R05.envelope_1","R06.envelope_1","R07.envelope_1","R08.envelope_1","D03.envelope_1"],"favored_condition":"C08","finding":"In 100\u2013500 Hz, Q01 peaks at 119.39 Hz, nearest to R08's 118.98 Hz; Q01 also has 59.69/29.90 Hz components. The numerical signature control also favors R08, but this single correlated envelope band is insufficient against the primary spectral mismatch.","priority":"medium"},"envelope_2":{"effect":"neutral","evidence_ids":["R01.envelope_2","R02.envelope_2","R03.envelope_2","R04.envelope_2","R05.envelope_2","R06.envelope_2","R07.envelope_2","R08.envelope_2","D03.envelope_2"],"favored_condition":"C05","finding":"In 500\u20131500 Hz, Q01 is led by 1.12 Hz with a strong 31.02 Hz line. No reference has this exact ordering; R02 supplies the closest low-frequency maximum and R04\u2013R06 supply approximately 30 Hz structure. The numerical control marginally favors R05, so evidence is ambiguous.","priority":"low"},"envelope_3":{"effect":"supports","evidence_ids":["R01.envelope_3","R02.envelope_3","R03.envelope_3","R04.envelope_3","R05.envelope_3","R06.envelope_3","R07.envelope_3","R08.envelope_3","D03.envelope_3"],"favored_condition":"C02","finding":"Q01's 1500\u20134000 Hz maximum is 358.27 Hz at only 0.0342; reference maxima occur at 0.51\u201358.57 Hz except Q01. This outside-reference peak location weakly supports different, although the low amplitude limits weight.","priority":"low"},"envelope_4":{"effect":"supports","evidence_ids":["R01.envelope_4","R02.envelope_4","R03.envelope_4","R04.envelope_4","R05.envelope_4","R06.envelope_4","R07.envelope_4","R08.envelope_4","D03.envelope_4"],"favored_condition":"C06","finding":"Q01's 4\u20138 kHz envelope is dominated by 120.61/120.51 Hz, whereas reference maxima are near 30, 61\u201362 or 91.7 Hz. R06 is the nearest numerical signature because it contains a strong 122.24 Hz component, but its maximum remains 61.12 Hz; the ordering mismatch supports different.","priority":"high"},"envelope_5":{"effect":"neutral","evidence_ids":["R01.envelope_5","R02.envelope_5","R03.envelope_5","R04.envelope_5","R05.envelope_5","R06.envelope_5","R07.envelope_5","R08.envelope_5","D03.envelope_5"],"favored_condition":"C05","finding":"Q01's 8\u201316 kHz maximum is 89.59 Hz, closely matching R05's 89.80 Hz maximum; R05 is also near in the numerical control. Other references differ, but this isolated high-carrier resemblance is correlated with the modulation map and is not decisive.","priority":"medium"},"estimators":{"effect":"supports","evidence_ids":["R01.estimators","R02.estimators","R03.estimators","R04.estimators","R05.estimators","R06.estimators","R07.estimators","R08.estimators","D03.estimators"],"favored_condition":"C04","finding":"Q01's Welch/DPSS maxima center around 80\u2013100 Hz, unlike reference estimator maxima centered around 30\u201370 Hz. Its integrated multitaper power 0.153 is near R04/R07, but absolute power is peak-normalization and response dependent; peak location supports different.","priority":"medium"},"fft":{"effect":"supports","evidence_ids":["R01.fft","R02.fft","R03.fft","R04.fft","R05.fft","R06.fft","R07.fft","R08.fft","D03.fft"],"favored_condition":"C08","finding":"Q01's strongest resolved peak is 89.6 Hz, followed by 59.7 and 29.9 Hz. Every reference is led at 43.9\u201345.9 Hz; R08 alone has a near-equal 89.3 Hz secondary line, but still has 44.6 Hz as primary. This is the strongest support for different.","priority":"high"},"harmonics":{"effect":"neutral","evidence_ids":["R01.harmonics","R02.harmonics","R03.harmonics","R04.harmonics","R05.harmonics","R06.harmonics","R07.harmonics","R08.harmonics","D03.harmonics"],"favored_condition":"C06","finding":"Q01 emphasizes 29/59/89 Hz spacing counts and has a 29.9 Hz candidate matching 16 listed peaks. R06 is nearest numerically and also has a strong 29.9 Hz family, but unrelated tones can create these pair spacings; no fundamental is established.","priority":"low"},"kurtosis_bank":{"effect":"contradicts","evidence_ids":["R01.kurtosis_bank","R02.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank","R05.kurtosis_bank","R06.kurtosis_bank","R07.kurtosis_bank","R08.kurtosis_bank","D03.kurtosis_bank"],"favored_condition":"C05","finding":"Q01 has modest high-frequency impulsiveness (median 0.0685, maximum 0.160), resembling R04/R05/R08 and sharply unlike R03/R06. This provides known-reference similarity and counters different, but does not identify a fault.","priority":"medium"},"levels":{"effect":"neutral","evidence_ids":["R01.levels","R02.levels","R03.levels","R04.levels","R05.levels","R06.levels","R07.levels","R08.levels","D03.levels"],"favored_condition":"C07","finding":"Q01 is unclipped, nonsilent and has no detected level changes, like R02/R03/R05/R07/R08. Its DC offset 0.0155 is closest to R06, but gain-normalized RMS and crest factor cannot establish class or severity.","priority":"low"},"modulation_map":{"effect":"supports","evidence_ids":["R01.modulation_map","R02.modulation_map","R03.modulation_map","R04.modulation_map","R05.modulation_map","R06.modulation_map","R07.modulation_map","R08.modulation_map","D03.modulation_map"],"favored_condition":"C05","finding":"Q01's map maximum is 0.0375, below all references' 0.0784\u20130.4404, despite the same 1,100 valid cells. This supports outside-reference modulation strength, though the map duplicates band-envelope evidence and the numerical control favors R05.","priority":"medium"},"orders":{"effect":"unavailable","evidence_ids":["R01.orders","R02.orders","R03.orders","R04.orders","R05.orders","R06.orders","R07.orders","R08.orders","D03.orders"],"favored_condition":null,"finding":"Orders are unavailable for Q01 and all references because no independent RPM trace was supplied; the 89.6 Hz line cannot be assigned a validated order.","priority":"unavailable"},"persistence":{"effect":"neutral","evidence_ids":["R01.persistence","R02.persistence","R03.persistence","R04.persistence","R05.persistence","R06.persistence","R07.persistence","R08.persistence","D03.persistence"],"favored_condition":"C07","finding":"Q01's maximum occupancy is 88.41%, exactly the reported R07 value and near R04/R05/R08, but its persistent population is centered differently. Chronology is discarded and overlapping frames are dependent, so this is neutral.","priority":"low"},"reassigned":{"effect":"supports","evidence_ids":["R01.reassigned","R02.reassigned","R03.reassigned","R04.reassigned","R05.reassigned","R06.reassigned","R07.reassigned","R08.reassigned","D03.reassigned"],"favored_condition":"C04","finding":"Q01 retains 99.9975% weight and shows a sharp stationary ridge near 90 Hz; references retain similarly high weight but concentrate their dominant low ridge near 45 Hz. This supports different but is correlated with FFT/STFT.","priority":"low"},"ridge":{"effect":"supports","evidence_ids":["R01.ridge","R02.ridge","R03.ridge","R04.ridge","R05.ridge","R06.ridge","R07.ridge","R08.ridge","D03.ridge"],"favored_condition":"C04","finding":"Q01's dominant eligible bin is 90 Hz for all 397 frames. R02\u2013R08 remain mainly at 40\u201350 Hz, R04 only intermittently reaches 80 Hz, and R01 switches broadly with median 50 Hz. The invariant 90 Hz ridge is outside the supplied set and decisively supports different.","priority":"high"},"spectral_kurtosis":{"effect":"neutral","evidence_ids":["R01.spectral_kurtosis","R02.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis","R05.spectral_kurtosis","R06.spectral_kurtosis","R07.spectral_kurtosis","R08.spectral_kurtosis","D03.spectral_kurtosis"],"favored_condition":"C07","finding":"Q01's maximum 0.552 at 7.32 kHz and median -0.0619 lie within reference ranges and are closest in maximum to R07. However, Q01 excludes 738 low-energy bins, more than every reference, and no significance threshold or bias correction is supplied.","priority":"low"},"stft":{"effect":"supports","evidence_ids":["R01.stft","R02.stft","R03.stft","R04.stft","R05.stft","R06.stft","R07.stft","R08.stft","D03.stft"],"favored_condition":"C08","finding":"Q01 is stationary like the references, but its brightest persistent low-frequency line is near 90 Hz rather than the common approximately 45 Hz line. The absence of drift is shared; the line location supports different.","priority":"medium"},"stft_detail":{"effect":"supports","evidence_ids":["R01.stft_detail","R02.stft_detail","R03.stft_detail","R04.stft_detail","R05.stft_detail","R06.stft_detail","R07.stft_detail","R08.stft_detail","D03.stft_detail"],"favored_condition":"C04","finding":"At 10 Hz bins, Q01 has median displayed PSD -95.66 dB, below every reference (-88.33 to -94.37 dB), and a persistent 90 Hz dominant line. The lower floor and shifted line support different, subject to acquisition response.","priority":"medium"},"synchronous":{"effect":"unavailable","evidence_ids":["R01.synchronous","R02.synchronous","R03.synchronous","R04.synchronous","R05.synchronous","R06.synchronous","R07.synchronous","R08.synchronous","D03.synchronous"],"favored_condition":null,"finding":"Synchronous averages are unavailable for Q01 and every reference because no independent RPM trace was supplied; phase-locked evidence cannot favor any condition.","priority":"unavailable"},"wavelet":{"effect":"neutral","evidence_ids":["R01.wavelet","R02.wavelet","R03.wavelet","R04.wavelet","R05.wavelet","R06.wavelet","R07.wavelet","R08.wavelet","D03.wavelet"],"favored_condition":"C01","finding":"Q01's 4.2 kHz-limited wavelet maximum is -30.58 dB and median -60.70 dB; the median resembles R01 and the maximum lies between R06 and the stronger references. Edge-masked cells and non-PSD scaling make this weakly diagnostic.","priority":"low"},"welch":{"effect":"supports","evidence_ids":["R01.welch","R02.welch","R03.welch","R04.welch","R05.welch","R06.welch","R07.welch","R08.welch","D03.welch"],"favored_condition":"C06","finding":"Q01's centroid is 99.29 Hz, between R01's 212.5 Hz and R02\u2013R08's 51.8\u201365.9 Hz, with a strong 80\u2013100 Hz region. Its flatness 0.000084 is reference-like, but the centroid is outside R02\u2013R08 and the 61.5 Hz bandwidth cannot resolve nearby FFT lines.","priority":"high"}},"limitations":"Only one peak-normalized recording represents each condition, so condition effects cannot be separated from microphone response, placement, reverberation, speed, load or acquisition regime. Correlated transforms are not independent votes. No calibrated distance threshold or predictive-accuracy evidence is supplied; nearest-reference distances are controls only. Orders and synchronous averages are unavailable. Cyclic and wavelet cells are masked, modulation maps omit cells, and Q01 spectral kurtosis excludes 738 low-energy bins. The judgment does not infer Q01's true label, fault identity, component or severity.","most_informative":["fft","ridge","bands","autocorrelation","envelope_4"],"priority_changes":"After adding Q01, ridge rises to decisive high priority because Q01 is fixed at 90 Hz while every reference is principally 40\u201350 Hz. Bands remains high but changes interpretation: it initially isolated R01, whereas Q01 nearly matches R01's split while conflicting with R01 in autocorrelation and ridge behavior. Autocorrelation remains high as essential counterevidence and specifically rules out the otherwise band-similar R01. Envelope_4 remains informative because Q01 introduces a 120.6 Hz maximum absent from reference maxima. Kurtosis_bank is reduced from high to medium because Q01's modest values resemble several references and do not resolve the dominant spectral conflict."},"stage":"decision","survey":"survey-8ca096c2c8dd560f","usage":{"completion_tokens":6220,"completion_tokens_details":{"accepted_prediction_tokens":0,"audio_tokens":0,"reasoning_tokens":186,"rejected_prediction_tokens":0},"latency_checkpoint":{"engine_tbt_ms":8,"engine_ttft_ms":1331,"engine_ttlt_ms":53391,"pre_inference_ms":2228,"service_tbt_ms":8,"service_ttft_ms":3675,"service_ttlt_ms":55732,"user_visible_ttft_ms":1447},"prompt_tokens":85057,"prompt_tokens_details":{"audio_tokens":0,"cached_tokens":3840},"total_tokens":91277}},{"baseline":{"category":"Wrong acceptance","dataset":"Ottawa","name":"Ottawa-excluded-D03","phase":"excluded","predicted":"C06","query_id":"D03"},"category":"Correct rejection","contract_sha256":"2bf2180d336fa50321e8d4f61c706a04e6eba287c306f425d53e2c261d46c036","dataset":"Ottawa","expected":"different","name":"Ottawa-excluded-D03","original_sha256":"b553a871870e104e67ea711476bcdb6cb03ebb3a42841c7770a9377f690e1f6e","phase":"excluded","predicted":"different","query_id":"D03","result":{"counterevidence":"The strongest counterevidence to rejection is Q01's long-lag autocorrelation, low-frequency tonal persistence, and several numerical controls that place it within the non-R01 cluster: the baseline is nearest R06, estimator/STFT-detail controls are nearest R04, and cepstrum is nearest R03. Q01 also shares about 29.9-Hz modulation with R04/R05 and an 89.6-Hz envelope_5 component near R05. These are partial, conflicting similarities rather than a coherent match because Q01's dominant 89.6-Hz FFT/ridge structure and 27.15% 100\u2013500-Hz power are absent from those references [Q01.C01.S0001,D03.autocorrelation,D03.envelope,D03.envelope_5,D03.ridge,R03.autocorrelation,R04.envelope,R04.estimators,R04.stft_detail,R05.envelope,R05.envelope_5,R06.autocorrelation,R06.welch].","decision":{"comparisons":[{"condition_id":"C01","differences":["Q01 is dominated by 89.6 Hz and has a 0.06698-s autocorrelation peak, whereas R01 is led by 45.9 Hz and its strongest positive autocorrelation peak is 0.000381 s.","Q01 has 71.71% below 100 Hz and 27.15% at 100\u2013500 Hz, versus R01's 68.27% and 26.65%, but Q01 has far less power above 500 Hz and lacks R01's large 0.148-FS DC offset and disturbance near 2.1 s."],"evidence_ids":["Q01.C01.S0001","D03.fft","D03.autocorrelation","D03.bands","D03.levels","R01.C01.S0001","R01.fft","R01.autocorrelation","R01.bands","R01.levels"],"similar":false,"similarities":["R01 is the closest reference in coarse <100/100\u2013500-Hz power fractions, and both have broader low-frequency distributions than R03\u2013R08."]},{"condition_id":"C03","differences":["R03 is dominated by 45.0 Hz and 96.19% of its Welch power is below 100 Hz; Q01 is dominated by 89.6 Hz, with only 71.71% below 100 Hz and 27.15% at 100\u2013500 Hz.","R03 has strong 15-Hz carrier-band modulation and broadly high-frequency impulsive kurtosis, unlike Q01's 119.4/1.12/358.3/120.6/89.6-Hz band leaders and modest kurtosis-bank maximum 0.160."],"evidence_ids":["Q01.C01.S0001","D03.fft","D03.bands","D03.envelope_1","D03.envelope_2","D03.envelope_3","D03.envelope_4","D03.envelope_5","D03.kurtosis_bank","R03.C01.S0001","R03.fft","R03.bands","R03.envelope_1","R03.envelope_2","R03.envelope_3","R03.envelope_4","R03.envelope_5","R03.kurtosis_bank"],"similar":false,"similarities":["Their autocorrelation lags and coefficients are close, and the cepstrum numerical control is nearest R03."]},{"condition_id":"C04","differences":["R04 is led by 44.9 and 74.8 Hz, while Q01 has a stable 89.6-Hz dominant ridge and FFT peak.","R04 retains 92.46% below 100 Hz and has 7.18% at 100\u2013500 Hz, unlike Q01's 71.71% and 27.15%; their band-envelope leaders also differ strongly."],"evidence_ids":["Q01.C01.S0001","D03.fft","D03.ridge","D03.bands","D03.envelope_1","D03.envelope_4","R04.C01.S0001","R04.fft","R04.ridge","R04.bands","R04.envelope_1","R04.envelope_4"],"similar":false,"similarities":["Estimator, reassigned, and STFT-detail numerical controls are nearest R04; both have moderate rather than R03/R06-like filter-bank impulsiveness."]},{"condition_id":"C05","differences":["R05 is dominated by 43.9 Hz with 95.86% below 100 Hz; Q01 is dominated by 89.6 Hz and has much more 100\u2013500-Hz power.","R05 has a very strong 29.9-Hz envelope_4 maximum of 0.487, whereas Q01's envelope_4 leads at 120.6 Hz with 0.242."],"evidence_ids":["Q01.C01.S0001","D03.fft","D03.bands","D03.envelope_4","D03.modulation_map","R05.C01.S0001","R05.fft","R05.bands","R05.envelope_4","R05.modulation_map"],"similar":false,"similarities":["Both show a leading broadband-envelope component near 29.9 Hz, an envelope_5 component near 89.6\u201389.8 Hz, and low positive filter-bank kurtosis; modulation-map and envelope_2 controls are nearest R05."]},{"condition_id":"C06","differences":["R06 is led by 44.9 Hz with a strong 29.9-Hz component, but Q01 is led by 89.6 Hz and its dominant STFT ridge is fixed at 90 Hz rather than 40\u201350 Hz.","R06 has only 7.11% at 100\u2013500 Hz and markedly higher high-frequency filter-bank kurtosis (median 0.352, maximum 1.043), versus Q01's 27.15%, 0.068, and 0.160."],"evidence_ids":["Q01.C01.S0001","D03.fft","D03.ridge","D03.bands","D03.kurtosis_bank","R06.C01.S0001","R06.fft","R06.ridge","R06.bands","R06.kurtosis_bank"],"similar":false,"similarities":["The uncalibrated full-band log-Welch baseline is nearest R06, and their autocorrelation lags are almost identical."]},{"condition_id":"C07","differences":["R07 is dominated by 45.9 Hz with a nearly fixed 50-Hz ridge; Q01 is dominated by 89.6 Hz with a fixed 90-Hz ridge.","R07 has 95.13% below 100 Hz and only 4.68% at 100\u2013500 Hz, unlike Q01's 71.71% and 27.15%; R07's 62.35/187.04-Hz high-carrier modulation also differs from Q01's 120.61/89.59 Hz."],"evidence_ids":["Q01.C01.S0001","D03.fft","D03.ridge","D03.bands","D03.envelope_4","D03.envelope_5","R07.C01.S0001","R07.fft","R07.ridge","R07.bands","R07.envelope_4","R07.envelope_5"],"similar":false,"similarities":["Both are steady, persistent tonal recordings and have modest filter-bank impulsiveness; the spectral-kurtosis control is nearest R07."]},{"condition_id":"C08","differences":["R08 is dominated by 44.6 Hz with 97.0% below 100 Hz; Q01 is dominated by 89.6 Hz with only 71.71% below 100 Hz.","R08's envelope_4 and envelope_5 leaders are 30.31 and 181.73 Hz, versus Q01's 120.61 and 89.59 Hz."],"evidence_ids":["Q01.C01.S0001","D03.fft","D03.bands","D03.envelope_4","D03.envelope_5","R08.C01.S0001","R08.fft","R08.bands","R08.envelope_4","R08.envelope_5"],"similar":false,"similarities":["Q01 and R08 have low positive filter-bank kurtosis and steady persistent spectra; envelope_1 and cyclic numerical controls are nearest R08 or nearly tied with R06."]}],"condition_id":null,"explanation":"Exploratory decision: different. Q01 falls outside the supplied references in the combined decisive pattern: a dominant 89.6-Hz line and fixed 90-Hz ridge, 27.15% of Welch power in 100\u2013500 Hz, and carrier-band modulation leaders that do not reproduce any single reference. R01 is closest in coarse band fractions but conflicts in autocorrelation, DC/transient behavior, and broadband distribution; R03\u2013R08 match the long-lag periodicity but are all dominated near 44\u201346 Hz and place 92.1\u201397.0% below 100 Hz. Numerical controls are mixed and uncalibrated, so visual and reported physical measurements\u2014not the controls\u2014drive rejection. This does not identify Q01's true condition, component, fault, or severity.","outcome":"different"},"decision_basis":"The decisive evidence is the cross-view consistency of Q01's 89.6-Hz dominant structure in FFT, physical STFT/ridge, and estimator views, together with its unusual band-power allocation and nonmatching carrier-band modulation. Correlated views were treated as one spectral pattern rather than independent votes. The conclusion is outside-reference evidence, not proof of a novel fault class.","decisive_diagrams":["fft","bands","ridge","autocorrelation","envelope_4"],"diagram_review":{"autocorrelation":{"effect":"contradicts","evidence_ids":["D03.autocorrelation","Q01.C01.S0001","R03.autocorrelation","R03.C01.S0001","R04.autocorrelation","R04.C01.S0001","R05.autocorrelation","R05.C01.S0001","R06.autocorrelation","R06.C01.S0001","R07.autocorrelation","R07.C01.S0001","R08.autocorrelation","R08.C01.S0001","R01.autocorrelation","R01.C01.S0001"],"favored_condition":"C06","finding":"Q01 peaks at 0.06698 s with coefficient 0.9767, closely matching R03\u2013R08's 0.06538\u20130.06838 s periodic cluster and especially R06's 0.06688 s, while strongly excluding R01's 0.000381-s peak. This argues against all-reference rejection but does not separate the fault conditions.","priority":"high"},"bands":{"effect":"supports","evidence_ids":["D03.bands","Q01.C01.S0001","R01.bands","R01.C01.S0001","R04.bands","R04.C01.S0001","R06.bands","R06.C01.S0001"],"favored_condition":"C01","finding":"Q01 has 71.71% below 100 Hz and 27.15% at 100\u2013500 Hz, close only to R01's 68.27%/26.65%; R03\u2013R08 instead have 92.12\u201397.00% below 100 Hz. Q01 nevertheless has much less power above 500 Hz than R01, so even the favored resemblance is incomplete.","priority":"high"},"cepstrum":{"effect":"neutral","evidence_ids":["D03.cepstrum","R03.cepstrum","R04.cepstrum","R06.cepstrum"],"favored_condition":"C03","finding":"Q01's first-bin cepstral amplitude is 0.1598, between R03/R04 and R06; no nonzero spacing peak is reported. The numerical control favors R03, but near-zero quefrency slope is weakly diagnostic.","priority":"low"},"cyclic":{"effect":"neutral","evidence_ids":["D03.cyclic","R04.cyclic","R06.cyclic","R08.cyclic"],"favored_condition":"C06","finding":"Q01's maximum squared cyclic coherence is 0.432 with only 38,497 valid cells, fewer than every reference. The numerical control is nearest R06 and R08, but differing masks and no significance threshold prevent acceptance evidence.","priority":"low"},"envelope":{"effect":"neutral","evidence_ids":["D03.envelope","Q01.C01.S0001","R04.envelope","R05.envelope","R06.envelope"],"favored_condition":"C04","finding":"Q01's broadband envelope leads at 29.90 Hz, resembling R04/R05 and differing from R03/R06 near 14.95 Hz, R07 near 30.61 Hz, and R08 at 44.65 Hz. This shared modulation is non-unique and may reflect beating.","priority":"medium"},"envelope_1":{"effect":"neutral","evidence_ids":["D03.envelope_1","R05.envelope_1","R08.envelope_1"],"favored_condition":"C08","finding":"Q01's 100\u2013500-Hz carrier envelope leads at 119.39 Hz, almost matching R08's 118.98 Hz; R05 is next in family at 102.45 Hz. The control also favors R08, but other bands do not preserve that match.","priority":"medium"},"envelope_2":{"effect":"supports","evidence_ids":["D03.envelope_2","R04.envelope_2","R05.envelope_2","R06.envelope_2","R08.envelope_2"],"favored_condition":"C05","finding":"Q01's 500\u20131500-Hz envelope is led by 1.12 Hz, with secondary energy near 30\u201331 Hz. No reference has 1.12 Hz as leader; the numerical control weakly favors R05 despite a different 29.29-Hz leader.","priority":"medium"},"envelope_3":{"effect":"supports","evidence_ids":["D03.envelope_3","R03.envelope_3","R04.envelope_3","R05.envelope_3","R06.envelope_3","R07.envelope_3","R08.envelope_3"],"favored_condition":"C07","finding":"Q01 leads at 358.27 Hz in the 1500\u20134000-Hz carrier band, unlike every reference leader (1.12\u201358.57 Hz). Distances are nearly tied and do not override the visible outside-reference leader.","priority":"medium"},"envelope_4":{"effect":"supports","evidence_ids":["D03.envelope_4","R03.envelope_4","R04.envelope_4","R05.envelope_4","R06.envelope_4","R07.envelope_4","R08.envelope_4"],"favored_condition":"C06","finding":"Q01 leads at 120.61 Hz, whereas references form approximately 30-Hz or 61\u201362-Hz groups. The numerical control favors R06, but the leading modulation is outside both reference groups.","priority":"high"},"envelope_5":{"effect":"neutral","evidence_ids":["D03.envelope_5","R04.envelope_5","R05.envelope_5","R07.envelope_5","R08.envelope_5"],"favored_condition":"C05","finding":"Q01's 89.59-Hz leader closely matches R05's 89.80 Hz, but not R03/R06 near 15 Hz, R07/R08 near 182\u2013187 Hz, or R04 at 273.98 Hz. This isolated match conflicts with other Q01 envelope bands.","priority":"medium"},"estimators":{"effect":"supports","evidence_ids":["D03.estimators","R04.estimators","R06.estimators"],"favored_condition":"C04","finding":"Q01's estimators peak around 80\u2013100 Hz, especially 90 Hz, whereas all references' estimator maxima lie mainly around 40\u201370 Hz. Although the distance control is nearest R04, the physical peak location remains different.","priority":"high"},"fft":{"effect":"supports","evidence_ids":["D03.fft","Q01.C01.S0001","R01.fft","R01.C01.S0001","R03.fft","R03.C01.S0001","R04.fft","R04.C01.S0001","R05.fft","R05.C01.S0001","R06.fft","R06.C01.S0001","R07.fft","R07.C01.S0001","R08.fft","R08.C01.S0001"],"favored_condition":null,"finding":"Q01's strongest FFT peak is 89.6 Hz at -6.75 dB FS, followed by 59.7 Hz; every reference is instead led at 43.9\u201345.9 Hz. Q01 contains 29.9/59.7/89.6-Hz spacing families, but no reference reproduces their amplitude ordering.","priority":"high"},"harmonics":{"effect":"neutral","evidence_ids":["D03.harmonics","R05.harmonics","R06.harmonics","R08.harmonics"],"favored_condition":"C06","finding":"Q01's largest peak-pair count is 11 near 29 Hz, comparable to R05/R06 and below most other references. Dense 29/59/89-Hz spacings are non-unique and do not establish a fundamental.","priority":"low"},"kurtosis_bank":{"effect":"neutral","evidence_ids":["D03.kurtosis_bank","R04.kurtosis_bank","R05.kurtosis_bank","R08.kurtosis_bank","R03.kurtosis_bank","R06.kurtosis_bank"],"favored_condition":"C04","finding":"Q01 has median/max complex excess kurtosis 0.068/0.160, resembling the low-impulsiveness R04/R05/R08 group and strongly differing from R03/R06. The control favors R05, but impulsiveness is not fault-specific.","priority":"medium"},"levels":{"effect":"neutral","evidence_ids":["D03.levels","Q01.C01.S0001","R03.levels","R05.levels","R07.levels","R08.levels"],"favored_condition":"C07","finding":"Q01 is unclipped, peak-normalized, steady, and has no >6-dB local changes, like R03/R05/R07/R08. RMS and peak levels cannot support acoustic severity; Q01's low DC offset differs especially from R01.","priority":"low"},"modulation_map":{"effect":"supports","evidence_ids":["D03.modulation_map","R04.modulation_map","R05.modulation_map","R06.modulation_map","R07.modulation_map"],"favored_condition":"C05","finding":"Q01's map maximum is only 0.0375, below every reference maximum (0.0784\u20130.4404), despite the numerical control favoring R05/R04. The 155 masked cells per map are missing data, not low modulation.","priority":"medium"},"orders":{"effect":"unavailable","evidence_ids":["D03.orders","R01.orders","R03.orders","R04.orders","R05.orders","R06.orders","R07.orders","R08.orders"],"favored_condition":null,"finding":"Order spectra are unavailable for Q01 and all references because no independent RPM trace was supplied.","priority":"unavailable"},"persistence":{"effect":"neutral","evidence_ids":["D03.persistence","R04.persistence","R07.persistence","R05.persistence","R08.persistence"],"favored_condition":"C07","finding":"Q01's maximum occupancy is 88.41%, exactly matching R07 and close to R04/R05/R08, but persistence discards chronology and restates the stable tonal structure.","priority":"low"},"reassigned":{"effect":"supports","evidence_ids":["D03.reassigned","R04.reassigned","R06.reassigned"],"favored_condition":"C04","finding":"Q01 retains 99.998% of reassigned weight and localizes a stable low-frequency ridge, but that ridge is around 90 Hz rather than the references' dominant 40\u201350-Hz structures. The control is nearest R04.","priority":"medium"},"ridge":{"effect":"supports","evidence_ids":["D03.ridge","R01.ridge","R03.ridge","R04.ridge","R05.ridge","R06.ridge","R07.ridge","R08.ridge"],"favored_condition":null,"finding":"Q01's dominant ridge is fixed at 90 Hz for all 397 frames. References have 40- or 50-Hz medians; even R01/R04 excursions do not produce a persistent 90-Hz ridge.","priority":"high"},"spectral_kurtosis":{"effect":"neutral","evidence_ids":["D03.spectral_kurtosis","R07.spectral_kurtosis","R04.spectral_kurtosis","R05.spectral_kurtosis"],"favored_condition":"C07","finding":"Q01's maximum spectral kurtosis is 0.552 at 7.32 kHz, close in magnitude to R07 and several others. Q01 excludes 738 low-energy bins\u2014more than any reference\u2014so direct ranking is weak without a significance threshold.","priority":"low"},"stft":{"effect":"supports","evidence_ids":["D03.stft","D03.ridge","R03.stft","R04.stft","R05.stft","R06.stft","R07.stft","R08.stft"],"favored_condition":null,"finding":"Q01 is temporally steady like several references, but its strongest persistent horizontal band is centered near 90 Hz rather than the references' approximately 44\u201346-Hz dominant family. This is correlated with FFT and ridge evidence.","priority":"medium"},"stft_detail":{"effect":"supports","evidence_ids":["D03.stft_detail","R04.stft_detail","R06.stft_detail"],"favored_condition":"C04","finding":"Q01's maximum/median PSD are -28.55/-95.66 dB, with the lowest median floor among the set. Its numerical distance is nearest R04, but the physical 90-Hz dominant structure remains unmatched.","priority":"medium"},"synchronous":{"effect":"unavailable","evidence_ids":["D03.synchronous","R01.synchronous","R03.synchronous","R04.synchronous","R05.synchronous","R06.synchronous","R07.synchronous","R08.synchronous"],"favored_condition":null,"finding":"Synchronous averages are unavailable for Q01 and all references because no independent RPM trace was supplied.","priority":"unavailable"},"wavelet":{"effect":"neutral","evidence_ids":["D03.wavelet","R01.wavelet","R04.wavelet","R06.wavelet"],"favored_condition":"C01","finding":"Q01's wavelet view is steady and low-frequency dominated, with median -60.70 dB; its control is nearest R01/R04/R06. The 4.2-kHz limit and correlated spectral content make this nondecisive.","priority":"low"},"welch":{"effect":"supports","evidence_ids":["D03.welch","Q01.C01.S0001","R01.welch","R01.C01.S0001","R03.welch","R04.welch","R05.welch","R06.welch","R07.welch","R08.welch"],"favored_condition":"C06","finding":"Q01's centroid is 99.29 Hz, between R01's 212.52 Hz and R03\u2013R08's 54.31\u201365.85 Hz, while its flatness 0.000084 is below all references. The uncalibrated baseline favors R06, but 41.02-Hz bins cannot resolve Q01's detailed low-frequency lines.","priority":"high"}},"limitations":"Only one reference exists per condition, so condition effects cannot be separated from microphone response, placement, reverberation, speed, load, and acquisition regime. All signals are peak-normalized; no acoustic-level or severity claim is valid. The 26 views are correlated transformations, not independent votes. Cyclic and modulation maps contain masked missing cells; wavelet edges are masked. Welch resolution is coarse, kurtosis/cyclic products lack significance intervals, and numerical distances have no acceptance threshold. Orders and synchronous averages are unavailable. No bearing geometry or independent speed is supplied, so no bearing-frequency, RPM, component, identity, or severity inference is made.","most_informative":["fft","bands","ridge","autocorrelation","envelope_4"],"priority_changes":"After adding Q01, FFT and ridge become the highest-priority evidence because the query's persistent 89.6/90-Hz dominance is outside every reference. Bands remains high but now gives conflicting evidence: it favors R01 coarsely while FFT/autocorrelation reject that match. Envelope_4 remains high because Q01's 120.6-Hz leader falls outside both frozen reference families. Autocorrelation stays high as the strongest counterevidence, linking Q01 to R03\u2013R08 but not distinguishing them. Kurtosis_bank is reduced from high to medium because Q01 merely joins an overlapping low-impulsiveness group. Numerical controls did not drive the decision."},"stage":"decision","survey":"survey-150908d5ac073e21","usage":{"completion_tokens":5705,"completion_tokens_details":{"accepted_prediction_tokens":0,"audio_tokens":0,"reasoning_tokens":190,"rejected_prediction_tokens":0},"latency_checkpoint":{"engine_tbt_ms":8,"engine_ttft_ms":847,"engine_ttlt_ms":46435,"pre_inference_ms":2044,"service_tbt_ms":8,"service_ttft_ms":3085,"service_ttlt_ms":48693,"user_visible_ttft_ms":1041},"prompt_tokens":76903,"prompt_tokens_details":{"audio_tokens":0,"cached_tokens":0},"total_tokens":82608}},{"baseline":{"category":"False rejection","dataset":"Ottawa","name":"Ottawa-known-D07","phase":"known","predicted":"different","query_id":"D07"},"category":"False rejection","contract_sha256":"13a094165af816af28fadfb730a21592510e12f772fde10244bec994132bf207","dataset":"Ottawa","expected":"C04","name":"Ottawa-known-D07","original_sha256":"eccdb7a81dc25b0785df30f3261e0912724b6e18307cc4c63c520d10ab979aec","phase":"known","predicted":"different","query_id":"D07","result":{"counterevidence":"Several controls force a nearest known reference: full-band Welch distance is smallest to R06 (3.003 dB), estimator/STFT-detail/reassigned controls favor R04, and the query shares approximately 30/60/90 Hz families with multiple references. Its filter-bank kurtosis maximum 1.995 is also close in magnitude to R03's 1.850. These are partial similarities, but they conflict with the query's unique 90.2 Hz dominant ridge, 70.90%/28.73% low-band split, 0.03324 s autocorrelation peak, weak 4\u20138 kHz envelope modulation, and unusually large 1.06 kHz spectral kurtosis.","decision":{"comparisons":[{"condition_id":"C01","differences":["Q01 lacks R01's broad high-frequency background, large 0.148 FS DC offset and irregular 40\u2013220 Hz ridge.","Q01 instead has a stationary 90 Hz ridge and much less power above 500 Hz."],"evidence_ids":["R01.fft","R01.bands","R01.autocorrelation","R01.ridge","D07.fft","D07.bands","D07.autocorrelation","D07.ridge","Q01.C01.S0001"],"similar":false,"similarities":["Their sub-100 Hz fractions are relatively close (R01 68.27%, Q01 70.90%) compared with R02\u2013R08, but the adjacent-band and line structures differ."]},{"condition_id":"C02","differences":["R02 is dominated by 45.1 Hz with 97.25% below 100 Hz and a 0.0665 s autocorrelation peak; Q01 is dominated by 90.2 Hz, has only 70.90% below 100 Hz and peaks at 0.03324 s.","R02's 4\u20138 kHz envelope maximum is 91.73 Hz/0.317, unlike Q01's 182.24 Hz/0.078."],"evidence_ids":["R02.fft","R02.bands","R02.autocorrelation","R02.envelope_4","D07.fft","D07.bands","D07.autocorrelation","D07.envelope_4","Q01.C01.S0001"],"similar":false,"similarities":["Both contain 30/60/90 Hz-related line and modulation families."]},{"condition_id":"C03","differences":["R03 is 45 Hz dominated and has strong 15 Hz modulation across several carrier bands; Q01 is 90.2 Hz dominated and emphasizes 30/60/90/120 Hz.","R03's high-frequency filter-bank impulsiveness is broad (median 0.506), whereas Q01 has a localized 7.49\u20138.74 kHz maximum with median 0.117."],"evidence_ids":["R03.fft","R03.envelope_2","R03.envelope_3","R03.envelope_5","R03.kurtosis_bank","D07.fft","D07.envelope_2","D07.envelope_3","D07.envelope_5","D07.kurtosis_bank","Q01.C01.S0001"],"similar":false,"similarities":["The filter-bank kurtosis maxima are similarly large (R03 1.850; Q01 1.995), indicating intermittency but not common fault identity."]},{"condition_id":"C04","differences":["R04 has strong 44.9/74.8 Hz lines, while Q01 has a dominant 90.2 Hz line and stationary 90 Hz ridge.","R04's strongest 4\u20138 kHz modulation is 30.41 Hz/0.376, versus Q01 182.24 Hz/0.078."],"evidence_ids":["R04.fft","R04.ridge","R04.envelope_4","D07.fft","D07.ridge","D07.envelope_4","Q01.C01.S0001"],"similar":false,"similarities":["R04 is nearest on several numerical controls (cepstrum, cyclic, kurtosis-bank, STFT detail, reassigned and wavelet), and both allocate relatively high power to 100\u2013500 Hz; nevertheless, the controls have no acceptance threshold."]},{"condition_id":"C05","differences":["R05 is dominated by 43.9 Hz with 95.86% below 100 Hz; Q01 is dominated by 90.2 Hz with 70.90% below 100 Hz.","R05 has very strong 29.9 Hz modulation in 4\u20138 kHz (0.487), absent from Q01's weak 182.2 Hz maximum."],"evidence_ids":["R05.fft","R05.bands","R05.envelope_4","D07.fft","D07.bands","D07.envelope_4","Q01.C01.S0001"],"similar":false,"similarities":["Both have broadband-envelope energy near 30 and 60 Hz and no reported level changes."]},{"condition_id":"C06","differences":["Although the baseline control ranks R06 nearest, R06 is 44.9 Hz dominated with a 40\u201350 Hz ridge and 92.12% below 100 Hz; Q01 is 90.2 Hz dominated with an invariant 90 Hz ridge and 70.90% below 100 Hz.","R06 emphasizes 15 Hz in several envelopes, unlike Q01's 30/60/90/120 Hz pattern."],"evidence_ids":["R06.fft","R06.bands","R06.ridge","R06.envelope_1","R06.envelope_3","R06.envelope_5","D07.fft","D07.bands","D07.ridge","D07.envelope_1","D07.envelope_3","D07.envelope_5","Q01.C01.S0001"],"similar":false,"similarities":["R06 is the forced nearest baseline Welch candidate (3.003 dB) and both have appreciable 100\u2013500 Hz content, but their dominant frequencies and temporal periodicity conflict."]},{"condition_id":"C07","differences":["R07 is dominated by 45.9 Hz with a 40\u201350 Hz ridge and 95.13% below 100 Hz; Q01 is dominated by 90.2 Hz with a 90 Hz ridge and only 70.90% below 100 Hz.","R07 has strong 62.35 Hz 4\u20138 kHz and 187.04 Hz 8\u201316 kHz envelopes; Q01's corresponding maxima are much weaker and at 182.24 and 90.20 Hz."],"evidence_ids":["R07.fft","R07.bands","R07.ridge","R07.envelope_4","R07.envelope_5","D07.fft","D07.bands","D07.ridge","D07.envelope_4","D07.envelope_5","Q01.C01.S0001"],"similar":false,"similarities":["Both show line families spaced around 30\u201331 Hz, but spacing alone does not establish a common fundamental."]},{"condition_id":"C08","differences":["R08 is 44.6 Hz dominated with 97.00% below 100 Hz and a 0.06721 s autocorrelation peak; Q01 is 90.2 Hz dominated with 70.90% below 100 Hz and a 0.03324 s peak.","R08 has strong 30.31 Hz 4\u20138 kHz modulation (0.436), whereas Q01's maximum is only 0.078 at 182.24 Hz."],"evidence_ids":["R08.fft","R08.bands","R08.autocorrelation","R08.envelope_4","D07.fft","D07.bands","D07.autocorrelation","D07.envelope_4","Q01.C01.S0001"],"similar":false,"similarities":["Both contain 30/60/90/180 Hz-related components, but with different dominant carriers and normalized strengths."]}],"condition_id":null,"explanation":"Exploratory judgment: Q01 is substantially outside every supplied single-record reference. The decisive observable combination is a persistent 90.2 Hz dominant component, 70.90% of Welch power below 100 Hz plus 28.73% at 100\u2013500 Hz, a 0.03324 s autocorrelation peak, weak/differently located high-band envelope modulation, and a localized spectral-kurtosis maximum of 3.043 near 1.06 kHz. No reference shares that complete pattern. Numerical controls provide forced nearest rankings\u2014especially R06 for baseline Welch and R04 for several derived views\u2014but do not drive acceptance and have no calibrated threshold.","outcome":"different"},"decision_basis":"Visual and reported numeric observations, rather than nearest-reference controls, drive the decision. FFT/ridge, band allocation, autocorrelation, high-band envelope structure and spectral kurtosis jointly place Q01 outside the supplied examples. These are correlated views of one signal and were not counted as independent votes.","decisive_diagrams":["fft","bands","autocorrelation","ridge","spectral_kurtosis"],"diagram_review":{"autocorrelation":{"effect":"supports","evidence_ids":["R01.autocorrelation","R02.autocorrelation","R03.autocorrelation","R04.autocorrelation","R05.autocorrelation","R06.autocorrelation","R07.autocorrelation","R08.autocorrelation","D07.autocorrelation","Q01.C01.S0001"],"favored_condition":null,"finding":"Q01's strongest positive peak is 0.03324 s at 0.982, unlike R02\u2013R08 at 0.0654\u20130.0684 s and R01 at 0.000381 s. A harmonic relation is possible, so no RPM is inferred.","priority":"high"},"bands":{"effect":"supports","evidence_ids":["R01.bands","R02.bands","R03.bands","R04.bands","R05.bands","R06.bands","R07.bands","R08.bands","D07.bands","Q01.C01.S0001"],"favored_condition":null,"finding":"Q01 allocates 70.90% below 100 Hz and 28.73% at 100\u2013500 Hz. R01 is closest below 100 Hz (68.27%) but has only 26.65% at 100\u2013500 Hz and much more above 500 Hz; R02\u2013R08 have 92.12\u201397.25% below 100 Hz.","priority":"high"},"cepstrum":{"effect":"neutral","evidence_ids":["R01.cepstrum","R02.cepstrum","R03.cepstrum","R04.cepstrum","R05.cepstrum","R06.cepstrum","R07.cepstrum","R08.cepstrum","D07.cepstrum"],"favored_condition":"C08","finding":"Q01 is dominated by near-zero quefrency bins (maximum 0.122), within the R02\u2013R08 range and numerically nearest R08; no reliable nonzero spacing peak is reported.","priority":"low"},"cyclic":{"effect":"neutral","evidence_ids":["R01.cyclic","R02.cyclic","R03.cyclic","R04.cyclic","R05.cyclic","R06.cyclic","R07.cyclic","R08.cyclic","D07.cyclic"],"favored_condition":"C04","finding":"Q01's maximum coherence is only 0.127 with 34,288 valid cells, below all references and with substantially heavier low-power masking; the numerical control favors R04, but unequal masks prevent a strong comparison.","priority":"low"},"envelope":{"effect":"supports","evidence_ids":["R01.envelope","R02.envelope","R03.envelope","R04.envelope","R05.envelope","R06.envelope","R07.envelope","R08.envelope","D07.envelope","Q01.C01.S0001"],"favored_condition":null,"finding":"Q01 emphasizes 30.10, 60.20, 120.30 and 90.20 Hz. Individual components occur in references, but none has this reported dominance ordering; modulation does not verify shaft or bearing frequencies.","priority":"medium"},"envelope_1":{"effect":"neutral","evidence_ids":["R01.envelope_1","R02.envelope_1","R03.envelope_1","R04.envelope_1","R05.envelope_1","R06.envelope_1","R07.envelope_1","R08.envelope_1","D07.envelope_1"],"favored_condition":"C02","finding":"In 100\u2013500 Hz, Q01 peaks at 120.31 Hz/0.317, not matching any reference maximum; its strong secondary 90.20 Hz component resembles R02/R03 families. The numerical control favors R02.","priority":"medium"},"envelope_2":{"effect":"neutral","evidence_ids":["R01.envelope_2","R02.envelope_2","R03.envelope_2","R04.envelope_2","R05.envelope_2","R06.envelope_2","R07.envelope_2","R08.envelope_2","D07.envelope_2"],"favored_condition":"C06","finding":"Q01's 500\u20131500 Hz envelope peaks at 30.10 Hz/0.200, close in frequency and strength to R06's 29.90 Hz/0.203 and R04's 29.90 Hz/0.239; the distance control instead marginally favors R05.","priority":"medium"},"envelope_3":{"effect":"neutral","evidence_ids":["R01.envelope_3","R02.envelope_3","R03.envelope_3","R04.envelope_3","R05.envelope_3","R06.envelope_3","R07.envelope_3","R08.envelope_3","D07.envelope_3"],"favored_condition":"C08","finding":"Q01 peaks at 30.10 Hz/0.0566 in 1.5\u20134 kHz, closest in frequency to R08's 29.80 Hz/0.0455; the full-signature numerical control favors R02.","priority":"low"},"envelope_4":{"effect":"supports","evidence_ids":["R01.envelope_4","R02.envelope_4","R03.envelope_4","R04.envelope_4","R05.envelope_4","R06.envelope_4","R07.envelope_4","R08.envelope_4","D07.envelope_4"],"favored_condition":null,"finding":"Q01's 4\u20138 kHz maximum is 182.24 Hz and only 0.0785. All references peak at approximately 30\u201392 Hz with amplitudes 0.212\u20130.487, making this a strong outside-reference feature.","priority":"high"},"envelope_5":{"effect":"neutral","evidence_ids":["R01.envelope_5","R02.envelope_5","R03.envelope_5","R04.envelope_5","R05.envelope_5","R06.envelope_5","R07.envelope_5","R08.envelope_5","D07.envelope_5"],"favored_condition":"C05","finding":"Q01 peaks at 90.20 Hz/0.144 in 8\u201316 kHz, nearest in frequency to R05's 89.80 Hz/0.164; the full-signature distance favors R02. This conflicts with the outside-reference evidence in envelope_4.","priority":"medium"},"estimators":{"effect":"contradicts","evidence_ids":["R01.estimators","R02.estimators","R03.estimators","R04.estimators","R05.estimators","R06.estimators","R07.estimators","R08.estimators","D07.estimators"],"favored_condition":"C04","finding":"Q01's integrated multitaper power is 0.176, close to R03 0.174 and R04 0.158, while its dominant estimator bins shift to 70\u2013110 Hz. The signature-distance control favors R04, providing limited counterevidence to rejection.","priority":"medium"},"fft":{"effect":"supports","evidence_ids":["R01.fft","R02.fft","R03.fft","R04.fft","R05.fft","R06.fft","R07.fft","R08.fft","D07.fft","Q01.C01.S0001"],"favored_condition":null,"finding":"Q01 is dominated by 90.2 Hz at -5.85 dBFS, followed by 60.2 and 30.1 Hz. Every reference is dominated near 43.9\u201345.9 Hz; none shows the query's dominant-line ordering.","priority":"high"},"harmonics":{"effect":"supports","evidence_ids":["R01.harmonics","R02.harmonics","R03.harmonics","R04.harmonics","R05.harmonics","R06.harmonics","R07.harmonics","R08.harmonics","D07.harmonics"],"favored_condition":null,"finding":"Q01's largest spacing count is 31 Hz (12), whereas all references have their largest count at 15 Hz. Q01 nevertheless shares 30/60/90 Hz families with several references; spacing candidates are not established fundamentals.","priority":"medium"},"kurtosis_bank":{"effect":"neutral","evidence_ids":["R01.kurtosis_bank","R02.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank","R05.kurtosis_bank","R06.kurtosis_bank","R07.kurtosis_bank","R08.kurtosis_bank","D07.kurtosis_bank"],"favored_condition":"C03","finding":"Q01 has a localized maximum of 1.995 at 7.49\u20138.74 kHz and median 0.117. Its maximum resembles R03's 1.850, but R03's positive impulsiveness is much broader with median 0.506; the numerical distance instead favors R04.","priority":"medium"},"levels":{"effect":"neutral","evidence_ids":["R01.levels","R02.levels","R03.levels","R04.levels","R05.levels","R06.levels","R07.levels","R08.levels","D07.levels","Q01.C01.S0001"],"favored_condition":"C03","finding":"Q01 is unclipped, nonsilent and has no reported level changes, like R02/R03/R05/R07/R08. Crest factor 7.07 dB is closest to R03, but peak normalization prevents level or severity claims.","priority":"low"},"modulation_map":{"effect":"neutral","evidence_ids":["R01.modulation_map","R02.modulation_map","R03.modulation_map","R04.modulation_map","R05.modulation_map","R06.modulation_map","R07.modulation_map","R08.modulation_map","D07.modulation_map"],"favored_condition":"C05","finding":"Q01's valid-cell maximum is 0.141 and median 0.00538, within the reference range; numerical distance favors R05. The 155 masked cells are missing, and this map repeats band-envelope evidence.","priority":"low"},"orders":{"effect":"unavailable","evidence_ids":["R01.orders","R02.orders","R03.orders","R04.orders","R05.orders","R06.orders","R07.orders","R08.orders","D07.orders"],"favored_condition":null,"finding":"Unavailable for Q01 and every reference because no independent RPM trace was supplied.","priority":"unavailable"},"persistence":{"effect":"neutral","evidence_ids":["R01.persistence","R02.persistence","R03.persistence","R04.persistence","R05.persistence","R06.persistence","R07.persistence","R08.persistence","D07.persistence"],"favored_condition":"C02","finding":"Q01's maximum occupancy is 95.47%, exceeding the 50.38\u201392.70% reference range and reflecting a highly persistent narrow population; the numerical signature control favors R02. Chronology is discarded.","priority":"medium"},"reassigned":{"effect":"contradicts","evidence_ids":["R01.reassigned","R02.reassigned","R03.reassigned","R04.reassigned","R05.reassigned","R06.reassigned","R07.reassigned","R08.reassigned","D07.reassigned"],"favored_condition":"C04","finding":"Q01 retains 99.995% of weight and has median pooled weight -84.87 dB, with stationary localized ridges. The distance control favors R04, but the query's dominant ridge frequency remains different and this view is correlated with STFT/FFT.","priority":"low"},"ridge":{"effect":"supports","evidence_ids":["R01.ridge","R02.ridge","R03.ridge","R04.ridge","R05.ridge","R06.ridge","R07.ridge","R08.ridge","D07.ridge"],"favored_condition":null,"finding":"Q01's strongest eligible bin is exactly 90 Hz in all 397 frames. R02\u2013R08 remain mainly at 40\u201350 Hz, R04 occasionally reaches 80 Hz, and R01 switches broadly; the 10 Hz grid precludes finer frequency interpretation.","priority":"high"},"spectral_kurtosis":{"effect":"supports","evidence_ids":["R01.spectral_kurtosis","R02.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis","R05.spectral_kurtosis","R06.spectral_kurtosis","R07.spectral_kurtosis","R08.spectral_kurtosis","D07.spectral_kurtosis"],"favored_condition":null,"finding":"Q01 reaches 3.043 near 1.06 kHz, well above all reference maxima of 0.511\u20131.255. However, 912 query bins are excluded versus 2\u2013544 in references, and no bias correction or significance threshold is supplied.","priority":"high"},"stft":{"effect":"supports","evidence_ids":["R01.stft","R02.stft","R03.stft","R04.stft","R05.stft","R06.stft","R07.stft","R08.stft","D07.stft"],"favored_condition":null,"finding":"Q01 is stationary but its persistent dominant horizontal population is around 90 Hz rather than the approximately 45 Hz population common to all references. It lacks R01's mid-record disturbance.","priority":"medium"},"stft_detail":{"effect":"contradicts","evidence_ids":["R01.stft_detail","R02.stft_detail","R03.stft_detail","R04.stft_detail","R05.stft_detail","R06.stft_detail","R07.stft_detail","R08.stft_detail","D07.stft_detail"],"favored_condition":"C04","finding":"Q01's median displayed PSD is -96.43 dB, lower than every reference (-88.33 to -94.37 dB), while its stationary line pattern yields the smallest numerical distance to R04. The latter is a forced control, not acceptance.","priority":"medium"},"synchronous":{"effect":"unavailable","evidence_ids":["R01.synchronous","R02.synchronous","R03.synchronous","R04.synchronous","R05.synchronous","R06.synchronous","R07.synchronous","R08.synchronous","D07.synchronous"],"favored_condition":null,"finding":"Unavailable for Q01 and every reference because no independent RPM trace was supplied.","priority":"unavailable"},"wavelet":{"effect":"neutral","evidence_ids":["R01.wavelet","R02.wavelet","R03.wavelet","R04.wavelet","R05.wavelet","R06.wavelet","R07.wavelet","R08.wavelet","D07.wavelet"],"favored_condition":"C04","finding":"Q01 shows stationary low-frequency power with maximum -29.29 dB and median -63.78 dB, broadly within R02\u2013R08 behavior; the numerical control favors R04. Edge-masked cells and the 4.2 kHz upper limit restrict comparison.","priority":"low"},"welch":{"effect":"supports","evidence_ids":["R01.welch","R02.welch","R03.welch","R04.welch","R05.welch","R06.welch","R07.welch","R08.welch","D07.welch","Q01.C01.S0001"],"favored_condition":null,"finding":"Q01's centroid is 94.86 Hz: between broad R01 (212.5 Hz) and R02\u2013R08 (51.8\u201365.9 Hz). Its 41 Hz-bin Welch peak region is shifted toward 90 Hz; the baseline distance still forces R06 nearest at 3.003 dB without an acceptance threshold.","priority":"high"}},"limitations":"Only one peak-normalized recording represents each condition, so condition effects cannot be separated from microphone response, placement, reverberation, speed, load or acquisition regime. Correlated transforms were not treated as independent votes. Masked cyclic/modulation/wavelet cells are missing evidence; spectral-kurtosis exclusions differ substantially. Orders and synchronous averages are unavailable. No calibrated distance or acceptance threshold exists, and the reference-only survey established no predictive accuracy. The result does not identify the unknown's true label, component, fault identity or severity.","most_informative":["fft","bands","autocorrelation","ridge","spectral_kurtosis"],"priority_changes":"After adding Q01, ridge and spectral_kurtosis rise to decisive status because Q01 has an invariant 90 Hz ridge and a 3.043 kurtosis peak near 1.06 kHz. Envelope_4 remains informative but is displaced from the top five by those clearer outside-reference observations. Kurtosis_bank drops from high to medium because Q01 combines an R03-like maximum with a different localized distribution, while autocorrelation and bands remain high priority."},"stage":"decision","survey":"survey-8ca096c2c8dd560f","usage":{"completion_tokens":5823,"completion_tokens_details":{"accepted_prediction_tokens":0,"audio_tokens":0,"reasoning_tokens":186,"rejected_prediction_tokens":0},"latency_checkpoint":{"engine_tbt_ms":7,"engine_ttft_ms":933,"engine_ttlt_ms":42976,"pre_inference_ms":2101,"service_tbt_ms":7,"service_ttft_ms":3206,"service_ttlt_ms":45251,"user_visible_ttft_ms":1105},"prompt_tokens":85032,"prompt_tokens_details":{"audio_tokens":0,"cached_tokens":0},"total_tokens":90855}},{"baseline":{"category":"Correct rejection","dataset":"Ottawa","name":"Ottawa-excluded-D07","phase":"excluded","predicted":"different","query_id":"D07"},"category":"Correct rejection","contract_sha256":"392a1aebfe5a5788bb9fbefdaf84b43f7d3ad17e50d96b9477663500cf9d8e4a","dataset":"Ottawa","expected":"different","name":"Ottawa-excluded-D07","original_sha256":"3f5a95886cf46fbf248bcc09ee32d73d66e390d793b69d2ea386e3d5bb0ac9ce","phase":"excluded","predicted":"different","query_id":"D07","result":{"counterevidence":"The numerical Welch control ranks R06/C06 nearest (3.003), and Q01 shares some approximately 30/60/90 Hz modulation structure with several references, especially R06. Q01's filter-bank kurtosis maximum 1.995 is also closest in magnitude to R03's 1.850. These are partial similarities, but they conflict with Q01's persistent 90.2 Hz dominant line/ridge, 0.03324 s autocorrelation lag, and 70.90%/28.73% split between 0\u2013100/100\u2013500 Hz, none of which is represented by a supplied reference.","decision":{"comparisons":[{"condition_id":"C01","differences":["Q01 is dominated by 90.2 Hz, whereas R01's principal FFT peak is 45.9 Hz.","Q01 has a 0.03324 s autocorrelation peak and 70.90%/28.73% power in 0\u2013100/100\u2013500 Hz; R01 has 0.000381 s and 68.27%/26.65%.","Q01 lacks R01's large 0.148 FS DC offset and reported level changes."],"evidence_ids":["R01.fft","D07.fft","R01.autocorrelation","D07.autocorrelation","R01.bands","D07.bands","R01.levels","D07.levels"],"similar":false,"similarities":["Both have much more 100\u2013500 Hz relative power than R02/R03/R05/R06/R07/R08 and comparatively elevated broadband content."]},{"condition_id":"C02","differences":["R02 is dominated near 45.1 Hz and has a 0.0665 s autocorrelation peak; Q01 is dominated at 90.2 Hz with a 0.03324 s peak.","R02 places 97.25% of Welch power below 100 Hz, versus 70.90% for Q01.","R02's 4\u20138 kHz envelope is strongest at 91.73 Hz, whereas Q01 is strongest at 182.24 Hz."],"evidence_ids":["R02.fft","D07.fft","R02.autocorrelation","D07.autocorrelation","R02.bands","D07.bands","R02.envelope_4","D07.envelope_4"],"similar":false,"similarities":["Both are stationary, unclipped, peak-normalized recordings with strong narrow harmonic structure."]},{"condition_id":"C03","differences":["R03's principal line is 45.0 Hz and ridge alternates at 40/50 Hz; Q01's principal line and ridge remain at 90.2/90 Hz.","R03 has 96.19% below 100 Hz, versus Q01's 70.90%.","R03 has strong approximately 15 Hz modulation in several carrier bands; Q01 is mainly organized around approximately 30/60/90/120 Hz."],"evidence_ids":["R03.fft","D07.fft","R03.ridge","D07.ridge","R03.bands","D07.bands","R03.envelope_3","D07.envelope_3"],"similar":false,"similarities":["Q01 and R03 both show high filter-bank impulsiveness maxima, 1.995 and 1.850 respectively, although their band patterns and medians differ."]},{"condition_id":"C05","differences":["R05 is dominated at 43.9 Hz with a 0.06838 s autocorrelation peak; Q01 is dominated at 90.2 Hz with a 0.03324 s peak.","R05 has 95.86% below 100 Hz, compared with 70.90% for Q01.","R05 has exceptionally strong approximately 30 Hz modulation in 4\u20138 kHz (0.487 ordinary envelope), while Q01's maximum there is only 0.0785 at 182.24 Hz."],"evidence_ids":["R05.fft","D07.fft","R05.autocorrelation","D07.autocorrelation","R05.bands","D07.bands","R05.envelope_4","D07.envelope_4"],"similar":false,"similarities":["Both show a strong broadband-envelope component near 30 Hz and high spectral persistence."]},{"condition_id":"C06","differences":["Although the Welch numerical control ranks R06 nearest, R06's dominant FFT line is 44.9 Hz and ridge is 40/50 Hz, unlike Q01's persistent 90.2/90 Hz.","R06 has 92.12% below 100 Hz and 7.11% at 100\u2013500 Hz; Q01 has 70.90% and 28.73%.","R06 reports eleven greater-than-6 dB level changes, while Q01 reports none."],"evidence_ids":["R06.fft","D07.fft","R06.ridge","D07.ridge","R06.bands","D07.bands","R06.levels","D07.levels"],"similar":false,"similarities":["R06 is the nearest baseline Welch control and shares prominent approximately 30 Hz modulation in the 500\u20131500 Hz envelope.","Both have elevated positive high-band filter-bank kurtosis compared with most references."]},{"condition_id":"C07","differences":["R07 is dominated at 45.9 Hz with a 50 Hz median ridge and 0.06538 s autocorrelation lag; Q01 is fixed at 90 Hz with 0.03324 s lag.","R07 has 95.13% below 100 Hz, unlike Q01's 70.90%.","R07's strongest 8\u201316 kHz modulation is 187.04 Hz; Q01's is 90.20 Hz."],"evidence_ids":["R07.fft","D07.fft","R07.ridge","D07.ridge","R07.autocorrelation","D07.autocorrelation","R07.bands","D07.bands","R07.envelope_5","D07.envelope_5"],"similar":false,"similarities":["Both contain dense approximately 30 Hz-related peak spacing, but spacing candidates are not established fundamentals."]},{"condition_id":"C08","differences":["R08 is dominated at 44.6 Hz with a 0.06721 s autocorrelation peak; Q01 is dominated at 90.2 Hz with a 0.03324 s peak.","R08 places 97.00% below 100 Hz, compared with 70.90% for Q01.","R08 has strong approximately 30 Hz modulation in 4\u20138 kHz and approximately 182 Hz in 8\u201316 kHz; Q01 reverses this tendency, with 182 Hz in 4\u20138 kHz and 90 Hz in 8\u201316 kHz."],"evidence_ids":["R08.fft","D07.fft","R08.autocorrelation","D07.autocorrelation","R08.bands","D07.bands","R08.envelope_4","D07.envelope_4","R08.envelope_5","D07.envelope_5"],"similar":false,"similarities":["Both are stationary and harmonic-rich, and their cepstra are dominated by near-zero quefrency."]}],"condition_id":null,"explanation":"Exploratory judgment: different. The strongest observable basis is a coherent but correlated set of frequency-domain findings: Q01 has a persistent 90.2 Hz principal line and 90 Hz dominant ridge, a 0.03324 s autocorrelation peak, and an unusually large 100\u2013500 Hz power fraction of 28.73%. Every non-R01 reference is instead dominated near 44\u201346 Hz with a 40/50 Hz ridge, approximately 0.065\u20130.068 s autocorrelation, and at least 92.12% below 100 Hz; R01 also fails to match Q01's 90 Hz dominance and autocorrelation. The nearest numerical control is R06, but no acceptance threshold exists and the decisive visual/numeric observations contradict acceptance. This does not infer Q01's actual condition or fault identity.","outcome":"different"},"decision_basis":"Visual and reported numeric observations\u2014not the numerical distance controls\u2014drive the decision. The decisive evidence is Q01's unmatched 90 Hz spectral/ridge regime, half-length autocorrelation period relative to six references, and outside-reference band-power distribution. Envelope-band carrier/modulation patterns provide additional conflict but are correlated summaries and are not interpreted as bearing frequencies or RPM.","decisive_diagrams":["fft","bands","autocorrelation","ridge","envelope_4"],"diagram_review":{"autocorrelation":{"effect":"supports","evidence_ids":["R01.autocorrelation","R02.autocorrelation","R03.autocorrelation","R05.autocorrelation","R06.autocorrelation","R07.autocorrelation","R08.autocorrelation","D07.autocorrelation"],"favored_condition":null,"finding":"Q01 peaks at 0.033238 s with coefficient 0.982, outside R01's 0.000381 s and the 0.0654\u20130.0684 s cluster of all other references; no RPM conversion is made.","priority":"high"},"bands":{"effect":"supports","evidence_ids":["R01.bands","R02.bands","R03.bands","R05.bands","R06.bands","R07.bands","R08.bands","D07.bands"],"favored_condition":"C01","finding":"Q01 has 70.90% in 0\u2013100 Hz and 28.73% in 100\u2013500 Hz. It is relatively closest to R01's 68.27%/26.65%, but Q01's remaining bands are much weaker and the pair is not a full-spectrum match; all other references have 92.12\u201397.25% below 100 Hz.","priority":"high"},"cepstrum":{"effect":"neutral","evidence_ids":["R01.cepstrum","R02.cepstrum","R03.cepstrum","R05.cepstrum","R06.cepstrum","R07.cepstrum","R08.cepstrum","D07.cepstrum"],"favored_condition":"C08","finding":"Q01 and all references are dominated by near-zero quefrency. Q01's first-bin value 0.122 is closest in scale to R08's 0.114, but no distinct nonzero-quefrency spacing is established.","priority":"low"},"cyclic":{"effect":"neutral","evidence_ids":["R01.cyclic","R02.cyclic","R03.cyclic","R05.cyclic","R06.cyclic","R07.cyclic","R08.cyclic","D07.cyclic"],"favored_condition":"C08","finding":"Q01 has lower maximum cyclic coherence, 0.127, than every reference (0.245\u20130.699) and only 34,288 valid cells; differing masks and no significance threshold make this weak outside-reference evidence.","priority":"low"},"envelope":{"effect":"neutral","evidence_ids":["R01.envelope","R02.envelope","R03.envelope","R05.envelope","R06.envelope","R07.envelope","R08.envelope","D07.envelope"],"favored_condition":"C02","finding":"Q01's broadband envelope is led by 30.10 Hz with 60.20, 90.20 and 120.30 Hz components, overlapping R02's approximately 30/45 Hz family and several references' 30 Hz structure; beating remains an alternative source.","priority":"medium"},"envelope_1":{"effect":"contradicts","evidence_ids":["R01.envelope_1","R02.envelope_1","R03.envelope_1","R05.envelope_1","R06.envelope_1","R07.envelope_1","R08.envelope_1","D07.envelope_1"],"favored_condition":"C08","finding":"In 100\u2013500 Hz, Q01 is strongest at 120.31 Hz; R08 is nearest at 118.98 Hz, while the other references peak from 3.88 to 102.45 Hz. Q01's stronger 90/120 Hz pair prevents a unique match.","priority":"medium"},"envelope_2":{"effect":"contradicts","evidence_ids":["R01.envelope_2","R02.envelope_2","R03.envelope_2","R05.envelope_2","R06.envelope_2","R07.envelope_2","R08.envelope_2","D07.envelope_2"],"favored_condition":"C06","finding":"In 500\u20131500 Hz, Q01's 30.10 Hz maximum closely resembles R06's 29.90 Hz and R05's 29.29 Hz; carrier-band evidence alone is ambiguous.","priority":"medium"},"envelope_3":{"effect":"neutral","evidence_ids":["R01.envelope_3","R02.envelope_3","R03.envelope_3","R05.envelope_3","R06.envelope_3","R07.envelope_3","R08.envelope_3","D07.envelope_3"],"favored_condition":"C08","finding":"In 1500\u20134000 Hz, Q01 peaks at 30.10 Hz, nearest R08's 29.80 Hz, but Q01's normalized maximum 0.0566 is weak and other references contain different low-frequency maxima.","priority":"low"},"envelope_4":{"effect":"supports","evidence_ids":["R01.envelope_4","R02.envelope_4","R03.envelope_4","R05.envelope_4","R06.envelope_4","R07.envelope_4","R08.envelope_4","D07.envelope_4"],"favored_condition":null,"finding":"Q01's 4\u20138 kHz envelope is strongest at 182.24 Hz with amplitude 0.0785, whereas references peak near 30, 61\u201362 or 91.7 Hz with amplitudes 0.212\u20130.487; this is clear outside-reference evidence.","priority":"high"},"envelope_5":{"effect":"neutral","evidence_ids":["R01.envelope_5","R02.envelope_5","R03.envelope_5","R05.envelope_5","R06.envelope_5","R07.envelope_5","R08.envelope_5","D07.envelope_5"],"favored_condition":"C05","finding":"Q01 peaks at 90.20 Hz in 8\u201316 kHz, nearest R05's 89.80 Hz; it conflicts with R03/R06 near 15 Hz, R01 near 61.7 Hz, and R02/R07/R08 near 182\u2013187 Hz.","priority":"medium"},"estimators":{"effect":"neutral","evidence_ids":["R01.estimators","R02.estimators","R03.estimators","R05.estimators","R06.estimators","R07.estimators","R08.estimators","D07.estimators"],"favored_condition":"C03","finding":"Q01's integrated multitaper power 0.176 is closest to R03's 0.174, but its smoothed maximum is centered near 80\u2013110 Hz rather than the references' 30\u201370 Hz; 60 Hz smoothing limits line interpretation.","priority":"medium"},"fft":{"effect":"supports","evidence_ids":["R01.fft","R02.fft","R03.fft","R05.fft","R06.fft","R07.fft","R08.fft","D07.fft"],"favored_condition":null,"finding":"Q01's principal peak is 90.2 Hz at -5.85 dB FS. Every reference's principal peak is 43.9\u201345.9 Hz; R01 is additionally much weaker at -16.69 dB FS. No reference reproduces Q01's dominant-line regime.","priority":"high"},"harmonics":{"effect":"neutral","evidence_ids":["R01.harmonics","R02.harmonics","R03.harmonics","R05.harmonics","R06.harmonics","R07.harmonics","R08.harmonics","D07.harmonics"],"favored_condition":"C03","finding":"Q01 emphasizes approximately 30 Hz multiples but its largest spacing-count bin is 31 Hz with count 12, unlike the references' usual 15 Hz-leading counts. Dense unrelated tones prevent fundamental or RPM claims.","priority":"medium"},"kurtosis_bank":{"effect":"contradicts","evidence_ids":["R01.kurtosis_bank","R02.kurtosis_bank","R03.kurtosis_bank","R05.kurtosis_bank","R06.kurtosis_bank","R07.kurtosis_bank","R08.kurtosis_bank","D07.kurtosis_bank"],"favored_condition":"C03","finding":"Q01's maximum 1.995 near 7.49\u20138.74 kHz is closest in magnitude to R03's 1.850, but Q01's median is 0.117 versus R03's 0.506 and their positive-band distributions differ.","priority":"medium"},"levels":{"effect":"neutral","evidence_ids":["R01.levels","R02.levels","R03.levels","R05.levels","R06.levels","R07.levels","R08.levels","D07.levels"],"favored_condition":"C03","finding":"Q01 is unclipped, non-silent and has no reported greater-than-6 dB changes, resembling R02/R03/R05/R07/R08; it lacks R01's large DC offset and R06's numerous changes. Peak normalization prevents severity claims.","priority":"medium"},"modulation_map":{"effect":"neutral","evidence_ids":["R01.modulation_map","R02.modulation_map","R03.modulation_map","R05.modulation_map","R06.modulation_map","R07.modulation_map","R08.modulation_map","D07.modulation_map"],"favored_condition":"C08","finding":"Q01's map maximum 0.141 lies near R08's 0.159 and R06's 0.111, but 155 cells are masked per map and this view is correlated with the five envelopes.","priority":"low"},"orders":{"effect":"unavailable","evidence_ids":["R01.orders","R02.orders","R03.orders","R05.orders","R06.orders","R07.orders","R08.orders","D07.orders"],"favored_condition":null,"finding":"Unavailable for Q01 and every reference because no independent RPM trace was supplied.","priority":"unavailable"},"persistence":{"effect":"neutral","evidence_ids":["R01.persistence","R02.persistence","R03.persistence","R05.persistence","R06.persistence","R07.persistence","R08.persistence","D07.persistence"],"favored_condition":"C05","finding":"Q01's maximum persistence is 95.47%, nearest R05/R08 at 92.70%, but its persistent population is centered around the distinct 90 Hz regime; chronology is discarded.","priority":"medium"},"reassigned":{"effect":"neutral","evidence_ids":["R01.reassigned","R02.reassigned","R03.reassigned","R05.reassigned","R06.reassigned","R07.reassigned","R08.reassigned","D07.reassigned"],"favored_condition":"C02","finding":"Q01 retains 99.995% weight and has maximum -12.36 dB, similar in scale to R02/R03/R05/R08, while its localized dominant ridge remains near 90 Hz; this largely repeats FFT/STFT evidence.","priority":"low"},"ridge":{"effect":"supports","evidence_ids":["R01.ridge","R02.ridge","R03.ridge","R05.ridge","R06.ridge","R07.ridge","R08.ridge","D07.ridge"],"favored_condition":null,"finding":"Q01's eligible dominant ridge is fixed at 90 Hz for all 397 frames. References remain at 40/50 Hz apart from R01's intermittent excursions; 10 Hz binning does not erase this separation.","priority":"high"},"spectral_kurtosis":{"effect":"supports","evidence_ids":["R01.spectral_kurtosis","R02.spectral_kurtosis","R03.spectral_kurtosis","R05.spectral_kurtosis","R06.spectral_kurtosis","R07.spectral_kurtosis","R08.spectral_kurtosis","D07.spectral_kurtosis"],"favored_condition":null,"finding":"Q01 has a maximum 3.043 near 1.06 kHz, exceeding all references' maxima of 0.515\u20131.255; however 912 low-energy bins are excluded and finite-sample bias is uncorrected.","priority":"medium"},"stft":{"effect":"supports","evidence_ids":["R01.stft","R02.stft","R03.stft","R05.stft","R06.stft","R07.stft","R08.stft","D07.stft"],"favored_condition":null,"finding":"Q01 is broadly stationary like the references, but its strongest persistent horizontal structure is near 90 Hz rather than the references' approximately 45 Hz principal structure.","priority":"medium"},"stft_detail":{"effect":"neutral","evidence_ids":["R01.stft_detail","R02.stft_detail","R03.stft_detail","R05.stft_detail","R06.stft_detail","R07.stft_detail","R08.stft_detail","D07.stft_detail"],"favored_condition":"C02","finding":"Q01's maximum -27.01 dB and median -96.43 dB are within or near reference ranges, but its 10 Hz-bin dominant structure differs; no unique transient chronology is visible.","priority":"low"},"synchronous":{"effect":"unavailable","evidence_ids":["R01.synchronous","R02.synchronous","R03.synchronous","R05.synchronous","R06.synchronous","R07.synchronous","R08.synchronous","D07.synchronous"],"favored_condition":null,"finding":"Unavailable for Q01 and every reference because no independent RPM trace was supplied.","priority":"unavailable"},"wavelet":{"effect":"neutral","evidence_ids":["R01.wavelet","R02.wavelet","R03.wavelet","R05.wavelet","R06.wavelet","R07.wavelet","R08.wavelet","D07.wavelet"],"favored_condition":"C06","finding":"Q01's wavelet maximum -29.29 dB and median -63.78 dB are closest in scale to R06/R07, while the persistent low-frequency band is shifted by Q01's 90 Hz dominance; the view is limited to 4.2 kHz.","priority":"low"},"welch":{"effect":"supports","evidence_ids":["R01.welch","R02.welch","R03.welch","R05.welch","R06.welch","R07.welch","R08.welch","D07.welch"],"favored_condition":"C06","finding":"Q01's centroid is 94.86 Hz, outside R02/R03/R05/R06/R07/R08 at 51.85\u201362.04 Hz but below R01's 212.52 Hz. The baseline control ranks R06 nearest, yet the approximately 61.5 Hz ENBW and absent threshold prevent acceptance.","priority":"high"}},"limitations":"Only one reference represents each condition, so condition effects cannot be separated from microphone response, placement, reverberation, speed, load or acquisition regime. All signals are derived peak-normalized; digital RMS cannot establish acoustic severity. Views from one signal are correlated rather than independent votes. Masked cyclic/modulation cells are missing data. Orders and synchronous averages are unavailable without RPM. No calibrated distance or acceptance threshold is supplied, and fault identity, component and severity are not established.","most_informative":["fft","bands","autocorrelation","ridge","envelope_4"],"priority_changes":"Relative to the frozen survey, ridge rises from medium to high because Q01 is fixed at 90 Hz while every reference is principally 40/50 Hz. Envelope_4 remains high but now supports an outside-reference conclusion because Q01 peaks near 182 Hz rather than any reference's 30/61/92 Hz maxima. Kurtosis_bank falls from high to medium because its maximum magnitude favors R03 while its median and band distribution conflict. Welch remains high as contextual evidence, but its numerical nearest-reference ranking does not drive the decision."},"stage":"decision","survey":"survey-74ed3ba6f4fe58a6","usage":{"completion_tokens":5441,"completion_tokens_details":{"accepted_prediction_tokens":0,"audio_tokens":0,"reasoning_tokens":161,"rejected_prediction_tokens":0},"latency_checkpoint":{"engine_tbt_ms":8,"engine_ttft_ms":912,"engine_ttlt_ms":42745,"pre_inference_ms":1910,"service_tbt_ms":8,"service_ttft_ms":2981,"service_ttlt_ms":44765,"user_visible_ttft_ms":1072},"prompt_tokens":75268,"prompt_tokens_details":{"audio_tokens":0,"cached_tokens":0},"total_tokens":80709}}],"surveys":[{"dataset":"Jin","name":"survey-aa9448d1719da544","result":{"diagram_review":{"autocorrelation":{"evidence_ids":["R01.autocorrelation","R01.C01.S0001","R02.autocorrelation","R02.C01.S0001","R03.autocorrelation","R03.C01.S0001","R04.autocorrelation","R04.C01.S0001"],"finding":"Strongest nonzero peaks differ: R01 68.0 \u00b5s/0.870, R02 249.4 \u00b5s/0.944, R03 181.4 \u00b5s/0.969, R04 249.4 \u00b5s/0.894; long-lag shapes also differ, but harmonic ambiguity precludes speed inference.","priority":"medium"},"bands":{"evidence_ids":["R01.bands","R01.C01.S0001","R02.bands","R02.C01.S0001","R03.bands","R03.C01.S0001","R04.bands","R04.C01.S0001"],"finding":"Relative Welch power strongly separates spectra: R01 is broadly distributed (0\u2013100 Hz 37.7%); R02 combines 0\u2013100 Hz 59.7% with >8 kHz 39.2%; R03 is 94.1% below 100 Hz; R04 is high-frequency weighted, with 61.1% above 8 kHz.","priority":"high"},"cepstrum":{"evidence_ids":["R01.cepstrum","R02.cepstrum","R03.cepstrum","R04.cepstrum"],"finding":"Largest reported nonzero-quefrency peaks are R01 22.7 \u00b5s (0.271), R02 45.4 \u00b5s (0.257), R03 181.4 \u00b5s (0.0878), and R04 68.0 \u00b5s (0.124); these may reflect spectral spacing or response/echo and do not establish RPM.","priority":"medium"},"cyclic":{"evidence_ids":["R01.cyclic","R02.cyclic","R03.cyclic","R04.cyclic"],"finding":"Maximum squared cyclic coherence is higher for R01/R04 (0.309/0.299) than R02/R03 (0.138/0.131), while medians are all about 0.003; no significance threshold is supplied and deterministic tones can contribute.","priority":"medium"},"envelope":{"evidence_ids":["R01.envelope","R01.C01.S0001","R02.envelope","R02.C01.S0001","R03.envelope","R03.C01.S0001","R04.envelope","R04.C01.S0001"],"finding":"Broadband modulation differs: R01 leads near 3.43 Hz with additional 17.27/20.71/31.11/162.32 Hz peaks; R02 has 2.93/6.46/22.73 Hz; R03 has weaker low-rate components plus 43.23/44.24 Hz; R04 is dominated by 13.23 Hz with 23.13/9.90 Hz companions.","priority":"high"},"envelope_1":{"evidence_ids":["R01.envelope_1","R02.envelope_1","R03.envelope_1","R04.envelope_1"],"finding":"In the common 100\u2013500 Hz carrier band, R01 has much stronger normalized 3.469 Hz modulation (0.599) than R02 3.265 Hz (0.280), R03 3.571 Hz (0.214), or R04 3.265 Hz (0.164); R02/R04 additionally show components near 169\u2013172 Hz.","priority":"medium"},"envelope_2":{"evidence_ids":["R01.envelope_2","R02.envelope_2","R03.envelope_2","R04.envelope_2"],"finding":"For 500\u20131500 Hz carriers, R01 is exceptionally strong at 3.469 Hz (1.262, with 6.939 Hz 0.979); R04 leads at 6.633 Hz (0.552), R02 at 6.531 Hz (0.350), and R03 instead at 81.327 Hz (0.239).","priority":"high"},"envelope_3":{"evidence_ids":["R01.envelope_3","R02.envelope_3","R03.envelope_3","R04.envelope_3"],"finding":"For 1.5\u20134 kHz carriers, R01 again leads at 3.469 Hz (0.844), whereas R02 peaks at 337.755 Hz (0.192), R03 at 187.347 Hz (0.213), and R04 at 171.735 Hz (0.178).","priority":"high"},"envelope_4":{"evidence_ids":["R01.envelope_4","R02.envelope_4","R03.envelope_4","R04.envelope_4"],"finding":"For 4\u20138 kHz carriers, normalized maxima are R01 3.469 Hz/0.205, R02 12.959 Hz/0.0816, R03 212.041 Hz/0.105, and R04 26.429 Hz/0.0787; amplitudes are lower than in several other carrier bands.","priority":"medium"},"envelope_5":{"evidence_ids":["R01.envelope_5","R02.envelope_5","R03.envelope_5","R04.envelope_5"],"finding":"For 8\u201316 kHz carriers, R01 peaks at 13.776 Hz (0.287), R02 at 6.531 Hz (0.352), R03 at 7.041 Hz (0.110), and R04 at 13.163 Hz (0.330); R01 also has a distinct 151.94 Hz component.","priority":"medium"},"estimators":{"evidence_ids":["R01.estimators","R02.estimators","R03.estimators","R04.estimators"],"finding":"Welch mean/median and 5-taper DPSS broadly preserve the reference-specific PSD shapes: R03 has the strongest low-frequency power, R04 the strongest ~16 kHz region, and R01/R02 intermediate mixtures; estimator differences exist but do not overturn those distinctions.","priority":"medium"},"fft":{"evidence_ids":["R01.fft","R01.C01.S0001","R02.fft","R02.C01.S0001","R03.fft","R03.C01.S0001","R04.fft","R04.C01.S0001"],"finding":"At 0.1 Hz bins and 0.15 Hz Hann bandwidth, R01 has 5.3 Hz plus 169\u2013193 Hz and ~16.035 kHz peaks; R02 has 5.3/8.2/50.1 Hz and a ~16 kHz cluster; R03 is dominated by 5.3 Hz with lines near 1.84\u20135.37 kHz; R04 is led by 50 Hz and a dense ~16.0 kHz cluster.","priority":"high"},"harmonics":{"evidence_ids":["R01.harmonics","R02.harmonics","R03.harmonics","R04.harmonics"],"finding":"Prominent-peak spacing counts differ: maxima are R01 3 Hz/count 14, R02 7 and 17 Hz/count 8, R03 1 Hz/count 12, and R04 1 and 13 Hz/count 18; listed ~49\u201350 Hz candidates occur in all records but can arise from unrelated tones.","priority":"medium"},"kurtosis_bank":{"evidence_ids":["R01.kurtosis_bank","R02.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank"],"finding":"Impulsiveness location differs markedly: R01 reaches 8.73 at 1.33\u20132.64 kHz; R02 reaches 1.95 at 14.41\u201315.72 kHz; R03 is near Gaussian overall with maximum 0.658 below 1.33 kHz; R04 is sub-Gaussian at low frequency and reaches 0.745 at 15.72\u201317.02 kHz.","priority":"high"},"levels":{"evidence_ids":["R01.levels","R01.C01.S0001","R02.levels","R02.C01.S0001","R03.levels","R03.C01.S0001","R04.levels","R04.C01.S0001"],"finding":"All 10 s records are nonconstant, unclipped, near-zero-DC, and peak-normalized to 0.95 FS. R03 has the largest RMS (0.343) and lowest crest factor (8.84 dB), versus R01 0.137/16.84 dB, R02 0.158/15.58 dB, and R04 0.195/13.78 dB; normalization prevents acoustic-level or severity interpretation.","priority":"medium"},"modulation_map":{"evidence_ids":["R01.modulation_map","R02.modulation_map","R03.modulation_map","R04.modulation_map"],"finding":"Across the five common carrier bands, map maxima rank R01 0.252, R02 0.142, R04 0.122, R03 0.0839; the carrier-specific envelope views show that these maxima occur at different modulation frequencies, so the scalar ranking alone is incomplete.","priority":"medium"},"orders":{"evidence_ids":["R01.orders","R02.orders","R03.orders","R04.orders"],"finding":"Unavailable for every reference because no independent RPM trace was supplied; no order-domain comparison or shaft-order claim is possible.","priority":"unavailable"},"persistence":{"evidence_ids":["R01.persistence","R02.persistence","R03.persistence","R04.persistence"],"finding":"Persistence distributions show different stable spectral populations, with maximum occupancy R01 28.0%, R02 33.0%, R03 42.8%, and R04 21.9%; chronology is discarded and overlapping frames are dependent.","priority":"medium"},"reassigned":{"evidence_ids":["R01.reassigned","R02.reassigned","R03.reassigned","R04.reassigned"],"finding":"Reassigned maps retain at least 99.76% of weight and localize the same persistent low-frequency and ~16 kHz structures seen in STFT/FFT; R03 is low-frequency dominated, while R04 has the strongest dense ~16 kHz ridge family.","priority":"low"},"ridge":{"evidence_ids":["R01.ridge","R02.ridge","R03.ridge","R04.ridge"],"finding":"Dominant-bin medians are R01 16.0 kHz, R02 50 Hz, R03 20 Hz, and R04 50 Hz; frequent switches to ~16 kHz occur in R02/R04 and rare switches in R03, but this strongest-bin statistic is source-selective rather than a validated speed trace.","priority":"medium"},"spectral_kurtosis":{"evidence_ids":["R01.spectral_kurtosis","R02.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis"],"finding":"Uncorrected maxima occur at R01 2.14 kHz/5.87, R02 290 Hz/22.07, R03 390 Hz/23.25, and R04 15.66 kHz/2.52. This sharply separates intermittent-frequency regions, although regime changes and background impacts remain alternatives.","priority":"high"},"stft":{"evidence_ids":["R01.stft","R02.stft","R03.stft","R04.stft"],"finding":"All records contain persistent narrow spectral structures, but R01 emphasizes periodic low-frequency activity plus ~16 kHz, R02 has a persistent ~16 kHz line with changing broadband floor, R03 is dominated by stable low-frequency content with several midband lines, and R04 shows dense persistent ~16 kHz energy plus lower lines.","priority":"medium"},"stft_detail":{"evidence_ids":["R01.stft_detail","R02.stft_detail","R03.stft_detail","R04.stft_detail"],"finding":"At 0.1 s frames and 10 Hz bins, physical STFTs confirm reference-specific stationary line patterns rather than major drifts: R01 periodic low-band structure, R02 low-band plus ~16 kHz, R03 strong low-band/midband lines, and R04 especially prominent ~16 kHz content.","priority":"medium"},"synchronous":{"evidence_ids":["R01.synchronous","R02.synchronous","R03.synchronous","R04.synchronous"],"finding":"Unavailable for every reference because no independent RPM trace was supplied; phase-locked waveform comparison cannot be made.","priority":"unavailable"},"wavelet":{"evidence_ids":["R01.wavelet","R02.wavelet","R03.wavelet","R04.wavelet"],"finding":"The Morlet views, limited to 4.41 kHz after resampling, show strongest periodic low-frequency activity in R01, comparatively weaker low/midband structure in R02, strong low-frequency energy with intermittent midband patches in R03, and repeating low-frequency patches in R04; they cannot assess the ~16 kHz features.","priority":"medium"},"welch":{"evidence_ids":["R01.welch","R01.C01.S0001","R02.welch","R02.C01.S0001","R03.welch","R03.C01.S0001","R04.welch","R04.C01.S0001"],"finding":"With 43.07 Hz bins and 64.60 Hz ENBW, centroids are R01 5.70 kHz, R02 6.44 kHz, R03 0.638 kHz, and R04 9.93 kHz. R03 is overwhelmingly low-frequency; R04 is most high-frequency weighted; R01/R02 retain strong ~16 kHz humps but differ in midband power.","priority":"high"}},"limitations":"This is a reference-only survey with one peak-normalized 10 s example per labeled condition; no query or query distances were supplied, so no classification or acceptance claim is possible. The four records may reflect different acquisition, microphone response, placement, reverberation, speed, or load, and gain normalization does not remove those effects. FFT, Welch, bands, STFT, persistence, reassignment, ridge, wavelet, envelope, modulation, cepstrum, kurtosis, and cyclic views are correlated transforms of each same signal rather than independent votes. Welch/STFT overlap reduces independence; persistence loses chronology; spectral-kurtosis values are finite-sample and uncorrected; cyclic coherence lacks a significance threshold. Orders and synchronous averages are unavailable for all references because RPM was absent. The wavelet upper limit excludes ~16 kHz evidence. Any apparent fault-specific predictiveness is tentative and cannot establish fault identity, affected component, or severity from one reference per class.","most_informative":["bands","fft","envelope_2","kurtosis_bank","spectral_kurtosis"]}},{"dataset":"Jin","name":"survey-4bf405b6bcc03509","result":{"diagram_review":{"autocorrelation":{"evidence_ids":["R02.C01.S0001","R02.autocorrelation","R03.C01.S0001","R03.autocorrelation","R04.C01.S0001","R04.autocorrelation"],"finding":"R02 has a 0.944 peak at 0.249 ms and a damped/slowly varying curve; R03 has the highest coefficient, 0.969 at 0.181 ms, with a broad smooth decay; R04 has 0.894 at 0.249 ms and conspicuous oscillatory lobes. The sub-millisecond peaks are not validated shaft periods.","priority":"medium"},"bands":{"evidence_ids":["R02.C01.S0001","R02.bands","R03.C01.S0001","R03.bands","R04.C01.S0001","R04.bands"],"finding":"Relative Welch power differs strongly: R03 is dominated by 0\u2013100 Hz (94.13%); R02 splits power among 0\u2013100 Hz (59.72%), 8\u201316 kHz (14.59%), and 16\u201322.05 kHz (24.64%); R04 is most high-frequency weighted, with 19.48% at 8\u201316 kHz and 41.59% above 16 kHz.","priority":"high"},"cepstrum":{"evidence_ids":["R02.cepstrum","R03.cepstrum","R04.cepstrum"],"finding":"R02 has the largest reported positive coefficient (0.257 at 45.35 \u00b5s); R03 is weaker overall (maximum 0.0878 at 181.41 \u00b5s); R04 is intermediate (maximum 0.124 at 68.03 \u00b5s). Shared peaks near 0.181, 0.249, and 0.431 ms limit uniqueness, and none establishes RPM.","priority":"medium"},"cyclic":{"evidence_ids":["R02.cyclic","R03.cyclic","R04.cyclic"],"finding":"With the same 10 Hz cyclic grid and 397 frames, R04 reaches markedly higher squared coherence (0.299) than R02 (0.138) or R03 (0.131), including a visible horizontal feature near cyclic frequency 140 Hz; no significance threshold is supplied and deterministic tones can contribute.","priority":"medium"},"envelope":{"evidence_ids":["R02.C01.S0001","R02.envelope","R03.C01.S0001","R03.envelope","R04.C01.S0001","R04.envelope"],"finding":"Broadband modulation differs: R02 reports peaks near 2.93, 6.46, 13.03, 19.49, 22.73, and 35.76 Hz; R03 emphasizes 0.20\u20133.33 Hz plus 43.23\u201344.24 Hz; R04 has a strong 13.23 Hz peak followed by 23.13, 9.90, 3.33, and 36.36 Hz. Broadband beating remains a possible source.","priority":"medium"},"envelope_1":{"evidence_ids":["R02.envelope_1","R03.envelope_1","R04.envelope_1"],"finding":"For the common 100\u2013500 Hz carrier band, strongest ordinary-envelope modulation is 3.265 Hz in R02 (0.280), 3.571 Hz in R03 (0.214), and 3.265 Hz in R04 (0.164); R03 uniquely has strong near-zero 0.102 Hz squared-envelope content, while R04 adds 9.90 and 13.16 Hz peaks.","priority":"medium"},"envelope_2":{"evidence_ids":["R02.envelope_2","R03.envelope_2","R04.envelope_2"],"finding":"For 500\u20131500 Hz, R02 is led by 6.531 Hz (ordinary 0.350), R03 by 81.327 Hz (0.239), and R04 by a substantially stronger 6.633 Hz line (0.552; squared 0.907). This is one of the clearest carrier-specific modulation differences.","priority":"high"},"envelope_3":{"evidence_ids":["R02.envelope_3","R03.envelope_3","R04.envelope_3"],"finding":"For 1.5\u20134 kHz, dominant modulation separates as 337.755 Hz in R02 (0.192), 187.347 Hz in R03 (0.213), and 171.735 Hz in R04 (0.178); secondary families also differ (R02 253.3/84.5 Hz, R03 21.2/208.5 Hz, R04 26.4/151.9 Hz).","priority":"high"},"envelope_4":{"evidence_ids":["R02.envelope_4","R03.envelope_4","R04.envelope_4"],"finding":"For 4\u20138 kHz, R02 is led by 12.959 Hz (0.0816), R03 by 212.041 Hz (0.105) with 56.53 and 28.27 Hz, and R04 by 26.429 Hz (0.0787) with 198.16 and 171.73 Hz. Peak amplitudes are modest and the transforms remain correlated with the underlying spectra.","priority":"medium"},"envelope_5":{"evidence_ids":["R02.envelope_5","R03.envelope_5","R04.envelope_5"],"finding":"For 8\u201316 kHz, R02 has strong 6.531 Hz modulation (ordinary 0.352; squared 0.770), R03 is much weaker and led by 7.041 Hz (0.110) with 63.57/212.04 Hz, and R04 is led by 13.163\u201313.265 Hz (0.330/0.312; squared 0.658/0.623).","priority":"high"},"estimators":{"evidence_ids":["R02.estimators","R03.estimators","R04.estimators"],"finding":"All three estimator families preserve the main PSD differences: R03 is strongest at 0\u201320 Hz and has integrated multitaper power 0.1178; R02 has 0.0250 with a strong approximately 16 kHz region; R04 has 0.0378 and comparable low-frequency and approximately 16.02 kHz maxima. Mean/median/multitaper disagreements are limited relative to these shape differences.","priority":"high"},"fft":{"evidence_ids":["R02.C01.S0001","R02.fft","R03.C01.S0001","R03.fft","R04.C01.S0001","R04.fft"],"finding":"At common 0.1 Hz bin spacing and 0.15 Hz Hann bandwidth, R03 has a dominant 5.3 Hz peak at \u22126.89 dB FS and additional lines around 1.84\u20135.37 kHz; R02 also peaks at 5.3 Hz but much lower (\u221218.08 dB) and has a dense approximately 16 kHz cluster; R04 is led by 50 Hz (\u221219.53 dB) plus a strong approximately 16.0 kHz cluster.","priority":"high"},"harmonics":{"evidence_ids":["R02.harmonics","R03.harmonics","R04.harmonics"],"finding":"Prominent-peak spacing patterns differ: R02's largest pair counts are at 7 and 17 Hz (8 each), R03's at 1 Hz (12) with several candidate families from 22.3\u201349.3 Hz, and R04's at 1 and 13 Hz (18 each) with only a reported 49.2 Hz candidate family. Unrelated clustered tones can create these spacings.","priority":"medium"},"kurtosis_bank":{"evidence_ids":["R02.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank"],"finding":"R02 has the strongest high-frequency impulsiveness, peaking at 1.949 in 14.41\u201315.72 kHz and 1.824 in 15.72\u201317.02 kHz; R04 has moderate high-frequency excess kurtosis up to 0.745 near 15.72\u201317.02 kHz but negative low-band values down to \u22120.559; R03 peaks at 0.658 below 1.33 kHz and is near zero or negative over most high bands.","priority":"high"},"levels":{"evidence_ids":["R02.C01.S0001","R02.levels","R03.C01.S0001","R03.levels","R04.C01.S0001","R04.levels"],"finding":"All records are 10 s, peak-normalized to 0.94998 FS, unclipped, and essentially DC-free. R03 has the largest normalized RMS (0.3434) and lowest crest factor (8.84 dB), R04 is intermediate (0.1945; 13.78 dB), and R02 has the lowest RMS and highest crest factor (0.1580; 15.58 dB); normalization prevents acoustic-level or severity inference.","priority":"medium"},"modulation_map":{"evidence_ids":["R02.modulation_map","R03.modulation_map","R04.modulation_map"],"finding":"Across the same five carrier bands, map maxima are R02 0.142, R04 0.122, and R03 0.0839. R02 concentrates strong low-frequency modulation especially in the highest carrier band, R03 shows more distributed mid-frequency features, and R04 shows low-frequency/high-band structure; 155 masked cells per map are missing evidence.","priority":"medium"},"orders":{"evidence_ids":["R02.orders","R03.orders","R04.orders"],"finding":"Unavailable for every reference because no independent RPM trace was supplied; no order-domain between-reference comparison can be made.","priority":"unavailable"},"persistence":{"evidence_ids":["R02.persistence","R03.persistence","R04.persistence"],"finding":"R03 has the most concentrated persistence population (maximum 42.82%), R02 reaches 33.00%, and R04 21.91%. Their persistent spectral locations mirror the PSD differences\u2014R03 low-frequency rich and R02/R04 with strong approximately 16 kHz content\u2014but chronology is discarded and overlapping frames are dependent.","priority":"medium"},"reassigned":{"evidence_ids":["R02.reassigned","R03.reassigned","R04.reassigned"],"finding":"All retain at least 99.76% of weight and localize persistent ridges already seen in STFT/PSD: R02 and R04 emphasize approximately 16 kHz, while R03 also shows several lower-kHz lines. Reported maxima/medians differ (R02 \u221218.40/\u221266.53 dB, R03 \u221211.79/\u221259.54 dB, R04 \u221220.04/\u221259.45 dB), but pooled weight is not PSD.","priority":"low"},"ridge":{"evidence_ids":["R02.ridge","R03.ridge","R04.ridge"],"finding":"Dominant-bin switching differs: R03 has median 20 Hz with only occasional jumps to approximately 16.17\u201316.18 kHz; R02 and R04 have median 50 Hz and frequent jumps to approximately 16.0 kHz. The strongest-bin ridge has no continuity model and is not validated speed.","priority":"medium"},"spectral_kurtosis":{"evidence_ids":["R02.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis"],"finding":"R02 and R03 show very large low-frequency intermittency peaks (22.07 at 290 Hz and 23.25 at 390 Hz, respectively), whereas R04's maximum is only 2.52 near 15.66 kHz. The estimates are uncorrected for finite-sample bias and do not identify a fault by themselves.","priority":"high"},"stft":{"evidence_ids":["R02.stft","R03.stft","R04.stft"],"finding":"All show largely persistent horizontal components over 10 s. R02 has a strong approximately 16 kHz ridge with weaker low-frequency bands; R03 combines strong very-low-frequency energy with persistent lines around 2\u20136 kHz and approximately 16 kHz; R04 has the clearest approximately 16 kHz ridge plus several weaker low-kHz lines. No major drift is evident at the displayed resolution.","priority":"high"},"stft_detail":{"evidence_ids":["R02.stft_detail","R03.stft_detail","R04.stft_detail"],"finding":"At identical 0.1 s frames and 10 Hz bins, physical STFTs confirm stable approximately 16 kHz energy in R02/R04 and stronger multiple low-kHz lines in R03. Map maxima are \u221231.72, \u221226.09, and \u221232.67 dB respectively; R02 has a lower median floor (\u221278.28 dB) than R03/R04 (about \u221271.8 dB).","priority":"medium"},"synchronous":{"evidence_ids":["R02.synchronous","R03.synchronous","R04.synchronous"],"finding":"Unavailable for every reference because no independent RPM trace was supplied; phase-locked waveform evidence cannot be compared.","priority":"unavailable"},"wavelet":{"evidence_ids":["R02.wavelet","R03.wavelet","R04.wavelet"],"finding":"Within the common anti-aliased 0\u20134.41 kHz range, R03 and R04 have higher median coefficient power (\u221272.36 and \u221271.66 dB) than R02 (\u221280.24 dB); R03 shows stronger persistent low-kHz structures, while edge-support masking leaves 144 cells per map unavailable. The approximately 16 kHz feature is outside this wavelet analysis bandwidth.","priority":"medium"},"welch":{"evidence_ids":["R02.C01.S0001","R02.welch","R03.C01.S0001","R03.welch","R04.C01.S0001","R04.welch"],"finding":"Welch shape sharply separates the references: R03 has centroid 638 Hz and 94.13% of power below 100 Hz; R02 has centroid 6.44 kHz with substantial high-frequency power; R04 has centroid 9.93 kHz and 61.07% above 8 kHz. Absolute integrated powers differ but are affected by peak normalization and acquisition response.","priority":"high"}},"limitations":"This is a reference-only survey: no query signal or query distances were supplied, so no classification or acceptance conclusion is possible. Each label has only one supplied recording, so observed differences may reflect acquisition regime, microphone response, placement, reverberation, speed, or load rather than generalizable condition effects. Peak normalization preserves spectral shape but invalidates acoustic-level and severity comparisons. FFT, Welch, bands, STFT, persistence, reassignment, ridge, wavelet, envelope, cepstral, kurtosis, and cyclic views are correlated transforms of the same signals, not independent votes. Order and synchronous evidence are unavailable for all references because RPM is absent; modulation-map masked cells and wavelet edge masks are missing data. No bearing geometry, independent speed, cyclic-coherence significance threshold, or estimator confidence intervals are supplied, so fault component, mechanism, and severity cannot be established from these references alone.","most_informative":["bands","fft","envelope_2","kurtosis_bank","spectral_kurtosis"]}},{"dataset":"Ottawa","name":"survey-8ca096c2c8dd560f","result":{"diagram_review":{"autocorrelation":{"evidence_ids":["R01.autocorrelation","R02.autocorrelation","R03.autocorrelation","R04.autocorrelation","R05.autocorrelation","R06.autocorrelation","R07.autocorrelation","R08.autocorrelation"],"finding":"R02\u2013R08 have strong long-lag periodic peaks at 0.0654\u20130.0684 s (coefficients 0.971\u20130.983), whereas R01's reported strongest peak is 0.000381 s with coefficient 0.905 and its curve is much less sinusoidal. This distinguishes R01, but harmonic ambiguity prevents an RPM interpretation.","priority":"high"},"bands":{"evidence_ids":["R01.bands","R02.bands","R03.bands","R04.bands","R05.bands","R06.bands","R07.bands","R08.bands"],"finding":"R01 has only 68.27% of Welch power below 100 Hz and 26.65% at 100\u2013500 Hz, unlike R02\u2013R08 with 92.12\u201397.25% below 100 Hz and 2.53\u20137.11% at 100\u2013500 Hz. Among R02\u2013R08, R06 has the largest 500\u20131000 Hz fraction (0.570%) and R07 the smallest (0.098%).","priority":"high"},"cepstrum":{"evidence_ids":["R01.cepstrum","R02.cepstrum","R03.cepstrum","R04.cepstrum","R05.cepstrum","R06.cepstrum","R07.cepstrum","R08.cepstrum"],"finding":"All records are dominated by the first very-low-quefrency bins rather than a clearly reported nonzero spacing peak; the first-bin magnitude is unusually high for R01 (0.526) and next highest for R06 (0.206), versus 0.098\u20130.150 for R02\u2013R05/R07/R08. The common near-zero structure is weakly diagnostic and floor-dependent.","priority":"low"},"cyclic":{"evidence_ids":["R01.cyclic","R02.cyclic","R03.cyclic","R04.cyclic","R05.cyclic","R06.cyclic","R07.cyclic","R08.cyclic"],"finding":"Maximum cyclic coherence differs from 0.245\u20130.397 for R02/R04\u2013R06/R08 to 0.461 R07, 0.588 R01 and 0.699 R03; valid-cell counts also vary substantially (44,209\u201362,565) because low-power cells are masked. Deterministic tones and unequal masking limit interpretation.","priority":"medium"},"envelope":{"evidence_ids":["R01.envelope","R02.envelope","R03.envelope","R04.envelope","R05.envelope","R06.envelope","R07.envelope","R08.envelope"],"finding":"Broadband modulation families differ: R02 emphasizes about 30.10/45.15/15.05 Hz, R03 and R06 emphasize about 14.95 Hz, R04 about 29.90 Hz, R05 about 29.29/58.48 Hz, R07 about 30.61/15.35 Hz, R08 about 44.65/29.80 Hz, while R01 includes 0.303, 40.0 and 9.29 Hz. These are modulation observations, not verified shaft or bearing frequencies.","priority":"high"},"envelope_1":{"evidence_ids":["R01.envelope_1","R02.envelope_1","R03.envelope_1","R04.envelope_1","R05.envelope_1","R06.envelope_1","R07.envelope_1","R08.envelope_1"],"finding":"In the common 100\u2013500 Hz carrier band, strongest modulation is record-specific: 3.88 Hz R01, 90.20 R02, 89.90 R03, 44.90 R04, 102.45 R05, 15.00 R06, 15.31 R07 and 118.98 Hz R08. Normalized ordinary-envelope maxima span 0.145\u20130.382.","priority":"high"},"envelope_2":{"evidence_ids":["R01.envelope_2","R02.envelope_2","R03.envelope_2","R04.envelope_2","R05.envelope_2","R06.envelope_2","R07.envelope_2","R08.envelope_2"],"finding":"For the 500\u20131500 Hz carrier band, strongest modulation is 6.53 Hz R01, 0.51 R02, 15.0 R03, 29.90 R04, 29.29 R05, 29.90 R06, 33.78 R07 and 14.90 Hz R08; R03/R04/R06 have especially strong normalized peaks (0.203\u20130.239).","priority":"medium"},"envelope_3":{"evidence_ids":["R01.envelope_3","R02.envelope_3","R03.envelope_3","R04.envelope_3","R05.envelope_3","R06.envelope_3","R07.envelope_3","R08.envelope_3"],"finding":"For the 1500\u20134000 Hz carrier band, R03 is conspicuous with 15 Hz modulation amplitude 0.193; other strongest peaks are much smaller (0.037\u20130.082) at 6.53 R01, 0.51 R02, 15 R04, 58.57 R05, 14.90 R06, 1.12 R07 and 29.80 Hz R08.","priority":"medium"},"envelope_4":{"evidence_ids":["R01.envelope_4","R02.envelope_4","R03.envelope_4","R04.envelope_4","R05.envelope_4","R06.envelope_4","R07.envelope_4","R08.envelope_4"],"finding":"The 4000\u20138000 Hz envelopes show narrow, differing families: strongest modulation is 61.73 R01, 91.73 R02, 61.02 R03, 30.41 R04, 29.90 R05, 61.12 R06, 62.35 R07 and 30.31 Hz R08. R05 and R08 have the largest ordinary-envelope peaks (0.487 and 0.436), while R06 is lowest (0.212).","priority":"high"},"envelope_5":{"evidence_ids":["R01.envelope_5","R02.envelope_5","R03.envelope_5","R04.envelope_5","R05.envelope_5","R06.envelope_5","R07.envelope_5","R08.envelope_5"],"finding":"For the 8000\u201316000 Hz carrier band, strongest modulation is 61.73 R01, 183.47 R02, 15.0 R03, 273.98 R04, 89.80 R05, 14.90 R06, 187.04 R07 and 181.73 Hz R08. R03 has the largest normalized peak (0.328), with R08 the smallest (0.093).","priority":"medium"},"estimators":{"evidence_ids":["R01.estimators","R02.estimators","R03.estimators","R04.estimators","R05.estimators","R06.estimators","R07.estimators","R08.estimators"],"finding":"Welch and DPSS agree on low-frequency dominance but differ through their 60 Hz smoothing: integrated multitaper power is lowest for R01 (0.0367), then R06 (0.1136), and highest for R08/R02/R05 (0.216/0.212/0.205). R01 additionally has a much higher median spectral floor than most references; absolute power remains affected by peak normalization and acquisition response.","priority":"medium"},"fft":{"evidence_ids":["R01.fft","R02.fft","R03.fft","R04.fft","R05.fft","R06.fft","R07.fft","R08.fft"],"finding":"The strongest resolved low-frequency peaks are 45.9 Hz R01, 45.1 R02, 45.0 R03, 44.9 R04, 43.9 R05, 44.9 R06, 45.9 R07 and 44.6 Hz R08, but secondary-line patterns differ markedly: R04 has a strong 74.8 Hz line, R03/R06 strong 30 Hz lines, R07 a 76.5 Hz line, and R08 nearly equal 74.4/89.3 Hz lines. R01's main line is much weaker after common peak normalization and its broad/high-frequency content is greater.","priority":"high"},"harmonics":{"evidence_ids":["R01.harmonics","R02.harmonics","R03.harmonics","R04.harmonics","R05.harmonics","R06.harmonics","R07.harmonics","R08.harmonics"],"finding":"All references contain dense peak-pair spacing evidence near 15 and about 29\u201331/44\u201346 Hz, but counts and families vary: the 15 Hz count ranges from 11 (R05/R06) to 24 (R07), while R02 has a strong 30.1 Hz candidate with eight matches and R06 has an 11-match 29.9 Hz family. These candidates may combine unrelated tones and do not establish fundamentals.","priority":"medium"},"kurtosis_bank":{"evidence_ids":["R01.kurtosis_bank","R02.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank","R05.kurtosis_bank","R06.kurtosis_bank","R07.kurtosis_bank","R08.kurtosis_bank"],"finding":"High-frequency filter-bank impulsiveness strongly separates R03 (median 0.506, maximum 1.850) and R06 (median 0.352, maximum 1.043) from R01/R07 (negative medians), with R02/R04/R05/R08 intermediate. R03's maximum is at 12.48\u201314.97 kHz; this indicates intermittency, not fault identity.","priority":"high"},"levels":{"evidence_ids":["R01.levels","R02.levels","R03.levels","R04.levels","R05.levels","R06.levels","R07.levels","R08.levels"],"finding":"No reference is clipped or silent. R01 has a large DC offset (0.148 FS), high crest factor (11.93 dB) and a brief level disturbance near 2.1 s; R04 and R06 report numerous level changes, while R02/R03/R05/R07/R08 report none. RMS differences are not severity evidence because all files are peak-normalized.","priority":"medium"},"modulation_map":{"evidence_ids":["R01.modulation_map","R02.modulation_map","R03.modulation_map","R04.modulation_map","R05.modulation_map","R06.modulation_map","R07.modulation_map","R08.modulation_map"],"finding":"Across the five common carrier bands, map maxima are largest for R05 (0.440) and R03 (0.232), lowest for R07 (0.078), and 0.097\u20130.169 for the others; the bright cells correspond to the same band-envelope peaks rather than independent evidence. Each map has 1,100 valid of 1,255 cells, so masked cells are missing.","priority":"medium"},"orders":{"evidence_ids":["R01.orders","R02.orders","R03.orders","R04.orders","R05.orders","R06.orders","R07.orders","R08.orders"],"finding":"Unavailable for every reference because no independent RPM trace was supplied; frequency peaks cannot be converted into validated orders.","priority":"unavailable"},"persistence":{"evidence_ids":["R01.persistence","R02.persistence","R03.persistence","R04.persistence","R05.persistence","R06.persistence","R07.persistence","R08.persistence"],"finding":"All references show persistent low-frequency populations and persistent narrow high-frequency lines; maximum occupancy differs from 50.38% R01 to 92.70% R05/R08. R01 has a broader level distribution, while R04/R05/R07/R08 contain highly persistent narrow populations; chronology is unavailable and overlapping frames are dependent.","priority":"medium"},"reassigned":{"evidence_ids":["R01.reassigned","R02.reassigned","R03.reassigned","R04.reassigned","R05.reassigned","R06.reassigned","R07.reassigned","R08.reassigned"],"finding":"All records retain at least 99.73% of reassigned weight and show stationary horizontal ridges. R01 has the least retained weight (0.9973) and a higher median pooled weight (-76.0 dB) than the others (-78.4 to -82.6 dB), consistent with its broader spectrum; this is correlated with STFT/FFT evidence.","priority":"low"},"ridge":{"evidence_ids":["R01.ridge","R02.ridge","R03.ridge","R04.ridge","R05.ridge","R06.ridge","R07.ridge","R08.ridge"],"finding":"R02\u2013R08 dominant bins remain almost entirely at 40\u201350 Hz, except R04 intermittently reaches 80 Hz; R01 switches broadly among 40\u2013100 Hz with occasional 210\u2013220 Hz selections. The 10 Hz STFT bin spacing and absence of a continuity/RPM model limit physical interpretation.","priority":"high"},"spectral_kurtosis":{"evidence_ids":["R01.spectral_kurtosis","R02.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis","R05.spectral_kurtosis","R06.spectral_kurtosis","R07.spectral_kurtosis","R08.spectral_kurtosis"],"finding":"R01 has the largest reported maximum (1.255 at 10 Hz); R02 peaks at 0.915 near 620 Hz, R03 at 0.872 near 15.13 kHz, and R04\u2013R08 peak at 0.511\u20130.600 in differing bands. Unequal low-energy exclusions (2\u2013544 bins), no bias correction and no significance threshold reduce direct comparability.","priority":"medium"},"stft":{"evidence_ids":["R01.stft","R02.stft","R03.stft","R04.stft","R05.stft","R06.stft","R07.stft","R08.stft"],"finding":"All references are predominantly stationary over 10 s with persistent horizontal spectral lines. R01 uniquely shows broader low-frequency energy and a visible mid-record disturbance, while R04/R06 show subtler amplitude-regime changes; no strong continuous drift is evident at the 0.1 s frame scale.","priority":"medium"},"stft_detail":{"evidence_ids":["R01.stft_detail","R02.stft_detail","R03.stft_detail","R04.stft_detail","R05.stft_detail","R06.stft_detail","R07.stft_detail","R08.stft_detail"],"finding":"At 10 Hz bin spacing, R01 has the highest median displayed PSD (-88.33 dB) and broadest background; R02\u2013R08 medians range from -90.35 to -94.37 dB. Horizontal lines are largely stationary, so this view mainly refines the correlated STFT/PSD differences.","priority":"medium"},"synchronous":{"evidence_ids":["R01.synchronous","R02.synchronous","R03.synchronous","R04.synchronous","R05.synchronous","R06.synchronous","R07.synchronous","R08.synchronous"],"finding":"Unavailable for every reference because no independent RPM trace was supplied; phase-locked waveform evidence cannot be assessed.","priority":"unavailable"},"wavelet":{"evidence_ids":["R01.wavelet","R02.wavelet","R03.wavelet","R04.wavelet","R05.wavelet","R06.wavelet","R07.wavelet","R08.wavelet"],"finding":"The 14 kHz auxiliary analyses cover only to 4.2 kHz and show broadly stationary low-frequency power in all records. R01 has the weakest maximum (-35.31 dB) but highest median (-60.65 dB) relative to several references, whereas R02/R03/R05/R08 have maxima around -26 to -27 dB; edge-masked cells are unavailable and wavelet power is not PSD.","priority":"low"},"welch":{"evidence_ids":["R01.welch","R02.welch","R03.welch","R04.welch","R05.welch","R06.welch","R07.welch","R08.welch"],"finding":"R01 is spectrally broadest, with centroid 212.5 Hz and flatness 0.00307, versus centroids 51.8\u201365.9 Hz and flatness 0.000076\u20130.000225 for R02\u2013R08. R02/R03/R05/R07/R08 are especially dominated by the lowest Welch bin region, while R04/R06 retain relatively more 100\u2013500 Hz power; 41.0 Hz bins and 61.5 Hz bandwidth cannot resolve the nearby FFT lines.","priority":"high"}},"limitations":"This is a reference-only survey: no query signal or query distances were supplied, so no classification or acceptance claim is possible. Each condition has only one peak-normalized recording, preventing separation of condition effects from microphone response, placement, reverberation, speed, load and acquisition regime; predictiveness is therefore tentative. Multiple diagrams are correlated transforms of the same signals and are not independent votes. RPM-dependent orders and synchronous averages are unavailable for all references. Cyclic maps contain unequal low-power masking, modulation maps contain masked cells, wavelets contain edge masks, and spectral-kurtosis exclusions differ. Numeric summaries were preferred over pixels. Several observations conflict with generic fault expectations\u2014for example, the nominally healthy R01 is the broadest-spectrum record and has the largest spectral-kurtosis maximum\u2014so no feature here establishes fault identity, component or severity.","most_informative":["fft","bands","autocorrelation","envelope_4","kurtosis_bank"]}},{"dataset":"Ottawa","name":"survey-ab7169c4a978e367","result":{"diagram_review":{"autocorrelation":{"evidence_ids":["R02.autocorrelation","R03.autocorrelation","R04.autocorrelation","R05.autocorrelation","R06.autocorrelation","R07.autocorrelation","R08.autocorrelation"],"finding":"All references are strongly periodic but differ slightly in strongest positive lag: R02 0.06650 s/0.9786, R03 0.06671/0.9789, R04 0.06683/0.9828, R05 0.06838/0.9717, R06 0.06688/0.9714, R07 0.06538/0.9790, R08 0.06721/0.9798. Harmonic ambiguity precludes an RPM interpretation.","priority":"medium"},"bands":{"evidence_ids":["R02.bands","R03.bands","R04.bands","R05.bands","R06.bands","R07.bands","R08.bands"],"finding":"All are low-frequency dominated, but 0\u2013100 Hz fractions range from 92.12% R06 and 92.46% R04 to 97.25% R02; R04/R06 have the largest 100\u2013500 Hz fractions (7.18%/7.11%), while R07 has notably little 500\u20134000 Hz power. These identical bands provide useful shape differences despite normalization.","priority":"high"},"cepstrum":{"evidence_ids":["R02.cepstrum","R03.cepstrum","R04.cepstrum","R05.cepstrum","R06.cepstrum","R07.cepstrum","R08.cepstrum"],"finding":"Reported maxima are all at the first nonzero quefrency (23.81 \u00b5s), indicating broadband/log-floor structure rather than a resolved long-quefrency family; amplitudes differ, highest R06 0.2056 and lowest R02 0.0984. No validated spacing or RPM follows.","priority":"low"},"cyclic":{"evidence_ids":["R02.cyclic","R03.cyclic","R04.cyclic","R05.cyclic","R06.cyclic","R07.cyclic","R08.cyclic"],"finding":"Maximum squared coherence differs substantially: R03 0.6995, R07 0.4607, R05 0.3967, R04 0.3118, R02 0.2813, R06 0.2509, R08 0.2446; valid-cell counts also vary because low-power cells are masked. Deterministic tones can produce this evidence and no significance threshold is supplied.","priority":"medium"},"envelope":{"evidence_ids":["R02.envelope","R03.envelope","R04.envelope","R05.envelope","R06.envelope","R07.envelope","R08.envelope"],"finding":"Broadband-envelope leaders differ: R02 30.10 Hz, R03/R06 about 14.95 Hz, R04/R05 about 29.9 Hz, R07 30.61 Hz with 15.35 Hz also strong, and R08 44.65 Hz. These are modulation observations, not independently established shaft or fault frequencies.","priority":"high"},"envelope_1":{"evidence_ids":["R02.envelope_1","R03.envelope_1","R04.envelope_1","R05.envelope_1","R06.envelope_1","R07.envelope_1","R08.envelope_1"],"finding":"For the common 100\u2013500 Hz carrier, strongest modulation is R02 90.20, R03 89.90, R04 44.90, R05 102.45, R06 15.00, R07 15.31, and R08 118.98 Hz; R07 has the largest normalized ordinary-envelope peak (0.3824).","priority":"high"},"envelope_2":{"evidence_ids":["R02.envelope_2","R03.envelope_2","R04.envelope_2","R05.envelope_2","R06.envelope_2","R07.envelope_2","R08.envelope_2"],"finding":"For 500\u20131500 Hz, leaders are R02 0.51, R03 15.00, R04 29.90, R05 29.29, R06 29.90, R07 33.78, and R08 14.90 Hz; R04 has the largest ordinary-envelope peak (0.2392), with R02 dominated by very-low modulation.","priority":"medium"},"envelope_3":{"evidence_ids":["R02.envelope_3","R03.envelope_3","R04.envelope_3","R05.envelope_3","R06.envelope_3","R07.envelope_3","R08.envelope_3"],"finding":"For 1500\u20134000 Hz, leaders are R02 0.51, R03 15.00, R04 15.00, R05 58.57, R06 14.90, R07 1.12, and R08 29.80 Hz. R03 is strongest in normalized amplitude (0.1926), whereas R02/R07 emphasize near-zero modulation.","priority":"medium"},"envelope_4":{"evidence_ids":["R02.envelope_4","R03.envelope_4","R04.envelope_4","R05.envelope_4","R06.envelope_4","R07.envelope_4","R08.envelope_4"],"finding":"The 4000\u20138000 Hz envelope is sharply differentiating: leaders are R02 91.73, R03 61.02, R04 30.41, R05 29.90, R06 61.12, R07 62.35, R08 30.31 Hz. R05 has the largest ordinary normalized peak (0.4869), followed by R08 0.4357.","priority":"high"},"envelope_5":{"evidence_ids":["R02.envelope_5","R03.envelope_5","R04.envelope_5","R05.envelope_5","R06.envelope_5","R07.envelope_5","R08.envelope_5"],"finding":"For 8000\u201316000 Hz, leaders are R02 183.47, R03 15.00, R04 273.98, R05 89.80, R06 14.90, R07 187.04, and R08 181.73 Hz. R03 has the largest normalized ordinary peak (0.3283), closely followed by R07 0.3079.","priority":"high"},"estimators":{"evidence_ids":["R02.estimators","R03.estimators","R04.estimators","R05.estimators","R06.estimators","R07.estimators","R08.estimators"],"finding":"With identical five-taper/60 Hz smoothing, integrated multitaper power ranges from R06 0.1136 to R08 0.2161; Welch mean/median retain reference-specific low-frequency shapes while DPSS broadens them. R04 uniquely has strongest DPSS bin at 60 Hz; others peak around 40\u201350 Hz.","priority":"medium"},"fft":{"evidence_ids":["R02.fft","R03.fft","R04.fft","R05.fft","R06.fft","R07.fft","R08.fft"],"finding":"At 0.1 Hz bins/0.15 Hz Hann bandwidth, dominant peaks are R02 45.1 Hz \u22124.29 dBFS, R03 45.0 \u22125.95, R04 44.9 \u22128.18, R05 43.9 \u22124.33, R06 44.9 \u22129.47, R07 45.9 \u22126.24, R08 44.6 \u22124.74. Secondary-line patterns differ strongly, e.g. R03/R06 have relatively strong \u224830 Hz, R04 has 74.8 Hz nearly comparable, and R08 has strong 74.4/89.3 Hz.","priority":"high"},"harmonics":{"evidence_ids":["R02.harmonics","R03.harmonics","R04.harmonics","R05.harmonics","R06.harmonics","R07.harmonics","R08.harmonics"],"finding":"Peak-pair histograms all favor about 15 Hz spacing but counts differ: R02 21, R03 17, R04 23, R05 11, R06 11, R07 24, R08 18. Candidate families and prominent peaks differ, but unrelated tones can create these spacings and no fundamental is established.","priority":"medium"},"kurtosis_bank":{"evidence_ids":["R02.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank","R05.kurtosis_bank","R06.kurtosis_bank","R07.kurtosis_bank","R08.kurtosis_bank"],"finding":"High-frequency complex excess kurtosis separates references: maxima/medians are R02 0.534/\u22120.006, R03 1.850/0.506, R04 0.748/0.077, R05 0.592/0.060, R06 1.043/0.352, R07 0.502/\u22120.082, R08 0.258/0.021. R03 is most broadly impulsive above about 7.5 kHz, followed by R06; this is not fault identity by itself.","priority":"high"},"levels":{"evidence_ids":["R02.levels","R03.levels","R04.levels","R05.levels","R06.levels","R07.levels","R08.levels"],"finding":"All 10 s peak-normalized records are non-silent and unclipped. RMS spans R06 0.3375 FS to R08 0.4680 FS; DC offsets span 0.0161\u20130.0538 FS. Reported >6 dB level changes occur only in R04 and R06, while R02/R03/R05/R07/R08 report none; normalization prevents acoustic-level or severity inference.","priority":"medium"},"modulation_map":{"evidence_ids":["R02.modulation_map","R03.modulation_map","R04.modulation_map","R05.modulation_map","R06.modulation_map","R07.modulation_map","R08.modulation_map"],"finding":"Across the same five carrier bands, map maxima are R02 0.1049, R03 0.2321, R04 0.1696, R05 0.4404, R06 0.1107, R07 0.0784, R08 0.1594. The maps compactly expose different cross-band modulation patterns, especially R05's strong 4\u20138 kHz \u224830 Hz component; masked cells are missing, not zero.","priority":"high"},"orders":{"evidence_ids":["R02.orders","R03.orders","R04.orders","R05.orders","R06.orders","R07.orders","R08.orders"],"finding":"Unavailable for every reference because no independent RPM trace was supplied; no order-domain comparison is possible.","priority":"unavailable"},"persistence":{"evidence_ids":["R02.persistence","R03.persistence","R04.persistence","R05.persistence","R06.persistence","R07.persistence","R08.persistence"],"finding":"Maximum frame occupancy differs (R02 73.8%, R03 54.7%, R04 86.1%, R05 92.7%, R06 72.5%, R07 88.4%, R08 92.7%), and the plots show differing persistent spectral ridges. Chronology is discarded and overlapping frames are dependent.","priority":"medium"},"reassigned":{"evidence_ids":["R02.reassigned","R03.reassigned","R04.reassigned","R05.reassigned","R06.reassigned","R07.reassigned","R08.reassigned"],"finding":"All retain more than 99.99% weight and show the same reference-specific stationary line structures seen in STFT/FFT; maximum pooled weights range from \u221215.29 dB R06 to \u221211.96 dB R08. This is correlated localization evidence, not an independent vote.","priority":"low"},"ridge":{"evidence_ids":["R02.ridge","R03.ridge","R04.ridge","R05.ridge","R06.ridge","R07.ridge","R08.ridge"],"finding":"Eligible dominant bins are confined to 40/50 Hz for R02, R03, R05, R06, R07 and R08; medians are 50 Hz for R02/R07 and 40 Hz for the others. R04 is distinctive, switching among 40/50/80 Hz with median 50 Hz. Ten-hertz binning and lack of continuity/RPM validation limit interpretation.","priority":"medium"},"spectral_kurtosis":{"evidence_ids":["R02.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis","R05.spectral_kurtosis","R06.spectral_kurtosis","R07.spectral_kurtosis","R08.spectral_kurtosis"],"finding":"Largest positive values occur at different frequencies: R02 620 Hz/0.915, R03 15.13 kHz/0.872, R04 18.86 kHz/0.511, R05 1.51 kHz/0.515, R06 20.83 kHz/0.600, R07 7.44 kHz/0.554, R08 16.46 kHz/0.582. Excluded low-energy-bin counts vary from 10 to 544 and finite-sample bias is uncorrected.","priority":"high"},"stft":{"evidence_ids":["R02.stft","R03.stft","R04.stft","R05.stft","R06.stft","R07.stft","R08.stft"],"finding":"The overview spectrograms show mostly persistent horizontal components in every reference rather than isolated broadband events; R04 and R06 exhibit more visible temporal variation, consistent with their reported level changes. This view is correlated with STFT detail, persistence and reassignment.","priority":"medium"},"stft_detail":{"evidence_ids":["R02.stft_detail","R03.stft_detail","R04.stft_detail","R05.stft_detail","R06.stft_detail","R07.stft_detail","R08.stft_detail"],"finding":"At common 0.1 s frames and 10 Hz bins, map maxima/medians are R02 \u221225.72/\u221294.37, R03 \u221226.69/\u221290.35, R04 \u221227.22/\u221294.03, R05 \u221225.85/\u221292.51, R06 \u221228.71/\u221292.18, R07 \u221227.61/\u221293.88, R08 \u221225.77/\u221292.31 dB. R04/R06 show the clearest time variation; fine FFT line separations are unresolved here.","priority":"medium"},"synchronous":{"evidence_ids":["R02.synchronous","R03.synchronous","R04.synchronous","R05.synchronous","R06.synchronous","R07.synchronous","R08.synchronous"],"finding":"Unavailable for every reference because no independent RPM trace was supplied; phase-locked waveform comparison cannot be made.","priority":"unavailable"},"wavelet":{"evidence_ids":["R02.wavelet","R03.wavelet","R04.wavelet","R05.wavelet","R06.wavelet","R07.wavelet","R08.wavelet"],"finding":"Morlet maps to 4.2 kHz show common low-frequency dominance; R03/R04/R06 display more visible temporal texture than the steadier references. Median coefficient powers differ from \u221265.11 dB R04 to \u221259.58 dB R05, but these coefficients are not PSD and this view is correlated with STFT evidence.","priority":"low"},"welch":{"evidence_ids":["R02.welch","R03.welch","R04.welch","R05.welch","R06.welch","R07.welch","R08.welch"],"finding":"Common Welch settings (41.02 Hz bins, 61.52 Hz bandwidth) show low-frequency dominance with integrated powers R02 0.2118, R03 0.1737, R04 0.1579, R05 0.2051, R06 0.1136, R07 0.1452, R08 0.2162 FS\u00b2. Spectral centroids range 51.85 Hz R02 to 65.85 Hz R04; R06 has the highest flatness (0.000225). Coarse bandwidth blends the narrow FFT lines.","priority":"high"}},"limitations":"This is a reference-only survey: no query signal or query distances were supplied, so no classification or acceptance decision is possible. Each condition has only one supplied reference, so observed differences may reflect acquisition regime, microphone response/placement, reverberation, speed, load or normalization rather than class-general signatures. All records are derived peak-normalized; absolute acoustic level and severity are not comparable. The 26 views are correlated transforms of the same signals, not independent votes. Orders and synchronous averages are unavailable for every reference due to absent RPM. Cyclic low-power cells are masked and therefore missing; no cyclic-coherence significance threshold is supplied. Welch/STFT resolution, smoothing, filter bands, edge exclusions, floors and wavelet scaling must be retained in any later comparison. Impulsiveness, modulation, harmonic spacing and nearest-reference similarity alone cannot establish fault identity, component or severity.","most_informative":["fft","bands","modulation_map","kurtosis_bank","spectral_kurtosis"]}},{"dataset":"Ottawa","name":"survey-150908d5ac073e21","result":{"diagram_review":{"autocorrelation":{"evidence_ids":["R01.autocorrelation","R01.C01.S0001","R03.autocorrelation","R03.C01.S0001","R04.autocorrelation","R04.C01.S0001","R05.autocorrelation","R05.C01.S0001","R06.autocorrelation","R06.C01.S0001","R07.autocorrelation","R07.C01.S0001","R08.autocorrelation","R08.C01.S0001"],"finding":"R01 is the clear outlier: strongest positive peak at 0.000381 s with coefficient 0.905, versus long, strongly periodic peaks at 0.06538\u20130.06838 s and coefficients 0.971\u20130.983 for R03/R04/R05/R06/R07/R08. Among the latter, lag differences are small and may reflect speed differences rather than condition.","priority":"high"},"bands":{"evidence_ids":["R01.bands","R01.C01.S0001","R03.bands","R03.C01.S0001","R04.bands","R04.C01.S0001","R05.bands","R05.C01.S0001","R06.bands","R06.C01.S0001","R07.bands","R07.C01.S0001","R08.bands","R08.C01.S0001"],"finding":"R01 has only 68.27% of Welch power below 100 Hz and unusually large fractions at 100\u2013500 Hz (26.65%) and above 500 Hz; R03/R04/R05/R06/R07/R08 have 92.12\u201396.995% below 100 Hz. Within that cluster, R04 and R06 have the largest 100\u2013500 Hz fractions (7.18% and 7.11%), while R08 has the largest <100 Hz fraction (96.995%).","priority":"high"},"cepstrum":{"evidence_ids":["R01.cepstrum","R03.cepstrum","R04.cepstrum","R05.cepstrum","R06.cepstrum","R07.cepstrum","R08.cepstrum"],"finding":"All references are dominated by excluded-near-zero short-quefrency structure rather than a reported nonzero spacing peak. The first-bin amplitude is much larger for R01 (0.526) than R06 (0.206) and R03/R04/R05/R07/R08 (0.101\u20130.150), but this is weakly diagnostic and affected by spectral slope and the log floor.","priority":"low"},"cyclic":{"evidence_ids":["R01.cyclic","R03.cyclic","R04.cyclic","R05.cyclic","R06.cyclic","R07.cyclic","R08.cyclic"],"finding":"Maximum squared cyclic coherence differs (R03 0.699, R01 0.588, R07 0.461, R05 0.397, R04 0.312, R06 0.251, R08 0.245), but valid-cell counts also differ substantially (47,116\u201362,565) because low-power cells are masked. Images mainly show sparse deterministic-tone/high-carrier structure; no significance threshold is supplied.","priority":"medium"},"envelope":{"evidence_ids":["R01.envelope","R01.C01.S0001","R03.envelope","R03.C01.S0001","R04.envelope","R04.C01.S0001","R05.envelope","R05.C01.S0001","R06.envelope","R06.C01.S0001","R07.envelope","R07.C01.S0001","R08.envelope","R08.C01.S0001"],"finding":"Broadband-envelope leaders differ: R01 has 0.303/40.0/9.29 Hz and several unrelated peaks; R03 leads at 14.95 Hz, R04 and R05 at about 29.9 Hz, R06 at 14.95 Hz, R07 at 30.61 Hz, and R08 at 44.65 Hz. These are useful acquisition-specific modulation patterns but can arise from beating and are correlated with the band envelopes.","priority":"medium"},"envelope_1":{"evidence_ids":["R01.envelope_1","R03.envelope_1","R04.envelope_1","R05.envelope_1","R06.envelope_1","R07.envelope_1","R08.envelope_1"],"finding":"In the 100\u2013500 Hz carrier band, strongest normalized modulation is reference-specific: R01 3.88 Hz (0.145), R03 89.90 Hz (0.250), R04 44.90 Hz (0.260), R05 102.45 Hz (0.261), R06 15.0 Hz (0.319), R07 15.31 Hz (0.382), and R08 118.98 Hz (0.146). R06 and R07 overlap near 15 Hz despite different labels.","priority":"high"},"envelope_2":{"evidence_ids":["R01.envelope_2","R03.envelope_2","R04.envelope_2","R05.envelope_2","R06.envelope_2","R07.envelope_2","R08.envelope_2"],"finding":"For 500\u20131500 Hz carriers, leaders are R01 6.53 Hz (0.054), R03 15.0 Hz (0.206), R04 29.90 Hz (0.239), R05 29.29 Hz (0.105), R06 29.90 Hz (0.203), R07 33.78 Hz (0.081), and R08 14.90 Hz (0.122). Thus R04/R05/R06 overlap near 30 Hz and R03/R08 overlap near 15 Hz.","priority":"medium"},"envelope_3":{"evidence_ids":["R01.envelope_3","R03.envelope_3","R04.envelope_3","R05.envelope_3","R06.envelope_3","R07.envelope_3","R08.envelope_3"],"finding":"For 1500\u20134000 Hz carriers, strongest modulation is R01 6.53 Hz, R03 15.0 Hz, R04 15.0 Hz, R05 58.57 Hz, R06 14.90 Hz, R07 1.12 Hz, and R08 29.80 Hz. R03/R04/R06 are not separated by the leading frequency, while R07 is distinctive but could reflect slow regime variation.","priority":"medium"},"envelope_4":{"evidence_ids":["R01.envelope_4","R03.envelope_4","R04.envelope_4","R05.envelope_4","R06.envelope_4","R07.envelope_4","R08.envelope_4"],"finding":"The 4000\u20138000 Hz band gives strong but overlapping families: R01 61.73 Hz (0.223), R03 61.02 Hz (0.241), R04 30.41 Hz (0.376), R05 29.90 Hz (0.487), R06 61.12 Hz (0.212), R07 62.35 Hz (0.378), and R08 30.31 Hz (0.436). It separates approximately 30-Hz and 61\u201362-Hz acquisition groups, not unique conditions.","priority":"high"},"envelope_5":{"evidence_ids":["R01.envelope_5","R03.envelope_5","R04.envelope_5","R05.envelope_5","R06.envelope_5","R07.envelope_5","R08.envelope_5"],"finding":"For 8000\u201316000 Hz carriers, leaders are R01 61.73 Hz, R03 15.0 Hz, R04 273.98 Hz, R05 89.80 Hz, R06 14.90 Hz, R07 187.04 Hz, and R08 181.73 Hz. R03/R06 overlap near 15 Hz and R07/R08 are nearby around 182\u2013187 Hz; all bands are available at the 42 kHz sample rate.","priority":"high"},"estimators":{"evidence_ids":["R01.estimators","R03.estimators","R04.estimators","R05.estimators","R06.estimators","R07.estimators","R08.estimators"],"finding":"All estimators confirm low-frequency-dominated spectra; integrated multitaper power ranges from R01 0.0367 through R06 0.1136, R07 0.1451, R04 0.1579, R03 0.1738, R05 0.2051 to R08 0.2161. Because all recordings are peak-normalized, these absolute differences are acquisition/shape evidence rather than acoustic severity; Welch and multitaper views are correlated.","priority":"medium"},"fft":{"evidence_ids":["R01.fft","R01.C01.S0001","R03.fft","R03.C01.S0001","R04.fft","R04.C01.S0001","R05.fft","R05.C01.S0001","R06.fft","R06.C01.S0001","R07.fft","R07.C01.S0001","R08.fft","R08.C01.S0001"],"finding":"At 0.1 Hz bins and 0.15 Hz Hann bandwidth, dominant peaks are R01 45.9 Hz at -16.69 dB FS, R03 45.0 at -5.95, R04 44.9 at -8.18 plus a strong 74.8-Hz peak, R05 43.9 at -4.33, R06 44.9 at -9.47 with strong 29.9 Hz, R07 45.9 at -6.24 with 76.5 Hz, and R08 44.6 at -4.74 with 74.4/89.3 Hz. R01 is spectrally much less dominated by the main low-frequency tone; the other references have overlapping line families shifted by operating frequency.","priority":"high"},"harmonics":{"evidence_ids":["R01.harmonics","R03.harmonics","R04.harmonics","R05.harmonics","R06.harmonics","R07.harmonics","R08.harmonics"],"finding":"All references show dense peak-pair spacings near 15, 29\u201331, and/or 44\u201346 Hz, so candidate families are non-unique. The largest pair count is 24 for R07, 23 for R01/R04, 18 for R08, 17 for R03, and 11 for R05/R06; unrelated tones can inflate these counts, and no candidate establishes a fundamental or RPM.","priority":"medium"},"kurtosis_bank":{"evidence_ids":["R01.kurtosis_bank","R03.kurtosis_bank","R04.kurtosis_bank","R05.kurtosis_bank","R06.kurtosis_bank","R07.kurtosis_bank","R08.kurtosis_bank"],"finding":"High-frequency complex excess kurtosis strongly separates some recordings: median/max are R01 -0.130/0.020, R03 0.506/1.850, R04 0.077/0.748, R05 0.060/0.592, R06 0.352/1.043, R07 -0.082/0.502, and R08 0.021/0.258. R03 and R06 are broadly more impulsive above about 7.5 kHz, whereas R01 is mostly negative; impulsiveness alone does not identify a fault.","priority":"high"},"levels":{"evidence_ids":["R01.levels","R01.C01.S0001","R03.levels","R03.C01.S0001","R04.levels","R04.C01.S0001","R05.levels","R05.C01.S0001","R06.levels","R06.C01.S0001","R07.levels","R07.C01.S0001","R08.levels","R08.C01.S0001"],"finding":"No reference clips and all are peak-normalized to about 0.95 FS. R01 has a large DC offset (0.148 FS), high crest factor (11.93 dB), and a brief level disturbance near 2.1 s; R04 and R06 report many >6 dB local changes, while R03/R05/R07/R08 report none. RMS differences (-12.37 to -6.59 dB FS) cannot support acoustic level or severity claims after normalization.","priority":"medium"},"modulation_map":{"evidence_ids":["R01.modulation_map","R03.modulation_map","R04.modulation_map","R05.modulation_map","R06.modulation_map","R07.modulation_map","R08.modulation_map"],"finding":"Across the five common carrier bands, map maxima are R01 0.0967, R03 0.2321, R04 0.1696, R05 0.4404, R06 0.1107, R07 0.0784, and R08 0.1594. R05 has the strongest concentrated common-band modulation, while R07 is weakest; 155 cells per map are masked/missing and must not be interpreted as low coherence.","priority":"medium"},"orders":{"evidence_ids":["R01.orders","R03.orders","R04.orders","R05.orders","R06.orders","R07.orders","R08.orders"],"finding":"Unavailable for every reference because no independent RPM trace was supplied; no order-domain comparison or shaft-order claim is possible.","priority":"unavailable"},"persistence":{"evidence_ids":["R01.persistence","R03.persistence","R04.persistence","R05.persistence","R06.persistence","R07.persistence","R08.persistence"],"finding":"All references show persistent low-frequency line populations plus weaker high-frequency lines. Maximum occupancy is R01 50.4%, R03 54.7%, R06 72.5%, R04 86.1%, R07 88.4%, and R05/R08 92.7%; chronology is discarded and overlapping STFT frames are dependent, so these are persistence descriptors rather than independent evidence.","priority":"medium"},"reassigned":{"evidence_ids":["R01.reassigned","R03.reassigned","R04.reassigned","R05.reassigned","R06.reassigned","R07.reassigned","R08.reassigned"],"finding":"All reassigned views retain at least 99.73% of weight and show stable narrow horizontal structures. R01 has the lowest retained fraction (0.9973) and lowest maximum weight level (-19.08 dB), versus maxima of roughly -11.96 to -15.29 dB for the others; these largely restate STFT/FFT differences.","priority":"low"},"ridge":{"evidence_ids":["R01.ridge","R03.ridge","R04.ridge","R05.ridge","R06.ridge","R07.ridge","R08.ridge"],"finding":"R01 has median 50 Hz with frequent switches up to 220 Hz; R04 has median 50 Hz and switches up to 80 Hz; R03/R05/R06/R08 have 40-Hz medians with maxima 50 Hz; R07 is nearly fixed at 50 Hz (minimum 40 Hz). The 10-Hz STFT-bin grid is coarse relative to the FFT peak differences, and no ridge is validated as RPM.","priority":"medium"},"spectral_kurtosis":{"evidence_ids":["R01.spectral_kurtosis","R03.spectral_kurtosis","R04.spectral_kurtosis","R05.spectral_kurtosis","R06.spectral_kurtosis","R07.spectral_kurtosis","R08.spectral_kurtosis"],"finding":"R01 has the largest reported maximum (1.255 at 10 Hz); the others peak at 0.511\u20130.872, generally at isolated frequency bins. Excluded low-energy bins vary widely (2 for R01 to 474 for R05), finite-sample bias is uncorrected, and no confidence threshold is supplied, limiting direct ranking.","priority":"medium"},"stft":{"evidence_ids":["R01.stft","R03.stft","R04.stft","R05.stft","R06.stft","R07.stft","R08.stft"],"finding":"All 10-s spectrograms show mostly persistent horizontal tonal bands rather than gross drift. R01 has visibly broader/higher relative broadband energy and a local disturbance around 2 s; R04/R06 show more level variability, while R03/R05/R07/R08 appear steadier. These observations correlate with levels, FFT and persistence and are limited by the displayed pooling.","priority":"medium"},"stft_detail":{"evidence_ids":["R01.stft_detail","R03.stft_detail","R04.stft_detail","R05.stft_detail","R06.stft_detail","R07.stft_detail","R08.stft_detail"],"finding":"With identical 0.1-s frames and 10-Hz bins, maximum/median PSD levels are R01 -32.32/-88.33, R03 -26.69/-90.35, R04 -27.22/-94.03, R05 -25.85/-92.51, R06 -28.71/-92.18, R07 -27.61/-93.88, and R08 -25.77/-92.31 dB. R01 has a lower peak but higher median floor, consistent with broader spectral power; normalization prevents acoustic-level interpretation.","priority":"medium"},"synchronous":{"evidence_ids":["R01.synchronous","R03.synchronous","R04.synchronous","R05.synchronous","R06.synchronous","R07.synchronous","R08.synchronous"],"finding":"Unavailable for every reference because no independent RPM trace was supplied; phase-locked waveform differences and synchronous rejection of asynchronous impacts cannot be assessed.","priority":"unavailable"},"wavelet":{"evidence_ids":["R01.wavelet","R03.wavelet","R04.wavelet","R05.wavelet","R06.wavelet","R07.wavelet","R08.wavelet"],"finding":"All Morlet views cover only up to 4200 Hz after resampling and show persistent low-frequency energy with varying texture. Median coefficient levels range from -65.11 dB (R04) to -59.58 dB (R05); R01 shows more pronounced time variation. Edge-masked cells and wavelet scaling make this supportive, correlated evidence rather than a direct PSD comparison.","priority":"low"},"welch":{"evidence_ids":["R01.welch","R01.C01.S0001","R03.welch","R03.C01.S0001","R04.welch","R04.C01.S0001","R05.welch","R05.C01.S0001","R06.welch","R06.C01.S0001","R07.welch","R07.C01.S0001","R08.welch","R08.C01.S0001"],"finding":"Using identical 41.02-Hz bins and 61.52-Hz ENBW, R01 has a much higher centroid/flatness (212.5 Hz; 0.00307) than R03/R04/R05/R06/R07/R08 (centroids 54.3\u201365.9 Hz; flatness 0.000113\u20130.000225). Integrated power spans 0.0361\u20130.2162 FS\u00b2 but is affected by peak normalization. Welch cannot resolve the closely spaced low-frequency FFT lines.","priority":"high"}},"limitations":"This is a reference-only survey: no query signal or query distances were supplied, so no classification, acceptance, fault identity, component, or severity conclusion is possible. Each condition has only one reference, so observed differences may reflect acquisition regime, microphone response/placement, reverberation, speed, load, normalization, or transient behavior rather than condition-general signatures. The seven references are all 10 s, 42 kHz, peak-normalized recordings, but gain invariance does not remove those confounders. Diagram views derived from the same signal are correlated, not independent votes. Numeric summaries were preferred over pixels. Masked cyclic/modulation cells and wavelet edge cells are missing data. Orders and synchronous averages are unavailable for all references because RPM was not supplied. Welch resolution is too coarse for many listed FFT lines, and kurtosis/cyclic products lack significance intervals.","most_informative":["fft","bands","autocorrelation","envelope_4","kurtosis_bank"]}},{"dataset":"Ottawa","name":"survey-74ed3ba6f4fe58a6","result":{"diagram_review":{"autocorrelation":{"evidence_ids":["R01.autocorrelation","R02.autocorrelation","R03.autocorrelation","R05.autocorrelation","R06.autocorrelation","R07.autocorrelation","R08.autocorrelation"],"finding":"R01 differs strongly: peak lag 0.000381 s and coefficient 0.905, versus approximately 0.0654\u20130.0684 s and 0.971\u20130.980 for R02/R03/R05/R06/R07/R08; lag-to-RPM conversion is not justified.","priority":"high"},"bands":{"evidence_ids":["R01.bands","R02.bands","R03.bands","R05.bands","R06.bands","R07.bands","R08.bands"],"finding":"Relative 0\u2013100 Hz power separates R01 (68.27%) from all others (92.12\u201397.25%); among the latter, R06 has the largest 100\u2013500 Hz share (7.11%), while R02/R08 have the largest 0\u2013100 Hz shares (97.25%/97.00%).","priority":"high"},"cepstrum":{"evidence_ids":["R01.cepstrum","R02.cepstrum","R03.cepstrum","R05.cepstrum","R06.cepstrum","R07.cepstrum","R08.cepstrum"],"finding":"All references are dominated by near-zero-quefrency bins; R01 has the largest first-bin value (0.526), R06 next (0.206), and no supplied summary establishes a distinct nonzero-quefrency spacing.","priority":"low"},"cyclic":{"evidence_ids":["R01.cyclic","R02.cyclic","R03.cyclic","R05.cyclic","R06.cyclic","R07.cyclic","R08.cyclic"],"finding":"Maximum cyclic coherence varies from 0.245\u20130.251 in R06/R08 to 0.699 in R03, but valid-cell counts differ substantially (44,209\u201362,565) because low-power cells are masked; deterministic tones and absent significance thresholds limit interpretation.","priority":"medium"},"envelope":{"evidence_ids":["R01.envelope","R02.envelope","R03.envelope","R05.envelope","R06.envelope","R07.envelope","R08.envelope"],"finding":"Broadband modulation differs: R03/R06 emphasize about 15 Hz, R02 about 30/45 Hz, R05 about 29.3/58.5 Hz, R07 about 30.6/15.35 Hz, R08 about 44.6/29.8 Hz, while R01 includes 0.303 and 40 Hz; beating can also produce these components.","priority":"medium"},"envelope_1":{"evidence_ids":["R01.envelope_1","R02.envelope_1","R03.envelope_1","R05.envelope_1","R06.envelope_1","R07.envelope_1","R08.envelope_1"],"finding":"In 100\u2013500 Hz, strongest modulation is reference-specific: R01 3.88 Hz, R02/R03 about 90 Hz, R05 102.45 Hz, R06 15.0 Hz, R07 15.31 Hz, and R08 118.98 Hz.","priority":"high"},"envelope_2":{"evidence_ids":["R01.envelope_2","R02.envelope_2","R03.envelope_2","R05.envelope_2","R06.envelope_2","R07.envelope_2","R08.envelope_2"],"finding":"In 500\u20131500 Hz, strongest modulation spans R02 0.51 Hz, R01 6.53 Hz, R03/R08 about 15 Hz, R05/R06 about 29\u201330 Hz, and R07 33.78 Hz; R03 has the strongest normalized peak (0.206).","priority":"medium"},"envelope_3":{"evidence_ids":["R01.envelope_3","R02.envelope_3","R03.envelope_3","R05.envelope_3","R06.envelope_3","R07.envelope_3","R08.envelope_3"],"finding":"In 1500\u20134000 Hz, R03 has a prominent 15 Hz peak (0.193), whereas R01 is strongest at 6.53 Hz, R02 at 0.51 Hz, R05 at 58.57 Hz, R06 at 14.90 Hz, R07 at 1.12 Hz, and R08 at 29.80 Hz.","priority":"medium"},"envelope_4":{"evidence_ids":["R01.envelope_4","R02.envelope_4","R03.envelope_4","R05.envelope_4","R06.envelope_4","R07.envelope_4","R08.envelope_4"],"finding":"The 4000\u20138000 Hz envelopes are highly differentiating: R05 and R08 are dominated near 30 Hz (0.487/0.436), R01/R03/R06/R07 near 61\u201362 Hz, and R02 near 91.7 Hz; R05 has the largest squared-envelope peak (0.845).","priority":"high"},"envelope_5":{"evidence_ids":["R01.envelope_5","R02.envelope_5","R03.envelope_5","R05.envelope_5","R06.envelope_5","R07.envelope_5","R08.envelope_5"],"finding":"In 8000\u201316000 Hz, R03/R06 are dominated near 15 Hz, R01 near 61.7 Hz, R05 near 89.8 Hz, and R02/R07/R08 near 182\u2013187 Hz; R03 has the largest squared-envelope peak (0.664).","priority":"high"},"estimators":{"evidence_ids":["R01.estimators","R02.estimators","R03.estimators","R05.estimators","R06.estimators","R07.estimators","R08.estimators"],"finding":"All estimators confirm low-frequency dominance, but integrated multitaper power ranges from 0.0367 in R01 to 0.216 in R08; R01 also has a much higher spectral median, while the 60 Hz smoothing limits narrow-line comparison.","priority":"medium"},"fft":{"evidence_ids":["R01.fft","R02.fft","R03.fft","R05.fft","R06.fft","R07.fft","R08.fft"],"finding":"R01 has a much weaker 45.9 Hz principal peak (-16.69 dB FS) and substantial non-low-frequency peaks, whereas R02/R03/R05/R06/R07/R08 have principal peaks near 43.9\u201345.9 Hz at -4.29 to -9.47 dB FS with distinct secondary-line patterns.","priority":"high"},"harmonics":{"evidence_ids":["R01.harmonics","R02.harmonics","R03.harmonics","R05.harmonics","R06.harmonics","R07.harmonics","R08.harmonics"],"finding":"All references contain dense approximately 15/30/45 Hz-related peak spacings, but counts and matched families differ: 15 Hz pair counts range from 11 in R05/R06 to 24 in R07; unrelated tones prevent treating any candidate as an established fundamental.","priority":"medium"},"kurtosis_bank":{"evidence_ids":["R01.kurtosis_bank","R02.kurtosis_bank","R03.kurtosis_bank","R05.kurtosis_bank","R06.kurtosis_bank","R07.kurtosis_bank","R08.kurtosis_bank"],"finding":"R03 is the clear high-band impulsiveness outlier (median 0.506, maximum 1.850), followed by R06 (median 0.352, maximum 1.043); R01/R07 are mostly negative, while R02/R05/R08 are near zero with localized positive bands.","priority":"high"},"levels":{"evidence_ids":["R01.levels","R02.levels","R03.levels","R05.levels","R06.levels","R07.levels","R08.levels"],"finding":"No reference clips or is silent; all are peak-normalized, so RMS is not acoustic severity. R01 has a large DC offset (0.148 FS) and three changes near 2.1 s, R06 reports numerous greater-than-6 dB changes, while R02/R03/R05/R07/R08 report none.","priority":"high"},"modulation_map":{"evidence_ids":["R01.modulation_map","R02.modulation_map","R03.modulation_map","R05.modulation_map","R06.modulation_map","R07.modulation_map","R08.modulation_map"],"finding":"Map maxima distinguish strong localized modulation in R05 (0.440) and R03 (0.232) from R07 (0.078), but 155 cells per reference are masked and the map is a correlated summary of the five envelope views.","priority":"medium"},"orders":{"evidence_ids":["R01.orders","R02.orders","R03.orders","R05.orders","R06.orders","R07.orders","R08.orders"],"finding":"Unavailable for every reference because no independent RPM trace was supplied.","priority":"unavailable"},"persistence":{"evidence_ids":["R01.persistence","R02.persistence","R03.persistence","R05.persistence","R06.persistence","R07.persistence","R08.persistence"],"finding":"All references show persistent low-frequency populations and recurring higher-frequency lines; maximum cell persistence ranges from 50.38% in R01 to 92.70% in R05/R08, but chronology is discarded and overlapping frames are dependent.","priority":"medium"},"reassigned":{"evidence_ids":["R01.reassigned","R02.reassigned","R03.reassigned","R05.reassigned","R06.reassigned","R07.reassigned","R08.reassigned"],"finding":"All retain at least 99.73% of weight and show stable narrow ridges; R01 has the lowest retained fraction and weakest maximum (-19.08 dB versus about -11.96 to -15.29 dB for others), largely echoing STFT/FFT differences.","priority":"low"},"ridge":{"evidence_ids":["R01.ridge","R02.ridge","R03.ridge","R05.ridge","R06.ridge","R07.ridge","R08.ridge"],"finding":"R02/R07 are predominantly at the 50 Hz eligible bin, R03/R05/R06/R08 switch between 40 and 50 Hz, while R01 has median 50 Hz but excursions to 220 Hz; 10 Hz binning and no continuity model limit interpretation.","priority":"medium"},"spectral_kurtosis":{"evidence_ids":["R01.spectral_kurtosis","R02.spectral_kurtosis","R03.spectral_kurtosis","R05.spectral_kurtosis","R06.spectral_kurtosis","R07.spectral_kurtosis","R08.spectral_kurtosis"],"finding":"R01 has the largest maximum (1.255 at 10 Hz), R02 reaches 0.915 near 620 Hz, and R03 concentrates positive maxima near 15.1 kHz; excluded-bin counts vary from 2 to 544 and finite-sample bias is uncorrected.","priority":"medium"},"stft":{"evidence_ids":["R01.stft","R02.stft","R03.stft","R05.stft","R06.stft","R07.stft","R08.stft"],"finding":"All show broadly stationary horizontal spectral structure over 10 s; R01 has comparatively stronger broadband/high-frequency content, while R06 shows more visible level variation, but this view is correlated with FFT, persistence and detailed STFT.","priority":"medium"},"stft_detail":{"evidence_ids":["R01.stft_detail","R02.stft_detail","R03.stft_detail","R05.stft_detail","R06.stft_detail","R07.stft_detail","R08.stft_detail"],"finding":"At 0.1 s frames and 10 Hz bins, maximum PSD ranges from -32.32 dB in R01 to about -25.72 dB in R02, while R01 has the highest median floor (-88.33 dB); no unique transient chronology is evident from the summaries.","priority":"low"},"synchronous":{"evidence_ids":["R01.synchronous","R02.synchronous","R03.synchronous","R05.synchronous","R06.synchronous","R07.synchronous","R08.synchronous"],"finding":"Unavailable for every reference because no independent RPM trace was supplied.","priority":"unavailable"},"wavelet":{"evidence_ids":["R01.wavelet","R02.wavelet","R03.wavelet","R05.wavelet","R06.wavelet","R07.wavelet","R08.wavelet"],"finding":"All wavelet maps show persistent low-frequency energy below the 4.2 kHz analysis ceiling; R01 has the weakest maximum (-35.31 dB) but highest median (-60.65 dB relative to several references), broadly mirroring spectral-shape differences.","priority":"low"},"welch":{"evidence_ids":["R01.welch","R02.welch","R03.welch","R05.welch","R06.welch","R07.welch","R08.welch"],"finding":"R01 is spectrally distinct with centroid 212.5 Hz and flatness 0.00307, versus centroids 51.8\u201362.0 Hz and flatness 0.000076\u20130.000225 for R02/R03/R05/R06/R07/R08; Welch resolution is only about 61.5 Hz ENBW.","priority":"high"}},"limitations":"This is a reference-only survey: no query evidence or distances were supplied, so no classification or nearest-reference claim is possible. Each condition has only one reference, preventing separation of condition effects from acquisition regime, microphone response/placement, normalization, load, speed and reverberation. All recordings are derived peak-normalized, so absolute levels and severity are not comparable acoustically. Diagram views derived from the same signal are correlated rather than independent votes. Orders and synchronous averages are unavailable for all references because RPM is absent; masked cyclic/modulation cells are missing evidence. Numeric summaries were preferred over image estimates, and fault identity, component and severity remain unestablished.","most_informative":["fft","bands","autocorrelation","envelope_4","kurtosis_bank"]}}],"technical_failures":0}