AI4Manufacturing/MFPT-annotated

Roles

Roles: reasoning view of MFPT — annot is the source label (inner_race / normal / outer_race), kept machine-parseable as the gold for verification and reward parsing; the model reads query + image, where the image is an envelope spectrum with the theoretical fault frequencies marked. The reasoning column is filled on all 78 records and is the SFT imitation target for this repo; the query enumerates the closed set of labels the answer must come from, and annot remains the string that is scored. metadata is provenance (split keys, source identities) and must never be fed to a model.

MFPT — bearing fault from the envelope spectrum, with reasoning

The MFPT reasoning track with the reasoning column filled. Same images, same queries, same gold labels, same splits: field-for-field identical to MFPT except reasoning (asserted at build time).

Records: 78 (splits {'train': 54, 'test': 24}); labels {'inner_race': 21, 'outer_race': 39, 'normal': 18}; evidence_tier {'confirmed': 78}.

Schema (7-field unified record)

field	meaning
query	the classification instruction (one of 30 deterministic paraphrases per representation)
image	the rendered signal image (bytes embedded)
annot	gold fault class: normal / inner_race / outer_race
reasoning	the half-template chain-of-thought for this record; it closes on FINAL ANSWER: <annot>
cate / task	C / T-C1 (signal fault classification)
metadata	JSON string: representation, file, window_idx, load_lbs, bearing, channel, fs, fr_hz, features, fault_freqs, published_fault_hz, computed_verdict, computed_snr, evidence_tier, line_evidence, image_sha256, split

How reasoning was produced

Half-template, not a language model. Each paragraph is assembled by a deterministic renderer from three pools drawn independently per record — 14 openers × 7 physics-knowledge blocks × 24 argument skeletons in use here — so the surface form varies widely while the argument stays checkable. No API, no sampling, no model output.

Every number is measured, never copied from the answer. All amplitudes, frequencies and peak counts come out of the record's line_evidence — the per-line measurement taken on the plotted envelope spectrum, over the BPFO / BPFI / BSF / FTF positions actually marked on its figure. The generator ran behind three gates, all of which had to be clean or nothing was written: (1) the paragraph ends on the gold, byte for byte; (2) every label-bearing field is deleted and the paragraph re-rendered — the bytes must be identical, which proves no number came out of the answer; (3) writing checks — nothing finer than 0.01 (the y axis cannot resolve it), no detector internals (family scores and thresholds are not on the figure and a reader cannot recompute them), no internal line keys, and numerals that agree with the families actually scored.

Re-checkable. The measurement the paragraph quotes ships with the record: metadata.line_evidence holds the per-line frequencies, amplitudes and SNRs and the noise floor, so every number in the text can be verified against the figure it was read off.

Closing. Every paragraph ends with a blank line and FINAL ANSWER: <annot>, matching the record's gold label exactly — re-verified for every record at build time.

Splits

train / test, following MFPT's own by-load-condition partition (leakage-safe: each file is wholly on one side, so windows never cross files). train holds the two baseline files, OuterRaceFault_1/2 and the first five *_vload_* files of each fault; test holds the held-out third baseline, OuterRaceFault_3 and *_vload_6/7.

Figures (style pool)

The spectra in this repo come from the figure-style pool: each record is drawn in one of 11 styles. What the style varies: colours, canvas, title placement and presence, tick spacing and minor ticks, grid, spines, fill, the fault-frequency marker drawn as a line or as a band, label placement and orientation, one colour per family plus a legend, font family, scientific notation, and tick direction. The style is picked by sha256("figstyle|" + image_path) % 11 -- deterministic, reproducible, and drawn independently of the opener / knowledge / skeleton pools.

What never varies is the content: the curve; the plotted band; which lines are marked (six on these figures -- the four families plus 2xBPFO and 2xBPFI); and the y-limits, autoscaled once per record and reused by every style, so a peak is the same height whichever style drew it. No threshold bar and no noise-floor line is ever drawn, and no marker is styled differently for being the answer. The point is that a model cannot pass by memorising one house plotting style. The -perception repo's images are unchanged.

What pins these images. Three separate things produced them, and each carries its own evidence -- none of them stands in for another:

• the signal and its manifest: forge_agent 0574e0e25d;
• the curve that is plotted, and the producer parameters behind it (demodulation band, plotted fmax, which lines are marked): forge_agent 733a683. This is not the code that drew the PNGs;
• the published bytes: figure_manifest_sha256 = 203e42bb5cd0491fa8a61faf569d0891c3f44db62d5b3b07b2961571e791c2db -- the sha256 of the pool's per-image sha256 manifest (_final_sha256.txt, one line per image, checkable with sha256sum -c).

The styling layer itself (_figstyle_corpus/_src/style_pool.py) lives on the data volume and is not under version control, so no commit identifies it and none is offered: a reader who checked out the commit above would get the same curve drawn some other way. The manifest hash is what these images should be checked against. provenance.json records all three pins separately.

Provenance & reproducibility

Generated deterministically by forge_agent/examples/mfpt/convert.py (0574e0e25d) → forge_model/MFPT/convert_mfpt.py (e023f6cb11); see provenance.json for the full record.

The MFPT test bearing's geometry (8 elements, d=0.235″, D=1.245″, 0° contact) is cross-checked against each file's own published fault frequencies: geometry-derived BPFO/BPFI agree to <0.001 Hz (BSF ~0.05 Hz rounding). FTF differs by convention — MFPT publishes the cage rate relative to the rotating inner race, forge_tools reports the standard train/cage frequency — reported, not reconciled. Labels come from the dataset's own filenames; the shaft rate (25 Hz), sample rate and load are read from each .mat. Evidence-gated, conflict-free release: a computed evidence_tier (from the label-independent envelope-spectrum detector) curates the published set so every image supports — and never fights — its label. On MFPT the gate keeps everything; that is a statement about the data, and it is re-measured on every build rather than assumed.

Caveats

• Small, clean source — MFPT ships 20 lab files (3 baseline, 3 + 7 outer-race, 7 inner-race) at a single 25 Hz shaft speed. The value is a groundable, load-varying bearing benchmark (loads 0–300 lb) that complements CWRU/XJTU/IMS, not raw record volume.
• No ball class — MFPT seeds only inner- and outer-race faults (+ healthy), so this is a 3-class task (unlike CWRU's 4). All four fault-frequency families are still marked on the figure, because ruling BSF and FTF out is part of reading it.
• Nothing is dropped, and that is a finding about the data, not a missing gate. The same evidence gate that removes most of IMS's and XJTU's windows removes none of MFPT's: every window is confirmed and every blind verdict matches its gold (baselines read healthy with zero false positives). Treat MFPT as the easy end of the corpus.
• Eval: the split is a condition split, not an unseen-bearing split — MFPT is one bearing type, so cross-bearing generalization cannot be measured here.
• FTF convention — metadata.published_fault_hz uses MFPT's cage-relative-to-inner FTF; metadata.fault_freqs uses forge_tools' standard train-frequency FTF. BPFO/BPFI (which drive the verdicts) agree in both.

Source & license

Source: Society for Machinery Failure Prevention Technology (MFPT) bearing fault data (orig. Eric Bechhoefer), obtained via MathWorks' RollingElementBearingFaultDiagnosis-Data distribution. License: CC BY-NC-SA 4.0 (Attribution — NonCommercial — ShareAlike): this derived dataset is redistributed under the same terms; non-commercial use only, with attribution to MFPT / Eric Bechhoefer and MathWorks.