TK2 Excitation Analysis
Use this skill for excitation-system recorder files in TK2 binary format, especially WFLC/TK-style files containing UAB, UBC, UCA, UFD, Uld, and IFD channels.
Quick Start
Run the bundled script first; it handles the fragile binary parsing and plotting:
python scripts/analyze_tk2.py INPUT.tk2 --output-dir OUT --mode single --phase UAB --kind step
Common modes:
# 5% or large step response, using UAB only
python scripts/analyze_tk2.py "01-5%阶跃.tk2" --output-dir out --mode single --phase UAB --kind step
# Pulse-blocking decay, using UAB only
python scripts/analyze_tk2.py "05-封脉冲.tk2" --output-dir out --mode single --phase UAB --kind decay
# Legacy three-phase average RMS, only if explicitly requested
python scripts/analyze_tk2.py "01-5%阶跃.tk2" --output-dir out --mode three-phase --kind step
Outputs:
*_curve.png: normalized voltage response curve.*_series.csv: per-cycle time, voltage RMS, UFD mean, IFD mean.*_metrics.json: sampling info and calculated metrics.
Parsing Assumptions
The script uses the recorder pattern observed in Manla excitation tests:
- Header starts at byte 0 and contains GBK/ASCII channel metadata.
- Waveform samples begin after the header; the default auto-detection tries to find the first offset where the remaining bytes divide into
6 * int16channel samples. - Effective channel order is:
1.UAB
2.UBC
3.UCA
4.UFD
5.Uld
6.IFD - Samples are little-endian signed 16-bit integers.
- Sampling rate is inferred from positive-going zero crossings in the chosen phase. For 50 Hz recordings, a 512-sample cycle gives
25600 Hz.
If auto-detection fails, rerun with explicit values:
python scripts/analyze_tk2.py INPUT.tk2 --header-bytes 8930 --sample-rate 25600 --output-dir OUT
Calculation Rules
For report work, prefer --mode single --phase UAB unless the user explicitly asks for three-phase average.
Single-phase mode:
- Compute one-cycle RMS from the selected phase (
UAB,UBC, orUCA). - Normalize response curves to the initial steady value.
Three-phase mode:
- Compute one-cycle RMS for
UAB,UBC, andUCA. - Average the three RMS curves.
For UFD, Uld, and IFD, calculate one-cycle arithmetic means.
Step Metrics
For --kind step, the script estimates:
amp_pct: final steady voltage change relative to initial value.Mp_pct: overshoot relative to the step amplitude.Tup_s: 10%-to-90% rise time.Tp_s: time from step start to peak.Ts_s: first time after which the response stays within a 5% band of final step amplitude.
Inspect metrics before copying them into reports. If the step happens outside the default detection window, set:
--step-window START END
--initial-window START END
--final-window START END
Decay Metrics
For --kind decay, the script estimates T_decay_1e_s from the time between:
- start of decay, detected when normalized voltage first drops below
0.999, and - crossing of
1/eof the initial value.
Use the result as the pulse-blocking open-circuit decay time constant when the recorder setup and test method match the report method.
Report Integration
When inserting results into a Word report:
- State the processing basis, for example:
按UAB单相有效值计算. - Keep the plotted caption consistent with the calculation mode, for example:
发电机5%空载阶跃响应试验（UAB单相有效值）. - Do not mix three-phase-average metrics with single-phase plots.
- Use exported CSV/JSON as auditable intermediates; use PNG for report figures.
