order op• データ種: signal → table
• 呼び出し: import acoustics; acoustics.order_spectrum(x, rate, rpm, samples_per_rev=64, revolutions=None, max_order=None, n_peaks=5) (または opsacoustics.get("order_spectrum"))
Amplitude against shaft order — the spectrum a run-up should be read in.
:func:angular_resample followed by an rFFT over a **whole number of
revolutions** (the record is cropped to that). Bin spacing is
`1 / whole_revolutions` in orders.
`revolutions` overrides how many whole revolutions to keep, and it matters
more than it looks. An order `o` lands exactly on a bin only when
`o * revolutions` is an integer; otherwise it straddles two and both read
low. Measured on the run-up below, which happens to cover 79.994
revolutions:
============== ========== ============= =============
revolutions resolution amp at o=1.0 amp at o=3.5
============== ========== ============= =============
79 (default) 0.012658 0.999967 0.636961
78 (even) 0.012821 1.000009 0.999371
============== ========== ============= =============
That 0.637 is the classic two-bin scallop loss, and nothing raises: the peak
is at the right order and 36 % too small, with a second peak of almost equal
height one bin away (measured 0.6370 at order 3.4937 and 0.6353 at 3.5063).
Cropping to an even number of revolutions puts every half-integer order
on a bin. The default is the largest whole number available; pass
`revolutions` when the order you care about is fractional.
Returns a dict: `orders, magnitude (single-sided, 2/N`),
`peak_order, peak_amplitude, peak_orders / peak_amplitudes`,
`resolution_order, whole_revolutions, samples_per_rev`,
`mean_rpm, max_order`.
Measured, and this is the whole argument for the operator. A 4 s run-up from
600 to 1800 rpm at 5 kHz carrying exactly two shaft-locked components
(orders 1.0 and 3.5, unit amplitude) plus one fixed 400 Hz resonance, read
with `revolutions=78`:
========================== ===================== ====================
quantity ordinary spectrum order spectrum
========================== ===================== ====================
order-3.5 peak amplitude 0.070203 (of true 1) 0.999371 (of true 1)
its -3 dB width 66.50 Hz (= 3.33 ord) 0.00000 order
400 Hz resonance amplitude 1.0000, one bin 0.0517, over 26.7 ord
========================== ===================== ====================
The ordinary spectrum recovers 7 % of the shaft-locked component's
amplitude, because the energy is spread over 3.3 orders' worth of bins; the
order spectrum recovers 99.94 % of it in a single bin whose -3 dB width
is one bin. The 400 Hz resonance goes the other way — sharp in hertz,
smeared across 26.7 orders after resampling. That reversal is the
diagnostic, and it is why both spectra are worth computing: what stays sharp
under angular resampling turns with the shaft, and what stays sharp under
ordinary transformation does not.
Raises `ValueError: everything :func:angular_resample` refuses (in
particular the aliasing refusal), a `revolutions` larger than the record
actually contains, and a `max_order` above the angular Nyquist
`samples_per_rev/2`.
• acoustic_condition_monitoring ファミリ ガイド
• サンプルデータ カタログ(DL URL / ライセンス) — 2-D は skimage.data(BSD/public)+ 合成、3-D は実データ源(Stanford/PDS 等)の DL URL。
• 演算子の来歴・参考文献 — この op 族の元になった研究/手法の出典。
• アルゴリズムの正典(著者・年)と用途は上記ファミリ使い方ガイドに記載。
• acoustic_condition_monitoring — py -3.11 examples/acoustic_condition_monitoring.py
table を入力に取れる)order)*Provenance: acoustics.py — ACOUSTICS operator registry. この per-op ノートは tools/opdocs.py md が自動生成(手編集しない)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.