StartupDetectionEvents¤

Equipment startups in a numeric signal (speed, current, temperature): by threshold crossing with optional hysteresis, by sustained positive slope, by an adaptive baseline, or across several signals; plus startup quality, phase tracking and failed startups.

ts_shape.events.engineering.startup_events · Engineering guide · API reference

Example¤

import numpy as np
import pandas as pd
from ts_shape.events.engineering.startup_events import StartupDetectionEvents

t = pd.date_range("2024-01-01", periods=1800, freq="1s")
current = np.interp(np.arange(1800), [0, 300, 360, 1200, 1230, 1799], [0.5, 0.5, 12, 12, 0.5, 0.5])
df = pd.DataFrame({"systime": t, "uuid": "motor_current", "value_double": current})

sd = StartupDetectionEvents(df, target_uuid="motor_current")
startups = sd.detect_startup_by_threshold(threshold=5.0, hysteresis=(5.0, 4.5), min_above="10s")
result = sd.assess_startup_quality(startups)
print(result[["start", "smoothness_score", "stability_score", "value_change"]])
failed = sd.detect_failed_startups(threshold=5.0, completion_threshold=10.0, max_completion_time="5min")

Methods¤

Method	Returns
`detect_startup_by_threshold(threshold, hysteresis=None, min_above='0s')`	`start`, `end` (= start + `min_above`), `method`, `threshold`
`detect_startup_by_slope(min_slope, slope_window='0s', min_duration='0s')`	runs with slope ≥ `min_slope`: `min_slope`, `avg_slope`
`detect_startup_adaptive(baseline_window=None, sensitivity=2.0, min_above='10s', lookback_periods=5)`	`adaptive_threshold`, `baseline_mean`, `baseline_std`
`detect_startup_multi_signal(signals, logic='all', time_tolerance='30s')`	`signals_triggered`, `signal_details`
`assess_startup_quality(startup_events)`	per startup: `duration`, `smoothness_score`, `anomaly_flags`, `value_change`, `avg_rate`, `max_value`, `stability_score`
`track_startup_phases(phases, min_phase_duration='5s')`	per phase: `phase_name`, `phase_number`, `start`, `end`, `duration`, `next_phase`, `completed`
`detect_failed_startups(threshold, min_rise_duration='5s', max_completion_time='5min', completion_threshold=None, required_stability='10s')`	`failure_reason`, `max_value_reached`, `time_to_failure`

Notes¤

Threshold: a startup begins where the signal rises to or above threshold (or the enter level of hysteresis=(enter, exit)) and counts only if it stays at or above the exit level for min_above.
Slope is in signal units per second, between consecutive samples by default; a positive slope_window takes it over that trailing window, which smooths noise.
Adaptive: threshold = baseline mean + sensitivity × baseline std, from the samples before each point (the last lookback_periods samples, or the span baseline_window), so a rising value does not raise its own threshold.
detect_startup_multi_signal takes {uuid: {"method": "threshold" | "slope", ...}}; logic="all" requires all signals within time_tolerance, "any" one of them.
A failed startup crosses threshold for at least min_rise_duration but does not reach completion_threshold (default 2 × threshold) within max_completion_time, or drops back before holding it for required_stability.
track_startup_phases phases use condition threshold (min_value), range (min_value, max_value) or slope (min_slope).