SteadyStateDetectionEvents¤

Steady-state intervals (rolling standard deviation below a threshold) and transient intervals of a signal, the share of time steady, and the operating band of each steady interval.

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

Example¤

import numpy as np
import pandas as pd
from ts_shape.events.engineering.steady_state_detection import SteadyStateDetectionEvents

t = pd.date_range("2024-01-01", periods=7200, freq="1s")
level = np.interp(np.arange(7200), [0, 1800, 2400, 5400, 6000, 7199], [150, 150, 180, 180, 160, 160])
df = pd.DataFrame({"systime": t, "uuid": "reactor_temp",
                   "value_double": level + np.random.default_rng(0).normal(0, 0.3, 7200)})

ss = SteadyStateDetectionEvents(df, "reactor_temp")
result = ss.detect_steady_state(window="5min", std_threshold=1.0, min_duration="10min")
print(result[["start", "end", "mean_value", "std_value"]])
summary = ss.steady_state_statistics(window="5min", std_threshold=1.0)
bands = ss.steady_state_value_bands(window="5min", std_threshold=1.0)

Methods¤

All arguments are keyword-only.

Method	Returns
`detect_steady_state(window='5min', std_threshold=1.0, min_duration='10min')`	`start`, `end`, `mean_value`, `std_value`, `duration_seconds`
`detect_transient_periods(window='5min', std_threshold=1.0)`	`start`, `end`, `max_std`, `duration_seconds`
`steady_state_statistics(window='5min', std_threshold=1.0, min_duration='10min')`	one row: `total_steady_seconds`, `total_transient_seconds`, `steady_pct`, `num_steady_periods`, `avg_steady_duration_seconds`
`steady_state_value_bands(window='5min', std_threshold=1.0, min_duration='10min')`	per steady interval: `mean_value`, `lower_band`, `upper_band` (mean ± std), `duration_seconds`

Notes¤

std_threshold is in signal units; set it above the sensor noise.
Intervals end at the next sample after the condition stops holding by default (interval_end="next_change"), so change-only signals keep their duration; "last_sample" ends them at the last flagged sample.
steady_state_statistics(..., as_frame=False) returns the deprecated dict.