Skip to content

DegradationDetectionEvents¤

Detect degradation patterns in time series signals: trend degradation, variance increases, level shifts, and composite health scores.

Module: ts_shape.events.maintenance.degradation_detection Guide: Quality Guide

When to Use¤

Use for predictive maintenance monitoring. Tracks gradual degradation in bearing temperature, motor current, or any signal that slowly drifts before failure. Ideal when you need early warning of equipment deterioration before alarm thresholds are reached.

Quick Example¤

from ts_shape.events.maintenance.degradation_detection import DegradationDetectionEvents

detector = DegradationDetectionEvents(
    df=sensor_df,
    timestamp_col="timestamp",
    value_col="bearing_temp"
)

# Detect slow upward drift in bearing temperature
trend_events = detector.detect_trend_degradation(window="7D", slope_threshold=0.05)

# Check variance increase against a known-good baseline
variance_events = detector.detect_variance_increase(baseline_window="30D", threshold=2.0)

# Composite health score combining all indicators
score = detector.health_score(
    trend_weight=0.4, variance_weight=0.3, shift_weight=0.3
)
print(f"Health score: {score}/100")

Key Methods¤

Method Purpose Returns
detect_trend_degradation() Rolling linear regression slope detection DataFrame of intervals with degrading trend
detect_variance_increase() Variance vs baseline comparison DataFrame of intervals with elevated variance
detect_level_shift() CUSUM-like mean shift detection DataFrame of detected level shift events
health_score() Composite 0-100 health score Float score combining trend, variance, and shift indicators

Tips & Notes¤

Set a meaningful baseline window

Use the first stable operating period as your baseline. A poorly chosen baseline will produce false positives from detect_variance_increase() and detect_level_shift().

Related modules

See Also¤