Skip to content

FailurePredictionEvents¤

Predict remaining useful life and detect escalating failure patterns.

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

When to Use¤

Use when a degradation trend has been identified and you need to estimate how much time remains before a failure threshold is reached. Pairs well with DegradationDetectionEvents for a complete predictive maintenance pipeline.

Quick Example¤

from ts_shape.events.maintenance.failure_prediction import FailurePredictionEvents

predictor = FailurePredictionEvents(
    df=sensor_df,
    timestamp_col="timestamp",
    value_col="bearing_temp"
)

# Estimate remaining useful life via linear extrapolation
rul = predictor.remaining_useful_life(threshold=95.0, window="7D")
print(f"Estimated RUL: {rul.days:.1f} days")

# Detect escalating threshold breaches
exceedances = predictor.detect_exceedance_pattern(
    threshold=80.0, window="1D", min_exceedances=3
)

# Time estimate to reach critical threshold
tte = predictor.time_to_threshold(threshold=95.0)
print(f"Time to threshold: {tte}")

Key Methods¤

Method Purpose Returns
remaining_useful_life() RUL via linear extrapolation to threshold Timedelta estimate of remaining life
detect_exceedance_pattern() Escalating threshold breaches over time DataFrame of windows with increasing breach frequency
time_to_threshold() Time estimate to reach a given threshold Timedelta or None if trend is stable

Tips & Notes¤

Combine with degradation detection

Run DegradationDetectionEvents.detect_trend_degradation() first to confirm a real trend exists. Calling remaining_useful_life() on a stationary signal will return unreliable estimates.

Related modules

See Also¤