AssetHealthIndexEvents

One 0-100 health score per asset and window, fused from several condition signals (vibration, temperature, motor current), each tagged with the failure mode it watches: alerts when the score falls below warning or critical, the signals that cost the most health, and a projection of when the score reaches a critical level.

ts_shape.events.maintenance.asset_health_index · Engineering guide · Maintenance guide · API reference

Example

import pandas as pd
import ts_shape
from ts_shape.events.maintenance.asset_health_index import AssetHealthIndexEvents

# three days at 10 min: vibration rises steadily, temperature and current stay put
kw = dict(n_points=432, freq="10min", start="2024-03-01")
df = pd.concat([
    ts_shape.make_timeseries(["vib:press-01"], baseline=2.0, noise=0.05, drift=3.0, **kw),
    ts_shape.make_timeseries(["temp:press-01"], baseline=50.0, noise=0.2, **kw),
    ts_shape.make_timeseries(["curr:press-01"], baseline=12.0, noise=0.1, **kw),
], ignore_index=True)

health = AssetHealthIndexEvents(
    df,
    asset_id="press-01",
    signals={
        "vib:press-01": "bearing",
        "temp:press-01": {"failure_mode": "thermal", "weight": 2.0},
        "curr:press-01": {"failure_mode": "electrical"},
    },
)
index = health.health_index(window="6h", baseline_window="12h")
print(index[["start", "asset_health_score", "dominant_failure_mode"]].tail(3))
alerts = health.health_events(window="6h", baseline_window="12h")
ranking = health.signal_contributions(window="6h", baseline_window="12h")
print(ranking[["signal_uuid", "failure_mode", "mean_health", "contribution_pct"]])
forecast = health.maintenance_forecast(window="6h", baseline_window="12h", critical_threshold=40.0)
print(forecast[["current_health", "decline_per_day", "recommended_maintenance_days"]])

The bearing (vibration) signal drives the loss of health, although temperature carries twice its weight.

Methods

Method	Returns
health_index(window='1h', baseline_window='24h')	per window: asset_id, asset_health_score, dominant_failure_mode, worst_signal_uuid, signals_available, signals_expected, confidence_level
health_events(window='1h', baseline_window='24h', warning_threshold=70.0, critical_threshold=50.0, only_crossings=True)	point events when the score enters warning or critical: asset_health_score, severity, threshold, dominant_failure_mode
signal_contributions(window='1h', baseline_window='24h')	one row per signal: failure_mode, weight, mean_health, min_health, weighted_deficit, contribution_pct, rank
maintenance_forecast(window='1h', baseline_window='24h', critical_threshold=50.0, min_confidence=0.5)	one row: current_health, decline_per_day, recommended_maintenance_days, forecast_time, confidence

Notes

• signals maps each signal uuid to a failure mode, or to {"failure_mode": ..., "weight": ...}; a bare string has weight 1. Weights must be positive.
• Each signal is scored per window against the opening baseline_window of its own data: up to a third of the score for mean drift, for growing variance and for the trend slope. Start the data at a known-good period.
• The asset score is the weighted mean of the signals present in the window. A signal without data in a window is left out rather than scored 0; confidence_level is the share of the configured weight that was present. dominant_failure_mode is the failure mode of the signal with the largest weighted health deficit.
• health_events reports only the window that enters a band by default; only_crossings=False reports every window below a threshold. critical_threshold must not exceed warning_threshold.
• maintenance_forecast fits a straight line through the score. confidence is the fit's R²; below min_confidence, and for a flat or improving score, the horizon columns are NaN.
• For one signal, DegradationDetectionEvents gives the same kind of health score per sample.