CUSTOMER STORY | MINING
SANY predicts mining truck engine failures within the hour, with an AUC above 98%
To reduce losses from unexpected mining truck breakdowns, SANY needed to predict engine failures. The Drift Report then showed when summer models decayed in winter.
Model accuracy (AUC)
>98%
Failure anticipation window
<1 h
Variables processed, for 107 breakdowns
355
OVERVIEW
The challenge.
Complex telemetry, and very few breakdowns to learn from.
355 correlated variables
The matrix contains 355 highly correlated variables for a target of only 107 historical breakdowns.
Temperatures, pressures, flows
The telemetry tracks temperatures, pressures and fluid flows across multiple sensors.
SOLUTION
What was done.
A robust model, then seasonal models.
Predictive model
A robust predictive model with an AUC above 98%.
Temperature thresholds
The model pinpoints the temperature thresholds that trigger an engine shutdown within the hour, so that preventive action is possible.
Drift Report
When summer models decayed in winter because of thermal baseline changes, the native Drift Report visualized the shift.
RESULTS
What changed.
Failures anticipated, and models kept valid across seasons.
Accuracy
AUC above 98%.
Anticipation
Failures are announced less than one hour in advance, which enables immediate preventive action.
Seasonal models
Engineers adapted and deployed season-specific models.
USE CASES BEHIND THIS STORY
Read the method.
Predictive Maintenance
Anticipate equipment failures from the data you already collect, with the reasons.
Model Drift Monitoring
See when a predictive model loses accuracy and needs an update.
Seasonal Model Recalibration
Adapt a model to predictable seasonal changes without rebuilding it each cycle.
SEE FOR YOURSELF
Test it on your own data.
Run TIMi on your own data with a free trial, or talk to us about your case.
