Case Study:
Predictive Maintenance
Business Goals:
Anticipate equipment failure.
Increase predictive maintenance.
Reduce reactive service calls by 30%.
Outputs & Benefits:
Variables & patterns most related to future failure;
Provides engineering insight into failure points.
Probability of failure within various time periods.
Model statistics like precision, recall, false positive rates, etc.
Anticipation & reduction of service calls:
Calls reduced by 30+%.
Costs reduced by 20+%.
Many on-demand calls transformed into anticipatory ones.
Data Sets & Models:
Log data from equipment.
Sensor readings.
Error, warning, and status codes.
Failure flag.
Machine ID#'s and diagram of manufacturing process.
Both random forests & neural network models trained on data.