Event-Log-Based Failure Prediction and Monitoring for After-Sales Engineering Systems
With the rapid development of information technology, an abundance of data
that record the events occurred in a system are now collected automatically
when the system is in use. It is generally believed that the event logs
provide rich information regarding the system working conditions and could
be used for condition monitoring, diagnosis, and optimal maintenance. This
talk will present the research works in: (1) System survival model fitting
using event logs to quantify the associations between system events and the
key failure. The survival model encodes the system events as covariates and
can be used to rigorously predict the failure probability based on the event
logs. And (2) Generic discrete events sequence monitoring, which is based on
hypothesis testing in survival analysis to test if the survival model fitted
from historical data can fully represent the present system characteristics.
The research results provide a quantitative foundation for optimal
maintenance service of after-sales equipment.