A failure-forecast method based on Weibull and statistical-pattern analysis

Abstract

This paper presents a method that combines Weibull analysis and statistical algorithms to forecast failures, and the experimental results applied to electronic systems. The Weibull analysis is a classical method to forecast failures using risk analysis and the mean-time-to-failure (MTTF) parameter. Statistical pattern analysis can forecast failures based on performance information from the component. Performance can be monitored by analyzing test data collected during the system or component's operating life. Weibull analysis uses failure datum to estimate the MTTF. Statistical pattern analysis uses test data to identify statistical patterns such as trend lines.