On-line process failure analysis and modeling

Abstract

Using techniques originally developed to statistically analyze and model a repairable system failure history, the authors present a procedure to implement these techniques in the context of a portable data acquisition and control system that drives a testing or production process. The process failures are described as being unwanted events recorded by the data acquisition and control instrumentation, which then modifies the process operating conditions. Statistical analysis and testing are performed on the time information related to these events to predict a confidence interval for the time to next process failure. Simplified formulas are introduced to allow the process computer to adjust the process parameters without ever requiring a numerical table.<>