ANALYSIS OF RELIABILITY USING REAL MASKED SYSTEM LIFE DATA Open Access
Downloadable ContentDownload PDF
Maintenance personnel often lack the resources to determine which specific subassembly of a complex electronic system has failed but can only isolate the cause of failure to a subset of subassemblies. In these circumstances, the most efficient manner for replacing the suspect subassemblies is based on the individual subassembly's probability of failure. The purpose of this study is to investigate the use of field failure data to determine a subassembly's probability of failure and to compare an iterative maximum likelihood estimation procedure (IMLEP) and a Bayesian methodology for handling masked data. The data base was queried for removed times, installed times, and a failed parts report for the item under study by equipment serial number and aircraft number. The time to failure and the subassembly that caused the failure was determined. Probability plotting was used to determine which probability distribution best modeled the subassembly failure times and the parameters of the assumed distribution were maximized. The cause of failure was randomly masked at levels encounter in field failure data. IMLEP and a Bayesian approach were used to maximize the distribution parameters from the masked data. The results of each approach were compared to the parameter estimates with unmasked data. This study indicates that competing risk theory can be applied to the field failure data to determine a subassembly's probability of failure. When the cause of a system failure is not known, the lifetimes of the suspect subassemblies can be modeled by a Weibull distribution. The performances of IMLEP and a Bayesian approach in estimating the subassembly lifetime distribution parameters under no masking conditions and low masking conditions are similar. However, a Bayesian approach may perform better when the competing risk responsible for a failure represents a smaller percentage of the total failures.