Early detection of reliability changes for a non-poisson life model using field failure data

The monitoring of changes in the pattern of failure field data is an important requirement of any reliability management program. Systematic monitoring of field failures also ensures that adverse changes in the production process are identified quickly in order to avoid facing serious reliability and warranty problems. These kinds of problems are mainly caused by unanticipated failure modes, unknown changes in raw material, changes in operating environmental conditions, etc. The major difficulty facing companies is attempting to isolate and detect changes in field reliability when a large installed base of products is already in existence. Metrics such as return rates do not adequately address the problem and a reduction in reliability will generally remain undetected until it becomes a serious warranty issue. The early detection of reliability problems through analysis of field data will save the manufacturer large amounts of money in warranty costs, improve product quality while also retaining customer goodwill. Significant research has been published on the monitoring and control of production processes using statistical process control techniques. However little of this research has been applied to the area of field reliability due to the complexities associated with field data. In addition, manufacturing companies face the nonPoisson failure process in practice, while many of the research papers mainly deal with constant rate parametric techniques. This paper develops and validates a technique for identifying small changes in field reliability irrespective of how large the installed base of product is. Upper and lower limits are identified for monthly returns. This research applied the early detection procedure for a lognormal failure process although the technique described could equally be applied to other nonconstant failure rate processes. Actual field reliability data obtained over a two-year period is used to illustrate and validate the technique. The method also allows traceability back to specific months in which the product was manufactured. This allows the manufacturer to quickly recognize any improvement or deterioration in field reliability and identify how changes in production have affected field reliability.