Degradation Modeling, Reliability Analysis, and Quality Improvement

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

Proposal ID: DMS-0204247 PI: Vijayan Nair Title: Degradation Modeling, Reliability Analysis, and Quality Improvement Abstract Degradation data are a very rich source of reliability information and offer many advantages over the analysis of time-to-failure data. This project will develop a flexible class of models for analyzing degradation data and related reliability inference. These results will be used to obtain efficient methods for the design and analysis of accelerated tests and reliability improvement experiments. In this work, time-to-failure is defined as the level crossing (first-passage time) of a specified degradation threshold. The first part of the project will consider models based on diffusion processes for analyzing degradation data with continuous sample paths. These models can accommodate a variety of degradation rates and shapes. They also lead naturally to a wide class of time-to-failure distributions. The inverse Gaussian distribution plays a central role, similar to the exponential distribution with constant hazard rates. The second part will study a class of degradation-based models for repairable systems data that is quite analogous to non-homogeneous Poisson processes with failure data. Multi-state degradation models will also be considered. This work is an interesting generalization of the formulation in traditional statistical process control. Degradation data allow for more informative accelerated tests and reliability improvement studies. Several research topics on design of accelerated degradation tests, analysis of data from reliability improvement experiments, and robust design studies will be pursued. There has been tremendous emphasis on quality and reliability improvement in industry, driven by global competition and increasing customer expectations. There is also continued pressure to reduce product development costs and cycle times. Design, development, and manufacturing of highly-reliable products in this environment raise many challenges. The focus within the reliability area has traditionally been on the collection and analysis of time-to-failure data. High reliability implies few failures, so reliability estimation and improvement for product and process design can be extremely difficult. Fortunately, recent advances in sensing and measurement technologies are making it feasible to collect extensive amounts of data on degradation and other performance measures associated with components, systems, and manufacturing processes. However, the lack of flexible models and methods inference has been a major deterrent to the widespread use of degradation data for reliability analysis. This project will develop new models and methods for analyzing reliability data and use them for quality improvement.