Accelerated Testing: Statistical Models, Test Plans, and Data Analyses by Wayne Nelson

Published by John Wiley & Sons in 1990 this 601 page book started my career in reliability engineering.

I didn’t know it at the time in the early ‘90s, yet my assigned task to create an accelerated test for a new product would spark an interest in cheating time. Wayne’s book helped make that first accelerated test successful.

The Accelerated Testing book is a compendium of different ways to conduct accelerate tests with a focus on the planning the test and analyzing the data. Wayne is a reliability statistician, and as he will tell you, not an engineer. Thus the book tends to focus on the math.

What I enjoy about the book is the math is not the dry academic derivation driven material, it is full of examples and immediately useful formulas. There is just enough explanation to help the math wonks pursue their interest, and enough practical information to allow engineers to develop and conduct meaningful experiments.

Book Details and Comments

The book is essentially a textbook complete with end of chapter exercises. It also is chock full of references for those interested in gaining a deeper understanding of accelerated testing.

Wayne organized the work in a typical fashion with an introduction, background material, then launches into the simpler life tests with a single stress before addressing more complex life testing methods.

He does tend to focus on life data distributions, like Lognormal and Weibull, which is understandable given their importance when describing life data.

Beyond the distributions there is also a full treatment of acceleration models including Arrhenius and many others.

It is the mixture of life distributions at use and accelerated conditions along with an appropriate means to translate between the two that form the heart of the book.

Other testing discussions include step stress and degradation approaches.

Analysis Techniques

Wayne describes analysis methods in detail, as they support test planning and data analysis. He starts with graphical methods, least squares when dealing with complete (everything has failed) data, and later in the book a detailed treatment of the maximum likelihood estimation method.

The analysis in large part depends on the data and objective and he details best practices from both a practical and statistical point of view. In nearly every section there are lists of assumptions to check, often with tools or approaches to verify assumptions. This is too often missing from books about life data analysis and accelerated testing. One can make serious errors by ignoring the underlying assumptions of your accelerated testing.

Another consideration for the analysis of life data or for test planning is when dealing with competing failure modes and other real world considerations when planning or analyzing your experiment. Besides a full chapter the advice on how to create a meaningful experiment is spread throughout the book.

Recommendation

If you are considering running an accelerated test, this old book is the one you need to be successful. Get it, study it, and apply the principles and approaches as appropriate.

My first reliability engineering project started with the last chapter of Accelerated Testing. Since then I’ve read and applied material from every chapter over and over. Highly recommended.

As you may be able to tell, I like this book. It is my goto resource. What are your favorite or goto references and why? Leave a comment and share your recommendation.