New analysis methods of multilevel accelerated life tests

The development of multilevel accelerated life stress tests has been hindered by the difficulty of performing adequate and satisfactory data analysis. When 2, 3 or 4 levels of stress exist, even a computer analysis can be formidable. Nelson (1990) showed that even a simple analysis of seven levels of a single stress can become quite a challenge. This paper shows a new computerized software approach that has recently become available to the reliability engineer. It makes the analysis of multilevel stresses much easier. Complete or suspended data with 2 different stresses and multiple levels can now be analyzed employing the maximum likelihood estimator to achieve best fit solutions. In addition, confidence estimates can be made and the detailed stress-life relationship of the important parameters may be investigated. The data analysis solutions will permit the reliability person a wider latitude in selecting test stresses, methods and conditions when testing products in the future.