Accelerated life test calculations using the method of maximum likelihood: an improvement over least squares

摘要

Reliability calculations from failure time data are often performed using least squares regression of the log of the mean time to failure (MTTF) vs. inverse temperature. It is demonstrated that this method produces estimates of relevant parameters like E_a and MTTF with confidence limits that are larger and more variable than necessary. In comparison, the method of maximum likelihood makes more efficient use of the data and as a result, does not suffer from the above difficulties. A previously published dataset is used to compare the two techniques and the maximum likelihood approach is shown to be superior.