Vehicle exhaust components and systems under fatigue loading often show multiple failure modes, which should be treated, at least theoretically, with rigorous advanced bi-modal and multi-modal statistical theories and approaches. These advanced methods are usually applied to mission-critical engineering applications such as nuclear and aerospace, in which large amounts of test data are often available. In the automotive industry, however, the sample size adopted in the product validation is usually small, thus the bi-modal and multi-modal phenomena cannot be distinguished with certainty.
The purpose of this paper is two-fold: (1) a large amount of historical fatigue data of several types of welded exhaust components belonging to several diverse programs are collected and statistically analyzed; the similarities and differences in the data patterns, the standard deviation of the cycles to failure in particular, of the different groups of the data are examined, and (2) with the statistical information obtained from the historical data and a previously developed Bayesian statistics approach, examples in design limit/curve construction are provided to demonstrate the values of the historical data in product validation.
