A methodology for predicting the variance of fatigue life incorporating the effects of manufacturing processes
Fatigue is the predominant failure mode for structures subject to dynamic loading. Experiment results have shown that fatigue life variance of nominally similar components could be so large that the average life is virtually meaningless for design a safe product economically without the associated fatigue variance. This phenomenon poses a significant challenge to the design of fatigue critical products such as aircraft where safety is of paramount importance and costs involved are enormous. Consequently, understanding and controlling the variance of fatigue life are instrumental in enhancing competitiveness of the product design and manufacturing. ^ Although it might be intuitively true that manufacturing processes have an impact on the variances of residual stress and fatigue life, no published literature documenting such findings has been found. By extending the methodology of surface integrity study, this thesis has identified experimentally that machining processes have significant influence on the variances of residual stress and fatigue life of the machined components. The significance is well appreciated by computing the ratio of average fatigue life over the fatigue life corresponding to 95% reliability. Under the machining conditions selected from a well-respected source, the drilled holes have a ratio of 1.87 while the internally ground holes have a ratio of 214.5. A Weibull distribution of fatigue life is assumed in the computation. It is thus suggested that the variance of fatigue life be considered as a new process capability facilitating the selection of manufacturing processes for fatigue critical products. ^ Based on the foregoing results, it is essential for any models predicting the variance and average value of fatigue life to incorporate the machining effects correctly. This thesis proposes a methodology for systematically predicting the variance of fatigue life. The variance of fatigue life can be analytically decomposed into individual components by this method. It is the first of its kind and has two implications. First, it provides a tool for pinpointing key driving factors of the variance of fatigue life, which is essential for the variance reduction of fatigue life. Second, the time consuming and costly fatigue tests to obtain critical variance information for reliability design may be divided into less time consuming tests for obtaining variance information for individual variables. Based on the variance prediction tool, a methodology for systematically incorporating manufacturing influence into the prediction of variance and average value of fatigue life is proposed and verified experimentally. The verification configuration is a structure with a central hole in the high cycle fatigue regime. The predicted fatigue life matches the actual average fatigue life well. The analytical partition of the variance of fatigue life into individual components reconfirms previous finding that the variance of residual stress plays a dominant role. The Chi-square test shows that good estimates of the variance of fatigue life are also obtained. ^
Major Professor: C. Richard Liu, Purdue University.
Engineering, Industrial|Engineering, Mechanical