Damage-Based Models for Step-Stress Accelerated Life Testing

Step-stress accelerated testing of mechanical components offers great potential for improvements in reliability demonstration bench testing. Most mechanical products are designed to operate for a long period of time and in such a case, life testing is a relatively lengthy procedure. Lengthy tests tend to be expensive and the results become available too late to be of much use. To reduce the experimental cost significantly and provide an efficient tool to assess the life distribution for highly reliable products, a step-stress accelerated test was developed. Step-stress testing is achieved by testing the component at stress levels greater than operational levels in a stepwise fashion to reduce the time-to-failure, or life. Nelson's cumulative exposure model has been widely used to determine the reliability of the test parts under step-stress accelerated testing. The Nelson model does not provide information pertinent to failure mechanism and sites. Further, the model assumes that the previous exposure stress history of a test part is accounted for by the cumulative exposure distribution, instead of the damage curve. Therefore, damage curve analysis (DCA), double linear damage rule (DLDR), and linear damage rule (LDR) are adopted to assess the experimental results from step-stress accelerated testing. Investigation into Nelson's cumulative exposure model and the three cumulative damage rules to predict fatigue reliability of mechanical products under step-stress accelerated testing was performed. The predictions by Nelson's cumulative exposure model and DCA are better than DLDR and LDR. Due to its simplicity of implementation, LDR is also capable of estimating failure life distribution reasonably well.