A failure-rate based multiple step-stress dependent competing risks model with applications to aerospace electrical connectors data

A novel failure-rate based multiple step-stress accelerated life testing (SSALT) model is proposed to accommodate dependent competing risks data within the progressive Type-I censoring framework. The cause-specific (latent) time-to-failure distributions conform to the flexible proportional hazard (PH) class of distributions that can capture monotonic and non-monotonic failure rate patterns. Dependency among the competing causes is addressed then assuming that the stress-specific time-to-failures follow the Marshall-Olkin bivariate proportional hazard (MOBPH) class of distributions. Employing the well-known power rule model as an acceleration factor and the principle of ML, the point estimates of all the model parameters are obtained. The power rule model facilitates the dimension of the optimization problem not growing up even with an increase in the number of stress levels. Interval estimates of the model parameters are computed based on the observed Fisher information matrix. To study the effectiveness of the proposed methods, extensive simulation experiments are performed. The proposed dependent competing risk SSALT model can be implemented quite effectively even when there are no cause-specific failures in one or more stress levels. The practical relevance of the proposed model is illustrated by a real-life example on Y11X-1419 aerospace electrical connector failures.