Optimal design of step stress accelerated degradation test with multistress considering stress weight

This paper presents a new optimization criterion considering stress weight based on D-s-optimality to solve the low estimation accuracy problem of some stress model parameters when using the existing optimization design criteria to optimize the multistress accelerated degradation test scheme. First, the accelerated degradation model and constraints are determined by analyzing the test characteristics of step-stress accelerated degradation test (SSADT) with multistress. Then, the weight of each stress is identified by fault tree analysis and potential improvement value methods. Combined with the tabu search-particle swarm optimization (TS-PSO) algorithm, the SSADT with multiple stresses of the optimization design process is given. Finally, taking the SSADT with multistress optimization design of motorized spindle as an example, the practicability of the method is verified, and the optimization results are compared with those under D-optimality and D-s-optimality to validate the effectiveness of the method.

Automated summary

This paper presents a new optimization criterion that considers stress weight based on D-s-optimality to solve the low estimation accuracy problem of certain stress model parameters when optimizing the multistress accelerated degradation test scheme using existing optimization design criteria. The accelerated degradation model and constraints are determined by analyzing the test characteristics of step-stress accelerated degradation test (SSADT) with multistress. The weight of each stress is identified using fault tree analysis and potential improvement value methods. Combined with the tabu search-particle swarm optimization (TS-PSO) algorithm, the paper proposes an optimization design process for SSADT with multiple stresses. The practicality of the method is verified through the optimization design of SSADT with multistress for a motorized spindle, and the effectiveness of the method is validated by comparing the optimization results with D-optimality and D-s-optimality.