Intelligence model-driven multi-stress adaptive reliability enhancement testing technology

This paper proposes an Intelligence Model-Driven Multi-Stress Adaptive Reliability Enhancement Testing (IMD-MSARET) technology, which aims to address the problems of rigid test design, low efficiency, and high consumption of test specimens in traditional reliability enhancement testing under multi-stress scenarios through efficient multi-factor test design and data analysis. The IMD-MSARET technology combines sequential test design and artificial intelligence techniques to dynamically construct and update the mapping relationship model between multiple stresses and failure characteristic during testing. In terms of mathematical models, we propose a Tuna Swarm Optimization-Gaussian Process Regression (TSO-GPR) model, which combines the global search capability of the tuna swarm optimization algorithm and the accurate prediction capability of Gaussian process regression, effectively handling the complex nonlinear relationships between multiple stresses and failure characteristic. In addition, we propose a three-factor step-by-step screening algorithm and scoring model to determine the optimal sequential test points. Case study verification shows that IMD-MSARET outperforms traditional methods, such as simple random testing and orthogonal testing, in terms of test efficiency, prediction accuracy, and test item consumption. The TSO-GPR model-driven IMD-MSARET is superior to GPR, TSO-SVM, support vector machine and Tuna Swarm Optimization-Backpropagation Neural Network (TSO-BPNN) in terms of accuracy, efficiency, and test item cost for constructing multi-stress limit envelopes.