Optimal design of step-stress accelerated degradation tests based on genetic algorithm and neural network

• Published in: Quality engineering Vol. 36; no. 1; pp. 66 - 79
• Main Authors: Liu, Gen, Wang, Zhihua, Bao, Rui, Mao, Zelong, Ren, Kunpeng
• Format: Journal Article
• Language: English
• Published: Milwaukee Taylor & Francis 02.01.2024; Taylor & Francis Ltd
• Subjects: Accelerated tests; Degradation; Design optimization; Error analysis; genetic algorithm; Genetic algorithms; multiple decision variables; Neural networks; optimal design; Optimization models; proxy model; Sensitivity analysis; step-stress accelerated degradation test
• ISSN: 0898-2112; 1532-4222
• DOI: 10.1080/08982112.2023.2225583

In this study, the optimal design of step-stress accelerated degradation tests is focused. An optimization model is proposed where an improved accelerated degradation model is involved to comprehensively consider the influence of accelerated stress and the measurement error. Then, a novel optimal design method is constructed, where multiple decision variables can be simultaneously optimized based on neural network and genetic algorithm. An effective sensitivity analysis method is further established to quantitively illustrate the influence of the predetermined model parameters on the optimal results. Finally, a case study is implemented, and a series of comparisons are implemented to demonstrate the effectiveness and rationality of the proposed method.