Abstract: Product life can be assessed effectively in the short period of time by using accelerated life test. Aimed at problems resulted from poor predictive ability of the previous accelerated life model, a method to establish accelerated life model for machine guns based on LS-SVM was proposed. It took the machine gun’s shooting ammunition quantity before the life end as the life feature and selected the test ambient temperature, barrel’s maximum temperature, shooting interval, maximum pressure in bore as the accelerated stresses. A genetic algorithm was adopted to determine the optimal parameters of LS-SVM. The prediction results show that the model established in this paper are better than the general transformation and BP neural network models obviously, and the LS-SVM method are effective on accelerated life prediction for machine guns.

Key words: system engineering, LS-SVM, neural network, genetic algorithm, accelerated life test