Fatigue life prediction for welding components based on hybrid intelligent technique

Abstract This paper details the fatigue life prediction model for welding components based on hybrid intelligent technology. We make use of the capabilities and advantages of rough set theory, particle swarm optimization (PSO) algorithm and BP neural network for establishing of the fatigue life prediction model. Firstly, rough set theory was used to deal with the original fatigue sample data; the minimum fatigue feature subset was obtained. Secondly, improved PSO algorithm was used to optimize the initial weighs and thresholds of the BP neural network, which resolves such problems as local extremum and slow convergence that exist in the traditional BP neural network. At last, minimum reduced subset was inputted into the optimized BP neural network to construct the novel fatigue life prediction model for welding components by the continuous training and adjusting. Sample data of the titanium alloy welded joints was used to verify the correctness and validity of the novel fatigue life prediction model, simulation results show that the fatigue life prediction model proposed in this paper has better fault tolerance, higher precision, and can fitting fatigue life value more accurately than traditional BP model. Consequently, the model based on hybrid intelligent technology can provide an effective new approach to predict the fatigue life of welded joints.