| Resistance spot welding technology is widely used in various industry fields including automobile, aerospace, nuclear energy, electronics, light industry, household appliances, due to its operation simplicity, low cost and high efficiency. Spot welded structure is often under multiaxial stress and complex loading conditions, so its fatigue analysis is very difficult. The prediction of spot weld fatigue life always attracts the material field concern. A large number of models and prediction equations have been proposed. Most of them confront many problems while the applications have been limited and no precise quantitative equation is widely accepted so far. Establishing of explicit equations can be avoided with the introduction of BP neural network. The relationship that hidden among the data is directly pulled out resulting in resolving these problemseffectively.In this paper, fatigue test under constant amplitude and two-level loading for spot welded specimens made of ST12 and BGXH50 low carbon steel sheets were carried out on MTS fatigue testing machine. Prediction method for fatigue life of spot weld specimen is given based on BP neural networks. The paper mainly concerned the following aspects:(1) BP neural network was used to study the fatigue life of spot welded specimen made of ST12. BGXH50 sheets under constant amplitude loading. The results indict simulated results matched experimental data well and small error is achieved. It showed that BP neural network can reflect the relationship of fatigue life of spot weld specimen under constant amplitude loading by comparing with the equivalent structural stress method.(2) Back propagation neural network is used to simulate fatigue life of spot weld specimen made of ST12 under two-level loading.Comparing the results of the experiment of fatigue life of spot weld specimen under two-level loading with the simulation results of Back-Propagation neural network, showed the method is feasibilty and accuracy of the method. |
PDF Full Text Request