Strength prediction of composite material bolted joints based on BP neural networks
To predict the strength of composite bolt connections,tensile tests were designed and the ultimate loads were recorded.The test variables were quantified as input parameters for the neural network model.A BP neu-ral network was used to train the model.Three reserved test results were used to validate the model's accuracy.The prediction errors for the three test groups were 6.13%,1.63%,and 3.34%respectively.The composition of the training groups was adjusted and the model was retrained.The results showed that the model had excellent predictive ability for new variable values.Finally,a finite element analysis model was established for comparison with the neu-ral network.The results indicated that compared to the finite element method,the neural network approach could predict the strength of composite bolt connections more quickly with the same precision.
万方数据
维普
