An artificial neural network model was established to predict the ultimate bearing capacity of aluminum alloy plate joints under compression-bending state. The input parameters of the network are composed of bolt diameter (R), flange plate thickness (T) and connecting plate thickness (L). The output parameter is ultimate bearing capacity. In this paper, firstly the finite element calculation results are compared with the experimental results to verify the accuracy of finite element simulation. Then 216 sets of data were obtained using the finite element program ABAQUS, which were used for training, verification and testing of the neural network model. In addition, the influence of sample size on the prediction accuracy of neural network is analyzed, and a structural optimization method combining finite element calculation and neural network prediction is proposed. The results show that the neural network model is accurate and effective in predicting ultimate bearing capacity, and the linear regression correlation coefficient is 0.98438. The data sample can be reduced to a certain extent to save the time costs and computing resources; It is reasonable and efficient to use the method of finite element parameter analysis and artificial neural network prediction to optimize the structure, which broadens the thinking of structure optimization.

建立人工神经网络模型预测压弯状态下铝合金板节点的极限承载力。网络的输入参数由螺栓直径（R）、法兰板厚度（T）和连接板厚度（L）组成。输出参数为极限承载力。本文首先将有限元计算结果与实验结果进行对比，以验证有限元模拟的准确性。然后利用有限元程序ABAQUS得到216组数据，用于神经网络模型的训练、验证和测试。此外，分析了样本量对神经网络预测精度的影响，提出了一种结合有限元计算和神经网络预测的结构优化方法。结果表明，神经网络模型对极限承载力预测准确有效，线性回归相关系数为0.98438。可以在一定程度上减少数据样本，节省时间成本和计算资源；采用有限元参数分析和人工神经网络预测的方法对结构进行优化，合理、高效，拓宽了结构优化的思路。可以在一定程度上减少数据样本，节省时间成本和计算资源；采用有限元参数分析和人工神经网络预测的方法对结构进行优化，合理、高效，拓宽了结构优化的思路。可以在一定程度上减少数据样本，节省时间成本和计算资源；采用有限元参数分析和人工神经网络预测的方法对结构进行优化，合理、高效，拓宽了结构优化的思路。