以Moldflow软件模拟得到的不同工艺参数下飞机机头雷达罩模型的翘曲变形量为训练样本，在雷达罩模型成型工艺参数与其翘曲变形量间建立反向传播（Back Propagation，BP）神经网络模型，然后采用遗传算法对工艺参数进行优化，得到使雷达罩模型翘曲变形量最小的工艺参数并进行试验验证。结果表明：在相同工艺参数下由BP神经网络得到的雷达罩模型翘曲变形量与采用Moldflow软件模拟得到的翘曲变形量相近，相对误差小于4%，证明了BP神经网络的可靠性；模拟得到雷达罩模型的最优成型工艺参数为注塑温度295℃、模具温度80℃、注塑时间0.75 s、保压时间8 s、保压压力125 MPa，此时翘曲变形量最小，为0.121 3 mm；在最优成型工艺参数下进行注塑成型后得到的雷达罩模型最大翘曲变形量为0.126 0 mm，试验结果与预测结果间的相对误差小于3.7%，验证了BP神经网络与遗传算法相结合方法的准确性。

Taking warpage deformation of the aircraft nose radome under different process parameters obtained by Moldflow software as training samples, a back propagation (BP) neural network model was established between the process parameters of the radome model and its warpage deformation values. Then genetic algorithm was used to optimize the process parameters, and the process parameters of the radome model with the smallest warpage deformation value was obtained. The results show that with the same process parameters, the warpage deformation value of the radome model obtained by BP neural network was similar to that simulated by Moldflow software, and the relative error was less than 4%, which proved the reliability of BP neural network. The simulated optimal molding process parameters of the radome model were injection temperature of 295 ℃, mold temperature of 80 ℃, injection time of 0.75 s, pressure holding time of 8 s, and pressure holding pressure of 125 MPa; the warpage deformation value was the smallest of 0.121 3 mm. The maximum warpage deformation of the radome model was 0.126 0 mm after injection molding with the optimal molding process parameters, and the ralative error between the experimental result and the predicted result was less than 3.7%, which verified the accuracy of the method of combining BP neural network and genetic algorithm.