【摘要】 为了准确预测双曲度板材成形回弹量，控制板材加工成形质量，利用ABAQUS有限元软件对板材成形回弹过程进行仿真，构建长短时记忆网络模型（LSTM）。针对麻雀搜索算法（SSA）容易陷入局部最优的问题，提出基于Circle混沌映射、反向学习、高斯与柯西变异扰动的改进麻雀搜索算法，优化LSTM模型的学习率、迭代次数、隐藏层神经元个数，并将模型与BP神经网络模型、LSTM模型和普通SSA-LSTM模型进行对比分析。结果表明，该模型对双曲度板材成形回弹预测达到整体最优预测效果，具有一定有效性和可行性。更多还原

【Abstract】 In order to accurately predict the springback of doubly curved plate forming and control the forming quality of the plate, the ABAQUS finite element software was used to simulate the forming springback process of the plate, and the long-term and short-term memory network model(LSTM) was constructed. Aiming at the problem that the sparrow search algorithm(SSA) is prone to local optimum, an improved sparrow search algorithm based on circle chaotic mapping, reverse learning, gaussian and cauchy variant perturbation was proposed. The learning rate, number of iterations and number of hidden layer neurons of the LSTM model were optimized. And the model was compared and analyzed with the BP neural network model, LSTM model and ordinary SSA-LSTM model. The results show that the proposed model achieves the overall optimal prediction performance on plate springback prediction, which has certain effectiveness and feasibility.更多还原