论文部分内容阅读
通过对4Cr5MoSiV1模具钢分别进行不同的加速电压、照射距离和轰击次数下的电子束表面改性试验,取得模具钢试样的实际耐磨性测试数据,并将其作为神经网络的训练样本和验证样本,建立了3×12×1三层网络模型进行模具钢电子束表面改性的神经网络预测,并对网络的预测精度进行分析。结果表明,该三层神经网络可进行较高精度的模具钢电子束表面改性神经网络预测。
The actual wear resistance test data of the mold steel samples were obtained through different electron beam surface modification tests at different accelerating voltage, irradiation distance and number of bombardment on the 4Cr5MoSiV1 tool steel, which were used as the training samples and verification of the neural network Samples, a 3 × 12 × 1 three-layer network model was established to predict the surface modification of die steel by electron beam and the prediction accuracy of the network was analyzed. The results show that the three-layer neural network can be used to predict the surface modification of the die steel electron beam surface with higher accuracy.