激光熔覆镍基熔覆层截面形貌预测

利用BP神经网络建立激光熔覆关键工艺参数(激光功率、扫描速度、送粉电压、送粉载气流量)与熔覆层截面形貌(熔宽、余高)的预测模型。以激光熔覆工艺参数为输入,熔覆层的截面形貌为输出,利用工艺试验数据对网络进行训练,实现对输入和输出的高度映射。结果表明,BP神经网络可以较好地对熔覆层形貌进行预测,同时双隐藏层BP神经网络模型预测结果误差波动更小,表现出优良的稳定性,最大预测误差相比单隐藏层神经网络大大降低。

Prediction of cross section morphology of Ni based cladding layer by laser cladding

BP neural network was used to prediction model of key parameters(laser power,scanning speed,powder feeding voltage,powder carrier gas flow rate)of laser cladding and cross section morphology(melting width and reinforcement height)of cladding layer.Taking the laser cladding parameters as the input and the cross-section morphology of the cladding layer as the output,the network was trained by the process test data to realize the high mapping between input and output.The results showed that BP neural network could better predict morphology of cladding layer and the prediction error of double hidden layer BP neural network model fluctuates was less,which showed excellent stability.Compared with single hidden layer neural network,the maximum prediction error of BP neural network was greatly reduced.