建立一个基于遗传算法的神经网络模型对焊道尺寸进行预测。输入层单元为电流、送粉速率、焊接速度、离子气流量;输出层单元为焊道宽度、高度、熔深。采用正交试验方法得出的25组数据作为训练样本,控制变量法得出的16组数据作为预测样本。结果表明:模型预测精度高、效率快,训练误差范围在4.1%内,预测误差范围在6.5%内。并提出将模型应用到单道多层和多道多层增材再制造工艺中的方法。 Establish a neural network model based on genetic algorithm to predict the weld bead size. Input layer unit for the current, powder feeding rate, welding speed, ion flux; output layer unit for the weld width, height, penetration. Twenty-five sets of data obtained by orthogonal test method were used as training samples and 16 sets of data obtained by controlled variable method as predictive samples. The results show that the model has high prediction accuracy and efficiency, the training error range is within 4.1% and the prediction error range is within 6.5%. And put forward the method to apply the model to single-channel multi-layer and multi-channel multi-layer additive material remanufacturing process.