激光熔覆得到的熔覆层面积与工艺参数存在一定的耦合关系, 且熔覆层中气孔缺陷的产生随工艺参数的改变而有所差异。基于试验及神经网络, 以单道单层熔覆层为研究对象, 建立了以加工工艺参数对熔覆层面积和熔覆层气孔率进行预测的基于BP神经网络的预测模型。依据输入量与输出量特点设计了神经网络结构, 通过激光熔覆试验采集熔覆层面积值与气孔率样本, 利用训练集训练BP神经网络, 建立输入量与输出量之间的映射关系, 并用测试集对训练好的预测模型进行检验。测试结果表明, 基于BP神经网络的预测模型对熔覆层面积的预测精度较高, 气孔率预测模型得出的预测趋势对于气孔率的预测具有一定参考价值。验证了这两个预测模型在理论和实践上的具有一定的可行性与有效性。

The area of the cladding layer obtained by laser cladding has a certain coupling relationship with the process parameters, and the generation of pore defects in the cladding layer varies with the process parameters. Based on the experimental and neural networks, a single-layer single-layer cladding layer is used as the research object, and a prediction model based on BP neural network to predict the cladding area and the porosity of the cladding layer is established based on processing parameters. The neural network structure was designed with the output characteristics. The cladding area and bubble rate samples were collected by laser cladding test. BP neural network was trained by training set to establish the mapping relationship between input quantity and output quantity, and the trained prediction model was tested by test set. The test results show that the prediction model based on BP neural network has higher prediction accuracy for the cladding area, and the prediction trend derived from the porosity prediction model has certain reference value for the prediction of porosity. It is verified that the two prediction models have certain feasibility and effectiveness in theory and practice.