基于GA-BP神经网络的镍基合金熔覆涂层形貌预测

Prediction of Ni-based Alloy Cladding Coatings Topography Based on GA-BP Neural Network

针对镍基合金在熔覆过程中成形质量难以控制的问题,通过建立3因素、4水平的正交试验参数表,在完成22组实验样本数据的基础上,建立经遗传算法优化的误差反向传播的神经网络模型,通过激光加工工艺参数预测熔覆层宏观形貌,形成两者之间的高度映射关系。采用极差分析法找出工艺参数对熔覆层宏观质量影响的大小关系并确定最佳工艺参数,为了适用于大面积熔覆的需要,引入熔覆层搭接率作为新的工艺参数来优化模型算法,并通过宏观质量分析确定适宜的扫描间距,利用5组不同搭接率的样本对模型的精度进行检验。实验结果表明,经过优化后GA-BP模型具有很高的预测精度,预测结果与测试样本之间的平均相对误差为3.951%,验证了该模型在理论与实际中的可行性,对提高镍基合金熔覆涂层的产品质量具有重要义。

To solve the problem that the forming quality of Ni-based alloys is difficult to control during cladding.Via establishing a 3 factors and 4 levels orthogonal test parameter table and on the basis of completing 22 sets of experimental sample data,build an error reverse propagation neural network model which is optimized by genetic algorithm.Predicting the macroscopic morphology of the cladding layer by laser processing parameters,form a high mapping relationship between the two.Optimum parameters can be attained by finding out the influence of process parameters on the macroscopic quality of the cladding layer using range analysis method.To meet the needs of large area cladding,introduce the cladding layer overlap rate as a new process parameter to optimize the model algorithm,and the appropriate scanning interval is determined by the macro-quality analysis.Using 5 sets of samples with different overlap rates to test the accuracy of the model.The experimental results show that the optimized GA-BP model has a high prediction accuracy,and the average relative error between the predicted results and the test samples is 3.951%,which verifies the feasibility of the model in theory and practice,and is of great significance for improving the product quality of nickel-base alloy cladding coatings.