基于粒子群优化算法-广义回归神经网络的磷化膜耐蚀性预测模型  

Prediction Model for Corrosion Resistance of Phosphating Film Based on Particle Swarm Optimization Algorithm and Generalized Regression Neural Network

作　　者：姚宏[1] 于七龙 那琳[1] YAO Hong;YU Qilong;NA Lin(Hebei Construction Material Vocational and Technical College,Qinhuangdao 066004,China;Northeastern University at Qinhuangdao,Qinhuangdao 066004,China)

机构地区：[1]河北建材职业技术学院,河北秦皇岛066004 [2]东北大学秦皇岛分校,河北秦皇岛066004

出　　处：《电镀与精饰》2021年第11期1-6,共6页Plating & Finishing

基　　金：河北省秦皇岛市科技局项目(201703A017)。

摘　　要：选取磷化液温度、磷化液游离酸度和磷化时间作为输入参数,耐点蚀时间作为输出参数,引入广义回归神经网络(GRNN)建立磷化膜耐蚀性预测模型,并分别采用果蝇优化算法(FOA)、粒子群优化算法(PSO)对平滑因子寻优进而优化预测模型。使用18组训练样本对优化后模型进行训练,9组检验样本用于优化后模型的预测准确度评价。结果表明:PSO-GRNN模型的预测值非常接近真实值,预测相对误差在[0.001,1.778]区间内,均方根误差最低、为0.682。与常规BPNN模型和FOA-GRNN模型相比,PSO-GRNN模型的预测准确度较高,对磷化膜耐蚀性预测效果良好。The generalized regression neural network(GRNN)was introduced to establish the prediction model for corrosion resistance of phosphating film taking the temperature of phosphating solution,free acidity of phosphating solution and phosphating time as the input parameters and the pitting resistance time as the output parameter.Fruit fly optimization algorithm(FOA)and particle swarm optimization algorithm(PSO)were used to optimize the smoothing factor and then the prediction model was optimized.18 groups of training samples were used to train the optimized model,and 9 groups of test samples were used to evaluate the prediction accuracy of the optimized model.The results showed that the predicted value of PSO-GRNN model was very close to the real value,the prediction relative error was within the range of[0.001,1.778],and the root mean square error was the lowest of 0.682.Compared with conventional BPNN model and FOA-GRNN model,the prediction accuracy of PSO-GRNN model was higher,and the prediction effect of PSO-GRNN model for the corrosion resistance of phosphate film was excellent.

关 键 词：磷化膜耐蚀性 耐点蚀时间 广义回归神经网络 果蝇优化算法 粒子群优化算法 

分 类 号：TG174[金属学及工艺—金属表面处理]