利用人工神经网络技术优化钛合金微弧氧化工艺  

Studies on Titanium Alloy Micro-arc Oxidation Process Optimized by Artificial Neural Network Technology

作　　者：牛宗伟[1] 李明哲[1] 

机构地区：[1]山东理工大学机械工程学院,山东淄博255000

出　　处：《电镀与精饰》2015年第4期10-14,19,共6页Plating & Finishing

基　　金：国家自然科学基金(51005140);山东省自然科学基金(ZR2010EQ037)

摘　　要：在正交试验基础上,利用MATLAB软件建立BP神经网络膜层耐蚀性能预测模型,通过网络模型对样本实验数据的学习,确定最佳网络结构,对钛合金微弧氧化膜耐蚀性能进行预测,并对微弧氧化工艺参数进行了优化。分析确定BP神经网络结构为4-7-1三层结构,该网络结构能够较好地掌握输入参数(电流密度、脉冲频率、占空比和氧化时间)与输出数据间(膜层腐蚀电位)的内在规律,网络的平均训练误差与平均预测误差分别为0.101%和0.596%,BP网络优化后,所得最佳参数Ja为15A/dm2、脉冲频率600Hz、占空比10%、氧化t为12min。Based on the data of orthogonal experiments,the prediction model of BP neutral network film corrosion resistance was established by using of MATLAB software; the optimal network structure was determined by training of the sample experimental data; the corrosion resistance of micro-arc oxidation coating on titanium alloy was predicted and the micro-arc oxidation process parameters were optimized.The results showed that a 4-7-1 three-tier neural network structure was determined by analysis; the network structure could bettergrasp the inherent law between input parameters（ current density,pulse frequency,duty cycle and oxidation time） and output data（ films corrosion potential）; the average training error and average prediction error of the network were 0.101% and 0.596% respectively.After the BP network optimization,the optimal parameters were as follows： current density was 15 A/dm2,pulse frequency was600 Hz,duty cycle was 10%,and oxidation time was 10 minutes.

关 键 词：钛合金 微弧氧化 神经网络 耐蚀性能 

分 类 号：TG174[金属学及工艺—金属表面处理]