基于神经网络的钛合金微弧氧化膜层性能预测模型的建立  

Establishment of a neural network model for prediction of the properties of micro-arc oxidation coating on titanium alloy

作　　者：李明哲[1] 牛宗伟[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：《电镀与涂饰》2014年第19期827-830,共4页Electroplating & Finishing

基　　金：国家自然科学基金(51005140);山东省自然科学基金(ZR2010EQ037);山东理工大学青年教师发展支持计划经费资助

摘　　要：采用微弧氧化技术在钛合金表面制备了陶瓷膜层。利用正交试验得到了不同工艺参数(电流密度、频率、占空比和氧化时间)与膜层性能(厚度、粗糙度和显微硬度)数据,借助MATLAB软件建立了由4个输入向量、13个隐含层节点和3个输出向量组成的BP人工神经网络模型。该网络能较好地掌握工艺参数与膜层性能之间的内在规律,并能高精度预测膜层的性能,3个性能参数的平均预测误差分别为4.1%、4.2%和2.4%,最大预测误差分别为8.2%、8.6%和3.1%。Ceramic coatings were prepared on titanium alloy by micro-arc oxidation. A BP artificial neural network model with 4 input vectors, 13 hidden layer nodes, and 3 output vectors was established by using MATLAB software. The coatings properties data（thickness, surface roughness, and microhardness） were obtained at different technological parameters（current density, frequency, duty cycle, and oxidation time） by orthogonal test. The inherent laws between technological parameters and properties can be better revealed by the network, and the coating properties can be predicted with high precision. The average and maximum prediction errors are 4.1% and 8.2% respectively for thickness, 4.2% and 8.6% for surface roughness, and 2.4% and 3.1% for microhardness.

关 键 词：钛合金 微弧氧化 神经网络 性能预测 

分 类 号：TG174[金属学及工艺—金属表面处理] TP183[金属学及工艺—金属学]