基于GA-BP神经网络的镁合金微弧氧化膜层厚度预测

发布时间：2018-01-25 00:18

本文关键词： 镁合金微弧氧化反向传播神经网络遗传算法膜层厚度　出处：《兵器材料科学与工程》2017年01期 　论文类型：期刊论文

【摘要】：为直观地检验膜层的质量,建立微弧氧化工艺参数(电流大小、脉冲宽度、氧化时间)与微弧氧化膜层厚度之间的反向传播(BP)神经网络预测模型,其结构为3-10-1(即3个输入神经元,10个隐含层节点,1个输出神经元)。采用遗传算法(GA)优化BP神经网络的初始权值和阈值,构建基于遗传算法神经网络的膜厚预测模型。用GA-BP神经网络对膜厚进行模型仿真,并将仿真结果与BP神经网络模型仿真结果进行对比。结果表明,GA-BP网络模型预测值的平均误差为1.65%,最大误差为9.75%,而BP模型预测结果的平均误差为8.62%,最大误差为13.68%。GA-BP神经网络模型预测精度要优于BP神经网络模型。
[Abstract]:In order to test the quality of the film directly, the neural network prediction model of the back propagation between the parameters of micro-arc oxidation (current, pulse width, oxidation time) and the thickness of the micro-arc oxide film was established. Its structure is 3-10-1 (that is, 3 input neurons, 10 hidden layer nodes and 1 output neuron). Genetic algorithm (GA) is used to optimize the initial weight and threshold of BP neural network. The prediction model of membrane thickness based on genetic algorithm neural network is constructed. The GA-BP neural network is used to simulate the membrane thickness, and the simulation results are compared with the simulation results of BP neural network model. The average error of the prediction value of GA-BP network model is 1.65, the maximum error is 9.75, and the average error of BP model is 8.62%. The maximum error is 13.68. The prediction accuracy of GA-BP neural network model is better than that of BP neural network model.
【作者单位】：江苏科技大学机械工程学院;
【分类号】：TG174.4;TP183
【正文快照】： 微弧氧化是通过在电解质溶液中发生阳极微等离子体击穿(火花放电或微弧),可在镁合金表面获得陶瓷膜层,该膜层具有良好的耐磨、耐蚀性能且与基体结合良好,从而极大改善镁合金的表面性能[1]。其中,微弧氧化膜的厚度对镁合金的表面性能有重要影响,故需通过所需膜厚选择合适的工艺

【相似文献】