RBF人工神经网络评估特征诊断铝锭脱模故障  

Fault Diagnosis of Aluminum Ingot Demoulding Based on RBF Artificial Neural Network Evaluating Feature

作　　者：黄将诚 沈廷杰 张弛 HUANG Jiangcheng;SHEN Tingjie;ZHANG Chi(College of Artificial Intelligence and Big Data,Chongqing College of Electronic Engineering,Chongqing 401331 China;College of Mechanical Engineering,Dalian University of Technology,Dalian 116024,China)

机构地区：[1]重庆电子工程职业学院人工智能与大数据学院,重庆401331 [2]大连理工大学机械工程学院,辽宁大连116024

出　　处：《机械制造与自动化》2020年第6期193-196,共4页Machine Building & Automation

基　　金：重庆市教育科学“十三五”规划项目(2018-GX-382);重庆市高等教育教学改革研究项目(193411)。

摘　　要：针对电解铝工艺过程中冷却后的铝锭在脱模过程中可能与容器发生粘接从而不能脱离的情况,建立了一种基于特征评估的铝锭脱模诊断模型。利用经验模态分解方法对采集到的敲击信号进行预处理,获得8个高频本征模态函数,从这8个高频本征模态函数和原信号中分别提取6个时域无量纲指标,通过敏感度数值的大小对这些特征进行排序,将这些特征按顺序输入到RBF神经网络中,选择出敏感特征并完成人工神经网络的训练。利用待测试信号的敏感特征对模型进行验证,可以从大量特征中选择出敏感特征从而降低人工神经网络的纬数,诊断正确率达到100%。Because the bonding of aluminum ingot with container wall may occur in the process of demoulding electrolysis aluminum,a novel model of fault diagnosis is established based on the feature evaluation.The original knocking signals are processed by empirical mode decomposition(EMD),thus providing 8 intrinsic mode functions(IMF).Then,time domain dimensionless features are selected from 8 IMF and original knocking signals and ordered by magnitude of sensitivity,those features are input to RBF neural network,thus selecting,the sensitive features and finishing RBF neural network training too.The model is verified by the test signal sensitive features,it is found that it not only can be used to choose the sensitive features from a large number of the features so as to reduce the number of weft artificial neural networks,but also to accurately diagnose the failure in aluminum ingot demoulding process and it diagnostic accuracy is 100%.

关 键 词：铝锭脱模 特征评估 RBF人工神经网络 经验模态分解 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]