模糊神经网络技术在铝电解故障诊断中的应用  

The Application of Fuzzy Neural Network in the Aluminum Eelectrolysis Fault Diagnosis

作　　者：李界家[1] 李旸[2] 吴成东[1] 

机构地区：[1]东北大学信息学院,辽宁沈阳110004 [2]沈阳建筑大学信息与控制工程学院,辽宁沈阳110168

出　　处：《沈阳建筑大学学报（自然科学版）》2005年第4期390-394,共5页Journal of Shenyang Jianzhu University：Natural Science

基　　金：建设部基金项目(03-2-117)

摘　　要：目的为了提高生产效率、降低成本、安全生产,通过对铝电解故障进行有效的检测和预报,减少铝电解过程中阳极效应、热槽、冷槽故障的发生.方法通过对铝电解故障发生机理和故障发生时相关特征量变化趋势的分析,基于模糊逻辑理论,建立了多级模糊故障检测模型,采用BP神经网络建立了故障分类模型,实现对铝电解故障的检测和预报.结果降低了模糊系统的维度,减少了规则数量,采用多级模糊与神经网络相结合的故障诊断预报的方法,提前了预报时间,提高了预报准确率.结论铝电解模糊神经网络故障诊断方法,有效地降低铝电解的故障发生率,降低了能耗,提高了铝的产量和质量,具有良好的应用前景.This paper, through the efficient detection and prediction of the faults, takes the anode effect and hot electrobath and cool electrobath fault for the research background in the course of aluminum electrolysis to aim at increasing productive efficiency, reducing cost, and producing safely. Through analyzing aluminum electrolysis fault generating principle and eigenvector variety tendency, based on fuzzy logic theory, this paper adopts multilevel fuzzy systems to establish the model of faults detection. Therefore, it has solved the dimension problem of fuzzy system, reduced the rules, and promoted the veracity of fault diagnosis. Based on neural network theory, it has established the model of fault classification based on BP neural network, realized the detection and forecasting aluminum electrolysis faults. The experimental result proves it effective to combine the fuzzy system with neural network technology in the aluminum ectrolysis fault diagnosis; the forecast time is advanced and the accurate rate of forecast is raised. This method of fuzzy neural network for the aluminum electrolysis fault diagnosis can reduce aluminum electrolysis fault generant rate and energy wastage, improve the output and quality, and therefore has an applicable prospect.

关 键 词：铝电解 故障诊断 故障分类 模糊系统 神经网络 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]