基于神经网络的化工过程故障诊断  

The artificial neural network based fault diagnosis of chemical process

作　　者：张卫华[1] 吴重光[1] 王春利[2] 李传坤[2] 

机构地区：[1]北京化工大学信息科学与技术学院,北京100029 [2]中国石油化工股份有限公司青岛安全工程研究院,化学品安全控制国家重点实验室,山东青岛266071

出　　处：《计算机与应用化学》2010年第7期987-991,共5页Computers and Applied Chemistry

摘　　要：由于化工过程的高复杂性及高危险性,且装置都是都是长周期连续运转,一旦出现故障会造成巨大的损失,因此对化工过程和设备进行早期和准确的故障检测与诊断,可以提高设备运行的安全性,避免发生重大安全事故,降低生产成本。人工神经网络具有非线性、大规模、并行处理能力强,以及鲁棒性、容错性、自学习能力强等特点,处理化工过程的复杂非线性问题,比其他方法都优越。本文描述了人工神经网络的基本原理,及近年来人工神经网络在化工过程故障诊断应用中的进展。以BP神经网络为例,分析和介绍了其结构和学习算法,说明了神经网络故障诊断的推理过程,并建议将神经网络与符号有向图(SDG)结合诊断故障。There were some necessaries to take the fault diagnosis in Chemical industry as its characteristics of high hazard and complexity.Earlier fault detection and diagnosis accurately to chemical process and equipments can increase safety of equipment operation,thus serious accidents are avoided and production cost can be decreased.The Artificial Neural Network（ANN） had the unparalleled superiority to deal with the problem of nonlinear chemical process compared with other ways because of its ability of self-adaptive and self-learning.It described the principle of ANN,and a review of recent progress of the application of ANN in chemecal process fault diagnosis had been presented.It taked BP network as an example to explain the inference process of ANN based fault diagnosis.and gave the the proposed algorithm to improve the BP network.At last,it presentd an hybrid fault diagnosis method using both Signed Directed Graph（SDG） and ANN.

关 键 词：故障诊断 神经网络 BP网络 符号有向图 

分 类 号：TQ015.91[化学工程] TP391.9[自动化与计算机技术—计算机应用技术]