人工神经网络在机械设备故障诊断中的应用  

Application of Artificial Neural Networks(ANNs) to failure diagnosis on the mechanical instruments

作　　者：王敬章 

机构地区：[1]中国石油阿姆河天然气公司

出　　处：《天然气工业》2009年第6期120-122,共3页Natural Gas Industry

摘　　要：油气田关键设备(各种机泵和压缩机)的运行状况直接关系到油气田开发的安全、平稳和高效运行。这些设备几乎是24h不间断运行,难免会出现故障甚至给生产造成损失。当前有多种神经网络被应用于这些设备的状态监测与故障诊断。对BP神经网络、径向基函数网络、概率神经网络、学习矢量量化网络、模糊神经网络和小波神经网络在机械设备故障诊断中的应用与研究进展进行了分析比较,阐述了各种网络的应用效果,分析了各种网络应用的优缺点。人工神经网络以其具有非线性、大规模、并行处理能力强、鲁棒性、容错性及自学习能力强等特点,在机械设备故障诊断中得到广泛的应用,应选择合适的神经网络对机械设备进行故障诊断,为油气田的安全、平稳和高效开发提供保障。The performance of essential equipments or instruments including various pumps and compressors will directly relate to the HSE management and working efficiency for the development of oil and gas fields. The long-run incessant operation of these equipments in fields will be not easy to survive from failure or accidents resulting in a big loss. At present,many kinds of artificial neural networks have been put into use of condition monitoring and failure diagnosis on these equipments. This paper introduced advantages and disadvantages of many different kinds of artificial neural networks including BP(Back Propagation),RBF-NN(Radial Basis Function Neural Network),PNN(Probabilistic Neural Network),LVQ (Learning Vector Quantization),FNN (Fuzzy Neural Network),and WNN (Wavelet Neural Network),focusing on the progress of research and application of those ANNs to failure diagnosis on mechanical instruments. The ANNs have been widely used in the failure diagnosis on the mechanical instruments because of their excellent features:non-linear,large-scale,parallel processing capability,robustness,fault-tolerance,and self-learning ability,etc. To select a proper kind of ANNs for failure diagnosis on the equipments or instruments will help ensure oil and gas fields to be developed under safe,stable and high efficient way.

关 键 词：人工神经网络 故障诊断 机械设备 应用 比较 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TH17[自动化与计算机技术—控制科学与工程]