神经网络和模糊专家系统在故障诊断中的应用

【作者】 孙增国；

【导师】 宋彤；

【作者基本信息】 大连理工大学 ， 检测技术及自动化装置， 2004， 硕士

【摘要】 随着生产过程的日趋复杂，如何提高大型复杂设备的可靠性和安全性问题已引起人们的极大关注。目前，以神经网络识别法和模糊识别法为代表的智能诊断技术在故障诊断领域得到了广泛的应用。 本文描述了人工神经网络和模糊推理系统的基本原理，分析了基于人工神经网络与模糊推理方法的故障诊断专家系统的设计思想，并介绍了系统结构及知识表示、知识获取和推理机制等方面的基本方法。在此基础上，构建了一种人工神经网络和模糊推理技术相结合的故障诊断方法。人工神经网络通过对部分测量数据的处理，实现系统的回路级故障诊断，输出各回路故障出现的可信度。模糊推理部分通过对神经网络得到的初步诊断结果和其他测量值的处理，实现系统的元件级故障诊断，并对最终诊断结果作出解释。该方法融合了神经网络自适应学习能力强和模糊专家系统知识表达明确的优点，简化了神经网络学习数据获取及模糊推理规则建立的过程。通过对热硝酸冷却系统故障诊断的仿真，证明了该故障诊断方法的有效性。更多还原

【Abstract】 With production process becoming more and more complicated, how to improve the dependability and security of the large-scale equipment has already aroused great concern from people. At present, the representative intelligent methods based on the artificial neural network(ANN) and fuzzy system have been extensively adopted in fault diagnosis.In this paper the design ideas of the fault diagnosis system(FDS) and the basic principles of ANN and fuzzy logic system are described, and the basic methods in systematic structure, knowledge showing, knowledge acquisition and reasoning mechanism are analysed in detail. On the basis, an artificial neural network is integrated with fuzzy system for fault diagnosis. ANN detects the loop faults sources through the partial data measured, and outputs the fault degrees of corresponding loops. Fuzzy system detects the element faults through the preliminary diagnosis results obtained by ANN connected with other correlative values measured, and interprets the final results. The FDS combines the adaptive learning diagnosis procedure of the ANN and the transparent knowledge representation of the fuzzy system, and simplifies the process of obtaining the ANN’s learning sample and establishing the fuzzy inference rules. Through the fault diagnosis simulation of a hot nitric acid cooling system, it has been proven that the fault diagnosis method is very valid.更多还原