Development of vibrational diagnosis system and creation of fault mode in rotating machinery

从理论上对事故进行预测或在故障初期发现事故是防止电厂或化工厂发生重大事故的最有效途径。为了实现上述事项，针对以下对象进行了研究。(1)从旋转机械功能和结构角度建立故障模式预测方法。(2)利用Bay理论建立了精确的估算方法，明确了由于旋转机械测点的不同而引起的振动测量数据的差异。(3)利用扩展系统或神经网络理论开发了自动诊断算法，建立了故障原因数据库，并利用矩阵计数法、模糊理论、Bay理论和神经网络系统建立了扩展诊断系统。即利用扩展系统技术建立专家对诊断过程的分析和诊断过程模型的开发。在实际应用中，编制了主诊断程序，在主诊断程序中加入了矩阵计数法、模糊理论、Bay理论和神经网络理论等不同类型的诊断算法，以便相互比较，从工程角度阐明了故障模式的产生及其与振动现象的关系。实现了Bay诊断技术和神经网络技术对故障的精确诊断。

It is a most effective way to protect the power plants or chemical plants from the big accidents that we predict the accident theoretically or find the accident at the initial stage of the faults. In order to achieve the above matters, the investigations were performed for the following objects. (1) Establishment of predicting method of fault modes from the viewpoints of function and stracture of rotating machinery. (2) Establishment of precise estimating technique by using the Bay's theory, in which the discrepancy of measurement vibrational data by the difference of measuring points of the rotating machinery are made clear. (3) Development of automatic diagnostic algorithm by using the expart system or neural network theory.As the results, the data base for the cause of the fault are established and the expart diagnostic systems by using the matrix counting method, fuzzy theory, Bay's theory and the neural network system are built up. That is, the analysis of diagnostic process by the specialist and the development of the modeling of the diagnostic process are built up by using the expart system technique. In practical case, main diagnostic computer program are finished, in which subsystem for different types of diagnostic algorithm, these are, the matrix counting method, fuzzy theory, Bay's theory and neural network theory can be added in order to compare with each other.In the diagnostic method, creation of fault modes and its relation to the vibration phenomena are clarified from the engineering point of view. Moreover realizations of Bay's diagnostic technique and the neural network technique to the precise diagnosis are accomplished.