基于数据挖掘的核动力装置故障数据处理及属性约简算法研究

Research on Faults Data Processing and Attributes Reduction Arithmetic Based on Data Mining for Nuclear Power Plants

根据核动力装置故障数据的特点,利用数据挖掘法强大的知识发现功能,提出了一种新的数据标准化方法——距离标准化法,对核动力装置不同工况、不同单位数量级的故障数据进行标准化处理。根据参量的特点,利用各报警值点进行数据离散化,为数据离散化断点数量的选择提供了参考。利用概念格的属性约简方法,进行了属性的约简处理,得出用于故障诊断的核心属性、相对必要属性和不必要属性。利用文献中的数据,进行了属性约简计算,将现有的属性正确分类,在形式背景确定的情况下,利用核心属性即可准确诊断故障。

A distance standardized method is proposed considering the features of fault data of nuclear power plants and based on the knowledge discovery function of the data mining method,to standardize the data under different condition and of different order magnitudes.According to the characteristics of parameters,the parameters are dispersed using the alarm values,as a reference for the choice of the break points of the discrete data.The data are reduced using the concept lattice attribute reduction method,and thus the core attributes,relative necessary attributes and unnecessary attributes for fault diagnosis are obtained.The data in literature are calculated and the attributes are classified accurately.When the formal context is affirmatory,the fault can be diagnosed exactly using the core attributes.