化工过程的故障聚类及诊断空间的分层递阶算法

Fault Cluster of Chemical Process and Hierarchical Structure Algorithm for Diagnosis Space

化工过程具有数据量大、数据采集传感器类型多、数据冗余严重等特点.本文对化工过程的测试数据进行聚类分析以获得故障模式的样本特征,由样本特征形成信息表并利用属性约简的方法对信息表进行精简以获得故障诊断规则集.在诊断推理算法中,通过由元件连接模型中发出的测试查询信号来观测各接收点的输出,对诊断空间进行分层递阶处理,从而缩小了故障诊断的范围,减少了诊断推理的工作量,提高了诊断效率.

Chemical process is of the features of massive data, multi-kinds of data sensors and serious data redundancy. Through the test data clustering for chemical process, sample characteristics of the fault patterns are obtained. The sample characteristics make an information form which will be produced a refined set of fault diagnosis rules. By sending signals and inquiring the outputs from receiver points, diagnosis space is caved into a hierarchical structure processing which may narrow the diagnosis range and reduce the load of inference.