基于融合相似度的制粉系统健康预警及故障诊断

Health Warning and Fault Diagnosisof Pulverizing System Based on Syncretic Similarity

反映火电厂制粉系统健康状况的参数众多,同时监视这些参数难度较大且故障诊断过程过于复杂。为此提出了一种适合工业系统的基于融合相似度的健康预警实现方法和故障诊断算法。此融合相似度由一种新型的基于主成分分析的加权马氏距离和加权正弦相似度融合而成。基于该方法形成的诊断系统无需大规模样本库或记忆矩阵并具有自学习能力,算法运行中可自行修正计算融合相似度所需要的中心参数。仿真结果表明,该方法结构简单可靠,诊断正确率高,实时性强,适合在线运用。

It is an arduous task to know health statusof a pulverizing system in power plant by monitoring these parameters simultaneously and the fault diagnosing process is complicated_4n approach based on syncretic similarity is presented which is suitable for industrial system＇s health warning and fault diagnosis.The syncretic similarity couples anew type of weighted mahalanobis distance based on principal component analysiswith an improved weighted sine similarity. The approachhas self-learning ability.Central parameterswhich are used to compute similarity can be modified along with the operation process.Simulation resultsshow that the method is suitable for online application because of its high accuracy, fast classification, high real-time performance, reliabilityand simple structure.