基于DDAGSVM的离心风机振动故障诊断*

摘要

In order to diagnoze the vibration fault occurred in the centrifugal fan running, to improve the economy and security of the fan running, according to the site operating experience and theoretical analysis of fan fault, a multi-classification method based on support vector machine-Decision Directed Acyclic Graph Support Vector Machine (DDAGSVM) method was used, the DDAGSVM topology structure of a 8 class problem was designed, the vibration fault of fan training signs of knowledge base was established, and the vibration fault of fan was diagnozed. This paper assumed that the fan fault rotor misalignment. The diagnostic results of DDAGSVM method showed that the DDAGSVM method can accurately diagnoze the common vibration fault of fan, and had the features of fast speed, high accuracy, and provided a new method for fault diagnosis of fan.%　　为了诊断离心风机在运行中发生的振动故障，提高风机运行的经济性和安全性，根据现场运行经验以及对风机故障的理论分析，采用了基于支持向量机的多分类方法——有向无环决策图（DDAGS-VM）方法，设计了一个8类问题的DDAGSVM 拓扑结构，建立了风机振动故障训练征兆知识库，对风机的振动故障进行了诊断。本文假设风机出现转子不对中故障，采用DDAGSVM方法诊断结果表明：DDAGS-VM方法能够准确的诊断风机常见的振动故障，并且具有诊断速度快、准确率高的特点，为风机故障诊断提供了一种新方法。