基于粒子群聚类分析与证据理论的船舶机械振动诊断

Vibration Diagnosis of Ship Machinery Based on PSO Cluster Algorithm and D-S Theory

粒子群算法避免了复杂的遗传操作,是一种有效的全局寻优算法,用于聚类分析可以更快地收敛于最优解;D-S证据理论提出了不同于贝叶斯主义和频率主义的构造性解释,非常适用于存在大量不确定性因素的故障诊断工作.将故障机械的振动信号按时域、频域、小波包域分解为多个参数空间.采用粒子群聚类分析算法对机械故障进行局部诊断,将局部诊断结果作为独立的证据体,构造相应的基本概率分配函数.结合融合诊断模型,将基于D-S证据理论的决策融合方法应用于船舶机械的故障诊断.对比试验表明,采用粒子群聚类分析与证据理论的方法能有效识别分油机的3种故障模式,验证了其在准确率和灵敏性方面的优势.

As an efficient global optimization algorithm,particle swarm optimization（PSO） avoids complicated genetic operation.Using PSO algorithm,the optimized solution of cluster analysis could be obtained.Different from the Bayesianism and Frequencism,the D-S evidence theory proposes a constructional explanation,and could be used in fault diagnosis which has a great quantity of uncertainty.Decomposing the vibration signal of the broken-down machine by waveform,frequency domain and wavepacket domain,the PSO cluster analysis is used to diagnose the mechanism fault and to achieve local conclusions.Based on the local conclusions of independent evidences,the corresponding basic probability assignment function is constructed.Combining the information fusion diagnosis model,the decision fusion method based on D-S evidence theory is presented to fault diagnosis of ship machinery.The contrast tests shows the method based on PSO cluster analysis and D-S evidence theory could recognize three different fault patterns effectively.Therefore,the accuracy and sensitivity advantage of the proposed method is verified.