基于小波奇异熵和相关向量机的氢气传感器故障诊断

Fault diagnosis of hydrogen sensor based on wavelet singular entropy and relevance vector machine

针对氢气传感器故障问题,提出了一种智能化的传感器故障诊断方法,可以对自身故障状态进行诊断和识别。提出了一种基于小波奇异熵(wavelet singular entropy,WSE)和相关向量机(relevance vector machine,RVM)原理的氢气传感器故障诊断方法,将小波变换和奇异熵两种分析思想相结合,提取信号的完备故障特征;利用小生境粒子群优化算法(niche particle swarm optimization,NPSO)对相关向量机的核参数进行优化,提高故障诊断的准确率。将提出的方法与其他成熟算法进行了比较,实验结果表明所提方法故障诊断识别率达到98%以上,解决了非线性、小样本条件下的传感器故障诊断问题,提高了传感器的可靠性。

Aiming at the fault problem of hydrogen sensor, an intelligent fault diagnosis method which can diagnose and distinguish the fault state of the sensor was proposed. The fault diagnosis method based on wavelet singular entropy and relevance vector machine was researched, the feature of fault signal was extracted completely by combining the theory of the wavelet transform and singular entropy. The niche particle swarm optimization algorithm was used to optimize kernel parameter of RVM, and the accuracy of the fault diagnosis was improved. The proposed method was compared with other mature algorithms. Results indicates that the fault diagnosis recognizable rate reaches 98%. It resolves the problem of sensor fault diagnosis under the condition of nonlinear and small sample, and promote the reliability of sensor.