基于HPSO-RBF神经网络的瓦斯传感器故障诊断  

Gas Sensor Fault Diagnosis Based on HPSO-RBF Neural Network

作　　者：王婷[1] 李国勇[1] 吕世轩 

机构地区：[1]太原理工大学信息工程学院,山西太原030024 [2]太原理工大学,煤矿装备与安全控制山西省重点实验室,山西太原030024

出　　处：《仪表技术与传感器》2015年第3期78-81,共4页Instrument Technique and Sensor

基　　金：山西省自然科学基金资助项目(2011011011-1)

摘　　要：针对瓦斯传感器常见故障,提出一种混合粒子群优化(HPSO)算法与径向基函数(RBF)神经网络相结合的传感器故障诊断新方法。文中首先采用HPSO算法对RBF神经网络模型参数进行了优化,在详细分析瓦斯浓度影响因素的基础上,建立了瓦斯浓度非线性预测模型。然后将瓦斯浓度预测结果与实际测量值相比较得到残差,并分析残差的变化趋势,从而实现对瓦斯传感器的故障诊断。实验结果表明,HPSO-RBF模型具有较高的预测精度,能够有效地诊断瓦斯传感器的故障状态。Aiming at the common faults of gas sensor,this paper proposed a novel fault diagnosis method for gas sensor,which combined Hybrid Particle Swarm Optimization（ HPSO） algorithm and Radial Basis Function（ RBF） neural network. Firstly,the model parameters of RBF neural network were optimized by the HPSO algorithm. Through analyzing the influencing factors of gas concentration in detail,a non- linear prediction model for gas concentration was established. Then,the residuals between predicted results and actual measured values were calculated,and residual＇s changing trend was analyzed,so as to realize fault diagnosis for the gas sensor. Experimental result indicates that the HPSO-RBF model has higher prediction accuracy and this method can diagnose faults of gas sensors effectively.

关 键 词：瓦斯传感器 混合粒子群优化算法 RBF神经网络 故障诊断 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]