基于ASGSO-SVR模型的瓦斯传感器故障诊断  

Gas sensor fault diagnosis based on ASGSO-SVR

作　　者：黄凯峰[1,2] 刘泽功[1] 王其军[2] 杨静[3] 高魁[1] 

机构地区：[1]安徽理工大学能源与安全学院,安徽淮南232001 [2]淮南职业技术学院信电系,安徽淮南232001 [3]安徽理工大学理学院,安徽淮南232001

出　　处：《煤炭学报》2013年第A02期518-523,共6页Journal of China Coal Society

基　　金：国家自然科学基金资助项目(51004003;50974004);安徽省优秀人才基金资助项目(2011SQRL198)

摘　　要：针对现行煤矿瓦斯传感器常见的卡死、冲击、漂移等故障,运用支持向量回归机建立多传感器数据融合的瓦斯浓度预测模型,详细研究影响该预测模型精度的相关参数选择方法,提出用ASGSO算法自适应优化支持向量机预测模型参数的算法,将模型预测结果与现场实测瓦斯浓度相比较得到残差δ,用于对瓦斯传感器故障的诊断。用现场监控数据对该方法进行离线仿真实验,得到残差信号的变化曲线。通过选择合理的阈值,判断传感器是否处于故障状态。结果表明,ASGSO算法参数优化对提高SVR预测模型的精度有很大帮助,此方法对瓦斯传感器的常见故障的诊断是正确和有效的。Abstract：For the common faults in the current coal mine gas sensor such as jamming, impact or drift, the gas concentration prediction model of multi-sensor data fusion was used, which was established by the support vector regression machine. Meanwhile, the related parameter selection method which influences the prediction model accuracy was worked up in detail and then the arithmetic was proposed to adaptively optimize the forecasting model parameters of the support vector machine through the Self-Adaptive Step Glowworm Swarm Optimization algorithm compared between the results of model prediction and the field measured gas concentration, the residual 8 for gas sensor fault diagnosis was got. Upon the field monitoring data got through this method, the simulation experiment in Matlab was done to get residual signal change curve. Fault diagnosis was implemented by fault threshold selection. The results indicate that the parameter optimization by the AGSO algorithm is helpful to improve the support vector machine regression prediction model precision and it is correct and effective for this method to the common gas sensor fault diagnosis. Key words： gas sensor ; diagnosis ; ASGSO algorithm ; support vector regression machine

关 键 词：瓦斯传感器 故障诊断 ASGSO算法 支持向量回归 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置] TD67[自动化与计算机技术—控制科学与工程]