强跟踪平方根容积卡尔曼滤波和自回归模型融合的故障预测-论文.pdf

《强跟踪平方根容积卡尔曼滤波和自回归模型融合的故障预测-论文.pdf》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、第31卷第8期控制理论与应用v01．31NO．52014年8月ControlTheory＆ApplicationsAug．2014D0I：10．7641，CTA．2014．30963强跟踪平方根容积卡尔曼滤波和自回归模型融合的故障预测杜占龙¨，李小民，郑宗贵，毛琼(1．军械工程学院无人机工程系，河北石家庄050003；2．第二炮兵研究院，北京100085)摘要：为了解决非线性系统中不可测量参数的预测问题，提出一种带有次优渐消因子的强跟踪平方根容积卡尔曼滤波(STSCKF)$1自回归(AR)模型相结合的故

2、障预测方法．利用AR模型时间序列预测法预测未来时刻的测量值，将预测的测量值作为STSCKF的测量变量，从而将预测问题转化为滤波估计问题．STSCKF通过在预测误差方差阵的均方根中引入渐消因子调节滤波过程中的增益矩阵，克服了故障参数变化函数未知情况下普通sCKF跟踪故障参数缓慢甚至失效的局限性，使得sTscKF能较好地预测故障参数的发展趋势．连续搅拌反应釜(CSTR)仿真结果表明，STSCKF的预测精度高于普通SCKF和强跟踪无迹卡尔曼滤波(STUKF)，验证了方法的有效性．关键词：强跟踪滤波；非线性滤波

3、；状态和参数联合估计；平方根容积卡尔曼滤波(SCKF)；故障预测中图分类号：TP273文献标识码：AFaultpredictionwithcombinationofstrongtrackingsquare-rootcubatureKalmanfilterandautoregressivemodelDUZhan—long¨，LIXiao—min，ZHENGZong—gui，MAOQiong(1．DepartmentofUAVEngineering，OrdnanceEngineeringCollege，Sh

4、ijiazhuangHebei050003，China；2．AcademeofSecondArtillerist，Beijing100085，China)Abstract：Todealwiththeproblemofprognosisofunmeasuredparametersinnonlinearsystems，weproposeafaultpredictionalgorithmwhichisacombinationofthestrongtrackingsquare—rootcubatureKalma

5、nfilter(STSCKF)withsuboptimalfadingfactorandtheautoregressive(AR)mode1．FuturetimevaluesofmeasurementvariablesareforecastedbyusingtheARmodeltimeseriespredictionmethod；andthen，thepredictedvaluesareusedasSTSCKFmeasurementvariables．Thus，theprognosticproblemi

6、stransformedintoafilterestimationissue．ThefadingfactorisintroducedintothesquarerootoftheSTSCKFpredictionerrorcovarianceforadjustingthegainmatrixinthefilterprocess．Asaresult，STSCKFeliminatesthedisadvantageofslowtrackingorevenunabletrackingoffaultparameter

7、sinconventionalSCKFwhenthetime—varyingfunctionsoffaultparametersareunknown，improvingthecapabilityforforecastingthevaryingtrendoffaultparameters．Simulationresultsonacontinuousstirredtankreactor(CSTR)showthatthepredictingaccuracyofSTSCKFishigherthanthatoft

8、heconventionalSCKForthestrongtrackingunscentedKalmanfilter(STUKF)，demonstratingthesuperiorityoftheperformancecapabilityoftheproposedmethod．Keywords：strongtrackingfilter；nonlinearfilters；stateandparameterjointestimation；squ