基于SVM预测器的传感器故障诊断与信号恢复研究.pdfVIP

第 18 卷 　第 2 期 传 感 技 术 学 报 Vol . 18 　N o . 2 2005 年 6 月 CHIN ES E J OU RNAL O F S EN SORS AND AC TUA TORS J u n . 2005 Research on Sensor Fault Detection and Signal Recovery Based on SVM 1 , 2 , 3 1 L I U D ong GE Yunj i an 1. I ns t it ute of I ntel l i g ent M achi nes , Chi nese A ca demy of S ciences , H ef ei 23003 1 , Chi na; 2 . D ep t. of A ut om at ion , Uni vers ity of S cience an d Technol ogy of Chi na , H ef ei 230026 , Chi na; 3. D ep t. of M at hem at ics an d p hy s ics , A nh ui I ns t it ute of A rchtect ure an d I n d us t ry , H ef ei 230022 , Chi na Abstract :Support vector machine ( SV M) i s a new machine lear ning met ho d ba sed on st ati stic lear nin g t he ory , it ha s excellent p erfor mance co mp ared wit h ot her nonlinear regre ssion , such a s neural net wor k . We int ro duced t he f undament al t heory of SV M regre ssion , p ropo sed a p redictor ba sedon SV M regre ssion for sen sor f ault det ection and signal recovery , and p re sent ed t he p rincip le of t he p redictor an d it s online algo rit hm . The simulatio n re sult s show t hat t hi s met ho d can det ect sen sor f ault an d recover sen sor signal suc ce ssf ully . Key words :support vector machine ; sen sor ; f ault det ection ; signal recovery EEACC :7230 基于 SV M 预测器的传感器故障诊断与信号恢复研究 刘 　东1 ,2 ,3 , 葛运建1 1. 中国科学院合肥智能机械研究所 , 合肥 23003 1 ; 2 . 中国科学技术大学自动化系 ,合肥 230026 ; 3 . 安微建筑工业学院数理系 , 合肥 230022