Modified On-Line RLS Identification for Condition Monitoring †


Abstract – The Recursive Least Squares (RLS) is usually utilized in controlapplications as in self-tuning strategy to estimate the plant discrete-time transferfunction. Furthermore, it can be used as a tool to continuously monitoring the operatingcondition of the plant under control. However, in such applications, the RLS should bealways in a “wake up” state so that it can estimate, in a few sampling time, the planttransfer function after any abrupt change in its dynamic.In this work, two modifications to the standard RLS are presented. The firstmodification is called the “switching forgetting factor” while the other is called the”resetting covariance matrix”. The two modifications are applied, under LabVIEWenvironment, on-line to estimate the proper transfer function of a DC motor as anexample to show their capabilities to monitor the motor operation. It is found that withthese modifications, the RLS can estimate the plant transfer function much faster incomparison to the standard RLS algorithm.