This paper proposes the use of three recursive system identification techniques for modeling a magneto-rheological (MR) damper. The results of the three models are compared to one another and to two parametric models that have been commonly used in the existing literature for modeling MR dampers. An MR damper has been built in-house and an experimental set-up has been fabricated as part of this work. The set-up is used for data collection and the data is used for building the recursive models. The results from the system identification models are compared to the measured data as well as to the parametric models. While the parametric models are seen to work well within limited bounds of input variables and operating conditions, these models can be used outside the range of these bounds only after carrying out a new characterization of the model parameters. The recursive system identification models, on the other hand, continuously update all model parameters as and when data becomes available, as demonstrated by the three recursive models presented in this paper. The advantages of the recursive models are conclusively established by lower measures of error, a better representation of hysteresis and saturation phenomena exhibited by the MR damper and significantly improved model tracking. This is a major improvement over the commonly used parametric models, thereby making the recursive models specifically conducive to adaptive control algorithms.
Recursive System Identification Using Outlier-Robust Local Models