System identification of a linear series elastic actuator using a recursive taguchi-based algorithm

This paper addressed a new system identification method which is simply applicable to identify both of the linear and nonlinear models. Having a rough and initial range of values for the parameters of the system, the Taguchi method is used to identify the model of the purposed system. The purposed method is implemented on a custom made linear series elastic actuator. Five parameters of the system are identified through 75 iterations and within three steps of system identification process. To use this model, first a rough estimation for the values of the parameters are estimated, using the Taguchi method and Signal to Noise Ratios (SNRs), the previous attempts are modified and a new range of the values for each parameter is offered. The changes of signal to noise ratio diagrams are used as the stop criteria for the process. Also, SNRs are used to determine the next level of values for the parameters. The presented method significantly reduced the system identification time process, and the model is identified using just 75 iterations. In addition, utilizing the capability of the Taguchi method, the purposed method robustly estimates the system model. In this paper, based on the physical laws, the structure of the model was known and only the parameter estimation was required; however, using the presented method, estimation of both of the structure and the model is possible. The method is simply applicable and the computational complexity is very low in comparison to the alternative methods.