Application of Taguchi design in system identification: A simple, generally applicable and powerful method. (February 2020)

• Record Type: Journal Article
• Authors: Norouzi Ghazbi, Somayeh; Akbarzadeh, Alireza; Akbarzadeh-T, Mohammad-R
• Is Part Of: Measurement. Volume 151(2020)
• Publication Date: 2020-02
• Subjects: System identification -- Taguchi -- Multi-Input Multi-Output system -- MIMO -- Nonlinear systems -- Non-parametric system; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2019.106879
• Languages: English
• ISSNs: 0263-2241

Abstract

Graphical abstract: Highlights: A simple and generally applicable system identification algorithm based on Taguchi method. Enables identifying a variety of systems including linear and nonlinear systems, mechanical or economic systems, etc. Effectively increase the model accuracy while significantly reduce the time required for the identification. Experimental results proved the efficacy of the method proposed. Abstract: This paper is a development of the authors' recently published idea on a simple and generally applicable system identification algorithm based on Taguchi method. In the previous paper, the effectiveness of the algorithm in identifying nonlinear systems was shown. This paper presents the modified algorithm with the capability of identifying Multi-Input Multi-Output (MIMO) systems, and systems with unknown parameters of quite different value-ranges. Regarding the former, a contribution analysis is introduced and utilized to determine that the model between which of the input-output pairs should be identified first. The results proved that this idea significantly results in less time-cost. Moreover, the study develops the identification algorithm by adding a step of breaking down the identification process to multiple steps, each identifying a group of parameters that are in about the same value-range. Utilizing Taguchi method that includes human intelligence in the loop, the presented method effectively saved the process time and led to a more accurate estimation. In the method presented, the algorithm identifies a group of parameters at each step starting with the most contributor ones rather than identifying all of them at once. To group parameters, a contribution analysis is employed. To prove the effectiveness of the method, without loss of generality, the algorithm is applied on a multi-input single-output system. The experimental test-setup used in this study is a custom-made linear series elastic actuator that is designed for lower limb exoskeletons in Ferdowsi University of Mashhad (FUM)-Robotic Laboratory.