A novel optimization algorithm for MIMO Hammerstein model identification under heavy-tailed noise. (January 2018)

• Record Type: Journal Article
• Authors: Jin, Qibing; Wang, Hehe; Su, Qixin; Jiang, Beiyan; Liu, Qie
• Is Part Of: ISA transactions. Volume 72(2018)
• Page Start: 77
• Page End: 91
• Publication Date: 2018-01
• Subjects: Cuckoo Search -- GMDA -- MIMO Hammerstein model -- Heavy-tailed noise -- Radial Basis Function neural network; Engineering instruments -- Periodicals; Engineering instruments; Periodicals; Electronic journals; 629.805
• DOI: 10.1016/j.isatra.2017.10.001
• Languages: English
• ISSNs: 0019-0578

Abstract

Abstract: In this paper, we study the system identification of multi-input multi-output (MIMO) Hammerstein processes under the typical heavy-tailed noise. To the best of our knowledge, there is no general analytical method to solve this identification problem. Motivated by this, we propose a general identification method to solve this problem based on a Gaussian-Mixture Distribution intelligent optimization algorithm (GMDA). The nonlinear part of Hammerstein process is modeled by a Radial Basis Function (RBF) neural network, and the identification problem is converted to an optimization problem. To overcome the drawbacks of analytical identification method in the presence of heavy-tailed noise, a meta-heuristic optimization algorithm, Cuckoo search (CS) algorithm is used. To improve its performance for this identification problem, the Gaussian-mixture Distribution (GMD) and the GMD sequences are introduced to improve the performance of the standard CS algorithm. Numerical simulations for different MIMO Hammerstein models are carried out, and the simulation results verify the effectiveness of the proposed GMDA. Highlights: A modified Cuckoo Search algorithm by introducing the Gaussian-mixture Distribution(GMD) and the GMD sequences called GMDA is firstly proposed in this paper. A novel method combined the advantages of GMDA and RBF neural network is applied to the MIMO Hammerstein model identification in the presence of heavy-tailed noise. Several simulation examples verify the effectiveness of the proposed method.