Evolutionary Computation in System Identification: Review and Recommendations

• Main Author: Md Fahmi, Abd Samad
• Format: Article
• Language: English
• Published: 2014
• Subjects: TJ Mechanical engineering and machinery; TA Engineering (General). Civil engineering (General)
• Online Access: http://eprints.utem.edu.my/id/eprint/12645/1/Md_Fahmi_Abd_Samad%2C_2014.pdf; http://eprints.utem.edu.my/id/eprint/12645/
• Institution: Universiti Teknikal Malaysia Melaka

Two of the steps in system identification are model structure selection and parameter estimation. In model structure selection, several model structures are evaluated and selected. Because the evaluation of all possible model structures during selection and estimation of the parameters requires a lot of time, a rigorous method in which these tasks can be simplified is usually preferred. This paper reviews cumulatively some of the methods that have been tried since the past 40 years. Among the methods, evolutionary computation is known to be the most recent one and hereby being reviewed in more detail, including what advantages the method contains and how it is specifically implemented. At the end of the paper, some recommendations are provided on how evolutionary computation can be utilized in a more effective way. In short, these are by modifying the search strategy and simplifying the procedure based on problem a priori knowledge.