Discrete-time system identification using genetic algorithm with single parent-based mating technique

Zainuddin, Farah Ayiesya (2024) Discrete-time system identification using genetic algorithm with single parent-based mating technique. Doctoral thesis, Universiti Teknikal Malaysia Melaka.

	Text (24 Pages) Discrete-Time System Identification Using Genetic Algorithm With Single Parent-Based Mating Technique.pdf - Submitted Version Download (544kB)
	Text (Full Text) Discrete-Time System Identification Using Genetic Algorithm With Single Parent-Based Mating Technique.pdf - Submitted Version Restricted to Registered users only Download (4MB)

Abstract

System identifiction (SI) is a methodology for developing mathematical models of dynamic systems using measurements of input and output signals. This research focuses on improving Genetic Algorithms (GA) for SI, specifically addressing inefficiencies in common crossover operators that limit search space exploration and lead to premature convergence. The study aims to enhance the performance of GA by introducing a novel Single Parent Mating (SPM) technique. The objectives are to enhance the performance of crossover operator for GA, simulate SI using GA with the SPM technique, and compare the performance of the modified GA to traditional GA in terms of prediction accuracy, convergence to global optimum, and model parsimony. The methodology encompasses data acquisition, GA program development, SPM technique implementation, and simulation using MATLAB. The study simulated single-input-single-output (SISO) models: ARX and NARX. Performance indicators included the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and Parameter Magnitude-based Information Criterion 2 (PMIC2). The findings showed that incorporating SPM with single-point crossover resulted in lower objective function (OF) values and improved error indices (EI) compared to traditional GA methods. This enhanced the GA's ability to avoid premature convergence and maintained a diverse solution set, leading to more optimal model selection. The study's results were validated using real-world data from industrial systems, including a hair dryer, an air compression system, and a flexible robot arm. In these cases, the SPM technique consistently outperformed traditional GA, demonstrating improved model fit and predictive accuracy. Rigorous validation tests, including autocorrelation and cross-correlation functions, confirmed the reliability and robustness of these models. These practical applications underscore the versatility and effectiveness of the SPM technique in enhancing GA for SI, proving its utility across different fields such as engineering, finance, and healthcare. The successful validation in real-world systems marks a significant milestone, showing that the SPM technique can significantly improve model optimization in diverse and practical contexts . This research makes several important contributions to the field of GA and SI. The introduction of the SPM technique represents a significant enhancement in the performance of GA, particularly in terms of improving genetic diversity and search optimization . The findings provide a robust framework for applying GAs to complex modeling problems, emphasizing the importance of effective crossover strategies and genetic diversity . The SPM technique offers researchers and practitioners a powerful tool for achieving faster convergence, better optimization, and more accurate models. This study not only advances the theoretical understanding of GA but also provides practical methodologies that can be applied to real-world problems, making it a valuable contribution to both academic research and practical applications.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Discrete-time systems, MATLAB, System identification
Divisions:	Library > Tesis > FTKM
Depositing User:	Muhamad Hafeez Zainudin
Date Deposited:	21 Jan 2026 08:00
Last Modified:	21 Jan 2026 08:00
URI:	http://eprints.utem.edu.my/id/eprint/29196
Statistic Details:	View Download Statistic

Actions (login required)

View Item