Design of Adaptive Fuzzy Modeling Based on Evolutionary Optimization Algorithms: A Comparative Study

Dynamic adaptive fuzzy modeling (DAFM) is a type of artificial intelligence (AI) that combines the principles of fuzzy logic and adaptive control. DAFM is used for various applications including intelligent control, signal processing, and decision-making. The main advantage of DAFM is its ability to handle systems’ complexity and nonlinearity with uncertainties. The parameters’ adaptation of DAFM represents a crucial task in the context of nonlinear fuzzy model element identification. In addition, its number increases depending on the number of membership functions used. So, in this paper, the DAFM membership functions’ adaptation is introduced based on different tuning approaches, which are gradient descent (GD), and evolutionary algorithms (EAs). The evolutionary optimization algorithms contain Genetic algorithm (GA), differential evolution (DE), and particle swarm optimization (PSO). A comparative study between the mentioned approaches has been applied to estimate the models of two systems, a simulated system and a practical model of a liquid-level rig. The simulation results provide a good identification process. Furthermore, the PSO technique provides results in perfect tracking, high accuracy, fast convergence, and low computational time. For instance, after a change occurred in the input, the PSO-based tracking converged twice as fast as DE and four times faster than GA.