Genetic-optimized neuro-fuzzy inference system (GONFIS) in nonlinear system identification

The combination of neural network and fuzzy inference system has widely used to imitate more precisely the behavior of nonlinear plants, with less computation effort. However, the derivative-based nature of adaptive networks causes some deficiencies. Therefore, in this paper, a novel approach that employ genetic algorithm, as a derivative-free algorithm, is proposed to enhance the capability of neuro-fuzzy systems. The benchmark Box-Jenkins nonlinear system identification problem, two well-known nonlinear plants modeling problem, and also magnetorheological (MR) damper identification, which is difficult due to the device complex behavior, are employed as the case studies to evaluate the effectiveness of the proposed approach. Results show high accuracy of the proposed approach to predict the plants behavior.