An IT2FNN with a Novel Hierarchical Learning Algorithm for Nonlinear System Modeling

Interval type-2 fuzzy neural networks (IT2FNNs) have been widely used for modeling in industrial processes, and the efficient parameter learning methods are crucial for obtaining accurate models. However, the problem of low computational efficiency and poor convergence performance still exists in existing learning methods. In this study, a novel hierarchical learning algorithm is proposed for constructing IT2FNN with better learning ability and generalization performance. Firstly, the improved recursive least square (RLS) method with an adaptive forgetting factor is designed to learn the consequent parameters. Next, the advanced stochastic gradient descent (SGD) method is adopted to learn the antecedent parameters. Finally, the proposed method is applied to real industrial processes and compared with various state-of-the-art IT2FNNs. Simulation results demonstrate the superiority of the proposed method.