ISA Trans. 2026 Mar 10:S0019-0578(26)00112-6. doi: 10.1016/j.isatra.2026.03.003. Online ahead of print.
ABSTRACT
This paper adopts a supervisory radial basis function neural network (SRBF-NN) that integrates a fractional-order proportional-integral (FOPI) controller within a hybrid feedforward-feedback system. To address saturation issues, i.e., in the Hammerstein system, the FOPI controller incorporates an anti-windup strategy (FOAWPI). This SRBF-FOAWPI controller effectively tracks desired inputs for both linear and nonlinear plants in MATLAB. The Lyapunov stability study provides the range of learning rate that influence the hidden layers. Additionally, it presents a statistical estimator with a fractional parameter, FraSU shrinkage, within the framework of discrete wavelet and packet filters. This approach effectively removes noise from the output of both simulated and real-time plants. Finally, it shows superiority in both transient and denoised performance over the VISU method by controlling its fractional wings.
PMID:41856878 | DOI:10.1016/j.isatra.2026.03.003
