Control of a pH Neutralization Process using Neural Network Approaches

This work presents an investigation into advanced control strategies for a nonlinear pH neutralization process. The study compares three control approaches: a traditional PID controller, a PID controller combined with a neural network (PID + NN), and a neural network-based PID controller with adaptive adjustment. A first-order plus dead-time (FOPDT) model is employed for parameter tuning, and simulations are conducted using Matlab to evaluate the performance of each method. Metrics such as overshoot, settling time, and Integral Squared Error (ISE) are analyzed. The results demonstrate that both the PID + NN and the adaptive neural network PID controller outperform the classic PID controller in terms of overshoot, settling time, and disturbance handling. In addition, details the process of training neural networks for parameter adjustment and response to changes, utilizing NARX networks and time-series training.