Abstract:
This thesis addresses the challenge of robust level control for a nonlinear process subject to significantdisturbances.Areal-timecontrolarchitecturewasimplementedusingaSiemensS7- 1200 PLC and MATLAB, connected via OPC UA. Three distinct controllers—a classical Proportional-Integral-Derivative (PID), a robust H-infinity, and an intelligent Recurrent Neural Network (RNN)—were designed and experimentally evaluated. The comparative analysis of their performance in setpoint tracking and disturbance rejection revealed that while the PID offers precision and the H-infinity controller guarantees robustness, the RNN provides a superior overall solution. The RNN matched the excellent disturbance rejection of the H- infinity controller while utilizing a significantly more efficient and smoother control signal, demonstrating the effectiveness of data-driven strategies for complex industrial processes.
