Comparing multi-control algorithms for complex nonlinear system: An embedded programmable logic control applications

Abstract

This paper examines the impact of multiple control algorithms, such as genetic algorithm (GA), artificial neural network (ANN), and proportional integral derivative (PID), on programmable logic controller (PLC) performance during a nonlinear thermodynamic process. The ANN was trained with data that modeled the thermodynamic process and then generated the control algorithm. GA was improved by applying the counter-premature algorithm (CPA) to address issues of pre-mature convergence, while the PID presents the current algorithm used to optimize the PLC in the existing testbed. Experimental evaluation of these models against the process set-points showed that all the algorithms were able to reject disturbance and follow the reference set points under different step changes, but each algorithm experienced different internal behaviors while trying to reject disturbance. Lastly, the result showed that while the improved GA was better than the PID, with a recorded slight overshoot due to the uncertainties of the thermodynamic process, the ANN achieved better control performance in terms of system stability than the other counterpart algorithms.