Intelligent Control of Interacting Multivariable Processes: A Machine Learning Approach for Quadruple Tank Systems

Abstract The Quadruple Tank Process (QTP) is a nonlinear multivariable process with complicated dynamics caused by interaction between control loops, transmission zero, and internal system instability. As a result, traditional control approaches do not match the entire functioning of the process particularly well. As a result, a series of experiments are carried out in the real-time quadruple tank laboratory process to investigate the relationship between process error and controller output. Based on these findings, a supervised machine learning controller model is developed using Rational Quadratic Gaussian Process Regression (RQ-GPR) and Medium Gaussian Support Vector Machine Regression (MG-SVR). According to analyses done with MATLAB software, the models have an accuracy of roughly 98%. It is demonstrated that the proposed models are credible to control the liquid levels in the quadruple tank by changing the feed flow rate of the pump