Intelligent Observer-Based Control of Chemical Process Concentration Using Neural Networks

This paper outlines the development of a control for the Reactor-Separator-Recycle (RSR) process, alongside presenting an Artificial Neural Network (ANN)-based estimator designed to forecast the output concentration in the Continuous Stirred Tank Reactor (CSTR) of the RSR system. The controller construction leverages empirical models derived through identification techniques that analyze the reaction curve. The predicted concentration is incorporated into the CSTR’s control loop. A series of tests are conducted to assess the system response to both temperature fluctuations and changes in initial concentration at the mixing junction. The effectiveness of this approach is evaluated using metrics such as Integral Squared Error (ISE), Integral Time Squared Error (ITSE), settling time (ts), and Total Variation (TV) for each segment of the RSR process.