Implementação da metodologia de self-optimizing control usando metamodelos kriging em linguagem python.

Abstract

One of the primary objectives of engineering is to operate processes optimally or near the optimum point, aiming to maximize or minimize an objective function that may take into account operational, environmental, and/or economic constraints. Ensuring optimal operation is only possible through a well-designed control framework to maintain the process within specifications. In this work, we propose a systematic procedure programmed in the Python language, utilizing the Jupyter Notebook development environment, for metamodeling, optimization, and application of the self-optimizing control (SOC) methodology. This procedure was developed with the goal of optimizing the implementation sequence of the selfoptimizing control methodology for simulated processes using Aspen Hysys. For analysis and validation of its functionality, the procedure is applied to two case studies. In the first case, a Python-developed model for a reversible and exothermic reaction in a CSTR reactor is analyzed to determine the controlled variable resulting in the least loss. The obtained results are compared with literature results, and all outcomes are numerically validated. In the second case study, the procedure is applied to analyze a high-purity distillation column modeled in the Aspen Hysys process simulator to determine the best control pairs. The obtained results exhibit coherence in terms of the process, thus validating the applicability of the procedure. The self-optimizing control methodology presents a promising approach for optimizing the operation of industrial processes. By employing metamodels like Kriging, it is possible to reduce the computational effort needed for optimization. In one of the most crucial steps of the SOC methodology, mathematical modifications were carried out to obtain metamodels. In this work, we demonstrate the computational and time gains achieved through the utilization of second-order regressors for Hessian estimation. Lastly, the study provides valuable contributions to the field of process control, furnishing an automated tool for applying the SOC methodology. The achieved results demonstrate the tool's effectiveness in determining the controlled variable in a CSTR reactor and identifying the optimal control pairs for a high-purity distillation column. Moreover, the numerical validation of results reinforces the reliability and utility of the proposed tool.