(186k) Optimization of Large Scale Industrial Process by Genetic Algorithms | AIChE

(186k) Optimization of Large Scale Industrial Process by Genetic Algorithms

Authors 

Rezende, M. C. A. F. - Presenter, State University of Campinas (UNICAMP)
Maciel Filho, R. - Presenter, University of Campinas, UNICAMP
Costa, A. C. - Presenter, State University of Campinas (UNICAMP)


The present work aims to employ the Genetic Algorithms to optimize an industrial chemical process, characterized to be difficult to optimize by conventional methods. The chemical process considered is the three phase catalytic slurry reactor in which the reaction of the hydrogenation of o-cresol on Ni/SiO2 catalyst producing 2-metil-cyclohexanol occurs. The kinetic model that represents the reaction is given by the reaction rate whose kinetic constant is a function of the temperature according to the Arrhenius law. In order to describe the dynamic behavior of the process, a non linear mathematical model is used which is composed of material and energy balances of the gas, liquid and solid phases as well as energy balance of the coolant fluid (heterogeneous model). This mathematical model forms a system of partial differential equations which are solved by discretization through orthogonal collocation, leading to a system of ordinary differential equations (ODEs). These ODEs are integrated by DASSL software, which is suitable for stiff systems. Since the process is multivariable, it have eight input variables, to know, linear velocity of gas, linear velocity of liquid, linear velocity of coolant fluid, concentration of hydrogen in the gas phase in the reactor feed, concentration of hydrogen in the liquid phase in the reactor feed, concentration of o-cresol in the liquid phase in the reactor feed, feed reactor temperature, feed coolant temperature, and generates two most important outputs: temperature at the exit of the reactor and o-cresol concentration in the liquid phase at the exit of the reactor. Due the high dimensionality and non linearity of the model, the solution of the optimization problem, through conventional algorithms, does not always lead to convergence. Furthermore, the dependence on initial estimate, a feature typical of the deterministic optimization algorithms, makes the application of such as algorithms in the considered process not suitable. Evolutionary methods, such Genetic Algorithms, have been largely used to successfully optimize several problems of engineering area. In this way, the Genetic Algorithms are chosen to be used in the present work in order to optimize the process. The aim of the optimization though the Genetic Algorithms is the searching of the process input that maximize the productivity of 2-methil-cyclohexanol subject to the environmental constraint of conversion greater than 90%. Another constraint important to be imposed is concerned to the dynamic temperature profile along the reactor, since high level peaks inside the reactor could result in catalyst in catalyst deactivation. In this way, the temperature profile is included as a process constraint. The optimization problem is then postulated considering the process and the environmental constraints. The results obtained with the optimization by Genetic Algorithms shows an improvement of twice the productivity, compared to the steady state at previous works using the same model.