MINLP Model for the Optimal Design of Fired Heaters

JUAN I. MANASSALDI

Santa Fe

Jornada; 37 Jornadas Argentinas de Informática; 2008

Sociedad Argentina de Informatica

In this paper, a mathematical optimization model for the optimal design of industrial furnaces/fired heaters is presented. Precisely, a detailed Mixer Integer Non Linear Programming (MINLP) model including operational and geometrical constraints is developed to get an efficient furnace design. Discrete decisions connected with the geometric design such as number of tubes in convection and radiation sections, number of shield tubes, number of passes and number of tubes per pass are modeled by using integer variables. Continuous variables are used for process conditions (temperatures, flow-rates, pressure, velocities, pressure drops among others). Large non-linear optimization problems are characterized by convergence problems which often lead to infeasible solutions. At the same time, the models involve nonlinearities/nonconvexities leading to local optimal solutions. Scaling on variables and equations as well as a systematic initialization procedure has been focused in order to guarantee the model convergence and avoid computational problems. The mathematical model and the solution procedure are implemented in General Algebraic Modeling System GAMS, Brooke [1]. Several applications are successfully solved by applying the proposed MINLP model. In this paper, two case studies are presented. The first one focuses the model validation by comparing the result output with the literature. In the second case the model is solved to minimize the total annual cost of the process. The proposed model and solution procedure are characterized by its robustness and flexibility.