On Solving General and Special Structure Medium-Scale MINLP Problems

Abstract: Mixed integer nonlinear programming (MINLP) problems are encountered in modeling a physical/industrial process consisting both nonlinearity and discrete selective parameters. There are variety of algorithms for solving MINLP problems most of which solve large-scale MINLP problems very slowly. In this research two parallelization scheme are suggested for solving general and special structure large-scale MINLP problems satisfyingly fast. Special structure refers to problems in which a group of variables are appeared in the objective function and constraints and others are appeared in some constraints separately. Moreover, it is proved that both algorithms have at least R-linear convergence. To show the efficiency of algorithms, numerical test results are presented and compared to available GAMS solvers. According to the numerical results, in almost 80 percent of tests both algorithms compute the solution faster than the best solver in GAMS.