A　robust　hybrid　genetic　algorithm　that　can　be　used　to　solve　process　synthesis　problems　with　mixed　integer　nonlinear　programming　(MINLP)　models　is　developed.　The　proposed　hybrid　approach　constructs　an　efficient　genetic　simulated　annealing　algorithm　for　global　search,　whereas　the　iterative　hill　climbing　method　as　a　local　search　technique　is　incorporated　into　genetic　simulated　annealing　loop　to　speed　up　the　convergence　of　the　algorithm.　To　efficiently　locate　quality　solutions　to　complex　nonconvex　MINLP　problems,　a　self-adaptive　scheme　is　developed　in　the　parallel　adaptive　genetic　simulated　annealing　to　maintain　a　better　tradeoff　between　the　global　and　local　search,　which　can　improve　the　convergence　speed　and　computational　efficiency　compared　with　some　other　hybrid　genetic　algorithm　methods.　For　the　evaluations,　we　use　some　well-known　standard　benchmark　functions.　The　computational　results　demonstrate　the　effectiveness　and　robustness　of　the　present　approach.　Further,　the　proposed　algorithm　is　tailored　to　find　optimum　solution　to　heat　exchanger　network　synthesis　problem.　The　results　presented　prove　that　the　proposed　method　is　a　very　efficient　global　optimization　tool　for　large　dimensional,　nonconvex　MINLP　problems.　©　2012　Curtin　University　of　Technology　and　John　Wiley　&　Sons,　Ltd.