This article improves the original genetic algorithm developed by He and Hui (Chem Eng Sci. 2007; 62:1504-1527) and proposes a novel global search framework (GSF) for the large-size multi-stage process scheduling problems. This work first constructs a comprehensive set of position selection rules according to the impact factors analysis presented by He and Hui (in this publication in 2007), and then selects suitable rules for schedule synthesis. In coping with infeasibility emerging during the search, a penalty function is adopted to force the algorithm to approach the feasible solutions. The large-size problems with tight due dates are challenging to the current solution techniques. Inspired by the gradient used in numerical analysis, we treat the deviation existing among the computational tests of the algorithm as evolutionary gradient. Based on this concept, a GSF is laid out to fully utilize the search ability of the current algorithm. Numerical experiments indicate that the proposed search framework solves such problems with satisfactory solutions. (C) 2009 American Institute of Chemical Engineers AIChE J,56: 2103-2121, 2010