Abstract: This paper first formulates the hybrid flow shop scheduling (HFSS) problem using an integer programming model and then develops a genetic descent algorithm (GDA) for it. The proposed GDA is constructed by combining the optimal job-ma-chine allocation rules with the appropriate genetic coding. This method can not only generate feasible initial solutions, but also guarantee the feasibility of solutions after genetic operations. In the meantime, neighborhood search is imbedded in the iteration process of the algorithm. In order to testify the effectiveness of GDA, simulation is done based on randomly generated 230 instances. Computational experiments show that. 1) for small size HFSS scheduling problems, the average deviation of GDA from the optimal solution is 0.01%; 2) for medium-large size problems, the performance of GDA is 10.45% better than that of NEH algorithm.