Abstract: An effective neural network-based approach to production scheduling is proposed in the paper, which is apt to solving complex job-shop scheduling problems with available time and due date constraints. In this approach, a constrained neural network is proposed to describe various kinds of processing restrictions,and an unreasonable starting time is tuned into a feasible scheduling solution ;and then a gradient search algorithm is applied to the feasible solution. This process is iterated until a satisfactory scheduling solution is obtained. The theoretic analyses, lots of simulation experiments and practical applications have manifested the approach's effectiveness.