Optimal scheduling for SMT assembly line using neural networks

This thesis presents a work about the development of an SMT shop floor scheduler and a neural network approach in solving the optimization model within the scheduler.; Surface Mount Technology (SMT) is the most significant development in electronics assembly since the advent of the printed circuit board. Associated with the advances in manufacturing technology, production control technology is required for development. This thesis work devotes to the development of a computerized SMT shop floor scheduler which aims to quickly generate a practical schedule based on production-related information. The goal is to minimize the total machine setup time subject to component availability and order due date, thus to improve the machine utilization rate, enhance the customer commitment, and hence improve the shop floor performance.; Operations research technique is employed in modeling the scheduling problem. A nonlinear mixed integer programming model is generated which is NP-hard. A Hopfield-like neural network is constructed to solve this model in a reasonable time. Software simulation of the neural network is carried out and the optimization model is solved successfully.; The usefulness of the shop floor scheduler and the efficiency of the neural network are demonstrated by examples. Compared with conventional manual scheduling process, computerized scheduler improves the machine utilization and reduces the job tardiness considerably. By employing neural network as an alternative to solve the NP-hard optimization problem, the scheduler may run fast enough to be used on-line.