Solving Feeder Assignment and Component Sequencing Problems for Printed Circuit Board Assembly Using Particle Swarm Optimization

Printed circuit board assembly (PCBA) is a process of connecting various electronic components through printed circuit boards (PCBs). Due to the need to assemble a lot of components and PCBs at the same time, the PCBA process tends to become the bottleneck in an assembly line. Many assembly firms have thus introduced automated PCBA machines to expedite this process. However, to best operate these machines, effective PCBA planning is still required. Some nature-inspired metaheuristics such as simulated annealing and genetic algorithm (GA) have been increasingly used for the PCBA planning. Also, we find that particle swarm optimization (PSO) has never been employed to deal with the feeder assignment problem (FAP) and component sequencing problem (CSP) at the same time, though it has been regarded as a good competitor to GAs. In this paper, we developed two PSO-based approaches to deal with the two problems simultaneously for a chip shooter machine. In addition, we have conducted experiments to compare the two PSO-based approaches with two GA-based approaches. The experimental results showed that PSO2, the PSO-based approach with sigmoid functions, outperformed others in terms of assembly cycle time. The comparison with an exact approach further shows that PSO2 has a high rate to find the optimal/near-optimal solution. Note to Practitioners-This paper was motivated by the problem of assembling electronic components through PCBs using automated surface mount machines. Besides the fact that a capable automated machine can directly improve the assembly productivity, we believe that good planning for PCBA is also important as a minor improvement on the assembly time for one PCB can lead to a significant cost reduction for a large batch. For PCBA, CSP and FAP are two essential problems and they can directly affect the assembly time of a PCB. To deal with the two problems, current practice mainly depends on engineers' experience or a simple heuristic rule instead of advanced approaches. Though some studies have proposed GAs to solve the two problems together, PSO is found with the potential to better deal with the two problems. This has prompted us to develop PSO-based approaches to deal with the two problems simultaneously. Our preliminary experiments showed that the PSO-based approaches have a better performance than the two genetic-based approaches. PSO2 is found being a good approach after comparing it with an exact approach.