On optimizing assembly plans for a robotized printed circuit board assembly center

The purpose of this research was threefold. First, a set of parameters was defined for use in analyzing the assembly planning problem in robotized printed circuit board (PCB) assembly centers. Second, a number of mathematical models were developed to optimize assembly plans for a robotized PCB assembly center. Third, an integrated heuristic method was developed to obtain near-optimal assembly plans.

The robotized PCB assembly center of interest consists of a moving X-Y table, a moving feeder carrier and a traveling pick-and-place robot. The assembly planning problem is one of determining the assembly sequence and assignment of feeders to optimize two prioritized objectives: minimizing assembly cycle time and minimizing total X-Y table travel time.

Three groups of parameters were defined through analysis of the problem. These are: (1) non-timing parameters, (2) timing parameters and (3) priority parameter. Based on the parameters, 10 different assembly environment classes were identified.

A number of mathematical models were developed. Among them, four models were of traveling salesman problem nature but with modification on the associated cost matrices. The remaining models were more complicated and were non-linear. These models were linearized to obtain 3 modules.

An integrated heuristic method of five stages was developed to quickly reach a near-optimal assembly plan. A set of experiments was performed to evaluate the performance of this integrated heuristic method.

When the number of feeders is no less than the number of components total delay can always be zero while total X-Y table travel is less than 1% above the best known solutions. When the number of components is greater than the number of feeders and total X-Y table travel time receives higher priority, the resultant assembly plan can have both performance measurers less than 1% above the best known solutions. For the remaining classes, the resultant assembly plans always have zero delay while total X-Y table travel time is no more than 3% above the best known solutions if Stage III and Stage V are used.