Optimizing the Cycle Time of a Multi-Spindle Pick-and-Place Machine

A typical Printed Circuit Board (PCB) assembly line comprises of three major process steps, namely solder paste printing, component placement, and soldering. A stencil printer is typically used to deposit adequate amount of solder paste at appropriate locations on the PCB. One or more component placement (pick-and-place type) machines are then used to populate the PCB. Finally, the entire assembly is passed through an oven for establishing the solder joint. It has been widely accepted that the component placement step is usually the bottleneck. Consequently, the cycle time of the placement machine has to be reduced in order to improve the throughput of the assembly line. Placement machines are expensive and hence their benefit-cost ratio can be improved by improving their cycle time. The objective of this research was to reduce the cycle time of a pick-and-place component placement machine with multiple spindles. The pick-and-place machine chosen for this study was the bottleneck in an assembly line at an electronics manufacturing facility. The primary functions of a placement machine are (1) to pick components from feeder slots, (2) check for any defects using the vision system, and (3) to place the components at appropriate locations. Two important decisions which affect the cycle time are (1) feeder assignment (component location along the feeder rack) and (2) the component placement sequence. A heuristic was proposed to determine the feeder assignments and the component placement sequence was determined by solving a multiple Traveling Salesman Problem (mTSP). The objective of the feeder assignment problem was to minimize the distance traveled by the head along the feeder rack (during the pick cycle). The objective of the component placement sequencing problem was to minimize the distance traveled by the head over the PCB (during the placement cycle). The individual placement tours are later sequenced such that the number of nozzle changes required is minimized. The time taken to populate a board (time to pick + place + nozzle changeover) using the proposed approach was compared to the software which was supplied by the vendor of the machine. Several ‘production boards’ were chosen for this experimental study. The proposed solution approaches outperformed the solutions suggested by the machine’s software for all the ‘production boards’ experimented with. The resulting improvements in cycle times demonstrate the effectiveness of the proposed approach.