Using diversity for optimizing cycle time for a pick-and-place machine

Assigning feeders and nozzles is an important step in the setup of pick-and-place machines. We study this feeder and nozzle assignment problem (FNAP) for a representative machine from Kulicke & Soffa (K&S). The goal is to assign nozzles and feeders so that a given set of printed circuit boards can be produced while minimizing cycle time. The heuristic currently used by K&S reliably finds feasible assignments but performs poorly regarding nozzle exchanges, which are strongly correlated with cycle time. To address this, we propose a novel algorithm that tries to find k feasible assignments that are as diverse as possible, for some parameter $$k>1$$ . By maximizing solution diversity, our approach increases the likelihood of finding initial solutions that can yield shorter cycle times after further optimization. On real-world instances, our algorithm achieves at least half the theoretical maximum diversity in over 70% of cases for k ≤ 10 . It also produces assignments with estimated nozzle exchanges roughly half those of the current heuristic in 95% of cases, while reducing the number of instances incorrectly classified as infeasible at only a small cost in runtime.