A Variable Neighbourhood Search for Component Pick-and- Place Sequencing in Printed Circuit Board Assembly

This work presents a heuristic for component pick-and-place sequencing to improve the throughput of a multi-head surface mount device placement machine for assembling printed circuit board. We present a Variable Neighbourhood Monte Carlo Search (VNMS), which employs variable neighbourhood search with an Exponential Monte Carlo acceptance criterion. VNMS is a descent-ascent heuristic that operates on three sets of neighbourhood structures that are based on three different local searchers. The first two sets use a steepest descent and Exponential Monte Carlo local search, respectively whilst the third set uses a random 3-opt operator. The solution returned by a local search, after exploring a neighbourhood structure, will be accepted based on the EMCQ (Exponential Monte Carlo with counter) acceptance criterion. The novelty of the VNMS approach (in the context of VNS) is the concept of three stages of neighbourhood search, using an EMCQ acceptance criterion (at the VNS level) and the shaking procedure (which is only applied when the local searchers cannot find an improved solution). Results show that the VNMS consistently produces good quality solutions.