Optimization of revolver head SMT machines using adaptive simulated annealing (ASA)

Among the various SMT assembly machines, revolver head machines are popular due to their small size, relatively high placement rate, wide range of parts they can place, and stationary feeder banks that can be replenished on-the-fly. These machines are capable of placing over 15,000 components per hour, thereby reducing PWB manufacturing cost. However, it is necessary to simultaneously optimize feeder set-up, nozzle set-up and placement sequence for the fastest placement rate to be achieved. This paper addresses development of optimization software for the Fuji NP-132, a dual station, dual revolver head, high-speed placement machine. Following a description of machine operation, a set of equations is derived for evaluation of total assembly time for a given board, based on the pick-place times of individual revolver head cycles. Feeder, nozzle and placement optimization problems are discussed in terms of the machine's degrees of freedom and physical constraints. An adaptive simulated annealing algorithm is proposed. Cheapest insertion and nearest neighbor path construction heuristics are used to generate placement sequences, while constraint satisfaction swapping heuristics are used to generate feeder and nozzle set-ups. This hybrid technique has the advantages of stochastic search for a global optimum and guided search which guarantees the feasibility of a solution, i.e. the feeder set-up fits in the machine's feeder bank and contains all parts required for the PWB. Experimental results are presented for the multi-dimensional combinatorial optimization problem for a set of Fuji NP-132 machine programs.