The Performance of ARTMAP in Pattern Recognition for a Flexible Vibratory Bowl Feeder System

Neural networks have been shown to exhibit good pattern recognition capability. By incorporating this capability using an ARTMAP neural network within a flexible vibratory bowl feeder system, the feeder can orientate parts of a part family with short changeover time from one part to the next; this is highly sought after in batch production. This paper investigates the influence of a set of parameters that characterised ARTMAP and the sampling rate on the pattern recognition capability of the system. The results show that the optimal performance can be achieved by selecting appropriate values for these parameters.