A programmable bowl feeder is developed which needs not require any mechanical tooling for identifying the orientation of each specific part being fed. The feeder is designed in such a way that an object can be discriminated and sorted according to every stable orientation. The operation of the bowl feeder is based on a 2-D image obtained from an array of fiber optic sensors positioned on the feed track. The binary image, acquired from a moving and vibrating part is processed through an algorithm developed in order to determine the orientation of an object. This algorithm has the signature processing function and the recognition function of comparing the signature with the reference one stored during a learning phase. The signature processing method is based on the division of the image with respect to a particular axis.
From the results of a series of the experiments, it is found that an object can be readily oriented with a reliability of nearly 100 percent. The whole processing time is less than 0.5 sec/part, so that feed rate of greater than 2 per second can be accomplished. Owing to the advantage of being highly programmable. The feeder is expected to be useful for feeding and sorting small parts prior to small batch assembly work.