A NEURAL NETWORK METHOD IS APPLIED FOR IDENTIFYING THE ORIENTATION OF INDIVIDUAL PARTS BEING FED FROM A PROGRAMMABLE BOWL FEEDER. THE SYSTEM IS DESIGNED IN SUCH A WAY THAT A PART CAN BE DISCRIMINATED AND SORTED ACCORDING TO EVERY POSSIBLE STABLE ORIENTATION WITHOUT IMPLEMENTING ANY MECHANICAL TOOLING. THE OPERATION OF THE PROGRAMMABLE BOWL FEEDER IS BASED ON A 2-D IMAGE OBTAINED FROM AN ARRAY OF FIBER OPTIC SENSORS LOCATED ON THE FEEDER TRACK. THE ACQUIRED BINARY IMAGE OF A MOVING AND VIBRATING PART IS USED AS INPUT TO A NEURAL NETWORK WHICH, IN TURN, DETERMINES THE ORIENTATION OF THE PART. SINCE THE OPERATION OF THE FEEDER IS HIGHLY PROGRAMMABLE, IT IS WELL-SUITED FOR FEEDING AND SORTING SMALL PARTS PRIOR TO SMALL BATCH ASSEMBLY WORK.