Smart parts feeder speeds agile assembly

Cambridge, MA--If the papers presented at recent manufacturing-technology conferences are right, the future belongs to the flexible. The ability to produce multiple products on one highly automated assembly line will help tomorrow's manufacturers compete in a global economy.

Although today's robots and automated test equipment have the precision and speed to make agile manufacturing possible, they're hamstrung by the need to present parts in the proper orientation for quick transport. Manufacturers must rely on manual palleting or on dedicated bowl feeders whose fixed tooling and long-lead design time prevent rapid assembly line changes.

Intelligent Automation Systems' FPF 2000 flexible parts feeder could eliminate that handicap. It accepts bulk loads of small parts, transfers them to a conveyor, then aligns and identifies correctly oriented parts for robotic pick-up. Requiring only a 486-based PC and motion-control card, it's easily integrated into existing assembly-line control schemes.

The feeder uses dual programmable parts-orienting arms to align or "singulate" parts on the conveyor and a low-cost, high-speed machine vision system to identify them. With an optional carousel, the feeder accommodates as many as six different kinds of parts, retaining handling instructions for each type in memory.

Although the part-selection process is asynchronous, so is any currently envisioned agile-manufacturing process. Further, the prototype feeder now running would probably need to be modified to accommodate larger or heavier parts. "What we've patented is the concept of flexible parts handling," explains Steven Gordon, Intelligent Automation's president.

Users teach the system to manipulate different parts via simple, on-screen instructions. To begin, the operator sets the height- and width-singulating arms' stepper motors to bring the parts into rough alignment on the conveyor. He or she then places a sample part on the conveyor's vision-staging area in the front of the conveyor beneath the feeder's Sharp GPB-2 machine-vision system.

Next, the menu-driven machine-vision learning begins. The system subtracts the part outline from the background image, forming a template which couples overhead images with side views of the part as seen in a 45 degrees mirror adjacent to the vision stage. It then generates a series of additional templates, electronically rotating the part's image in small increments over 360 degrees of rotation. In that way, the machine learns to recognize parts without the aid of a human engineer.

Using those templates in a proprietary, machine-vision algorithm, the system identifies and locates parts to with 0.002 inches and 0.5 degrees in approximately 350 msec.

In a demonstration using high-aspect-ratio, asymmetrical parts, the system dumps parts from their transfer bucket onto a vibrating platform at the rear of the feeder from which they fall onto the dc-motor controlled conveyor. The system singulates, identifies, and halts the conveyor for robot pick up flawlessly.

Parts not conforming to templates, i.e. upside-down or incorrect parts, fall from the conveyor onto a second belt from which they return to the bucket and are returned for a another go-round. Timers for the conveyors and bucket-transport system coordinate this movement.

Gordon acknowledges that the FPF 2000 will be more expensive and slower than fixed-purpose bowl feeders, but he says, that's not a proper comparison. Bowl-feeder operations require additional time and money to create special tooling to present parts correctly. Their tooling needs costly modifications if the product being made is redesigned and it must be replaced if the line changes to a different product. "The FPF only needs to be retaught," says Gordon. "This could help change the economics of automated assembly for lower-volume products."

Additional details...Contact Steven Gordon, Intelligent Automation Systems, Inc., 142 Rogers St., Cambridge, MA 02142, (617) 354-3830.