Vision system certifies auto-bolt inspection
Automotive supplier replaces manual inspection with automated vision system for improved throughput and yield of wheel bolts.
In the highly competitive automotive industry, parts inspection for ensuring high quality is of paramount importance. For example, a failed bolt can cause the loss of a wheel and result in a serious auto accident. ISO Sorting & Packaging LLC (Shelby Township, MI) is working to prevent such a failure by replacing the manual sorting of automotive bolts, washers, and other metal parts with a three-camera vision system to quickly, easily, and inexpensively detect defective parts.
The vision system has been in use for about two months, says Mike Carten, ISO general manager. During that time, "we have done daily calibrations, and the vision system has found every defective part. It runs almost unattended. We think we can run four or five such systems with one operator," Carten says. Furthermore, inspection speed also has been boosted by a factor of 12 compared to that of manual inspection.
The Model 343B vision system inspects 3-in. bolts from Retina Systems Inc. (Seymour, CT). ISO has ordered several more systems. The Model 343A system is slated to accommodate 6-in. bolts. Carten claims that ISO is the sole provider of sorted and packaged wheel bolts, nuts, and washers to a major parts supplier for several auto manufacturers.
Floyd Moir, Retina Systems president, says that the Model 343B system incorporates a nested dial assembly and three cameras to inspect a range of bolt types and sizes at speeds as fast as 350 parts per minute. The cameras use matrix and linear array imaging, combined with custom lighting techniques.
Auto manufacturers have long demanded stringent quality control from their suppliers. "Most of them have zero-tolerance programs in place and mandate 100% inspection of all fasteners," Moir points out. For this reason, Retina Systems now guarantees 100% certified part sorting and inspection. ISO previously used both ultrasonic and eddy current technologies to inspect fasteners, but neither method could meet the requirement to rotate parts for crack detection the way the Retina vision system does, according to Moir.
FIGURE 1. During bolt inspection (bottom), the bulk feed hopper maintains a constant flow of bolts to the vibratory bowl feeder. In turn, the bowl feeder orients and feeds bolts to the vibratory in-line feeder. This feeder presents bolts to the nested dial plate, where the bolts are held by their heads and then indexed through the three-camera inspection sequence (top). Good bolts are placed in the good-bolt eject chute, after which a conveyor automatically indexes empty boxes to be filled with good bolts to preset counts.
The bolts to be inspected are first moved from the Model 343B vibratory feeder bowl, which is supplied by AFM Engineering Corp. (Dayville, CT), placed in an in-line orientation, and then fed, hanging by their bolt heads, down a slotted track into a nested dial assembly (see Fig. 1). The nests are cutouts in a circular plate that hold one bolt each; the dial plate is driven by an Intelligent Motion Systems Inc. (Marlborough, CT) MicroLynx stepper motor so that different views of the bolts can be captured by the three cameras (see Fig. 2).
The bolts are inspected for features such as head height and diameter, shank length, thread pitch, and head cracks. The bolts are held by their heads in the nests as they are presented to the vision system by the dial plate. This plate rotates continuously or indexes, via the stepper motor, depending on the part number entered into the system. When linescan imaging is being performed, the nested dial moves in an indexed manner. As bolts move along the dial, fiberoptic gate sensors trigger various strobe lights. This process delivers images into associated vision modules for image capture.
The two matrix cameras are located at 90° to each other on the periphery of the nested dial plate and capture images of the bolts' profiles (see Fig. 3). These cameras from Sony Electronics Inc. (Park Ridge, NJ) provide a resolution of 640 × 480 pixels. Retina Systems' light-emitting diode (LED) illumination is used to capture images of features such as head height, head diameter, overall length, shank length, major/minor thread diameter, thread pitch, and presence of a locking patch. For this application, the locking patch is a nylon coating applied to prevent loosening of the bolt during use on an automobile wheel.
Next, a gripper picks up the bolt, another inspection setup spins the bolt, and the third or linescan camera captures a profile view. This 1024-pixel camera, supplied by PerkinElmer Optoelectronics (Fremont, CA), inspects for features such as damaged threads or thread cracks. Xenon-type strobe lights, also from PerkinElmer, illuminate this setup.
Images from the cameras are captured and processed by frame grabbers from Data Translation Inc. (Marlboro, MA). Moir says that the Retina machine's imaging sequences and lighting are controlled by a Pentium 4 PC, which is purchased from multiple suppliers. The PC incorporates a passive backplane, which will accept advanced frame grabbers and processors as they become available.
Based on parts blueprints, all the bolt dimensions and features to be inspected by the vision system are entered into a database using proprietary software developed by Retina Systems. The company also developed the vision system's image-processing software, as well as the nested dial system and conveyors built into the overall bolt-inspection machine.
All three cameras are connected to the frame grabbers, which connect to the computer's PCI bus. Communications required between the vision system and the Allen-Bradley (Milwaukee, WI) programmable logic controller are performed using discrete digital logic circuits.
Proprietary software runs the vision system. The nested dial speed and lighting levels are controlled by the PC and are entered into the system when the part number is loaded. "The feed rate and the dial speed are balanced by the software so that no operator is needed for the system once the part number is entered," Moir adds.
If the dimensions calculated during image processing do not match the part-profile tolerances previously entered for the bolt being inspected, then the bolt is rejected. ISO's Carten says the "go/no-go" decision takes place after the third (linescan) camera's image is processed. Acceptable bolts are delivered through a chute into a box on a packaging conveyor. Defective bolts continue to move around the dial and are mechanically removed.
"The best feature of the system is its fail-safe protection," Carten says. "It won't allow a bad bolt to get into the good-bolt bin undetected. If that happens, sirens and lights activate, the dial is cleared, and the machine can't be started until the bad bolt is removed," he adds. In addition, Carten notes that the vision system enables real-time dimensional analysis of the bolts being inspected over the Internet. This enables remotely located bolt manufacturers to identify and correct problems early in their manufacturing processes.
The system's speed is related to the bolts being inspected. For example, Retina's Moir says, bolts that do not require inspection for head cracks move at a rate of 250 to 350 per minute because they do not need to be lifted and rotated. The movement rate for rotated bolts is 60 to 75 parts per minute.
ISO's Carten says the highest rate for manual sorting was 25 bolts per minute. "With the vision system, we now run continuously at upward of 300 bolts per minute and average about 280." More important, he adds, "during early operations, they have found every bad part."
Finally, he points out that whereas manual sorting resulted in "a lot of false rejects, and we didn't know why the part bolt was rejected," the vision system logs each reject and the reason it failed inspection. "We sort for everything on the blueprint, and we can show exactly why a bolt failed. We can also count the number of failures for each flaw logged, and then notify the vendor about problems early in its manufacturing process," Carten concludes.
By Lawrence J. Curran, Contributing Editor