Multiple cameras sort brake shoes
Machine-vision systems have long proved themselves cost-effective successors to manual inspection in the automotive industry. Morse Automotive Corp. (Chicago, IL) is one of the latest companies to reap the benefits of an automated vision and sorting system. The company uses a five-camera vision system to inspect used truck brake shoes that are then refurbished for sale by its Morse Heavy Duty (Memphis, TN) division.
Capturing images from five cameras at various angles enables the precise identification of used truck brake shoes.
By Lawrence J. Curran,Contributing Editor
Machine-vision systems have long proved themselves cost-effective successors to manual inspection in the automotive industry. Morse Automotive Corp. (Chicago, IL) is one of the latest companies to reap the benefits of an automated vision and sorting system. The company uses a five-camera vision system to inspect used truck brake shoes that are then refurbished for sale by its Morse Heavy Duty (Memphis, TN) division. The vision system was integrated and installed by Walsh Automation (Montreal, Canada), a division of Invensys plc (London, England).
Morse manufactures heavy-duty brake shoes for the after-market. The division receives thousands of used truck brake shoes that must be correctly identified so that the customer who returned them can be credited with the proper value. The brake shoes are then sorted before they can be refurbished and returned to the retail market.
Correctly identifying and processing the brake-shoe cores is a critical part of the inspection process, according to Bob Wilkes, Morse Automotive vice president for product development. He says, "Prompt and proper identification of these cores is necessary to determine the proper issuance of credits to customers." Furthermore, the brake shoes must be sorted and stored for quick and easy retrieval by part number for subsequent "build-to-order" remanufacturing.
Identification and sorting was performed manually before the Walsh Automation vision system was installed. The company's previous reliance on experienced workers with the ability to identify brake-shoe types proved chaotic when workers' assignments were changed. Moreover, manual inspection proved to be expensive.
Walsh Automation's solution was to custom-design and oversee the installation of a machine-vision system implemented with five cameras and software algorithms that simulated worker knowledge in identifying and sorting brake shoes. Wilkes points out that several years before installing the current machine-vision system at Morse Heavy Duty, Walsh Automation had installed a vision sorting system in a Morse Automotive brake-shoe facility in Arkansas. "This [Arkansas] system has been working for eight years and has greatly improved our overall operation by reducing labor costs, reducing errors associated with manual part recognition, and speeding issuance of core credits to customers," he reports.
Morse Automotive considered investigating systems from companies other than Walsh Automation, but it could not find a firm willing to undertake the task of identifying the brake cores the way Morse Heavy Duty wanted the job done. When manual sorting was done, the inspection process relied on an individual employee's familiarity with many different brake-shoe types. "In some cases, rivet size was the only variable that determined one core from another," Wilkes notes. The job also required the use of the most experienced employees in the operation because they were the most knowledgeable about core recognition. "The process was subject to many errors because of uncontrollable variables. Also, this was an unacceptable use of experienced employees whose talents could be better used in other value-added operations," he adds.
Jean-Francois Dupont, project director at Walsh Automation, says the vision and sorting system his company integrated for the Morse Heavy Duty brake-shoe inspection application incorporates five JAI AS (Glostrup, Denmark) CV-M10 progressive-scan CCD cameras linked to two Meteor II frame grabbers from Matrox Imaging (Dorval, Canada). One camera acts as the master and is linked to one of the frame grabbers. The other four cameras are slaved and synchronized to the master and are supported by the other frame grabber (see Fig. 1).
A motherboard from ASUSTeK Computer Inc. (Taipei, Taiwan) incorporates dual 600-MHz Pentium microprocessors that direct the operation of the vision system. The motherboard is assembled for Walsh Automation by SPM Micro (Montreal, Canada). The overall vision and sorting system line, which includes a washing station, conveyor belt, and selection bins or baskets to collect sorted and classified truck shoes, is under the control of a Series 90-30 programmable logic controller from GE Fanuc Automation (Charlottesville, VA).
FIGURE 1. Morse Heavy Duty brake-shoe inspection system uses five JAI CV-M10 progressive-scan CCD cameras linked to two Matrox Meteor II frame grabbers. Two ASUSTeK Computer 600-MHz Pentium microprocessors direct system operation and run Windows NT, Visual C++, and image-processing software developed by Walsh Automation. Lighting for the vision station is basic backlighting, helped by fluorescent and quartz lamps. The vision and sorting system line, which includes a washing station, conveyor belt, and selection bins to collect sorted and classified truck shoes, is controlled by a GE Fanuc Automation Series 90-30 programmable logic controller.
The image-processing software was developed by Walsh Automation. The Morse Heavy Duty application, as are all the Walsh Automation applications, "was custom-designed from our library," Dupont says. Lighting for the vision station is basic backlighting, aided by some direct lighting using both fluorescent and quartz lamps, he adds.
How it works
In operation, pallets of random models of used truck brake shoes from a single customer are manually placed in the washing station to remove as much dirt as possible. After washing, the shoes are loaded onto a flat conveyor belt, on which they move along to the vision station. Sorting by brake shoe model is needed because the amount of credit given to each Morse customer for the shoes is based on the model. Dupont says that 80% of the sorting into specific models is done by the profile camera, which takes the first images at the vision station (see Fig. 2).
FIGURE 2. Typical screen shot of a previously unrecognized truck brake shoe shows a profile view (upper left) and a view of the brake shoe web (the width between the two horizontal ribs), as well as the rivet holes (bottom left). In operation, the data and images of this shoe are crosschecked against all model numbers from the customer to make sure it isn't already in the database. If it isn't, the technician enters the data and assigns an appropriate model number.
An overhead camera then captures images of the brake shoe's web size, which is the distance between the two horizontal ribs (see Fig. 2). Two more cameras take photos of each end of each shoe. The fifth camera records images of the shoe's rivets, "because the size of the rivets may be the only way to positively identify the shoe model," explains Dupont. Different imaging software modules are used for each imaging angle.
The images taken from all the cameras are stored in a host PC database. These images hold the model number associated with that image of each manufacturer's brake shoes that Morse Heavy Duty accepts for re-manufacture. After a positive identification (ID) is made, based on a comparison of the images captured live on the line with those in the database, the shoe is routed to the bin or cage that matches that model number. At the end of the inspection run, the system prints out a report showing how many of each model number were processed. Based on the vision system data, Morse Heavy Duty issues credit to its customer.
Dupont points out that if a brake shoe being inspected is not recognized as one in the database, it may be because the shoe is damaged. Accordingly, this shoe is rejected and routed to a separate basket. The customer would get no credit for that shoe. If, however, the shoe core isn't recognized because it's new to the system and there's no match for it in the database, the technician manually enters the new shoe information into the database at the end of a shift. This entry includes the model number and associates this number with the already captured images.
A major design challenge that confronted Walsh Automation in developing the vision and sorting system was how to cope with either dirt or water left on a brake shoe after washing. This problem was particularly bothersome because the shoe rivets could be obscured, preventing accurate imaging. One correction technique involved the addition of air jets between the washer and vision station to blow away as much water as possible. For another correction, says Dupont, "We also did a lot of work on the recognition algorithm."
It was also important for Walsh Automation to provide a user-friendly interface for the vision system so that it could be operated by truck shoe sorters who are unskilled in computer hardware or software. Accordingly, the host PC runs Windows NT and the user interface is written in Visual C++; both contributed to the design of an easy-to-use graphical user interface (GUI).
Lastly, Walsh Automation provides close support to the Morse Heavy Duty facility. "We see the production operation live from Montreal," Dupont says. "We're linked to the vision and sorting system by modem and to the facility by a direct phone line."
Adapting to change
As with any major factory-floor change, many managers and workers are not immediately ready to endorse a new vision and sorting system. "Most workers do not accept change easily," Morse Automotive's Wilkes points out. For example, during the initial design phases, the plant management wanted to keep the old sorting conveyor belt in close proximity to the new vision and sorting system as a "just-in-case-it-doesn't-work fallback position."
Wilkes adds that it was important to appreciate employees' feelings. Before the installation of the Walsh Automation system, the highest degree of automation at the plant was in the form of programmable logic controllers (PLCs) on individually operated machines. "Some fears were understandable because the Morse Heavy Duty employees weren't convinced they had the technical expertise to operate and maintain such a system," he explains.
"Employees who operated the manual sorting line were curious and skeptical of the new system," he adds. "They said it wouldn't work, but as the vision system started to take shape, and vision tests were performed by Walsh Automation, initial fears started to wane and genuine curiosity grew. Everyone wanted to know how it worked and tested with shoes they had trouble recognizing manually."
Learning the system was a challenge, primarily because Morse Heavy Duty is historically a low-technology operation, including the turnover typically associated with that type of industry. "But the training did not require above-average computer skills. In fact, we have persons without any prior computer skills operating and teaching the system, with little support. With continuing support from Walsh Automation, both on site and via telephone, the system continues to operate at the designed level," Wilkes maintains.
The only change Morse Heavy Duty made to the installed system was in report generation. "We added our customers' part numbers to the system, so we did not have to cross- reference part numbers. This improved our process of issuing credits on core returns."
Morse Heavy Duty has been using the truck brake shoe vision and sorting system for more than a year. Its use has "improved our identification of core returns, increased throughput, and reduced labor costs. An additional benefit of processing cores in this manner is the improved accuracy of accounting for actual core returns instead of relying on the invoice quantity from customers," Wilkes concludes.
ASUSTeK Computer Inc.
GE Fanuc Automation
Charlottesville, VA 22911
DK-2600 Glostrup, Denmark
Dorval, Quebec, Canada H9P 2T4
Morse Automotive Corp.
Chicago, IL 60609
Anjou, Quebec, Canada H1J 2T3
Montreal, Quebec, Canada H4B 2M8