Smarter Machine Vision Systems Proliferate in Automotive Industry

In the automotive industry the trend is toward more automated inspection to ensure the right parts are installed, installation is correct and fit and finish is acceptable.

The watchdogs that accomplish these tasks are a new generation of more compact, powerful and faster machine vision instruments, which can identify small defects as well as large and, most importantly, are far and away less costly and complex to install and operate than their forebears.

These systems do not necessarily work alone, either. Increasingly, machine vision systems are integrated with robotics to help perform traditional robot activities like body panel cutting, welding assembly and painting and part picking and placing, as well as new jobs like serving as a transport platform so the vision system can move around its target for tasks such as surface inspection or measurement.

Since the production and inspection of automotive parts, subassemblies, frames, panels and accessories and car or truck assembly has long been dominated by highly trained human technicians and engineers, who needs visually competent machines?

Virtually any manufacturing operation can benefit because even the most capable human workers react relatively slowly and experience declines in efficiency over long working periods. Historically, human inspectors, at best, have been only about 85 percent effective in catching problems. In addition, certain locations on or in a car under assembly, which need inspection, are quite difficult for a human to see. And there is a time factor, too. On some of today's most up-to-date final assembly lines, a car is completed every 45 seconds.

Furthermore, car manufacturing can be dangerous. Metal cutting, welding and assembly all have hazards. Even now, some manufacturers are removing humans from cutting and welding work.

The transition to vision-equipped robots will be accelerated because manufacturers now have to cope with new and tougher ergonomic regulations by the U.S. Occupational Safety and Health Administration, Washington, DC. Similar rules already exist in Europe (and in the United States) in the form of ISO 9000 requirements and encourage automating heavy lifting and repetitive motion tasks.

While no one really knows exactly how many machine vision systems are installed in U.S. automotive plants, industry veteran, Nello Zuech, long-time author of AIA's annual industry report and president of Visions Systems International, a consultancy based in Yardley, PA, believes opportunities are growing. 'Total automotive use, including Tier 1 and 2 suppliers, could run to about 64,000 machine vision systems within the next five to seven years,' he predicts.  The annual market study he produces for the Automated Imaging Association forecasts average growth in automotive applications of 20% over the next five years.

In current applications at Ford Motor Co., Dearborn, MI, fixed vision systems are used mainly to check car and truck bodies near the end of the line for quality of parts fit (such as windshields), presence of all parts, proper application of adhesives and lack of gaps between body parts. Distribution of machine vision systems now at Ford ranges from two to five in the body shop, 12 in subassembly and others down the line in final assembly operations, reports Valerie Bolhouse, a technical specialist at Ford, who is enthusiastic about the potential for machine vision inspection.

Although machine vision has been available for almost two decades, its use in auto building is really just starting. Even with many systems already in place at Ford, Bolhouse believes, a lot more are needed 'to achieve the desired level of automation.'

Machine vision is still in early stages of adoption in the automotive industry because the first machine vision systems were not really designed for on-line inspection, particularly of moving objects. Early systems also were a bit bulky, heavy and expensive, ranging from $40,000 to $60,000 each, a largely unaffordable price for an industry where many systems are needed. In addition, setup and operation were complex and technical capability and flexibility were limited. Furthermore, performance claims by some suppliers were exaggerated and performance did not live up to expectations. Although vision system vendors now have a much better understanding of the automotive industry's needs and systems are much more powerful, these early disappointments have not been completely forgotten and still haunt vendors a bit today, say industry observers.

Spurring interest among automakers is a growing number of complete machine vision systems in the $5,000 to $6,000 price range, the principal segment of systems now being bought by automakers. And prices continue to decline as suppliers streamline units to target specific applications. There are even a few simpler units priced below $5,000, which may find niche uses in automotive inspection. It's no longer one machine does everything. Instead, units are designed for specific functions. This simplifies setup and installation and reduces the need for time consuming and costly custom programming.

Advances in computing also have been a positive force. Formerly, each vision unit had to have its own processor. Now, multiple arrays can be driven by a single processor. 'We have a 16-camera array now available, entirely driven by a single processor,' reports Mike Johnson, Automotive Branch manager at PPT Vision, Inc., Minneapolis, MN.

Capabilities are greater too. For example, PatFind, an advanced vision software tool created by Cognex Corp., Natick, MA, is able to determine the X, Y, and theta coordinates of a part despite angle, scale or appearance variations, and then fires the coordinates to a robot for action, asserts Matt Smith, this vision company's automotive expert.

On the robotics side, Ed Roney, manager of development for Fanuc Robotics North America, Inc., Rochester Hills, MI, who has been involved with machine vision since 1984, provides robots equipped with integrated machine vision systems both to guide automated assembly operations and perform on-line inspections. Fanuc has 150 vision-guided robots installed at automotive companies. The units are equipped with X-Y-Z axes and rotate around each axis to accomplish difficult scanning jobs and record three-dimensional measurements.

Roney now is receiving queries about whether very slim vision-equipped robots could be used for assembling automotive components. 'Sure,' he says, 'but the vision systems must have more power and a finite field of view in the subassembly of smaller parts. 'Currently, we have a vision unit capable of 6 degrees of view. This narrow scale, not a broad one, is needed for such specialized applications,' he believes.

Ford's Bolhouse is researching use of vision in vehicle assembly, not only for initial subassembly, but also for installation of subassemblies and components in the frame. A problem here, she notes, is that most assembly conveyors are not yet accurate enough in configuration to permit automated targeting by vision-equipped robots. However, she also is conducting automated assembly research to determine exactly what is needed. The location is a tough one, and she notes - as ever in this industry - that although the vision field has done a lot to provide more robust cameras, more toughness is needed.

'All approaches will have their place,' declares PPT's Johnson. There will always be fixed vision scanners as well as robotic-mounted units. The fact is different automakers and different operations call for different solutions.

And so it goes. Vision engineers are learning more about what their counterparts in the automotive industry need today and are striving to meet the ever changing demands for tomorrow. Their automotive engineering customers are beginning to understand what vision systems alone and in combination with robots can do to reduce labor expenses and other assembly costs, while at the same time improving the quality of their automobile and truck products. It should be a profitable enterprise, which also will include big benefits for the end-use customers.