Printer Friendly

Getting into machine vision.

Getting into machine vision

Suppose that you find yourself in this position regarding machine vision (MV):

You've seen MV demonstrated at several trade shows, and you've read a few articles about it. You find yourself interested in the technology, and think there's at least one good application for it in your plant.

At the same time, though, you've been holding back. You're at home with machining, mechanical engineering, and metallurgy, but MV involves optics, illumination, electronics, and computer science--disciplines that are largely foreign to you.

Still, you haven't given up on MV. You feel that it works, and that it's here to stay. You know, from testimony you've read and heard, that MV has achieved remarkable benefits for some companies. And, perhaps most important, you remain intrigued. Skeptical and uncomfortable, but intrigued.

Has any of this rung a bell? If so, be assured that you are not unique. Many of your fellow readers find themselves in the same situation vis-a-vis MV.

How can you move into and adopt this new technology with more confidence, and with minimum risk to your job, reputation, and the company you work for?

To obtain some answers, we interviewed seven of the top US experts on MV. Three of these are independent consultants, people who have no axe to grind for particular product lines or technical approaches. The other four are executives in successful, well-established companies supplying MV systems engineering and installation.

Following is a report on what these experts told us, in the form of a nine-step guide.

It's not only what but how you learn

Step 1. According to our experts, your first step toward successful adoption of MV should be to learn the basics of the technology. This sounds so simple and obvious that you may be tempted to flip the page, but hold a minute. The experts stress that how you go about learning the basics may be just as important as the content of what you're trying to learn.

Without exception, the experts advise that your primary source of information should be seminars, clinics, and particularly hands-on workshops. It's important that you see the various types of equipment, lighting setups, and methods of parts presentation.

Today the best clinics and workshops are being offered by two MV organizations within the Society of Manufacturing Engineers (SME), Dearborn, MI. The organizations are the Machine Vision Association of SME (MVA/SME), for individual members; and the Automated Vision Association of the Robotics Industries Association (AVA/RIA), for vendor companies. Both the MVA/SME and AVA/RIA publish lists of upcoming clinics and workshops. For a circle number, see the box.

You can choose between two types of sessions. One, usually lasting for a full day, offers illustrated lectures on the basics of MV technology, plus hands-on work with equipment from one or two vendors. The second type is a more intensive workshop; this usually lasts three days, and gives you hands-on experience with equipment from 8 to 10 vendors.

"I strongly suggest that the novice go to a short, introductory clinic first,' says Nello Zuech. A leading MV consultant, he's president of Vision Systems International, Yardley, PA.

"If you don't understand the basics before you go to one of the longer workshops, you'll come away confused. Better to take Vision 101 before Vision 102.'

Before attending your first short clinic or seminar, you may want to begin learning the basic terms and concepts by reading introductory literature. An excellent multipurpose source is Machine Vision Systems--a Summary and Forecast, second edition, published by Tech Tran Corp, Naperville, IL. This 220-pg softbound manual and directory contains not only an illustrated introduction to MV fundamentals and a glossary of terms, but also up-to-date listings of system and component supplier companies, research organizations, consultants, books, conference proceedings, and periodicals. For a circle number, see the box.

You can also learn from sales engineers representing system supplier companies that you have already contacted. During your first three-day workshop and beyond, it's a good idea to let the company experts help you learn. But Nello Zuech offers a word of caution:

"If you are a novice in MV, avoid relying solely on sales engineers for your technical training,' he stresses. "Many prospective users of MV have become victims--not intentionally, but out of naivete on both sides.

"MV is a young technology, and many MV salesmen are still relatively inexperienced. They tend to promote only their own company's products and methods, which may not be right for your applications.'

Several consultants now offer a valuable service for first-time users of MV. For a fee, the consultants will visit your plant and put on seminars covering the broad basics of MV, including the various kinds of equipment, systems, and image analysis. If you and your company have developed a serious interest in MV, this may be a worthwhile service for you.

Step 2. Get other people involved. As several of our experts point out, it's best not to try to go it alone in the adoption of a new technology such as MV. You may play the role of MV's champion in your company, but you will need support and cooperation from others in the company.

An effective way to accomplish this is to sign up a consultant for a one-day inplant seminar. Recruit plant staffers and managers from all the disciplines that may become involved--not only manufacturing engineering, but also plant engineering, industrial engineering, material handling, methods engineering, and inspection.

Putting on a seminar in your own plant will be much easier and more cost-effective than attempting to send all concerned indivduals to out-of-town seminars and workshops.

"Many companies are getting started by setting up internal MV groups,' says George Gagliardi, MV specialist with Arthur D Little Inc, Cambridge, MA. "To ensure success, it's critical that the company make available time and resources so all involved people can learn what MV is, and what it can and can't do. Eventually the group should understand MV well enough to write fairly firm specs for a system.'

Keep it simple

Step 3. Our experts advise that the first on-line application of MV in your plant should have the following characteristics:

Simplicity. By this they mean that the application should be straightforward, relatively easy from the technical point of view, and preferably of a type that has already been solved by other users and system suppliers.

"Too often, the beginner suggests a very complex, technically difficult application for his first one,' says Perry West. A nationally recognized MV consultant, he's president of Automated Vision Systems, headquartered in Campbell, CA. "Gradually, though, people are coming to realize that they shouldn't start that way.'

And what constitutes a "simple' application for MV? "For one thing, the inspection criteria should be consistent from part to part,' says Ralph D'Iltalia, national sales manager for Automatix Inc, system supplier in Billerica, MA. "For another, the parts and their key features should have contrast that can be illuminated, acquired, and analyzed within realistic cycle times.

"Then too, look for a situation in which the MV system can operate in a friendly environment,' D'Iltalia continues. "By that I mean, for example, there shouldn't be grease on the parts, and no overlapping of parts on the belt. Flat parts should always be presented flat, not sometimes standing on edge.'

Above all, don't look at the first MV system as a way to solve the toughest inspection or inetrology jobs in the plant, the ones you've never been able to adequately solve with conventional means. MV may not be able to solve these problems either, and if your first MV project fails, the results could include damage to your reputation as well as a lasting aversion to MV by all in your company.

Quantifiability. You should be able to express the desired objectives numerically. In addition, it is a good idea to pick an application in which the demonstrated results can be cost-justified.

"You could probably justify an MV system easily if, for instance, it formerly required five hours to check metal parts on an optical comparator or coordinate measuring machine, but MV would enable you to check the same parts in three minutes,' says Larry Klein, field engineer with View Engineering Inc, Simi Valley, CA. He works out of the company's regional office in Carnegie, PA.

"The case becomes even stronger when you can show that MV captures and transmits data automatically to a database for process monitoring, comparison, and reporting.'

Repeatability. As stated earlier, the first application should preferably be one that someone, somewhere has solved before. An MV system is not the sort of thing you can simply buy off the shelf and slap into place. Each general type of application, and each individual case, requires a great deal of investigation, evaluation, and development work.

"You can help the prospective system vendor to amortize the high development costs, and hence to lower the total system cost, if you pick a relatively simple application that could be repeated within one or more company plants,' says James West. He's president of MVA/SME, and also vice president of Perceptron Inc, system supplier based in Farmington Hills, MI.

"This is also one way to ensure getting attention from the top system vendors,' West continues. "Our company, like other established, profitable MV system vendors, looks for repeat business within your plants. If you come to us and propose an application for MV, and it's a highly difficult challenge technically --and it's one that likely couldn't be repeated in the company--then frankly we may turn the job down. The custom development work would probably be so time-consuming and expensive that, one, we couldn't recover enough of our costs, and two, you wouldn't want to pay the price.'

Do you know your process variables?

Step 4. Analyze the proposed application thoroughly. Once you have picked a likely candidate for your first MV application, study what your present inspection and measurement people are doing. This may sound easy, but it isn't. People perform many little operations without thinking about them, often not aware of all that they're doing. And even in allegedly precise functions such as inspection, many human judgments are highly subjective.

Often, in fact, the decision to pass or reject a part is based on Kentucky windage rather than on numerical parameters. It's important that you learn the bases for the parameters, and establish numerical values.

"Many people don't really understand their process variables,' observes Nello Zuech. "With people running the show, you don't need to know all the variables. You can teach people to ignore all data except those that are critical, and to apply their experienced judgment. An MV system won't ignore the variables, however, and it has no expertise to apply.'

If the application involves gaging, select a sampling of parts that will be truly representative of your production runs. "A key question to ask yourself is, would you be willing to buy an MV system solely on the basis of the samples you selected?' says Perry West. "If your answer is no, then the selection is inadequate.'

For gaging applications, selecting samples usually doesn't present major problems, West observes. Flaw detection is another story, though.

"Sometimes it's very difficult to quantify flaws,' he says, "and to establish tolerance ranges for go or no-go. But you had better make the effort to quantify before you proceed too far with the project.'

Step 5. Pin down your management objectives. Thus far we've been looking at preliminary technical matters, but your desire to adopt MV probably springs from a need to achieve certain management objectives--and not, we trust, from a mere infatuation with the technology. Your next task is to clearly identify the management objectives, set them down on paper, and prepare to communicate them to prospective system suppliers.

Most companies become interested in MV because they want to gain a higher level of quality control. Sometimes this is gained through off-line inspection; other times, through on-line, 100-percent inspection.

Inspection, identification, and sortation of objects moving by on a conveyor belt is a boring, tedious, fatiguing task for humans; one that often drives them to seek other work. For that reason, some companies want to implement MV systems to retain the services of valued employees.

Perhaps you want to achieve accurate, real-time identification, tracking, and control of parts moving through a process. An MV system that can identify parts by shape, color, or texture--or that can read alphanumeric or bar code symbols on the parts--will take you a long way toward achieving that goal.

If you have pinpointed reduction of labor as one of your goals, you may want to heed a cautionary note from Perceptron's James West. "Don't expect to make a one-for-one trade between a human and a machine,' he says. "The two creatures are quite different from one another; each has its peculiar capabilities and limitations.

"Most MV systems in place today have not been justified on the basis of laborsavings,' he adds, "and it's not likely that future systems will be justified on that basis.'

Concerning establishment of management objectives, Nello Zuech urges that early in the process ask yourself, "Is it really the capabilities of MV that I want, or is it rather a more extensive, accurate control over the process itself?'

Zuech says that when a company calls him in to evaluate a possible application, he starts with the premise that the company may not need MV, or even a human inspector. "Better statistical process control, and better control over the mechanical process and its variables, may obviate the need for MV,' he states. "Sometimes looking at the situation this way will enable the company to use a simple MV system rather than a complex one, or to substitute a very simple sensing system for MV.'

The hardest part

Step 6. Learn all you can about the vendors, especially their experience and capabilities.

According to the Tech Tran directory mentioned previously, there are now some 94 US companies offering MV systems design and installation. If you add manufacturers of components--cameras, image processors, etc--plus consultants and system integrators, the total number of companies involved in MV tops 150.

Let's assume, though, that you're interested only in the 94 companies that supply systems. Your mission now is to find 10 or 12 that have the experience and capabilities needed to solve your problem.

It's not easy; in fact, this could be the most difficult step in the procedure. "For one thing, MV systems work is highly application intensive, so most companies have been gravitating toward niches,' points out Gary Wagner, vice president of Penn Video, a systems supplier based in Akron, OH. "The successful companies specialize in certain MV functions and in certain industries.'

It is important to note that the notion of technology transfer doesn't fully apply here. "The typical MV tasks in auto assembly plants, for instance, are quite different from those in electronics manufacturing or packing,' observes A D Little's George Gagliardi. "It's not easy for a system supplier to make the leap from one industry to another. So, unless you're tackling a relatively easy, common type of problem, beware of vendors who claim they can do anything for anybody.'

This means that you cannot simply look up the name of an MV company in the Yellow Pages, call them up, and expect to have your problem solved--at least not satisfactorily. Generally speaking, you need to find the companies that have been specializing in your industry and the kinds of applications in which you are interested.

Step 7. Narrow your list. Once you have lined up 10 or 12 prospective vendors, you're ready to narrow the field to the likeliest three or four. Ask questions such as:

Will the supplier take full, singlesource responsibility for the system, or will he take only part and open the door to finger-pointing later on?

Does the supplier have a large enough staff to provide adequate back-up service?

What about spare parts and components? Does the supplier carry stocks, or does he order everything from his suppliers as needed?

What kind of image does the supplier present in workshops and at trade shows? Does he impress you as being professional, or is he all razzle-dazzle and sales talk?

When the supplier's sales engineers make claims, do they sound reasonable, and are they properly qualified?

What do the supplier's customers say about him? As you'd expect, many users of this young, powerful technology will not reveal details of their installations, but at the same time some are willing to discuss the supplier company, its methods, and the degree of satisfaction experienced. If a prospective supplier won't give references, caveat.

Is the supplier willing to work with you in a phased program, which would include criteria for final acceptance and buy-off?

And then there is the question you probably won't ask outright, but for which you certainly ought to seek an answer: How long will the prospective supplier be around? With so many companies jockeying and scratching for survival in MV, there is bound to be a shakeout. Has the prospective supplier established itself in one or more application /industry niches (but not so many as to stretch credibility)? And what is the financial condition of the company?

The phased program

Step 8. Launch a phased program with the vendors. Assuming now that you have narrowed the list of prospective system suppliers to three or four, your next step is to set up and put into action a suggested five-phase program.

In Phase one, the suppliers come to your plant and study in detail the proposed application and your management goals. The suppliers will attempt to learn as much as possible about the manufacturing processes involved. If your company is to be a first-time user of MV, this phase may extend over a long period of time.

"All MV supplier companies that are reasonable understand the need for two-way education,' says Perceptron's James West. "The customer usually needs to learn about machine vision, and we need to learn the customer's processes. We often anticipate spending a full year with a customer before making a sale. In some of the more complex jobs, two years have elapsed between initial visits and the signing of the purchase order.'

If your company has had extensive education and previous experience in MV, the first phases probably won't take that long. Says Gary Wagner of Penn Video, "Today in the automotive industry, for example, it is not unusual to receive a signed order within 60 days after initial contact. But these are not exceptionally difficult jobs, and many people in automotive have become quite knowledgeable about MV.'

Phase two. After the system suppliers have completed their learning period, ask them to submit initial plans and cost estimates. Doing this will help to further narrow the list of prospective suppliers.

What you do not want to do is to whip up rough sketches, shotgun requests for quotes to a dozen companies, and then base your selection mainly on price. "We still see a lot of this,' says Perry West, "but we recommend against the practice. It is very expensive for a supplier to put together a proposal, for one thing. For another, this is a closely knit industry, and the suppliers soon find out who is shotgunning requests for quotes. If you're shotgunning, and a supplier gets a request from you, he may not give it the care and attention that it deserves.'

Phase three. Narrow the field to one or perhaps two prospective suppliers, and ask them to perform feasibility studies. In most cases, the supplier does this by setting up a simulation of your proposed system in his lab. The simulation should be as realistic as possible, using your company's parts and fixtures, along with a conveyor and other handling devices that match yours closely.

"Some system suppliers will perform feasibility studies free of charge,' points out Perry West, "while others will charge for them. Don't reject a supplier because he wants to charge; this is money well spent. It is very costly to end up with an MV system that doesn't perform as desired.'

Phase four. This phase of your program involves asking for firm quotations from one or two finalists, with specs based on the system as demonstrated and approved. In the fifth phase, you set up acceptance test criteria; these become part of the final purchase document.

Step 9. Consider a learning system. This step applies if you intend to plan and install MV systems yourself, using in-house staff and technicians. This may be a desirable procedure if your company is large and you anticipate implementing a considerable number of systems.

In this step, you would purchase a fairly sophisticated MV system, set it up in your lab, and use it to learn basics and application details. The vendor will assist you in phasing-in the system.

"We do not advise this, however, for companies that plan to put in only one or two systems,' says Gary Wagner "Most of the companies that have tried it are still in the lab playing with the system.'

If you hastily team up with the first two or three system vendors you encounter, you could get lucky, or you could incur difficulties and disappointment. Following the steps outlined here, though, will ensure an excellent chance for success in your first MV installation.

Photo: Designing an MV system for a user, a vision engineer experiments with various lenses and optical filters. Experts say that companies planning to install large numbers of MV systems should consider setting up their own vision lab for training of company staff members. This procedure does not usually work well for the company that plans to install only a few systems, however. Photo courtesy Penn Video.

Photo: Most successful MV system suppliers have gravitated toward niches within certain industries. For example, Perceptron Inc specializes in metrology in auto assembly plants. This photo shows a complete body inspection system; using over 100 cameras, it is reported to be the largest MV system installed to date.

Photo: Learning the relationships between types of applications (shown in circles) and the basic types of image-processing technologies will be part of your early indoctrination in MV. Note, for instance, that mathematical morphology, lower left, lends itself to inspection of finished surfaces, sub-pixel (very high resolution) gaging, and 3-D robot guidance. Applications such as part-defect detection and 3-D robot guidance (center) may call for any one of the four technologies, depending on the nature of the individual case. At lower center, the "SMD Inspection' stands for Surface Mounted Device Inspection, a procedure used in the manufacture of printed circuit boards. Drawing courtesy Machine Vision International, Ann Arbor, MI.

Photo: Inspection for foreign objects or unwanted features is a fairly common type of application for MV. In this case, a system using 2-D edge-detection technology examines metal standoffs for the presence of screws that have caught in the threads. At upper left is the initial scene; at upper right, the digitized version of the scene. Lower left, the standoffs are isolated; and lower right, the screws are highlighted so they can be found readily and disposed of. Photo by David Mittendorf, Arthur D Little Inc.
COPYRIGHT 1985 Nelson Publishing
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1985 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Quinlan, Joseph C.
Publication:Tooling & Production
Date:Jul 1, 1985
Previous Article:Installation of FMS seen as progression.
Next Article:A machine-tool family.

Related Articles
Machine-vision systems - what can they do for you?
Machine vision: reaching for maturity.
Focusing on vision.
Vision thing: machines provide it for control.
Machine vision focuses on quality.
Machine vision market celebrates growth in NA.
Machine vision sorts out poultry defects.
New seminars. (Informalia).
New standards in high-performance vision. (DVT).

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters