Guidelines When Undertaking a Machine Vision Project

While one can find things to consider when undertaking a machine vision project in my book ‘‘Understanding and Applying Machine Vision’‘ published by Marcel Dekker and available at a discount through the AIA’s web site, in the course of the past year or so I have been writing a series of articles based on email ‘‘interviews’‘ with representatives of machine vision companies typically participating in one or another industry-specific solutions. In each case I have asked the specific question: What advice would you give to a company investigating the purchase of a machine vision system?

What follows is a distillation of their comments, which should serve as a guideline of what to do and what to consider as one deploys machine vision.

1. Develop an understanding of the basics of machine vision. Regardless of the manufacturer that you elect to work with, this will help you to become familiar with many of the general terms and techniques used. Customers must understand the properties of the different vision tools to assure that the system purchased fits the application.

2. Companies should look for an integrated solution that addresses the application. Working with companies offering turnkey systems with a track record related to the specific application pose the least amount of risk. This will make the first experience more rewarding.

3. If the application does not involve a ‘‘catalog solution’‘ choose an important but fairly easy first application.

4. Do not spend all of the lead-time trying to get the very lowest capital equipment price but instead spend the time putting together the strongest team possible. The technology will only be as good as your people make it.  Your people will either make it wildly successful, mediocre or a failure.  Success always boils down to the capabilities and commitment of the people at the site.  So, for every machine vision project appoint a dedicated project team of one mechanical, one electrical and one supervisor.  Make it their job to have a wildly successful project. Get as many people involved that are cognizant of the requirements of the application. This might include line operators, line supervisors, quality engineers, process engineers, etc.

5. Clearly define how the system will be used and by whom; make sure your organization understands the commitment in people, training, and processes required to get value out of the system; ideally plan the vision system at the same time the manufacturing process is being developed.

6. Before even communicating with a machine vision vendor, if one is not purchasing a ‘‘catalog solution’‘ a comprehensive functional specification should be prepared.

7. Take the time to research the basics of how communications, interfacing, integration, etc. will work.

8. In preparing a functional specification a good Team is important. Drawing on the collective experience of all team members should lead to a comprehensive understanding of the application, especially the specific and measurable variations in your process. It is critical to know the total variation if you are considering an automated solution as a retrofit to your current process. It is important to analyse in depth the complete production quality requirements. Specifications should reflect an understanding of needs – reflect the complexity of parts, not only shape complexity but also reflectance complexities.  They must also reflect shop floor conditions and potential issues such as vibration isolation. This has to be reflected in the specification so that all prospective vendors know all the issues surrounding the application and recognize what measures are required to validate system performance at run-off time; otherwise the end-user and integrator will face a longer implementation and acceptance time in the plant where both lose profitability.
a. While the specification should be as detailed as possible it should also be realistic.
b. Full production workflow analysis is important too. Machine vision-based solutions can be integrated at several points along the production process.

9. Don’t buy more than you need.  We see numerous customers with a performance list (and/or wish list) that is far in excess of their primary applications.  All this excess can greatly impact your price, cost of ownership and the ROI. Of course, one must think of future application needs, but see if the platform is upgradeable to accommodate future improvements that will cost less at a later date.

10. Considering a portfolio solution can improve the cost of ownership for the customer.  Be sure to include the cost of ownership and system flexibility in your considerations. These often impact the bottom line more than feature-based investigation of machine vision solutions.   Also, a realistic analysis of the cost of passing along defects instead of finding and fixing them as early as possible will demonstrate the economics of a machine vision system.

11. Make a global implementation plan, select your entry points in accordance with short term goals working to your long-term goals.

12. Understand where the application success lies. Is it in the vision system, the robot application (or other complementary equipment) or both?  Consider support from the company you are purchasing your system from. How much support can they really supply and for how long? If you are buying the vision system and the robot (or other complementary equipment) separately, who then will be responsible for the overall success of the project? 

13. If you already have worked on previous machine vision applications with a system-engineering firm and have been pleased with the results, it might make sense to involve them as a member of your team early in the specification process to insure a suitable solution is delivered as the final product.

14. Where no company can be identified with a track record in the specific solution required, identify suppliers with at least some experience addressing some of the generic aspects of the application. In other words, if the application is surface inspection for blemishes, look for a company that has delivered systems to address surface inspection applications especially in the materials you are using and ideally with a range of similar size and shaped parts. This will help with pricing and deployment since suppliers with a track record in this area will be able to give good prices and deliveries.

15. Give prospective vendors as much sample product for evaluation as they will tolerate. Get to know the prospective suppliers: visit their facility; meet their engineers; see their vision lab; learn about similar projects they have completed.

16. Find someone with experience. Experience is more than knowledge of the tools. It includes experience in vision applications like specialized lighting and also successful project development and implementations. A supplier that can provide a formal engineering development process, with formal specifications, design reviews, testing, change management, documentation and support structure is valuable. It is important to get a demonstration of their ability with your product or a similar one, even if you have to pay for it. It helps prove the technology and test-drives the company relationships.

17. Try to eliminate the technical risk up front. Ask prospective vendors for free pre-sale demos that cover the worst-case range of variation in the product, lighting, and presentation to cameras. Always compare scanned raw images from different systems, look for a system with highest quality and resolution of the images.

18. Selection of the vision supplier is a critical step – look at the capabilities of the supplier, including depth of experience in vision technology, experience in related applications, knowledge of the true total cost of integration and ability to respond quickly to specific needs.

19. Look for a supplier that provides engineered solutions as opposed to an integrator. A good solution requires:
a. Complete understanding of the application including significant testing of samples.
b. Selecting the correct illumination approach.
c. Optimum selection of cameras.

20. Deal with a company in whom you have confidence of their longevity and commitment. The companies having the flashiest advertisements do not necessarily provide the best technology. Buy from a company that is specialized in manufacturing machine vision systems instead of ‘‘department stores’‘ that offer everything from material handling, etc. - a manufacturer cannot be best in everything.

21. Look at the supplier closely. What level of support and engineering capability do they bring to your needs? Do they understand your process? Have they worked in your industry before? Can they meet the needs of an entire organization or are they just a local or regional supplier? Also, do not be fooled by magazine ads for low price. What costs will you internally have to absorb to make you application successful? These are often hidden costs and if the personnel in the plant change then it usually means the application will suffer.

22. Select a company that can provide a total vision system, one that includes all of the handling, vision, lighting, reject, and measurement capabilities. Also look for a company that develops its own adaptable software rather than relying on limited canned solutions, and finally, look for a company who has the ability to provide the after sale technical service.

23. Establish validation procedures to assure the machine system is really meeting your specification of inspection: speed, set-up time, accuracy, etc.

24. Confirm that the machine vision company will provide the service and support required to insure ongoing performance. Is the machine vision company only interested in selling hardware, or will they become your partner for the current and future requirements? Ask for references and confirm the quality of the company and its personnel and products. Demand access to the installation where he has delivered his technology and learn as much as you can about what went right/what went wrong.  Get your ‘‘techie’‘ to talk with their ‘‘techie’‘.  Machine vision is a partnership.  It’s not a simple plug-in. It’s an ever expanding and changing entity. Get a written money back agreement guaranteeing that the system can be returned if the specifications are not achieved.

25. Questions to as prospective suppliers:
a. How long he expects to support the technology?
b. What kind of service is available?
c. Does the vision supplier support his own installations or hire outside contractors to do the servicing?
d. Is the system expandable and do I get new software upgrades as they are introduced?  At what cost?
e. Look at the financial stability of the company. How long have they been in business, are they growing and profitable?

26. Make certain the vendor has displayed commitment to the market.

27. It is important to make sure that the system can grow together with changing requirements. Look for machines with upgrade capability since the sensing technology will improve all the time. When PC’s replaced microprocessors, many applications were no longer supported (how many gas stations still offer leaded gas; how many video stores provide BETA videos; and have you tried to buy a stylus for your hi-fi lately?). The point is that technology advances and a good vision system will accommodate that.

28. Always go with 100% inspection as anything else is just sampling and will not guarantee the quality of your product.

29. Think about your ROI when investigating these systems. Although they initially cost more the ROI can be quick.

30. Make sure you have qualified people on staff.  Once you have a little machine vision, you want more.  Same principle as installing a LAN.  There, too, you need a network supervisor.  So, with machine vision, you need a person who can follow the technology and demand the most from his relationship with the vendor.

31. Don’t buy on price.  Although there are some basic, less sophisticated applications for machine vision, generally speaking, the more advanced the technology, the greater the capabilities.  A sophisticated vision platform will allow for expansion, whereas a simple defect recognition system probably won’t.  The simple system may do the job today, but may be obsolete in two years.

32. Buy a service contract.  When the vendor has a sufficient number of vision products installed in a given region, it is much cheaper to do periodic tune-ups, upgrades, etc. when he can send his maintenance people on one trip to do multiple jobs.  Without a service contract, you’ll be paying loaded costs every time you call for service.  Most vendors offer reasonably priced service contracts that provide peace of mind.

33. Value ease-of-use (GUI)

My thanks to all those who have contributed to my articles over the last year and, therefore, indirectly to this article.