End Users Need to be Smart about Integrator Selection

Setting up your first machine vision application can be a daunting task full of potholes. Do you keep the job in house, or go with an experienced integrator? And if outside help is warranted, how does the novice end user pick the right expert for the job?

The Automated Imaging Association, with help from its members, has put together a list of do’s and don’ts when it comes to deciding whether a project warrants outside help, and how to find that help if necessary.

Mitigating risk
Automated inspection systems are created to improve efficiency; however, the introduction of any complex solution – machine vision or other mechanical or process control system -- comes with some risk.

‘‘Machine vision is still a risky thing to do,’‘ said Brian Durand, executive officer of BD Automation (Burnsville, MN). ‘‘I’m a fan of machine vision, but there is still a fair amount of risk that the customer is taking on. We mitigate that risk through our experience…Yes, [a company] could probably figure it out on [their] own, given adequate time and resources…[but] is machine vision your core competency? Why not partner with a company that already has the experience to make your project successful?’‘

Durand’s comments address the first question an end user needs to ask: How much do I know about machine vision, lighting, optics and pattern matching algorithms? If the answer is little or none, then the end user faces two basic choices: a long ramp time with significant research along the way and the possibility that the system still will not function, or working with an experienced integrator.

This is not to say that going with an integrator eliminates the need for diligence and work on the end users’ part. Experts agree that the customer is every bit as important to a successful specification as the integrator.

Know your needs
The search for an integrator starts with groundwork needed to build a specification. If the end user follows certain steps designed to fully understand the needs of the application, then a good specification, integrator partnership and final solution are likely to follow.

‘‘End users, either with or without an integrator channel, need to do a proper analysis of the application. Start with an analysis of the company’s need for machine vision, the company’s capacity to utilize and support the technology and then come up with a functional specification,’‘ explained David Dechow, president of Aptura Machine Vision Solutions (Wixom, MI).

An analysis should include identifying a champion for the machine vision project within the company; the champion should pull together a project team that includes management, finance, union operators if applicable, maintenance, IT and engineering disciplines as required by the application. By identifying the company’s capacity in these various areas, the project lead can qualify the company’s estimated level of acceptance of a new technology. The project lead should identify the expected return on investment related to the machine vision system and the targeted manufacturing process.

‘‘After these steps,’‘ Dechow said, ‘‘the end user is ready to begin the specification.’‘

Know your product
System specifications start with the product and manufacturing process. End users need to fully understand all the types of defects associated with the product and have full understanding of the manufacturing process and all associated systems (e.g. PLCs, conveyors, etc.). ‘‘At a recent vision seminar, I asked how many people in the audience had worked with vision integrators before and most had,’‘ said BD Automation’s Durand. ‘‘Some had projects that went bad, and when I asked why, what they all told me was that it went wrong defining the pass/fail criteria and expectations to the vendor.’‘

End users should provide samples of the product and defects to all candidate integrators. When combined with the company’s background information, Durand said that an end user can start to qualify potential integrators by seeing if the return proposals include careful part evaluation that confirm the pass/fail criteria, careful definition of the project scope and what is included; an itemized list of responsibilities for each part; a complete description of mechanical, electrical and data interfaces to existing systems that make up the manufacturing process, a fixed price and complete schedule of deliverables.

After carefully considering the proposals of various integrators, the end user should ask the integrator to prove their technical skills through previous jobs and show operator and maintenance documentation from and demonstrate their long-term support of the project through references from existing clients, according to Aptura’s Dechow. Durand goes so far as to suggest that the integrator should have a fully outfitted demonstration and evaluation lab for testing the system and previewing sample components and systems for the application in question. Lastly, does the integrator have a plan in case the application requires engineering support beyond their abilities? This is often illustrated by the integrator's list of corporate partners, which should include as many software, hardware and third party engineering resources as possible.

How much experience is enough?
The flexibility of machine vision systems dictates that one RFP can generate many different types of solutions. In some cases, a newcomer to the field can determine which integrators seem to understand the application best by comparing the different solutions and focusing on those that are similar in tone. However, for challenging or new applications, sometimes it is best to think outside the box.

‘‘The ‘Best Case’ may be to find [an integrator] with specific experience in your industry, but experience also relates to how well the integrator understands the various technologies that go into machine vision,’‘ said Durand. ‘‘Maybe a certain algorithm applied by the semiconductor industry could be applied to an application in the food industry. Integrators familiar with the food industry exclusively might not be as open to new ideas, but again, they will know how to package cameras for wash down environments, for example.’‘

According to Dechow, ‘‘I’d stop short of asking if the integrator has done this application before.  What the end user must know is does the integrator have experience in the fundamental technologies involved in integration machine vision, not necessarily a deep knowledge of that exact application based on previous experience.’‘