Integrators agree that the machine vision integration process is not a simple one, but it can be easier if customers have done their research and understand the process.
Calling for reinforcements can speed the integration process along, but projects still take time, and it is important not to underestimate a project’s complexity. Ned Lecky, owner of Lecky Integration (Little Falls, NY), likens it to a project in which someone says they have developed a 300-mile-per-gallon car. “If the only important part is the engine and that’s the part you have problems with, it is not a simple problem,” Lecky points out. If the engine does not work, the car will not be going anywhere, let alone 300 miles on a gallon of gas.
The same goes for vision systems. What may look like a simple DIY project to the manufacturer may turn out to be much more complicated and require an integrator’s outside assistance. And with project costs that can range from $10,000 well into six figures, success is not something to be taken lightly.
Integrator SelectionEven if the end user has worked on one or two integration projects, it may be preferable to trust the vision system to someone who has done 100 of them. That’s where an integrator or consultant comes in. Perry West, president of Automated Vision Systems Inc. (San Jose, CA), works with systems integrators and clients to help the machine vision project go smoothly. He has end user clients that work with system integrators as well as clients who are system integrators, and he advises his clients to consider several criteria before selecting an integrator. Because circumstances vary, he says each criterion should be weighted depending on the situation.
To start, West says customers should consider if the systems integrators’ technology is an appropriate fit for a particular application. “Systems integrators typically work with a number of select principles they know well,” West says. “If they go outside that, they are just another smart person trying to solve a problem.”
Consider geography as well: is the integrator local? There are people on both sides of the issue; West suggests giving it some weight but still recommends picking a highly qualified person vs. a local one.
Another aspect to consider is continuing support.
Certainly other elements are important. Does the systems integrator have a sound business? Usually a company’s purchasing department can gather some data that will help decide that.
Along with a sound business and machine vision expertise, Lecky says it is important to have a personal fit as well.
“From a customer perspective, it’s kind of like going to the doctor,” Lecky says. “You should select someone you are comfortable with personally and professionally.”
Lecky also suggests talking to the integrator’s previous customers and asking, “If you were going to do another project, would you work with this person again?”
It also is a good idea to have some basic understanding of how the system works. Customers should know what is going on and need to be prepared for the machine vision process. It may be difficult to discern the best person for the job if customers have no experience with machine vision. “[Integrators] all promise the moon and stars, and only some can deliver,” says Leonard Ginsburg, a systems integrator and president and chief engineer of Production Robotics Inc. (San Leandro, CA).
Success Is...Once the integrator is on board, West recommends defining the project with a statement that begins “Success is …” and can be completed in 25 words or less. This information keeps both customers and integrators on track. “The CEO wants to know what this system is going to do for me,” he says. “They don’t want an overabundance of information. What I find that statement does is it keeps everyone focused on what the project goals are and prevents expectation creep, the real bane of all systems integration projects.”
West will help customers provide three things that are important to give to a systems integrator to start: a good specification, an acceptance test procedure and a good set of samples that the integrator can use to develop the application.
The ProcessAfter the research is done and an integrator-and possibly a consultant-is hired, what next? For the best results, experts recommend treating the integrator as an extension of the internal team-rather than the enemy. After all, both parties want the project to succeed and it has a better chance of doing so if the project is supported on all sides.
The relationship is an important one and working with people will create more variables than the technology. If programmed correctly, the machines will behave in a predictable fashion, but people are much less straightforward, Lecky points out.
When working with an integrator, an open relationship can lead to less delays or production problems. If something deviates from the outcome expected at the start of the process, it is much easier on everyone involved if the integrator is comfortable discussing the problem with the client-instead of discovering a problem and ignoring it for two months.
Customers want the project to be successful and under budget, so it makes sense to do everything possible to help the integrator along the way. “All this technology stuff is great,” Lecky says, “but at the end of the day, it’s people working together. People have a broad range of behavior and technology is only one small part.”
TroubleshootingEven with the most careful specification, selection of technology and research, machine vision projects, like life, are fraught with unknown variables. In that case, honest communication can prevent small issues from ballooning into big ones or break down a big problem into smaller ones.
To be truly honest, every application in machine vision has some component of wishing an element had gone smoother, says David Dechow, president of Aptúra Machine Vision Solutions LLC (Lansing, MI).
“I’m willing to go out on a limb and say that I don’t think any application goes absolutely worry-free with absolutely no glitches,” Dechow says, though the issues may simply be minor ones.
Fred Glover, president of Visicon Inc. (Los Gatos, CA), an independent system integrator business, also has seen his share of machine vision problems. “It’s easy for both the end user and the systems integrator to underestimate the front end, the lighting and the optics, part placement and part handling. That whole end is frequently under-engineered,” Glover says.
Right the First TimeEnd users should be sure they can describe the problem to the integrator and determine specific goals to hit to be successful. A goal that “everything will be running much better in the factory” is too nebulous to nail down, Lecky says. Instead, it would be better to say: the machine will run for eight hours and will inspect X number of parts or this many o-rings, and make no more than three mistakes.
Dechow says he often hears customers say, “We would like to put a camera here on our production line.”
However, a camera is the last step, Dechow says, and first the integrator will need to analyze the application. Dechow will instead start over and ask, “What do you want to do, and why do you want a camera?”
Post-ProjectAfter every project, it is a good idea to sit down with the integrator, with the goal to review any potential issues that came up during the project-and there always are some-and learn from them.
When machine vision first came out, West says it was useful only in discrete manufacturing, where parts were relatively well-fixtured. Today, the technology allows for much more flexibility in placing the machine vision components.
Along with improving technology, the relationship between integrator and end user should be carefully built.
With better education today, both sides have a little better trust and the working relationship has improved. “The road is not smooth,” West says, “but the potholes are not as deep as they used to be.” V&S
Ask the Right QuestionsJust as Ned Lecky, owner of Lecky Integration (Little Falls, NY), remembers projects that went smoothly, he also recalls those that did not.
The customer’s system was designed to look at one circuit board, and the company was having problems differentiating between the orange and pink part on the board. Is there a part on the board? Yes or no? If so, is it orange or pink?
It all seemed to work pretty well, and Lecky says his company priced the project so that they would break even on the first machine, but they would make money on the other nine machines they would need.
Unfortunately, the machine did not work very well. There was a lot of variation of orange in the orange parts, and variations of pink for the pink ones, so much so that even people couldn’t tell them apart. “My failure was not insisting on a complete range of parts to differentiate,” Lecky says.
Instead, the system should have been looking at part numbers, but they never went down that road. When the customer described the need to differentiate between orange and pink, he should have asked them to describe the application at a higher level.
They went back and forth on the issue, and by then, the factory was not really making those parts and they did not need the machine.
Despite investing a lot of time and resources on the project, the company simply paid them and donated the machine to a local university.
For more information on the companies mentioned in this article, visit
Aptura Machine Vision Solutions
Automated Vision Systems Inc.
Production Robotics Inc.