Industrial Automation Insider: What You Need to Think About – and Be Prepared to Change – When Spec’ing Out a Machine Vision System

How do you know what you need from a machine vision system? Or if machine vision is even the right imaging technology for what you need to see? Do you need deep learning optical character recognition (OCR) or some other AI-centric vision tools? How do you ensure your machine vision system will work well with your robotics components or RFID or any other connected technologies on the line? And how do you get the lighting right? (Because it’s all about the lighting when trying to snap the perfect picture and keep your line operators comfortable.)

These are some of the questions I asked my guests on this episode of the Industrial Automation Insider. 

Nick Munger and Joel Nernberger are both application engineers for Tri-Phase Automation and spend their days supporting every aspect of vision-centric projects, from scope development and product selection to component integration, programming, and orchestration with other non-vision technologies on the line or shop floor.

They’ve come to recognize some best practices for scoping out and spec’ing out machine vision projects as well as some practices or ways of thinking that tend to complicate efforts and inhibit a return on investment (ROI). 

So, tune into this 20-minute episode now to learn everything you need to know and what you need to do when you’re trying to figure out which vision technology is going to help you accomplish your goals – including the top 3-4 things you must look at closely before deciding on system components.

You’ll also hear:

• Why it’s so important to uncover all the nuances of the application up front and conduct a feasibility study for the hardware and software components you’re leaning toward. 

• How the discovery and scoping phases typically unfold for machine vision (and other vision-based automation) projects. 

• What types of questions you need to ask – or expect to answer – when you call an automation engineer for assistance with a machine vision implementation. (There’s a lot of information you should come to the table with if you want to walk away with a well-designed solution and clear action plan for implementation. A well-articulated business problem is one of them. There are several other data points to jot down before you pick up the phone, though.)

• The importance of keeping an open mind as you go into – and through – the discovery and scoping/spec’ing phases of the project, even if you are a machine vision or automation expert and have used this type of technology before. (For example, you may not know you need to bolt on tools such as deep learning OCR/AI until the nuances of the application are well understood.

We also had some off-camera discussions that I thought would be helpful to you, so I’ve shared excerpts of those below:

How do you ensure everyone is accounting for the different human, technology, or process variables that play into business improvement projects?

Joel: For new or one-off systems, it is important to build in as much flexibility as possible and filter out as much potential noise as possible. For flexibility, that means going with higher power, fully licensed systems so you aren’t limited by what tools you have access to use and how fast you can process images. That also means having adjustable camera and lighting mounting brackets. For filtering out noise, that may mean adding polarizers and using specific wavelength lighting to minimize the effects of ambient light and changing products and surface finish. For OEM and repeat systems, we can dial in on the most cost-effective hardware, software and mounting. For the human element, adding an HMI display showing the vision results goes a long way to having operators and supervisors understand and work with the vision system vs potentially tripping up the vision inspections.

Nick: As Joel mentioned, having flexibility in the vision tools is so important to ensure that the job can be done. Project planning and avoiding those unfortunate “uh-oh” moments are why a scope of work is so important. We need to clearly define what the application requirements are, what technology we will use, and how we will try to limit the number of variables in the project. 

What types of questions should be asked during the discovery phase? How do you avoid a situation during the implementation or testing phase where you realize someone forgot to share an important piece of information that will impact system performance or the user experience?

Joel: Ask all the questions. If it isn’t spelled out and you don’t know, ask. Don’t make assumptions. The best practice is to have total understanding of the project as it relates to the vision inspection. That is very difficult to do for the customer, where they may not appreciate the complexities that can come with vision inspections, and it’s difficult and time-consuming for vendors to insert themselves into more project discussions to ensure they have a full idea of what is involved. We had an application where the customer had experience working with machine vision. We made assumptions about how they would execute the installation based on that prior experience, and they made assumptions about what the vision system was capable of. That resulted in some steps backward, and a longer road to the finish.

Nick: Yes, new information can surface that might impact system performance. Having a vision system with more capabilities than needed can help with this. It is also important to understand the project during the discovery phase. This ties back to having a quality scope of work and an effective feasibility test. It is important that we keep our customers happy by guiding them through what they need from the very start of the project.

We know many imaging systems are working in conjunction with RFID or scanning systems or sensors and certainly different execution systems to gather, analyze, share or action data – or maybe they should be working with those technologies if they aren’t today. Even if someone is coming to you only for an imaging system, are you considering what else may be in that environment either today or in the future to support the information flow or workflow? 

Nick: Future proofing our vision system is a conversation needed with each customer. Will the customer eventually add ERP or manufacturing execution system (MES) integration? Will there be additional input/output (I/O) needed for a new robot or external control? Are there new marks that will show up on a label that we will need to inspect? These are great questions to ask. That being said, automation is developing faster than it has previously, and there may be new technologies available that we may not be able to plan for.  At Tri-Phase Automation, we work with great vendors like Zebra that are consistently pushing for the best and newest technology. Incorporating this tech into our solutions is an important part of our business.

Plus, with Tri-Phase Automation being a value-added distributor with a very diverse line card, we are always on the lookout for additional ways to help our customers. We serve as an extension of our customers’ engineering teams and keep the systems-oriented and future-proof mindset during project development. It is rare to see only a Zebra camera installed – typically there are other automation components such as PLCs, HMIs, servos, and sensors that are needed to fully implement a Zebra system. We are also very competitive with control panels. We have a UL508A certified panel shop, a schematic design team and, of course, our engineers. By offering all these value-add services, we can bring a more complete solution to complement the Zebra vision system.

###

Editor’s Note: 

You can reach Joel, Nick, and other members of the Tri-Phase Automation team here if you’d like a consult or have technology questions.

There’s also a transcript of the full podcast episode available here, and you can download the MP3 version to listen offline later if you’d like: