Where do vision systems fit in the process?

Jan. 24, 2018
Whether it’s assisting with assembly or improving inspection, vision systems need to be maintained

A Control Design reader writes: As a production engineer at a medical device manufacturer, I have been tasked to investigate the use of vision systems to assist assembly and to improve our quality and inspection processes, but I'm feeling overwhelmed with the options. I'm considering throwing it over the fence to a system integrator or machine builder.

Where do I start adding vision systems to the variety of automation equipment assembling handheld medical devices, with lots of different parts and packaging, in our plant? What are some of the low-hanging-fruit applications that are easy to integrate to existing control systems and equipment? Additionally, how can I ensure the vision-sensor or vision-system software is supportable by internal personnel?


Start with a match

Full vision systems can be complicated and overwhelming. For those just getting started, a simple vision sensor with onboard display that doesn’t require a PC is a great option and often can be set up in a matter of minutes. These sensors are ideal for simple verification applications, especially where the components being verified are in small quantities. A device with integrated lighting makes it even easier to set up a basic vision inspection with no additional devices required.

Pattern-match inspections are a good starting point since they simply require capturing an image of a correct condition. When the sensor detects a part that doesn’t match, the sensor sends a fail output to the line, and the product is rejected. For example, a vision sensor can be used to verify the presence of a label on a product. Another good application is complete assembly verification. For example, a vision sensor can also be used to verify that all correct components are present in a completed medical assembly.

The next step after a vision sensor would be a smart camera, which usually requires a PC but provides additional capabilities such as feature measurement and flaw analysis. Look for a camera with intuitive software that can handle complicated inspections while still being user-friendly. Smart cameras are ideal for critical gauging applications—for example, verifying whether a needle is straight or bent—that require more precision and higher resolution than a vision sensor.

Mark Lampert / senior business development manager / Banner Engineering

Ask the right questions

Figuring out how to get started with your first machine-vision project can seem like a daunting task. Believe it or not, you’re already on your way simply by asking the right questions.

The first step is to invite a machine-vision expert in your area to your plant to identify the assembly processes, quality inspections and packaging operations where vision systems can add value. Automation-equipment distributors with in-house vision expertise and who can provide no-charge consulting and application feasibility can be a valuable resource in this initial plant survey. Together, you can evaluate potential applications according to several primary criteria:

  • Where are your pain points? In other words, which processes are most costly to you in terms of material scrap, slow throughput, machine downtime, low rate of tool utilization and regulatory and safety compliance?
  • What are the metrics you use to measure your manufacturing quality, efficiency and cost effectiveness? What are your goals for improving in these areas? For example, you probably have goals to decrease device assembly errors or to increase line speeds by some percentage.
  • Which potential vision applications are relatively simple, and which are complex?
  • Can you retrofit machine vision on existing machines? Will you need new automated handling equipment and mechanical fixtures?

The answers to these questions will help you estimate the return on investment for each potential vision project. Of course, the ideal first vision application yields the best ROI in the shortest period of time with the equipment you already have and for the lowest cost. Fortunately, these applications are often the simplest, most affordable and easiest to support—the low-hanging fruit—and can be solved with a fully integrated and contained point-and-click vision sensor.

Checking the presence or absence of a part in an assembly is one example of a simple application that can be configured and deployed relatively quickly. Another example is using vision-based 1D and 2D barcode readers to comply with unique-device-identification (UDI) traceability requirements, even for shiny metal parts that are common to medical devices.

After you have that first successful machine-vision installation under your belt and your automation engineers, line operators and management see the value that machine vision can deliver on your plant floor, you’re ready to move on to the next application.

For more complex applications, such as in-line precision measurement, verifying adhesive and sealant coverage or inspecting part surfaces for defects, you may need to engage the services of a system integrator or machine builder. Choose an automation supplier with deep experience in developing and installing systems sold by the leading machine-vision companies. These companies have a wide range of vision systems to fit nearly any performance and cost requirement, as well as worldwide field engineering and technical support standing behind the integrator to provide additional assistance, if needed.

Robb Robles / principal product manager, Vision Products Business Unit / Cognex

Get up to speed

Vision systems are complicated; lighting can be the first thing to cause problems with vision systems, and then there is the pixel count per image size. I would suggest looking for the quick-success project, where you are simplifying the data entry using barcode scanning for quality control, and work your way up to quality inspection of parts before or after the assembly process. Most vision systems require some training from the vendor. If you use an integrator, they will set up the system so that you will have to call them in to make adjustments. Go directly to the vendor and get the training, and then set up a one-month on-site assistance from the vendor for the first project. Then arrange periodic review visits with the vendor every six months. I have seen people come up to speed quite quickly with Cognex vision systems, and your alternate would be Keyence. I have also used Teledyne Dalsa, and an untrained person could operate and get an image on their own without training in less than a couple months on the Dalsa.

David Arens / senior project engineerr / Fresenius Kabi

Aid to visual

It’s assumed that every step in the process adds cost and that any improperly executed step or defective part will require scrapping or rework.

Since management has suggested that vision might be worth looking into, it’s a guess that there are escapes that cause scrap or rework. Making a chart of the type of escape against the cost should help to create a prioritized list of areas that need addressed. This could be incoming inspection, final inspection or intermediate inspection steps along they way.

The low-hanging fruit may not be the highest priority item; a simple inspection task may be best to ease into machine vision. There are many companies that sell smart cameras that can be trained on a good part or have a simple interface for programming using icons.

Capturing an image suitable for automated inspection may require some skill with lenses and lighting.

The only way to assure that your vision system is supported by internal personnel is to have a person who is interested in machine vision assigned to the task and make sure that person is trained.

The Advanced Imaging Association (AIA, www.visiononline.org) is a global association for vision information and has a wealth of resources including webcasts. It also hase conferences around the world, including the upcoming Vision Show, April 10-12, where a wide range of classes will be offered, as well as a trade show where you can interact directly with vendors. I will be teaching one of those classes, and my company is a member of AIA.

Using machine vision to “assist in assembly” is a little tricky because that implies real-time interaction between an image, the part and the assembler. This would be called “aid to visual” and could be as simple as a camera that shows the proper image up on a large screen.

In some cases, aid to visual can become augmented reality, where a projection system, or in some cases glasses, helps to identify assembly order and correct part placement.

Working with a system integrator is a great way to get your inspection needs met without relying on in-house talent; however, the system integrator needs to have experience and competency in your type of inspection and needs to be physically close enough to swing by for updates, software tweaks and hardware repair or adjustments.

If you have an individual who is to be the expert in machine vision, you could ask the system integrator to train that individual, not just in the operation of the system, but the planning and implementation process front to back.

Robert Tait / partner / Optical Metrology Solutions

Separate vision logic

Those of us who have been in the industry a long while like to joke around that machine vision is an exact art and subtle science. Successful machine vision integration is a careful blend of lighting/optics, controls and software development. Much like I wouldn’t know my away around building a hot rod or where to begin, digging in and getting your hands dirty is the only way to gain experience.

Having an educational background in racing engine design can take your hot rod from good to the set of “The Fast and the Furious.” The same can be said with having an educational background in optics. But you don’t need a degree in optics to get started. The AIA has put together a collection of courses that are available online for free, and they’re a great place to start to gain context of what is involved and the science behind putting together a vision system.

Pretty much all lighting, optics and camera vendors will allow for a 30-day evaluation on pieces of equipment, which at times can take a couple of weeks to come in. When time is of the essence, it’s best to work with an integrator that specializes in machine vision versus a machine builder, since they are more likely to have a well-stocked lab of lights, optics and cameras. With the expertise, an evaluation can often be turned around in a short period of time. As well, with the experience of installing many varieties of systems under various conditions, they will know what external factors can impact a system’s success.

Established machine-vision integrators have an engineering process, provide documentation and segment the design so that the vision logic is separated from the rest of the user interface, camera acquisition and communications. Understanding the underlying vision tools and the logic flow of the vision inspection, it is possible for a vision system to be maintainable from in-plant resources—much like debugging PLC ladder logic is maintainable from in-plant resources, provided that those resources have received training in the software package, they have easy access to a laptop with the software installed, and the logic has not been locked out by the integrator.

The degree of success of internal resources maintaining the system will require training time on the software away from the facility and having a laptop that can be taken home to play with outside of normal working hours. Depending on the vision-tool platform, formal training can take 2-5 days. Some vision-software vendors have free online training videos, as well; most vision integrators have structured hands-on training classes where you come out to the integrator’s facility. Some vision platforms may also require a separate license to run the machine-vision development environment when not attached to a camera.

My recommendation on your first vision project:

  • work with a regional machine-vision integrator, so that you are getting the best support possible
  • quiz them on their engineering process
  • inquire about their vision lab
  • get a sample of their documentation
  • bundle software training in the turnkey quote; this way it’s in the project’s budget
  • verify that you will have open access to the vision logic
  • have an off-line image playback system where archived images can be used to play around in the vision logic.

Robert T. Couture / CTO and founder / 4th Vector Technologies

Justify the cost

Oftentimes, what is installed is a vision system to catch a bad part or assembly prior to its being packed up and shipped to a customer. There are customer-satisfaction and cost issues there, and sometimes fines, as well. A vision system is requested to catch the bad part just prior to packaging in order for the part to be fixed and reinserted into the production line or discarded. However, in this approach, albeit a solution the customer benefits from, the manufacturer is still spending lots of money to identify, fix and reinsert the part a second or third time. We have been asked to participate in applications with similar demand requirements, but our approach is to back up in the production line where the problem occurred and identify a critical identifying feature that may lead us to the best solution; which is to eliminate the failure altogether.

Think of a syringe-filling operation, for instance. Let’s say a label is being applied to the syringe, and, as the label is wrapped around the tube, the overlap of the label distorts or covers over certain pharma code information which is critical for FDA approval. This information can not be distorted; it must be accurate. So, you add a vision system to identify an incomplete code. Now the customer no longer receives distorted information, and the FDA requirements are in compliance. But the cost of this solution is enormous. The vision system itself can cost thousands of dollars, and the time to program and code the system for all variability can be a long process, as well.

However, if you simply can ensure the printing of the label was accurate enough to land the code in an area of the label which cannot be distorted or covered by the label overwrap, you’ve solved the problem completely. Photoelectric sensors, which can be as little as $100 or as much as $600, can ensure the accuracy of your print-and-apply registration and remove the failure from the process completely. No more cost to the process, except to pick the right sensor for the job—one that is accurate and repeatable. Once we’ve shown the customer this approach, the need to consult with vision-system integrators is reduced.

Don’t get me wrong. There are legitimate needs for vision systems. Product variability that cannot be controlled or manipulated by mechanical or electronic means is often the overriding factor to the vision-system requirement.

But, when you are thinking about light, which is what enables the vision system to see the difference between a bad part and a good part, then a simple or smart photoelectric sensor may also be considered.

Another example: A small over-the-counter (OTC) bottle needs to be inside a small box. There are many ways to determine if the bottle is in the box. Look at it overhead during the insertion process with a camera. Weigh it along the path of travel with an expensive scale. Or, with a powerful photoelectric-sensor system, you can light up those paper boxes with infrared light and immediately detect the shadow of the bottle inside. This can be accomplished at insertion or any other place down the line. Again, this solution would be under $250 and removes the need for some kind of wasteful solution that actually adds cost to their process every time a bad part is identified.

This is another cost issue with vision systems. Do I hire a vision tech, an integrator or a consultant? Can my current staff engineers and technicians understand and maintain the equipment long-term? The brand of vision system or camera you employ for the solution will greatly impact the ease-of-use issue. There are very many to choose from, and they all have their differences. Some are fantastic technologies, but are difficult to maintain and use, and they are normally the expensive ones. There are some that aren’t so smart, but are easy to use and maintain; but give them future variability, and you now have to buy all the upgrades.

Photoelectric sensors, if they can be used to solve these types of application requirements, are much easier to use and maintain, and most of your maintenance techs and engineers have a lot of sensor experience. This should not be a concern, if you utilize a more straightforward technology.

Tim Kelley / vice president of U.S. sales / Tri-Tronics

Let the integrator handle it

The best place to start is with a professional integration company that specializes in machine vision. They will have the knowledge and experience across multiple platforms and vendors to provide you the most suitable solution. They can also educate you and bring you along so you can try future projects with less support.

An experienced integrator will walk the prospective customer through the very beginning of vision and vision inspection. They will explain how vision works how it could be utilized to fit their specific needs. A plant tour would allow the integrator to identify those low-hanging fruit at that point.

Charlie Mayton / director of vision and operations / Global Controls

Straightforward tools

Vision inspection can be used in almost every step of the medical-device assembly process, from incoming inspection of materials, parts pre-assembly, during assembly and post-assembly to packaging inspection. This includes a wide range of applications, such as shape precision measurement, orientation check, presence/absence, parts placement, surface scratches inspection, barcode reading, OCR and packaging inspection.

There are many factors to be taken into consideration before making a decision about where to add vision inspection and what objective you want to achieve. It is important to identify the bottleneck in the manufacturing process first for both quality and productivity perspectives. At this time a vision solution professional can offer expertise in determining the application requirements to meet both technical and financial objectives.

Low-hanging-fruit applications can differ from one scenario to another depending on factors such as complexity of the original assembly lines, existing communication devices and protocols, working space constraints and of course the budget.

In most cases the inspections before and after assembly are relatively simpler and more straightforward since these inspections, such as presence/absence, measurement, 1D/2D/OCR reading or packaging inspection, can be independent. Any inspection interacting with parts assembly is often more complicated.

It is critical to choose vision software that is easy to use with a straightforward user interface, designed for quick and easy deployment and equipped with powerful vision tools and communication protocols. Operators can master it quickly and easily with little on-site maintenance needed in the future.

Steve Zhu / director of sales / Teledyne Dalsa

About the author: Mike Bacidore
About the Author

Mike Bacidore | Editor in Chief

Mike Bacidore is chief editor of Control Design and has been an integral part of the Endeavor Business Media editorial team since 2007. Previously, he was editorial director at Hughes Communications and a portfolio manager of the human resources and labor law areas at Wolters Kluwer. Bacidore holds a BA from the University of Illinois and an MBA from Lake Forest Graduate School of Management. He is an award-winning columnist, earning multiple regional and national awards from the American Society of Business Publication Editors. He may be reached at [email protected]