Industry 4.0 isn’t anything new for the manufacturing engineers at Glidewell. The Newport, CA-based manufacturer of crowns, bridges and other dental products has been applying the concept since before it had a name.
“We are an extreme high-volume manufacturer, but everything we make is custom—a lot size of one,” says Dave Leeson, vice president of engineering at Glidewell. “We might make 100,000 unique products every week.
“The only possible way to pull that off is to have a deeply connected manufacturing system. All the teeth we make look similar, but they’re actually different, so just making sure the right tooth gets to the right person in a reasonable time frame is a massive job. Traceability is vital.”
As a result, all of Glidewell’s machines are interconnected. Every product carries a unique identification code that’s scanned at each step in the manufacturing process. At any given time, manufacturing personnel know exactly where each product is in the system and how well each production line is performing.
Industry 4.0 at Glidewell involves a lot more than just connectivity. Today, the company is applying a variety of cutting-edge technologies to improve its products and processes. For example, the company is employing artificial intelligence to create crowns and bridges that are a perfect match for each patient.
Glidewell’s process for manufacturing a dental prosthesis begins with a 3D scan of the patient’s teeth or a cast of the teeth. From there, the company leverages an artificial intelligence technique, known as generative adversarial networks (GANs), to create a perfect 3D model of the tooth to be made. Developed by Ian Goodfellow, a research scientist at Google Brain, GANs is a machine learning technique in which two neural networks contest with each other in a zero-sum game where one agent’s gain is another agent’s loss.
The technique gained notoriety as a tool for creating “deepfake” videos on the internet, but it can also be adapted to work with 3D data to customize production of physical products, a concept that Goodfellow has dubbed “GANufacturing.” Glidewell is the first company to use GANs to make better teeth. Dentists often spend considerable time and effort creating custom dental prostheses. Not only does a new prosthesis have to fit a 3D shape that works with the patient’s other teeth, but it also must work well with the overall pattern of the person’s bite. As a result, a prosthesis typically needs to be tested on a dental model and ground to fit. Through GANufacturing, Glidewell can generate a near-perfect, realistic and functional tooth that needs little or no post-processing.
And, because the company has made millions of prosthetics over the years, it has a wealth of data to feed into the software. “It’s one of the most advanced ways to generate 3D data for manufacturing,” says Leeson.
Glidewell is even using AI and machine vision to error-proof manual operations. “Sometimes, we need people to transfer parts from one process to another because they’re too difficult for a robot to handle,” says Leeson. “In those cases, we use AI-driven vision systems to track the movement of those parts to ensure they are not misplaced.”
Glidewell is also using quality control data to provide direct feedback into its manufacturing processes. The company inspects 100 percent of the products it makes with 3D geometric scanning. “We compare that data with the design data to make a pass-fail decision,” explains Leeson. “If the product fails the inspection, it will automatically return for additional machining without any need for human intervention.”
All the manufacturing data for every product Glidewell makes is stored in a database that can be accessed from anywhere—on web-based dashboards on the plant floor or on an engineer’s mobile phone. What’s more, engineers can analyze trend data to make adjustments to manufacturing processes as necessary. The company can trace problems to specific operators or specific machines or tools.
“We’re confident enough in our data that, if the system detects a pattern of failures, it will automatically take that machine off-line and generate an order for a maintenance technician to see what’s wrong,” says Leeson. “We are proactively preventing scrap.”
The company has so much data on its manufacturing processes that it knows exactly how long each process should take and when a part should be expected at a particular station. If a part fails to show up within the expected time limit, it’s automatically scrapped and a replacement is queued up at the start of the line.
“Things can happen on an automated line that you can’t predict—perhaps a robot has dropped a part,” says Leeson. “We don’t want a situation where a person has to track down what went wrong. We’re very concerned with on-time delivery, so sometimes it’s better to just make another one. Maybe that’s not the most efficient thing to do, but in our business, it’s necessary. Unlike other manufacturers, we can’t buffer our line.”
Medical Device Manufacturing 4.0
Glidewell is not alone in applying advanced digital technologies to its assembly lines. Many medical manufacturers are embracing digital transformation to increase efficiency, decrease lead times, become more agile, and meet regulatory requirements.
One such company is Zimmer Biomet, a manufacturer of artificial hips, knees and other orthopedic implants. Like Glidewell, the Warsaw, IN-based company is a high-volume manufacturer of highly customized products. And, like Glidewell, fast, on-time delivery is critical.
“We use Industry 4.0 technology to ensure the right parts get to the right place at the right time,” says Abhijit Balan Mepadan, project manager for Zimmer Biomet. “We use data to analyze our performance. When a customer places an order, we can predict how much time is needed to make that part. From there, we can backtrack to order the parts and raw materials to make sure we meet the expected delivery date.”
That capability was impossible when orders were managed on paper, he adds.
Digitization is also enabling the company to improve traceability. “If there’s ever a field issue with one of our products, we know where it was made, when it was made, who made it, which machines were used, and what inspections were done,” says Mepadan.
That’s no small thing for a company with manufacturing operations in more than 40 countries worldwide.
Now, the company is striving to make its manufacturing operations more proactive rather than reactive. “We want to create a more predictive environment where our systems talk to us automatically. What issues can we foresee in, say, the next three days? What data can help our managers make better decisions?” he says.
For one manufacturer of continuous glucose monitoring systems, Industry 4.0 played a key role in helping to cope with a massive increase in production volume. “We started out making 40 million units per year,” says Ram Bulusu, senior director of global information technology and operations technology for the company. “Today, it’s more like 200 million. Most of the systems we had in place were not really designed to handle that kind of volume.”
Industry 4.0 has provided myriad benefits for the company, such as optimizing product costs, reducing labor costs, increasing overall equipment effectiveness (OEE), improving quality, and getting new models out faster. “It’s made our operation more predictable, repeatable, scalable and flexible,” he says.
At Dentsply Sirona, a manufacturer of dental products in Charlotte, NC, Industry 4.0 is about connecting machines and gathering data to gain insights into the manufacturing process and to get ahead of problems before they occur. Dentsply has long been collecting data from sophisticated equipment like five-axis milling machines, but now the company is extending that concept to simpler machines, such as parts cleaning equipment.
“We’ve determined that water temperature is critical to cleaning our parts, so we use a digital thermometer to monitor water temperature over time,” explains Dan Ron, senior process engineer at Dentsply. “If we see the temperature is trending downward, that might be a sign that it’s time to replace the heater.”
Dentsply is collecting data on manual processes, too. “The biggest value of Industry 4.0 will be the ability to collect data from manual operations,” says Ron. “It’s easy to collect data from machines.”
The company began implementing digital work instructions several years ago. Now, every workstation is equipped with a computer connected to the company’s ERP system. When a project arrives at a workstation, workers scan a bar code to get step-by-step instructions on how accomplish the task at hand. This has eliminated errors and sped up production, since workers aren’t wasting time flipping through binders of instruction sets.
An additional benefit is that managers can now collect data on manual processes without having to visit the line with a stopwatch and clipboard. “Coming from a lean and Six Sigma background, cycle time is a big deal,” says Ron. “We need to know how long an operator will take to complete a task so we can better manage the workload of the entire facility. We can look at four operators performing the same task, and maybe one person is working faster than everyone else. Why is that? Maybe he’s cherry-picking, or maybe he’s figured out a way to do the task better that everyone else can learn from.”
It also helps the company comply with regulations. “In medical device manufacturing, it’s important to say what you do and do what you say,” says Ron. “Digital work instructions allow us to control our documentation.”
Getting to a point where manufacturing lines run themselves and artificial intelligence helps design parts doesn’t happen overnight. It takes considerable planning, leadership, and investment in people and technology. It also takes some incentive. Industry 4.0 is not for everyone.
“We’re more advanced than a lot of companies because the nature of our product has forced us to make that investment,” says Leeson. “Any company that embarks on that journey will need that same drive, too, because it’s not an easy road and it requires an awful lot of investment.”
Bulusu agrees. “The medical device industry can be slow to change,” he observes. “It often takes a crisis to for new ideas to be accepted. If everything is going well, management will not want to spend the money.”
Bulusu recalls an earlier stint at another medical device manufacturer. “Our OEE was 45 percent,” he says. “When we implemented Industry 4.0 technologies, we doubled our OEE to 90 percent. It took six months and a lot of collaboration, but when you can deliver success like that, management will be more receptive to new ideas.”
Assemblers that want to pursue Industry 4.0 are well-advised to unite disparate departments—design, manufacturing, IT—under common leadership, says Leeson. “We had to work closely together to make it happen, and, honestly, we stumbled quite a few times,” he admits. “Different teams had different priorities, and we didn’t succeed until all of those teams were put under common leadership with a common vision.
“To accomplish something like this requires a huge spectrum of skills,” he continues. “Our control engineering and electrical engineering folks had to learn a lot more about modern communication protocols so they could better understand the world of our IT people. And, of course, the reciprocal was also true. Our IT people needed to better understand the engineering world.”
Working together, Glidewell’s team determined things like which PLCs were best for connectivity and what hardware and software provided the best security.
The situation was the same at Bulusu's company. “At first, the automation team handled all the manufacturing; the IT team handled all the data; the quality team was doing its own analyses; and our suppliers were doing their own thing. It really wasn’t meshing well,” he recalls.
To create a true smart factory operation, Bulusu unified these groups under a single model he dubbed manufacturing and supply chain quality operations technology (MASCQOT). Bulusu’s concept expands on the ISA-95 standard for developing an automated interface between enterprise and control systems. But, whereas ISA-95 is a vertical model, MASCQOT is more horizontal. This has enabled his company to communicate more effectively with the manufacturing systems of its suppliers.
A major benefit of MASCQOT is that it enabled the company to integrate quality management into the system. At many medical device companies, manufacturing and quality are different functions, Bulusu explains. The two departments might share information, but it might not be timely or adequate enough to meet FDA reporting requirements or to improve manufacturing. With both quality and manufacturing part of one system, Bulusu's company has the ability to trace adverse events to potential manufacturing issues on a specific line. For example, if several people in a particular region complain of developing a rash from wearing a monitor, engineers might be able to trace the issue to, say, an out-of-spec batch of pressure-sensitive adhesive at one plant.
Communication and teamwork are important, agrees Ron. “Before implementing Industry 4.0 technology, work with your IT department first,” he counsels. “The last thing you want is to install something only to have the IT department tell you that it’s not allowed on the network.”
Mepadan advises assemblers to temper expectations for digitization. It’s a good idea to start with a small project and grow from there. “You shouldn’t expect a lot of automation from day one—that’s just not possible,” he warns. “You have to take it in small bites and slowly build toward the end goal.”
Manpower can be an issue. “The biggest challenge with implementing Industry 4.0 is the skill shortage,” says Bulusu. “To be successful, we need people with manufacturing experience, IT experience and operations technology experience.
“At most companies, these are different people and different organizations, and they don’t have transferable skills. At my company, I brought the IT and OT functions together under the same umbrella. Now, the people who are programming the PLC and SCADA are working alongside the people who run the ERP and MES systems. There’s no gap in knowledge or connectivity.”
Deciding how much data to collect is another issue to resolve. “In the beginning, we tried to be very economical with the data we collected,” recalls Leeson. “Initially, we only collected what we thought would be immediately useful.
“What would happen, however, is that someone would have an idea or a problem would arise, and we wouldn’t have the data to support it. Now, our philosophy is, let’s collect and store all the data we can from a process. We may not need it now, but we might want it in the future. To do meaningful work with AI, you need hundreds of thousands of data points.”
Even when manufacturers have all their ducks in a row, hardware and software may not be mature enough to accomplish everything they want.
“I work with Amazon, Google Cloud and Microsoft Azure, and I can tell you that these cloud offerings are very good, but they’re all designed for information management rather than real-time control systems,” Bulusu points out. “If you have machines running in Ireland and your data center is in Germany, you will have some latency. The cloud offerings have not yet caught up to the needs of manufacturers, but they are getting there.
“Manufacturing applications need improvement, too. Whether you look at Siemen’s Camstar or Rockwell’s FactoryTalk, they’re not yet fully cloud ready. By themselves, they’re great applications. But they need to be designed for the cloud in such a way that there’s no latency.”
For Bulusu, a surprising stumbling block is related to accounting. “The financial models that we have used historically to roll out significant manufacturing technologies are based on capital investment,” he says. “You buy so many servers, so many sensors, things like that. You invest in capital, you calculate depreciation and that’s how determine your ROI.
“However, with the cloud ecosystem, it’s based on monthly subscription fees. You pay so much to Amazon or Google to run things in their cloud. The problem with that is, it’s treated as operating expenses rather than capital investment, so your profit and loss statement is affected. Our financial models have to be updated to include that.”
Cybersecurity must also be addressed early on.
“I’m less concerned about sending information to and from the cloud than I am with operations technology,” says Bulusu. “SCADA systems and PLCs are designed to be stand-alone. They have not yet evolved to be networkable. If something needs to be updated, someone has to physically connect with that device.
“We have responded to that problem by creating a targeted OT cybersecurity program. Where we cannot put in technical controls, we are putting in procedural controls. It’s just not realistic to replace all the PLCs on your shop floor with devices that are cyber-ready.”
For Glidewell, the next step in its Industry 4.0 journey is to find a way to integrate customer feedback into its data stream. “We have all this customer feedback data, and we are trying to map that onto our manufacturing process data to improve our lines,” says Leeson.
The company’s initial work in this area has already paid dividends. Glidewell performs full dimensional inspections on 100 percent of the product it turns out. But, it didn’t always. “Initially, we scanned one in three crowns. It’s not a cheap process, and even one in three is a pretty significant sampling percentage,” says Leeson. “But, when we compared customer feedback data with our inspection data, we discovered that the rate of customer returns or complaints was slightly higher on the parts we didn’t scan than the ones we did.
“By merging the two data streams, we were able to justify the investment in more scanners for 100 percent inspection. For us, even a quarter percentage point increase in customer satisfaction is massive.”
At Bulusu's company, the goal is to use real-time analytics to alert manufacturing managers to potential problems before they become actual problems.
“We are working on ways to use real-time analytics to establish trend alarms,” explains Bulusu. “For example, we measure the thickness of an enzyme on our glucose sensor. Let’s say our measurements tell us that the thickness is off by 0.1 micron, but it’s still within our specification. Then, our measurements indicate that the next batch coming down the line is off by 0.2 micron, but again, it’s still within our specification. We would like to use real-time analytics to tell us, if this trend continues, we will be making out-of-spec product within six more runs; we need to make an adjustment.
“In a fully mature Industry 4.0 model, we would like to have a self-aware manufacturing system that diagnoses and resolves issues like that with little or no human intervention. We would like to have a system that says, ‘This part needs to be replaced soon, plan for some downtime next Tuesday.’ We’re not there yet, but that’s where Industry 4.0 can take you.”