Digital Twins Deliver Better Machines Across Product Lifecycle
| By: KristinLewotsky,, Contributing Editor
Simulation has been an indispensable technology in machine design for decades. In recent years, mechatronics has streamlined and improved machine design by applying modeling and simulation in a cross-disciplinary framework to test design concepts and root out problems in the early stages. Each technology represents an advance over previous machine design and operation, but much of the process still involved trial and error. Today, the digital twin takes virtual representation to a new level by providing an evolving model of the machine that can be used to dramatically improve the design, construction, and operation of a machine throughout its entire lifecycle.
The digital twin goes beyond conventional modeling and simulation such as CAD and finite-element analysis. It includes aspects of manufacturing execution systems, product lifecycle management, predictive maintenance, and the industrial Internet of things. The results go far beyond the benefits of any of the techniques individually to provide an accurate virtual representation of equipment in the real world.
"A digital twin is a living replica of that thing that you're modeling so that as the physical system develops, the digital twin changes," says Dave Vasko, director of advanced technology, Rockwell Automation (Milwaukee, Wisconsin).
"I see the digital twin as the coming together of three data models: the traditional product-lifecycle management one, the business one, and the operational one," says Andrew Hughes, principal analyst at LNS Research (Antwerp, Belgium). The framework delivers powerful insights at all levels of the machine lifecycle. "Machine makers are looking for much more than just being able to simulate. They're looking for the ability to use the digital twin for fundamental design, for manufacturing process management, for the manufacturing itself. They want the ability to redesign or to adjust to new requirements. They want this ability for existing and new products or just to be more and more efficient in their manufacturing."
The technology has attracted significant interest already in the industry. In a recent LNS market survey, 75% of discrete manufacturing executives had a digital twins strategy in place at some level, anywhere from having a digital twins initiative in the budget to actively using the technology. The most common use cases included reducing manufacturing costs, reducing unplanned downtime, boosting throughput, increasing safety, and testing out new design ideas. Organizations anticipate spending between $500,000 to multiple millions of dollars on digital twins initiatives over the next five years.
From cradle to grave
The digital twin can be applied at a number of levels within an organization. Digital twins can be created for products, for processes, and for machines. In terms of machines, the digital twin concept can be applied to components like bearings, to assets like palletizers, or to entire systems such as packaging lines. For this article, we'll restrict the discussion to machines.
The digital twin is intended to provide a detailed representation of the form and function of a physical system throughout its lifecycle. It can be applied in:
- Fundamental design
- Virtual commissioning
- Customer support
"We think the maintain and operate phases may have the greatest value," says Vasko. "We think the greatest opportunity is taking information that was developed during the design phase and then using that in the other phases…in the installation and commissioning, in the operate and maintain phases.
The design phase
Previously, modeling at the design phase involved using CAD renderings to identify potential collisions and make sure that all the components would fit. The digital twin approach is far more extensive. It doesn't just reveal whether the physical model is viable. It makes it possible to determine whether the system actually works in motion, identifying collisions and other issues before the system is built. Because the work is model-based, teams can try out dozens of alternatives with minimal additional cost.
The commissioning phase
In traditional machine building, the physical system needs to be assembled before the controls engineers can exercise the equipment. The physical equipment can't be exercised until the frame is built, the electronics are added, and the controls architecture put in place. In this scenario, an issue with a linear guide can delay the entire project, delaying subsequent steps such as controls engineering. The controls engineers may be able to simulate the results of controls in advance, but that provides only approximate behavior.
Virtual commissioning makes it possible to perform many of these steps digitally, testing out trajectories and control architectures even as mechanical parts of the machine are being designed and constructed. The digital twin starts with the CAD model, augmenting it with real-world data to more accurately model the actual machine. At that point, the exact same software is installed on the digital twin as is used on the physical machine. This allows the controls engineers to test out different scenarios and gather data on performance before the physical machine is ever finished.
"In the virtual world, you can do a lot of planning, testing, and programming before you actually build that machine on the floor," says Zachary Gray, strategic market development manager for machine tool systems at Siemens Industry (Elk Grove Village, Illinois). "That doesn't mean that it's going to be completely perfect when you get to the floor, but you can do a lot of the work before then, which reduces time and effort for the controls engineering."
Gabe Manescu, senior applications engineer, Siemens Industry (Troy, Michigan), points to a customer that sought to reduce commissioning time by 30% via virtual commissioning. "The very first pilot project reduced commissioning time by 72% and the next one was 78%.”
The increase in speed highlights one of the benefits of the digital twin approach. A digital twin module constructed for one machine can be applied to the next, and augmented by additional modules for specific functions. The process continues with each machine built until the OEM has developed a library of modules that can be efficiently assembled into new machines.
The days of a dedicated machine churning out a single product 24/7 for years are drawing to a close. Manufacturers want to be able to produce multiple products with a single piece of equipment. Needs and designs change over time, and they need the fastest possible means of responding to demand. A "living" virtual model makes it possible to develop and test entirely new recipes and processes before the work is ever done with the physical machine. Engineering teams can try out different modifications and programs, troubleshooting any issues well in advance. The result is greater flexibility and faster changeovers with reduced downtime and lower risk overall.
One of the core benefits of the digital twin is that it doesn't just model the machine as designed and built. Augmented by a steady stream of sensor data from the physical machine, the virtual machine mirrors the evolving condition of the real machine. That makes it possible to predict to a very granular level when the actual machine needs maintenance. An array of predictive maintenance technologies already exists. The digital twin is not intended to supplant them but to take advantage of the granular input to help maximize operational equipment effectiveness (OEE).
|Figure 1: With the digital twin, operators can get the exact experience of running the machine without risk, enabling them to explore various features and options. (Photo courtesy of Siemens Industry)|
The approach offers other benefits over the lifecycle of a machine. Although the frame of a machine may last for decades, the electronics are typically upgraded every 10 years. Having a mature digital twin enables OEMs to explore options off-site before any purchasing decisions are made. “If you need to retrofit that machine in the future, you can test out your controls program without taking that machine off-line," says Gray. "Another use-case is variability with automation systems on that machine in the virtual world and not doing it on the shop floor.”
The benefits of this granular data go beyond just predictive maintenance. It can provide insights that can be used to modify future designs. Having the digital twin makes it possible to explore different options in a risk-free environment with the goal of improving performance and lifetime. "Predictive maintenance is good but the next logical step is to determine if a machine is failing all the time, why is it failing?" Vasko asks rhetorically. "Is that part undersized? Are we using it beyond specification? Is there a better way to design the control system to give it a softer start that would give it more longevity? You’re using information you've learned from the digital twin to not just predict when it's going to fail, but to reduce the probability of it failing."
Having an accurate physical model of the machine is useful for nonproduction applications, as well. A virtual machine can give a realistic experience to operators during training sessions, enabling them to explore different options and tasks without risk to the machine, product, or themselves (see figure 1).
In sales and support
|Digital twin of CNC machine shows the production of a part with full HMI display on the left-hand side.|
An accurate working model that can be exercised with different recipes and controls software is also useful for sales calls. A digital twin of a CNC machine would include the virtual numerical control kernel, for example. The customer could load the part program for a product into the digital twin of the machine tool and see exactly how the model would perform (see video).
Building the machine is just one part of the OEMs role. They typically also provide service and support to their customers. An accurate virtual model makes it possible to monitor equipment at a distance to assist with maintenance and fine-tuning performance. "The machine vendor would have access to the digital twin of the machine that is running at the customer site and be able to warn of potential problems," says Hughes. "They can also analyze asset performance using real-life data on a virtual machine. This can assist with introducing a new product or slightly new processes, and so on."
Finally, the digital twin can help with troubleshooting systems at a distance, for example in the case of collisions. Trying to replicate conditions with the real machine introduces the possibility of further damage. The virtual machine lets the engineers explore various scenarios to determine the source of the problem without endangering equipment or personnel. "You now have an environment where you can do that analysis without the risk of crashing that machine again," says Manescu. "It happened to me as an OEM a few times by trying to reproduce a crash, the process was so dynamic it was quite a big problem." In the machine tool world, for example, tool changes take place at high speed. In robotics, path is sometimes speed-dependent. "Testing at low speeds doesn't guarantee that the behavior will be identical at high speeds. That's why having this digital twin to check back on when something bad happens solves a lot of issues.”