AI-Powered Digital Twins Raise Hopes for Better Predictive Maintenance
Machine learning and data science add new expectations and complexities.
Hexagon digital twins of an infrastructure project as point cloud models. Image courtesy of Hexagon.
December 3, 2024
As digital counterparts of physical systems, processes and equipment in the field, digital twins rely on a steady stream of data—such as machine uptime and downtime, equipment failure and maintenance records—to replicate the real-world conditions of their physical twins. Without this data link, a digital twin cannot accurately predict the pending failures of its physical twin and issue timely warnings. Therefore, as digital twin adoption rises, users are also confronted with the challenge of establishing a data thread that ties the digital to the physical.
“The significance of digital threads will grow immensely,” says Christian Kehrer, business development director, Altair. “[Since they are] already making an impact in optimizing product development, businesses will increasingly use digital twins to monitor real-time product performance.”
In DE 24/7’s “Technology Outlook 2024” survey, when we asked 220 respondents to identify technologies that they believed would have the biggest impact on product design and development over the next 5 years, 23% of respondents picked digital twins, and 64% checked artificial intelligence (AI). For their understanding of digital twins, 58% of the respondents said they knew what it was, an encouraging sign. However, 24% said “no,” and 18% said “unsure,” indicating the need to raise awareness.
The Role of AI
Keshav Sundaresh, senior director, digital transformation, Altair, expects to see “the next generation of digital twins, which are more human and user-centric, offering a dynamic and interactive experience.” The key to his vision is a live data thread, enabled by the sensors that are monitoring and recording the real-world conditions of the machines and equipment.
“With the advancement of IoT [internet of things] sensors capturing real-time data, engineers can now compare live data with synthetic models, resulting in more accurate insights and intelligent outcomes. This fusion of live and synthetic data allows for deeper analysis and improved product performance, making digital twins more intuitive and effective in real-world applications,” says Sundaresh.
“AI reduces the time and cost in keeping a digital twin up to date,” says Johannes Maunz, vice president of AI at Hexagon. “For example, it takes 20 minutes to fly a drone over a construction site and then post-process the data. Previously, this would have taken hours and hours.”
The ease of data collection made possible by drones and sensors is a blessing and a curse. The downside is, the amount of data available for analysis is too voluminous and far more granular than what can be processed with human intelligence. This is why, among digital twin adopters, you can expect to see a corresponding rise in the adoption of AI-powered data parsing and analysis.
“AI enhances predictive maintenance and digital twins in several ways, most notably by its ability to improve data analysis, optimize processes and provide actionable insights, ultimately leading to increased operational efficiency and decreased costs,” says Dale Tutt, vice president of Industry Strategy, Siemens Digital Industries Software. AI also reduces the complexity in deploying a digital twin.”
According to the industrial technology provider Hexagon’s “Digital Twin Industry Report,” capturing the digital-twin maturity survey responses of more than 650 executives, “The vast majority (80%) of leaders say AI has made them more interested in digital twins … The most prevalent use cases are the processing of front-end data (59%) and enhancing the user experience (56%).”
“AI is driving awareness and improvement in data collection and data quality,” says John Renick, senior strategy enablement consulting lead, Hexagon’s Asset Lifecycle Intelligence division. “This enables a shift in maintenance to be more proactive and supports creation of digital twins with meaningful ROI.”
Renick doesn’t buy the argument that AI requires significant process change. Instead, he points out, “Improving existing processes by augmenting them with AI increases the adoption, value realized, and provides the ability to start small and scale.”
Augmenting Physics-Based Simulation
In some predictive tasks, the 3D model of the physical asset—such as the CAD model of an industrial machine—plays an important role. They offer digital twin users the opportunity to use physics-based simulation to understand the movement, wear and tear, and fatigue of the equipment over time. But a new approach is gaining ground.
“A new technological advancement is poised to revolutionize engineering: geometric deep learning,” says Fatma Kocer, vice president, engineering data science, Altair. “This cutting-edge AI approach builds on recent breakthroughs in integrating AI with simulation software to accelerate decision-making and product development cycles.
“Geometric deep learning takes these advancements a step further by training machine learning models using existing simulation data and learning about 3D shapes at a level of understanding comparable to human perception of everyday objects,” Kocer continues. (For more on this topic, turn to “How Well Does Your FEA Program Know You?,” page 20.)
This month, industrial titan Siemens acquired Altair, paying $10 billion to add the latter’s simulation software titles to the Siemens software portfolio. To put the price tag in perspective, in the beginning of 2024, semiconductor design software maker Synopsis paid $35 billion to acquire Ansys, a leading engineering-simulation software developer. Siemens’ acquisition signals a tighter integration of Altair solvers into the cloud-friendly Siemens Xcelerator offerings [“Siemens to Buy Altair in $10.6B Deal”].
The encroachment of AI-augmented digital twins raises new questions about the role of human experts. According to Hexagon’s digital twin report, “While over half of respondents (53%) say their digital twins can make at least some decisions previously made by humans, a greater percentage (59%) say they are informing human decision-makers.” Therefore, leaving routine monitoring and alerts to AI and delegating human experts to evaluate the AI-proposed solutions might become the norm.
According to Tutt, “AI’s capabilities expand what humans can do with predictive maintenance and digital twins, making it a true game-changer in industry.”
More Altair Coverage
More Hexagon Coverage
More Siemens Digital Industries Software Coverage
Subscribe to our FREE magazine, FREE email newsletters or both!
About the Author
Kenneth Wong is Digital Engineering’s resident blogger and senior editor. Email him at kennethwong@digitaleng.news or share your thoughts on this article at digitaleng.news/facebook.
Follow DE
