Credit: Pixabay/CC0 Public Domain
Imagine if you could peer into the future of a machine—track its wear and tear, predict when it might fail, and fine-tune its performance—all without touching it.
That's the promise of digital twin modeling: a virtual model that evolves and adapts with its real-world doppelganger.
Unlike static simulations or 3D models, digital twins—developed using expert knowledge—are constantly updated by live data collected from the physical asset they represent, whether that's a wind turbine, a car, or even a human heart.
This real-time feedback allows engineers to track performance, predict faults before they occur, and plan for short and long-term maintenance with better accuracy.
It's reported that 29% of global manufacturing companies have either fully or partially implemented a digital twin strategy.
Associate Professor Pietro Borghesani, from UNSW's School of Mechanical and Manufacturing, says in industries where safety, reliability and cost-efficiency are vital, digital twin modeling is extremely valuable in asset management.
"A digital twin doesn't just simulate, it lives with the machine," he says. "You can use the digital history of your machine to control how the asset degradation is evolving and then use that knowledge to streamline your operations.
"Instead of being reactive to what happens to the asset, it allows us to plan—but with better accuracy."
From wind turbine blades to heartbeats
Digital twins are already being used to monitor complex systems in the manufacturing and energy sectors. However, the technology isn't confined to large, physical machines.
It's widely used by construction companies to design and build phases of a structure to uncover issues before they develop and become costly.
In the medical industry, biomedical researchers are also experimenting with digital twins of human organs, such as the heart, to better understand disease progression and to personalize treatments.
"The same principles apply: feed in patient-specific data, update the model continuously, and simulate future outcomes," says A/Prof. Borghesani
"The only thing that changes is the physics. For a machine, we use dynamics and vibration analysis. For a human heart, it's about biology and medicine."
Data is king—but expertise still rules
The backbone of a successful digital twin is data, and lots of it. But even with advancements in sensors and Internet of Things (IoT) devices, collecting enough high-quality data can still be a barrier.
Professor Zhongxiao Peng, who leads the Tribology and Machine Condition Monitoring Research Group at UNSW, says to build a good digital twin, you need both data and strong fundamental knowledge of how the system works.
"In many critical applications, it is often difficult to collect large amounts of data, especially under different operating or faulty conditions," she says.
"For example, if you are using a digital twin to predict wind turbine failures, you don't want a lot of data because that would mean that many wind turbines have failed—which I know is a little contradictory.
"In these cases, human expertise steps in to build the models from fundamental physical principles. Then the digital twin can be fine-tuned with whatever data is available."
Digital twins aren't just useful for predicting failures or reducing downtime. They also serve as digital repositories of institutional knowledge.
A/Prof. Borghesani says one of the biggest advantages for companies is that you can embed the expertise of an experienced engineer into the digital twin. In that way, the expertise doesn't walk out the door when they leave the organization.
"This is especially appealing to medium-sized firms, which often struggle to recruit or retain highly specialized staff. A well-designed digital twin acts as both a performance monitor and a training tool for the next generation of engineers."
Scaling up: From machines to cities
As the scale of assets grows, so do the challenges of deploying digital twins.
For large-scale systems, such as complex road networks or energy infrastructure, traditional sensors can be costly or impractical to install across every data point.
A/Prof. Borghesani says approaches such as 'crowd sensing' may offer a solution to this problem.
"Researchers have explored the idea of using everyday activities to collect live performance data," he says.
"Imagine you could help collect data on road infrastructure just by driving your normal route to work. There have been studies that have done just that, with multiple vehicles collecting data to estimate road roughness."
A/Prof. Borghesani acknowledges that issues around privacy and data ownership have limited the adoption of this idea but says other data sources such as publicly available satellite images could be used as sensors too.
"Digital twins can also be programmed to interpolate between missing data points, if you're limited in how much data you can collect," he adds.
The AI tension
One of the biggest barriers of adoption is the lack of people-expertise. However, artificial intelligence (AI) has the potential to automate the labor-intensive process of building a twin by learning directly from data.
"It's reverse engineering how we would usually build a standard digital twin," says A/Prof. Borghesani.
"With traditional models, you understand the physics and the reasoning because you start with the expertise.
"With AI, it's often a black box. It's not expertise you're putting into the software; it's a lot of historical data. And the program will gather the expertise from the data."
However, this model raises issues about transparency.
"Your AI model might tell you that the system will fail in three months, but without insights into why the algorithm came to this conclusion, engineers and operators would often lack the confidence to take action," says A/Prof. Borghesani.
"We're exploring hybrid approaches, where physics-based models are combined with machine learning in targeted ways.
"It's about enhancing, not replacing, the expertise."
As new technologies come to market at an even faster rate, A/Prof. Borghesani says the capability of digital twin modeling has become more important.
He says the transition to decentralized infrastructures across the industries such as energy, logistics and transport have seen digital twins become an integral function of operations.
"In the past, you may have only had a few dozen power plants to manage. But now you have thousands of wind turbines and millions of solar panels feeding into the same energy grid," he says.
"Whether it's keeping a wind farm running smoothly, or infrastructure safe and efficient, digital twins are emerging as a key tool."
Provided by University of New South Wales