INTERVIEW
Breathing life into the digital twin
Daniel Reed from MxD talks to us about advancing technology with the digital twin.
By Luke Morris
MxD is a not-for-profit public private partnership that matches $1 of federal investment with $1 of private industry investment. Its aim is to advance digital manufacturing technology for the entire manufacturing sector across the USA.
We spoke to Technical Program Manager, Daniel Reed, about how MxD helps businesses and how they built a digital twin that’s revolutionizing the production and use of medical devices.
Can you tell us a bit about what you do at MxD?
Our focus is around digital manufacturing, cybersecurity and the digital thread. Today I’m in our 22,000 square foot future factory space in Chicago. Here, our members bring some of their most advanced technologies to showcase it and teach others about it.
We also have a lot of the projects that MxD scopes, funds and executes on the factory floor. We have a real variety here – everything from demonstrators for educating the public about manufacturing. to our advanced wireless testbed where companies can test their products on every kind of wireless protocol available.
We host events here too, so it’s a truly multifunctional space.
It’s certainly very different from a traditional factory, isn’t it?
Absolutely. The old image of a factory is a dirty, dangerous place, but this is far from that. It’s bright, gleaming and shiny. Modern factories are very advanced and combine both physical manufacturing technology and digital data technology. Collecting and understanding data about their operations is what’s going to give manufacturers a competitive advantage. It’s used to drive insights and decision-making and enhance automation. That’s not about replacing humans but automating the more dangerous or dirty jobs so people can focus on more important activities and improve overall efficiency.
So, how do you go about helping businesses collect operational data?
Most factories have been around for decades and it’s not feasible to buy all new digitally enabled equipment. So, we show companies how to attach something as simple as sensors to their legacy equipment without making any direct changes to processes.
For instance, we pointed a webcam at an analog dial gauge and used visual recognition and Python code to translate the image into a digital reading of the dial that is recorded.
Another example is a Bridgeport milling center that we’ve added sensors to, allowing us to track the status – is it on or off? Is it cutting? Is it waiting? Is it spinning but not cutting?
By capturing this data, a manufacturer can gain a lot of insight into machine usage. This drives decisions such as whether it should be upgraded, replaced or even removed if it’s not being used often.
OK, let’s talk about digital twin models. First of all, what’s your definition of a digital twin?
I think of a digital twin as a simulation that’s enhanced with the data from the real world – in this case, from those sensors that we just talked about. Without that data, you can’t perfectly simulate the real world. The digital twin improves simulation by capturing data from the real physical system and learning what normal operation looks like. You can then use that enhanced model to understand all kinds of different things with simulations.
Normal operating conditions change throughout the life of a product, so you can use a digital twin to see how it will behave in different scenarios. If you decide you want to use it for something slightly different, you can test the viability first. Which brings us nicely onto our Lungs in the Loop demo.
Yes, we’re excited to learn about this! How did the project come about?
In the early days of the COVID-19 pandemic, there were huge numbers of patients showing up to hospitals with respiratory issues. The nation’s stock of ventilators was much lower than required, but these are complicated medical devices – you can’t just quickly produce thousands of new ones.
The existing ventilators weren’t approved for use by more than one patient on each, but doctors were having to make difficult decisions when patient numbers were at their highest. As they’d never been used like this before it was impossible to say when it was or wasn’t safe to have two patients attached to the same ventilator.
We partnered with Siemens to take a holistic look at the product design of a ventilator. We built a digital twin of the machine operating with both one and two patients attached. We then ran simulations to show which scenarios were safe and which weren’t. We then detected anomalies that were occurring and identified potential patient deterioration in certain situations. Ultimately, this enabled us to prototype a redesigned, improved ventilator in a much quicker timeframe.
What does this digital twin look like?
It’s essentially a one-dimensional simulation of the ventilator operation and how gas flows through it. We took the operating conditions of the system and then put sensors in the operating line at critical points to see how much air or breath was flowing in real time. So, by using data from the real physical product and then simulating airflow, we were able to see how it performed based on its expected operation conditions.
We call this a live twin as it’s running in real time along with the digital twin. For the physical system we used lung simulators to simulate the patients’ breathing. We had a device that introduced a small leak into the line so we could confirm this was detected and an alarm was triggered.
Our simulations showed that for two patients to be ventilated on the same machine, they must be in similar lung condition. Their lungs must be able to take in air at the same pressure. With the physical system we simulated one patient getting sicker and not getting enough air, but the ventilator didn’t understand this as it wasn’t designed to work that way. However, the digital twin understands it as it’s learned over time what normal operating conditions look like. It could then produce an alarm saying that you may need to move one of the patients.
Presumably, the more operational data you gather, the better insights you get?
Yes, if you deploy the digital twin across an entire fleet then you can collect even more data to help with predictive maintenance, you can then spot the signs that a machine is approaching breakdown and intervene before it affects performance. Designers can learn from the data too by seeing how it operates in different conditions and take all the learnings to apply to future designs.
This is very sensitive data though, as it’s real people’s health information. We need to extract the value from it without risking patient privacy. That’s why cybersecurity is a key part of what we do too.
Thanks for sharing your experiences with us, it’s really interesting to learn about the work MxD is doing. Finally, what are your thoughts on the future of the digital twin?
Simulation technology is evolving fast. Previously, we needed dedicated computing resources but now you can run simulations on a laptop or even a cell phone. And digital twin models I think are on the same path, albeit slightly behind.
Digital twin models started as completely physical copies of live systems, which is obviously incredibly expensive. But as computation costs come down and technology advances, we’re able to do more and more virtually. This will enable even more experimentation in the future and I’m sure we’ll see impressive results across a wide range of fields.
“We partnered with Siemens to take a holistic look at the product design of a ventilator. We built a digital twin of the machine operating with both one and two patients attached. We then ran simulations to show which scenarios were safe and which weren't.”
Daniel Reed