Digital twins help manufacturers make optimal decisions faster and more consistently. Think of a smart home app, which is a digital twin of your home’s environment.
When you have a real-time visualization of your home’s environmental conditions, you can act accordingly: say, remotely crank up the heat when your mom (who’s always cold) calls to let you know she’s dropping by later.
Digital twins serve a similar function in a manufacturing context, by replicating real-world systems digitally. Their potential value is enormous, but they also require a large upfront investment of time and resources. Because of that, it’s essential to understand the basics of how to approach them.
Let's look at the “three Cs” of digital twins: complexity, connectivity and consequences. By the end, you should have a sense of what kinds of problems digital twins are good at solving and how to prioritize digital twin projects to maximize ROI.
Complexity: Does the ROI justify the investment?
Digital twins tend to work well for problems that are…
- High stakes, meaning a lot of revenue or savings is on the line.
- Complex, meaning there are many layers or moving parts.
- Measurable, meaning you can get reliable data for the component parts.
- Recurring, meaning you have lots of opportunities to understand what to measure and what possible data models you can construct.
- Sufficient and complete, meaning the data available is enough to fully represent the physical system.
For problems that don’t meet this complexity threshold, there are probably lower-cost and lower-lift solutions that will get you the results you want.
But for problems that are sufficiently complex, a digital twin can deliver significant ROI. Imagine a digital twin of your factory floor, for example. Maybe you’re wondering whether efficiency would improve if you moved equipment into a different configuration.
Testing that in real life would be difficult: time- and labor-intensive, with no guarantee of improvement. But a digital twin would let you project alternative setups, identify the one that’s most efficient, and continue to monitor that setup in real life to determine whether the projected scenario is delivering the ROI you were expecting.
Connectivity: Can the data flow from the physical to the digital world?
I mentioned above that a system has to have measurable components to work as a digital twin. Those measurements—aka the data involved—must also be able to flow from the physical world to the digital, into a single source of truth, in something close to real time.
In many cases, this is possible thanks to digital investments manufacturers have made in the last decade: equipment that connects to the internet, digitized employee badges, etc. Just as often, though, getting adequate data requires additional investment: in temperature and humidity sensors, for example.
It’s also important to note that “sufficient” data doesn’t have to mean owned data. Many successful digital twins contain a combination of proprietary data (say, sensors from owned trucks and inventory) and purchased data (say, on weather patterns).
Combined, they might fuel a supply chain digital twin that helps an organization predict disruptions and so minimize their impact.
Another key consideration for data: not only must it be able to travel from the physical world to the digital; it must also be able to paint a full picture of the system being represented digitally. Because of the need for complete data, the minimum viable product (MVP) of a digital twin is by definition more complex than the MVP of other intelligent products.
Consequences: Can the digital twin deliver ongoing insight?
To justify the upfront investment required to build a digital twin, it has to be able to offer valuable, actionable insights on an ongoing basis. This is why digital twins work well for systems that are ongoing and dynamic.
Take the digital twin of a factory floor. Let’s imagine it tracks equipment performance via connected equipment and environmental conditions via temperature and humidity sensors. When it’s first live, the digital twin shows that humidity is higher in some locations than others.
Upon further investigation, it’s clear that the equipment in those areas requires servicing more often than it should.
A quick adjustment to environmental conditions can prevent the costs of unplanned downtime and extra maintenance.
Further analysis of the digital twin reveals other conditions that trigger excess wear and tear. Over time, the manufacturer is able to update its preventative maintenance protocols to consider not only guidelines for each piece of equipment but also environmental triggers known to affect performance.
The new approach reduces unplanned downtime by 25 percent, a substantial savings for the organization.
Build the right twin, build the twin right
Broadly speaking, scoping a digital twin requires industrial leaders to first identify the right digital twin to build (i.e., the system that both can be digitized and whose digitization will allow for substantial ROI) and then determine how to build it.
This sounds simple, but I often see attempts that are both too ambitious (e.g., in how much data to include) and not ambitious enough (e.g., in how many decisions to impact). The ideal digital twin is technically pragmatic but strategically bold.
That can be a difficult balance to strike, but with the big-picture view in mind, it gets much easier. And when an organization can build the right digital twin and build that twin in an effective way, the business outcomes can be significant.