One of the consequences of today’s fast-paced digitally oriented lifestyle is that our inboxes are overloaded with buzzwords. Terms that seem to be synonyms are misused and misunderstood, and their value is compromised to accommodate the monetization needs of poorly educated market players. For instance, “digital twins” is a buzzword that has been often confused with the term “simulators.” These terms are indeed very close but not identical. Here are some examples that explain why.
Hardware and software vendors use a number of methods for simulating physical processes and real devices to simplify and speed up development processes. Imagine that we develop a wide product line of devices that interact with the real-world environment. Let’s say these are meteorological devices and weather stations, and we have quite a lot of them. On the one hand, their amount and variety do not allow us to equip each engineer’s workplace with its own set of devices, as the workplaces would turn into warehouses. On the other hand, we cannot recreate different climatic conditions at every workplace, including variations in temperature, pressure, humidity, wind, etc.
To solve this “small issue,” we develop software simulators that can transfer data to your software by simulating the signals of real sensors that would measure the real-life weather parameters. The term to be used here is “simulators” (sometimes also called “software-in-the-loop,” SIL). To develop and test a device that displays current weather parameters, a set of manually created data or historical data collected by various weather stations in real conditions is sufficient.
Now let’s take it to the next level and try to understand how our weather station will behave over the next ten years by measuring not just atmospheric parameters but also exposure to them: heat, direct sunlight, freezing temperatures… What stress load will the device’s mounting hardware bear under a blizzard or heavy rain? To answer such questions, you will need to create a digital twin.
Going back to the weather station’s development, imagine that after analyzing the simulation results, we found out that, after a certain period of time, the blades of the turbine, which is used to measure wind speed, are deformed in some specific and predictable way. This introduces distortions into the measurement results that we send to the customer. We can adjust software algorithms so that, considering deformations, our weather station continues to send reliable results. But if the models show that after some time the blades are damaged in an unpredictable way, then all that our software can do is add an “inaccurate result” flag to the message about the measured wind speed. Most likely, we can still be guided by the fact that the values of the stronger wind will differ from those of the weaker wind, but we should not trust the numerical values. And it means it’s time to conduct weather station maintenance and replace the defective parts.
This is also applicable to other industries, for example, to one of the domains in which we have the most expertise: medical device software development. In our projects, we widely use simulators of medical equipment (automatic infusion pumps, for instance), which dramatically speed up and significantly simplify the development and testing process. However, if a model of a human body is connected to an infusion pump model to study living occlusion thresholds, how electronics and software work under the specific conditions of the patient’s blood vessels, how emergency conditions are treated, how they change when the patient’s body position changes, and how all this is influenced by mechanical parts wear, the system is not just a simulator but a digital twin.
More complicated cases are also possible. For example, let’s say we want to build a hydro power station and analyze what may happen to it in the future or calculate the lifespan of a vehicle engine. We can still test the engine using a real unit (although this can be time consuming) and test the infusion system by building a system of tubes, tanks, and pumps. However, the idea to build several hydro power stations, destroy them, and compare which one is more reliable sounds, to put it mildly, impossible.
So, as simple as that, the difference between simulators and digital twins is significant. In general, both systems create a simulated environment to develop real devices. But simulators are defined by the fact that historical data (statistics accumulated over a certain period) is sufficient, while digital twins show how our device will behave in the future, considering all real-life conditions. Digital twins require a huge amount of accumulated data obtained during measurements of many indicators of the object in the real world. The analysis of such data gives accurate information about the system performance and helps us draw conclusions about how to change the product itself and/or the manufacturing process.
