Executive Summary
While many readers report a willingness to consider implementing digital twins into their work, a smaller number are currently using this technology for sensors, monitors, modeling or simulation. Some of the barriers preventing readers from using digital twins include costs, lack of training and the need for extensive data.
Q.1. In your field, have you incorporated virtual twins/digital twins or clones for sensing and monitoring a specific part/component interface? If yes, please describe in short detail. If no, what application do you envision using it in for the future?
We employ modeling and simulation tools to assist lubrication failure analysis.
No, we do not use them and cannot envision us moving toward them in the future.
Not yet. I would like to monitor the gear/lubricant interface in the future.
Numerical simulation of mechanical seals.
Yes—simulate an atomic force microscope tip-cantilever system by a digital twin.
Yes, we did and it is not a new thing in the nuclear industry. We have simulators like the aerospace industry. We perform dynamic and interactive modeling on system and component behaviors using finite element and other stress analysis programs.
I think in design and in uncertain ventures into new materials or applications, these tools are useful. For routine monitoring, these tools are not mature enough.
We use digital artifacts to simulate the motion of parts to optimize fixation of testing specimens in our lab tests.
Not yet.
Our particular group doesn’t use much simulation in general, but I envision such digital twins would be very appropriate for high value components/systems where the cost of failure or unplanned shutdown is unusually high.
No. No plan to use in our applications any time soon.
Would you consider implementing digital twins in your company in the future?
Yes
73%
No
27%
Based on an informal poll sent to 15,000 TLT readers.
No—can’t envision any applications in the work that I do.
I do not plan to use it yet.
No. Frictional heating.
We believe that computer-aided simulation will improve the mold designing for metal pressing and injection process of plastic and rubber in the future.
Yes. Indeed, a data scientist on my team is working on a development of a data-driven artificial intelligence (AI) model to predict the wear in grinding media. This may not be quite interesting to you as it is not in the field of lubrication. However, it is an example of practical application of digital twins in the field of abrasive wear.
We have not yet utilized these technologies and may adopt them in online tribological analysis equipment in the future.
No, not suited for my research area.
No, have no plans to use this in the near future.
Grease deterioration is determined using color and odor sensors.
Yes, we are working on digital twins, mainly in the area of blast furnace and steel making operation.
Not yet.
No. Future applications are in optimizing materials, such as 3D materials for improved heat transfer in brakes or clutches. Another possible application is in studying fatigue effects in gears and bearings.
No, I anticipate this will be helpful in the prediction/simulation of lubricant formulation performance that could be used in conjunction with a bench test.
Yes, I was leading a large R&D project on the application of AI/machine learning (ML) tools in crankcase lubricant development.
For lubricants formulating, especially metalworking fluid complex formulation.
No. Turbines, gas compressors and pumps might be future applications.
Yes. We use finite element analysis (FEA) and continuum fluid dynamics (CFD) methods to simulate the flow field in bearings and seals, and we monitor the temperatures in the test to benchmark our simulations. During the daily operation, we keep monitoring the bearing pad surface temperature to ensure safety operation.
Yes, in electric motors.
Using optical sensors and molecular dynamic simulation tools for observing the engine oil lubricant degradation in heavy-duty commercial vehicle testing.
We have not yet used digital twins. However, I could imagine running them in parallel with tribometer tests in order to be able to evaluate influencing factors digitally.
Q.2. What are the current challenges encountered in implementation of high-fidelity digital twins in the field?
Modeling tribological system is a very complex task and comes with an expensive computational cost.
Financial benefits or potential applications for fluid failure analysis.
Challenges are 1.) quantifying the value a digital twin can add relative to the cost (ROI), 2.) implementation hurdles from a computational energy usage and 3.0 Talent/skill gap.
It’s a multi-physical problem, with different physical phenomena to consider.
The detailed geometry at interface.
1.) Reliability of the technology. 2.) Understanding of the technology by the users; don’t want it to be a “black box”where the users are blind, or worse, indifferent, to what is going on with the technology. 3.) Ongoing sustaining of the technology (costs, hardware and software, refresh training of users).
Lack of detailed mechanistic understanding in many cases.
Software user-friendliness.
What type of sensor(s) have you utilized to couple a tribological physical asset with its virtual/digital twin?
Thermal
53%
Optical
20%
Electrical/electronic
43%
Acoustic
22%
Pressure
37%
None of the above
39%
Based on an informal poll sent to 15,000 TLT readers. Total exceeds 100% because respondents were allowed to choose more than one answer.
Cost, sustainability.
Reproducibility of real and repeatability of material models applied to digital twins.
Implementation cost and long-term sustainment of keeping the twin relevant and up to date with new designs.
Technology application is at high end.
Cost factor.
I’m not familiar with this application.
Main challenge is reliable and extensive dataset as these models require a considerable amount of data to be trained properly and hence be used with reasonable reliability.
Not suitable for micro research area.
Grease is an opaque fluid and difficult in sampling. Additionally, sensitivity may vary due to dirt on the sensor.
Getting skilled people to work in a steel plant is a challenge.
Doubt on reliability; adaptability to different standards.
The challenges encountered in implementation of high-fidelity digital twins may be successfully dealt with by the use of limiting AI.
The lack of quality data to validate and develop digital twins. Production of data isn’t an interesting research topic and gets little attention in funding programs, but it is the first step toward digital models that actually work.
Simulating mechanochemistry is a real challenge! Within a lubricated contact you have multiple chemistries interacting and one chemistry against another that are generally fighting for the contact surface. This is hard to simulate and ultimately impacts friction and wear.
1.)
Communication barriers between data scientists, engineers and lubrication experts. 2.) Secrecy/incomplete information/lack of quality data regarding component design and lubricants formulations. This makes it difficult to properly train ML models that seriously impact their predictive power. 3.) Overselling AI. There’s a widespread belief that AI is just a marketing gimmick.
What type of simulation/modeling tool do you implement for virtual/digital twin applications?
Finite element analysis (FEA)
34%
Continuum fluid dynamics (CFD)
36%
Molecular dynamics (MD)
22%
Density functional theory (DFT)
10%
None of the above
50%
Based on an informal poll sent to 15,000 TLT readers. Total exceeds 100% because respondents were allowed to choose more than one answer.
I am in China and here the main challenge is for the tasks. When digital twins are implemented, many people lose their jobs, so the implementation is neglected.
My current company is just distribution lubes, so we can only advise end-users to try.
The simulations are quite time consuming, and the accuracy highly depends on the engineers’ proficiency and depth of their understandings.
Accuracy/validity.
Coordination of devices and digital interface.
We do not yet have any experience with digital twins. I therefore see a relatively large barrier to entry.