Reducing risk with digital twins

9 mins read

By Richard Sturt, Solutions Architect Manager at Rockwell Automation

Building a new factory or making major modifications can be a risky business. If it’s similar to previous projects, then it is relatively low risk. But what if it is a new innovative design, new configuration or involves integrating new software systems. Then there is an element of risk that needs to be considered. New market demands and environmental challenges are forcing manufacturers to develop completely new concepts like electric vehicles and direct to customer logistics. Simulating the manufacturing process in advance can help to reduce risk.

So how does building a digital twin help?

A digital twin is a software model of key parts of a manufacturing plant. It can be used to test different concepts, simulate production runs and critically, test software before going live. It’s common practice to have 3D CAD drawings with realistic rendering of the proposed plant. You may even be able to move around your virtual plant with a virtual reality headset and see what it is going to look like. The limitation is that these are static models. They only show you what the plant will look like. They don’t show you how it actually works.

Dynamic modelling

The latest simulation products allow you to simulate the movement of product and equipment with realistic movement and timing. Critically it shows the interaction between different pieces of equipment to understand bottlenecks and throughput of different plant configurations. With a digital twin you can test those “what-if” calculations. Can you increase production with a change in layout? How many robots are required? What impact does a short stoppage on a piece of equipment have? It’s much better to better to identify and correct issues at this stage rather than later at build, or worse still, commissioning stage.

How does this compare to traditional methods? You can look at the performance of equipment and calculate likely performance with spreadsheets etc. If your plant is complex these calculations can get very difficult, and it is not easy to visualise what is going on. A realistic digital twin will help to give you more confidence in the design process. You can even collect statistics from virtual production runs to understand equipment utilisation and efficiency.

At design stage you are unlikely to have detailed designs of all the proposed equipment to use in an initial simulation. The early concepts can be tested using "black box" equipment with particular production characteristics, capacity, delays etc. These can be replaced by more detailed representations at a later stage. Simulating the system will give a better understanding of the proposed production and reduce the risk of making expensive mistakes.

Software Testing

The next stage is virtual controls testing to reduce commissioning time. With a complex build all the physical equipment is not going to come together until near the end of the project. A digital twin with motors, sensors, robots and realistic physics can be linked to the control software that is being developed. You can start testing software with the digital twin in parallel with the actual physical system being built. Commissioning is always on the critical path, and this allows you to get a good part of it done early. The sooner you identify problems, the sooner you can correct them.

Most modern plants will link together ERP, Manufacturing Execution Systems and Control Systems to all work together with minimal manual intervention. How do you test all this without a real plant? And if you do have a real plant, is it cost effective? Is it OK to run a day’s production through the system to see if the ERP, MES, control system interaction works correctly? That could be thousands of products that you can’t easily sell. If you find problems, you will need to re-run the tests with more delays and more waste. Finding problems at this stage is time consuming and expensive. The more comprehensive your test procedures, the longer this will take. But skimping on testing different scenarios could lead to problems later.

With a realistic digital twin, you can do a lot of testing while the physical plant is being built. You can run a day’s production, a week, change products, whatever makes sense for your business. If there are problems, you can just re-set and start again. You are only wasting virtual product, so it doesn’t matter. You can iron out software issues before you start with production for real.

Expanding existing system

Modifications and enhancements to existing plants are more common than new greenfield projects for many companies. This can actually be more difficult than building from scratch, you have to interrupt live production to make changes. This could be changing the physical layout, expansion or new software system. The changes typically have to fit into short downtime periods, so they are high pressure. Again, a digital twin can help. You can build a digital twin of the existing system and add the proposed modifications. This can be tested with a copy of the current control system with the proposed software modifications. Improving testing before you cutover to the new system will give you have a higher confidence that it is going to work.

Digital twin definition & model fidelity

The phrase “digital twin” is not very tightly defined. It is used for models as described in this article but also collections of data from a running plant that are used in future decision making and combinations where actual data is fed back into models to improve accuracy. This article focusses on models that are built with appropriate physics, that are capable of interacting with other software systems either in a real-time environment for control system testing or faster/slower than real-time for analysis.

The fidelity (accuracy and detail) of the model needs to match the application that you are using it for. In some cases, you may only need a low fidelity model to understand plant layouts. For example, you could just represent a packaging machine as a box that has a certain speed. In other applications you may need really accurate physics. How fast can you accelerate a bottle on a conveyor without it falling over? For this the model takes into consideration the exact dimensions of the bottles, centre of gravity etc. Modelling large systems with high fidelity takes a lot of effort and computing power so it’s important to understand what you are trying to achieve with the digital twin and model it at the appropriate fidelity. Software tools like Emulate3D from Rockwell Automation allow you to simulate at the fidelity that is appropriate for your application. It’s possible to have high accuracy physics for very large-scale systems if you application requires it. Equally you can simulate systems with simple “black boxes” with appropriate performance parameters if that is more suitable.

Digital Twin for training

A digital twin can also be used a training tool. New operators can train on the virtual plant to understand how it runs. It is also possible to simulate faults in the system to see if they can be easily found and corrected. For example, you can easily simulate a motor not starting or a switch failing. How quickly can the problem be identified and corrected? This can be an excellent tool to support passing on the knowledge from experienced staff to new recruits.


A digital twin can be a really useful tool to reduce risk in critical applications. With advances in technology, computing power and ease-of-use they are increasingly being used in new build and expansion projects. They can be used to reduce risk at several stages of a project including initial proof-of-concept; predicting performance through to software testing to speed up commissioning to get the plant running faster. If there is a level of risk associated with a project, whether it’s a late start-up or not meeting performance criteria it may be worth considering how a digital twin can help.To secure this buy-in requires a business outcome-based reason or evidence to embark on any project. It cannot be technology for technology's sake. It cannot be because we think it would be better, it must be based upon the company facing a certain challenge, whether it be time to market, or product to market times or costs. By implementing or looking at a certain strategy, we believe we can overcome that and improve time to market. It has to be based around business sense and what is important to senior leadership or the owner of the SME, because without that, why would they even entertain it when they could probably pick another three things that they think are more important?

Building on rapid successes

Making the project achievable and measurable is vital, but it is also important that it is a quick payback. Only by having something that is measurable, but also delivers results very quickly, can they see the benefit and advance to the next project. Digital transformation projects should be small and quick but add up to a larger sum that takes the business in the right direction. There needs to be a plan about what you are trying to achieve and where you are trying to go, because without that it cannot fit with the KPI measurements and the business outcomes.

Larger organisations may have a Lean or Six Sigma champion who will recognise this concept of low hanging fruit and be able to oversee this transformation as a by-product of their job and involve everyone in the plan. But if you are an SME, you may not have implemented that. So, the ethos is the same, small, quick wins, create bigger gains while having that awareness of what can be done. If you have got that Lean or Six Sigma experience, then it is not hard to implement these transformations based upon that principle.

On crucial consideration when developing these smaller, pilot projects is that they must be scalable. You must consider carefully who your technology provider partners are going to be and make sure that they can not only provide wins, but to take that to scale. Without the ability to scale securely and efficiently you could be left with a Blackbox solution that could add further risk into your business operations.

Connecting data silos

When we do look at the digital factory it is about pulling multiple threads together and connecting them together with other enterprise elements such as finance and logistics packages. It is about bringing those together and utilising the data within them rather than building or buying brand new systems. It is about understanding that and making sure that any investment you make in the future has that connectivity for when you are ready to use it.

Throughout the SME manufacturing community there is wide divergence of companies’ digital infrastructure. Some have made excellent progress on the digital journey while others are still fundamentally analogue. However, that is not a roadblock because it is just a starting point. It is about taking leadership on a journey and investigating the desired business outcomes. Armed with that understanding, the right area to make the investment can be selected to ensure a successful result.

For companies that have no network infrastructure and nothing connected, the first step might simply be connecting their machines and ensuring that they have a digital backbone in place so that they can connect the relevant silos of data. There is no wrong place to start but if you do not start then you will face problems. Just because there is not a fully functioning, fully automated, top to bottom system does not mean you should shy away from the opportunities. Simply select battles that are going to give you the biggest win to start off with and chip away over time.

Lacking in skilled workers

One challenge that the digital transformation can help with is the ongoing skills shortage that all of industry is suffering from. With the changing industrial landscape, it is often hard to imagine what skills are going to be needed in the next five years. An equally important problem or risk for too many of these companies is they have got a wealth of knowledge tied up in their existing workforce which needs to be retained in some way. You would like to think that those people will stay with you forever, but they might not, they might retire, or they may choose to go somewhere else. Companies must look at ways that they can both upskill and capture the existing workforce and skill set, but also be attractive to the future workers.

If you consider the existing skill set, it is not difficult to start to capture how somebody does something so you can use it to train people in the future. Not only are you making that person's role more relevant and giving them a purpose to help out with future training of staff, you are also capturing knowledge and experience in case they decide to leave before you have replaced that skill set.

Clearly you also need to have people with the tools to build to drive that forward. There has been a tendency in the past to try and replace departing workers with someone with a similar skill set. With the growing digital influence, it is important that new workers, while being trained in traditional skills, can work with modern technologies to keep driving innovation for the company.

It is crucial that SMEs look at their plan to partner because they will not have all the skill sets on site. They will not need all the skill sets on site permanently and they will need them as and when they require them. This partner could be a system integrator or technology supplier.

The final piece of the puzzle is funding. There are several government initiatives available to provide funding to SMEs for innovation, such as the Manufacturing Made Smarter Industrial Strategy Challenge Fund. If they are aware of that help or work with a partner who can support them, it helps speed up adoption.