What’s your experiences with digital twins in industry? Opinions welcome

[planingkoala]

Hi to all, I work in the food industry and, while I may be mistaken, I’ve noticed that digital twins are becoming an increasingly common topic of discussion. I don’t yet know what my company's plans are, but I wonder if there really is as much practical benefit to them as they seem at first glance.

The idea is that a virtual representation of machines and processes is created before anything comes to a standstill in actual production. I can already see how expensive it can be when production grinds to a halt due to errors. At the same time, being a chronic sceptic, I’m wondering whether this actually helps with day-to-day operations.

I’d therefore be interested in hearing about experiences and general opinions. Does anyone have any experience with this, or at least an opinion on the matter? If so, do you think digital twins are a step forward or more of a fad?

[Fretum]

We created a digital twin once. Or rather, we had one created. We didn’t have any experience with it, so we had machineering do it for us. They have their own software (iPhysics) and create digital twins.
One of our customers wanted a digital twin; otherwise, we wouldn’t have got the contract.
The more complex the situation, the more sense it makes. You can identify problems early on and even train staff on the machines before they’ve even been set up.
Especially if you often have to make lots of adjustments to your machines, it can really make sense.
One of our employees is now looking into this in more detail and has completed training on the software, or is about to do so. But I think we’ll be using it more often in the future too.

[planingkoala]

Thanks – that’s exactly the kind of real-world example I was looking for! I find it particularly interesting that it didn’t stem from a pure enthusiasm for technology, but from a customer request. From that point on, it often becomes more than just a pipe dream and ends up in a quote or a contract. I can well imagine that this makes more sense with complex systems. Otherwise, particularly where processes are always identical, I would take a close look at the cost-benefit ratio. It’s great that there’s training before the actual installation, especially when every hour of downtime hurts later on. In our food industry, the first question would probably be whether something like this is worth it for new lines or retrofits. What do you think?

[Fretum]

As mechanical engineers, we build the machines and technical systems that our customers require. If, over time, adjustments are needed (new customers, new products, etc.), it is naturally easier if you can plan this in advance using software and immediately see the impact of the changes. Of course, that would be the responsibility of the respective company, so they would need to take care of it. As machine builders, we cannot know exactly what the companies need. It certainly helps us too, and it can be done much more quickly and cost-effectively.
It will all make sense. If you only have one customer or one product, you don’t switch production over. Then no problems will arise. If you change the processes frequently and keep having new settings, then problems will arise. This is certainly a good and cheaper way of finding out without disrupting operations or even having to shut down.

[planingkoala]

Yes, the point that the exciting thing isn’t the perfect simulation itself, but rather the question of whether changes can be tested safely, certainly makes sense. When several departments are involved, a model like this can help to ensure that engineering, control systems and production are all working towards the same goal. However, perhaps I’m worrying too much, but I’m still cautious about the phrase faster and cheaper because that’s only true if the model is properly maintained. Otherwise, you’re planning on an attractive but flawed basis. I also wanted to ask you how much work it took to develop the first model into something you could really work with.

[Fretum]

The idea sounds very convincing, but I assume the quality of the results depends heavily on how accurate and up to date the model is. If the model is not maintained properly, I can imagine that it could lead to wrong assumptions and decisions.
Since I have not worked with such a system myself yet, I am also curious about the effort required to create the first usable model.
You’ll probably need to familiarise yourself with the software and get some training. As we only needed it once, Machineering did it for us and we didn’t have to worry about those steps. You might want to ask them how long it would take to get to know the software.

[planingkoala]

So, the crux of the matter is whether it’s properly maintained in day-to-day operations. It’s not as though the assumptions will eventually become detached from real-world operations – and that’s precisely what I’d be wary of. It’s also clear that a one-off, external development for a single project is, of course, simpler than the internal development of software, model logic, and the maintenance effort involved. Once things become more concrete for us, we’ll see how long it takes to produce the first usable model, and what happens following changes to the production line or formulation. In any case, the topic has become much more tangible for me as a result. Thank you for your honest assessment.