How digital twins and AI become innovation drivers

What benefits can machine builders achieve with digital twin concepts - and how do their customers benefit from them? Can you name some practical examples?

Guido Reimann: With the digital twin, for example, very large savings potentials can be realized through virtual commissioning. Although the topic has been in focus for some time, it has by no means been implemented across the board. However, customers can benefit greatly if many things can be simulated and tested in advance, such as the entire functionality, sensor technology, and control technology. This can significantly shorten the commissioning phase - and production can start much faster. In addition, customers can already check in advance whether a new machine or component fits into a brownfield environment or a networked system. There are now also services related to training, where machine operators can better practice and learn processes in the simulation based on data from the digital twin. Here, the combination with generative AI (GenAI) plays an important role: Instructions and information on questions are immediately available in natural language, without having to search through long documentation, such as how to specifically replace a spare part.

What preliminary work is typically involved in bringing digital twins into practice?

Reimann: The topic of digital twins has been discussed and advanced in the industry for several years. We see that companies are often a few steps ahead with the digital twin of their own products compared to their processes. However, data is always the core, and without it, progress cannot be made. There is still room for improvement here: data quality has been a huge issue for many years, but unfortunately, it is often somewhat neglected. To access the data, companies need to strategically address this and complete some "homework." It is important to have a culture where data, similar to financial resources and raw materials, is seen as a crucial factor for one's business model. This applies not only to its expansion but also to using the resources deployed much more efficiently. In maintaining data quality, AI can also make an important contribution today to simplify these tasks.

How could AI help improve data quality?

Reimann: Without a good data foundation, meaningful simulations, for example for planning processes in order processing, cannot be run. For example, processing times for machines were often recorded at some point. With AI that accompanies these processing processes, they could be automatically kept up to date (for example, as part of the digital twin data model). Language models (large language models - LLM) help service maintainers or service technicians, for example, to document in natural language which parts were replaced, which errors were displayed or found, and what the solution was.

At the same time, this AI accompanies the processes up to the automated spare parts ordering or invoice creation for the service deployment. Generative AI based on LLM can be fed with product documentation, parts lists, and other relevant information. The more data is collected, the better the AI learns about relationships and can provide tips that may also help the customer solve problems faster themselves!

Why is it so important for the mechanical and plant engineering sector that generative AI is actually integrated into products and services? Where do you see the greatest leverage and how far along are companies in this regard?

Reimann: A joint, recent study with PwC shows: Generative AI (GenAI) is seen by 52 percent of the 247 companies surveyed as a potential game changer for the industry. We have looked at 45 real application scenarios. Just a handful of these use cases bring such decisive advantages that they should be implemented as quickly as possible to position oneself at the forefront. These use cases could result in an increase in operating margins by a total of 10.7 percentage points. This would correspond to an additional profit of 28 billion euros for the entire German mechanical and plant engineering sector based on the current situation.

It is also clear: The world around us does not stand still. Especially for an export-oriented industry, this means that competitors in other countries are also dealing with the technology. Those who do not do this have correspondingly worse cards. With GenAI, customer benefit comes even more to the fore, as customers increasingly have high demands due to regulations, market dynamics, or diversification needs. For this, the value-adding processes must not only be enriched with AI but sometimes also fundamentally rethought.

What impact can GenAI have with regard to skilled workers?

Reimann: Often, with long-lasting products, the equipment is older than the service technicians who are supposed to maintain them. Here, language models help make historically collected experiential knowledge from many sources available through natural language communication. It has now been shown: If access to a specific pool of documents and information sources is restricted, then hallucinations, i.e., errors of the AI, can also be avoided. GenAI can also improve the onboarding of new employees. A current practical example: A smaller machine builder has completed many projects over the years but has hired a lot of new staff due to natural turnover.

The "newcomers" do not know the old projects, but there is good, digitally available project documentation. With a language model, they now receive well-prepared information and experience from comparable previous projects immediately based on targeted questions. Also, in the area of offer creation, where high speed is increasingly becoming a success factor, or in inquiries about prices or delivery times of an order, GenAI accelerates the processes, for example, with pre-prepared emails by accessing ERP data. Humans may only need to critically review the suggestions once more.

How do circular economy and sustainability benefit from digital twin models and AI?

Reimann: First of all, let me preface this: especially in the area of circular economy, companies will need to focus much more on cooperation and networks in the future. It is particularly important for component manufacturers without direct contact with machine operators that all parties work more closely together - and above all, share data. It is only with the help of these lifecycle data that we gain insights into how products are used and operated or what service cases exist. Then, products can be designed in the future to match real usage, and resource consumption in terms of energy, raw materials, or spare parts can be minimized.

Additionally, manufacturing companies can benefit from digital twins and AI-based analyses regarding the use of machines and systems: through optimization suggestions for workpiece processing, conducting maintenance intervals, or additional operational assistance: these are also measures that can contribute to extending the lifespan of machines and thus conserve resources.

When products are finally subjected to "dismantling" and recycling, information is again needed to correctly assign the individual materials and use the appropriate processes. Here too, the digital twin is the tool of choice: It ideally combines all important lifecycle information and thus enables targeted recycling. In addition, raw materials can be sorted even better automatically with AI in image analysis.

It represents a significant economic factor for Europe if we not only manage to reduce resource consumption but also make more of the raw materials "built into" products reusable. But for this, we need appropriate machines and suitable equipment because otherwise, recycling cannot be in a balanced cost-benefit ratio: An important opportunity for our industry.