Artificial intelligence in industry: quiet but powerful!
It's nice when artificial intelligence writes e-mails, but the benefit is limited. For process industries such as chemicals, pharmaceuticals and energy, as well as mechanical and plant engineering, the potential lies elsewhere entirely. But what applications can engineers and technicians actually expect? And how realistic are the forecasts?
The Gartner Hype Cycle shows the various phases that new technologies go through. According to the report, generative AI will have passed the peak of inflated expectations in August 2024.The peak of the hype continues to have an impact today: up to 40% of working hours – and thus jobs – in cognitive occupations could be replaced by AI in the future. The World Economic Forum came to this conclusion in spring 2023, triggering shockwaves throughout the economy. However, anyone familiar with the hype cycle mechanisms knows that new technologies follow the same pattern in the public perception. A topic is hyped until it reaches the peak of inflated expectations – after which it descends into the valley of disillusionment. The inventors of the hype cycle – the American Gartner Institute – predicted this for generative AI in August 2024.
But the topic is by no means off the table – because although no realistic assessment of the use of AI has yet reached the public, one thing has long been clear: the potential is huge. Most companies are now checking whether and how AI can take over tasks before hiring new employees. At the OPEX Forum hosted by the consulting firm Conor Troy Consulting in October 2024, participants discussed examples of how companies are already checking whether a job can be eliminated by using AI before hiring someone for it. According to market researchers, the savings for the manufacturing sector could reach into the hundreds of thousands, and in some cases millions, per year.
The use of AI in industry is not new. For years, companies have been using AI to analyse data from production and supply chains, optimise processes and ensure quality. However, recent advances in generative AI are opening up completely new dimensions of application, for example as chatbots and virtual assistants that support customer support and product design.
Artificial intelligence will support engineers in plant design in the future.Applications in the process industry: advances in chemicals and pharmaceuticals
The process industries expect the main benefits of using AI to be advances in plant operation. In the future, assistance systems could support plant operators in making decisions about the optimal operation of processes by evaluating historical data and learning from it – while at the same time preserving experiential knowledge. This data could also be used for predictive maintenance to avoid unplanned downtime. But can't AI then control the entire plant? So far, specialists have rejected this hope: in the KEEN research project, industrial companies and scientific institutions in Germany spent three years investigating the technical and economic potential of AI in the process industry along the product life cycle – and found that the data quality in safety-relevant processes, such as in the chemical industry, is not yet sufficient. However, the project also showed that AI methods do offer added value: Above all, they can support plant operators in making decisions by providing real-time analysis and making engineering tasks more efficient.
Further potential lies dormant in the pharmaceutical industry: Gartner's market researchers estimate that by 2025, more than 30% of all new active pharmaceutical ingredients will be discovered using generative AI techniques. For example, pharmaceutical giant AstraZeneca is already making extensive use of AI to optimise the predictive modelling of the physical and chemical properties of active pharmaceutical ingredients and to predict the performance of formulated products during manufacturing. The company has already managed to reduce development lead times by 50% and cut the use of active pharmaceutical ingredients in experiments by 75%. In manufacturing, AI-powered digital twins at AstraZeneca simulate the relationship between drug substance properties, process conditions and product quality to optimise operating conditions.
Potential in plant engineering: support in planning should alleviate shortage of skilled workers
AI is also increasingly proving to be a valuable productivity factor in plant engineering. The shortage of skilled workers, which is worsening due to demographic change, could be alleviated if assistance systems relieved the workload of engineers in the preparation of offers, in planning and in the operating phase. The automation of processes such as collision control in 3D models, the creation of assembly sequences with a schedule, and the automatic recording of costs on the construction site are among the potential application scenarios reported by the plant manufacturer Standartk essel Baumgarte reported at the Engineering Summit congress in October 2024: AI-supported systems can be used in the planning phase to automatically create initial drafts for 3D layout plans and to detect possible collisions (‘clashes’) at an early stage. In addition, AI tools can iteratively improve the static calculation and layout planning in order to optimise load transfer.
Engineering software providers such as Aucotec are already implementing assistance systems based on large language models by scouring central databases for correlations. Even ‘non-intelligent’ drawings such as P&I diagrams are migrated by artificial intelligence into a readable data format that not only recognises the equipment used in existing plants, but also the process engineering relationships.
AI in mechanical engineering: quality control, production optimisation and AI automation
In mechanical engineering, AI is now an established part of quality assurance and production control. Companies like Bosch are using generative AI for faster production processes, while AI-based image recognition in optical inspection reduces the error rate and increases production quality. Schäffler and Siemens are also already developing generative AI for production machines. They are using three key developments in AI automation:
- Prompt-based programming for industrial co-pilots: Language models such as ChatGPT enable employees to use natural language commands to control machines and improve communication. This function is revolutionising the interaction between humans and machines and redefining industrial automation.
- New programming paradigms in robotics: Autonomous robots adapt movements independently (error-based programming) or optimise them for specific tasks (skill-based programming). This significantly improves the efficiency and adaptability of robots.
- Synthetic training and digital twins: Digital twins create virtual test environments for robots. This allows engineers to simulate scenarios, identify problems early on and save on hardware costs by testing in virtual worlds.
From the valley of tears to the plateau of productivity
The transition to an AI-supported industry is in full swing, but it is not without its challenges. Many companies are faced with the task of improving their data quality, as the requirements for AI applications demand a high level of consistency and accessibility. In addition, specific expertise is needed to integrate AI, which is still lacking in many companies. However, continuous training of the workforce and closer collaboration between IT and operational technology are creating the basis for data-centred and AI-based applications in production. This means that AI in industry could soon reach the next stage in the hype cycle – and climb out of the valley of disillusionment and into the plateau of productivity.
