AI and DB to drive smarter process industries

By Dr. Alicia Martinez de Yuso, Research Office Technician at ZLC and Dr. Mustafa Çagri Gürbüz, ZLC Professor.

Leer versión en español


ZLC has joined with four other European institutions in an exciting new project exploring the applications of Artificial Intelligence in the process industries.

‘Artificial Intelligence and Big Data for Process Industry Users, Business Development and Exploitation’, or the ‘AI-CUBE’ for short, is a two-year Collaboration and Support Action programme funded from the European Commission Horizon 2020 budget, and is a response to a call from SPIRE. SPIRE stands for Sustainable Process Industry through Resource and Energy Efficiency, and is a public-private partnership bringing together the major European process sectors – cement, ceramics, chemicals, engineering, minerals and ores, steel, non-ferrous metals and water.

SPIRE is well aware that its industries, though long-established, face a number of acute new challenges. Ever-increasing global competition demands fast and efficient, market-minded innovation. Process industries are central to the new ‘circular economies’ that must be built, and this requires unprecedented levels of linkage with upstream operators and downstream users. The industries need to formulate innovative materials, for functionality and for life cycle performance, and new approaches to eco-design and material usage as well as revised business models that reflect service rather than mere material consumption. And to remain competitive all this requires the transformation of process plants into fully optimised, smart, innovative and safe units that are fully integrated in their environment.

In the modern world responses to these challenges will necessarily be data-driven. The process industries are not short of data, but there are problems. Physically, process plants, which have long been highly automated, have thousands, even tens of thousands, of sensors and other data-points: it is not possible for people, unaided, fully to understand the implications of the information available. The industries have a vast and ever-expanding corpus of structured knowledge in patents and other literature, but a shortage of the ‘basic science’ skills to master this knowledge base. Automation has also eroded individual and collective ‘know-how’ – the intuitive and pragmatic understanding of processes, equipment and materials.

Artificial Intelligence and Big Data (AI&BD), in all its forms (including, for example, machine learning, reasoning, predictive analysis, computer ‘vision’ and other senses, natural language processing) should offer solutions to these challenges through its ability to collect and analyse data and other observations and make decisions or take actions beyond specifically programmed responses.

There are many examples of how Artificial Intelligence and Big Data technologies have been successfully applied, in operations, in product development, in process control and elsewhere..

  • AI-powered automation is used in apple harvesting, increasing efficiency and resolving labour issues in a physically demanding task.
  • AI is used in sales to identify and prioritise leads with the greatest likelihood of conversion into sales.
  • In fashion industries, new product design ‘personalisations’ of colours, patterns and styles can be proposed, using AI to analyse customer tastes and purchasing behaviours. Similarly, machine learning allows the supply of personalised cosmetic products, based on customer photographs and skin tones, in just three days.
  • Lightweight aircraft components such as partitions have been developed using ‘reinforcement learning’ based on patterns found in nature.
  • AI is being used in purchasing and inventory control to identify on the one hand obsolete parts, and on the other, parts at risk of supply shortage, leading to lower inventory levels, scrapping losses, and disruption costs.
  • Near-time failure risk in wind turbines is monitored, maintenance regimes improved, and increased uptime secured, using machine learning.
  • Machine learning is also used to predict demand and pricing for scrap and finished steel, improving profitability.
  • Energy efficiency campaigns are targeted at the household level by analysing usage data, demographics and weather forecasts using machine learning.

Applications such as these, in process industries or in other situations with challenges analogous to those faced by processing, are developing very fast, but SPIRE maintains that European process industries are lagging in the development and implementation of AI&BD technologies, both in general, and at the level of specific sectors and activities.

Previous Horizon 2020-funded SPIRE projects (some of which ZLC has been involved with), have focused on areas such as advanced process control and scheduling, process and catalyst design, and digital platforms for industrial symbiosis, but the consortium believes that major steps forward are needed. SPIRE aims to contribute to the EU strategy on AI, as well as the process industries’ own goals, by developing the ability to identify targets for AI based on process industry realities, and to collaborate with relevant digital partnerships to jointly develop the solutions that will have maximal positive impact for society. The AI-CUBE project aims to deliver the required understanding and insight.

The principal deliverable, the AI-CUBE itself, is simple in concept – a 3-dimensional conceptual matrix whose three axes are the various AI and Big Data (BD) technologies, the different industry sectors or users, and the application areas, activities and processes. This will reveal where and how AI is being successfully utilised and, by default, where there are deficiencies either in the available technologies themselves or in their adoption and implementation by particular sectors or in specific process areas. That will then generate a roadmap and guidelines for further AI/BD development, transfer and use in the process industries.

That is simple in concept but is going to be very complex in practice. The project partners will need to develop a sound taxonomy in order to classify the technologies and look inside the processes – a feature of AI technologies is that they aren’t necessarily what they appear to say on the shrink-wrap, and technologies have many fine differences with big implications. We will need to trawl through the literature, including of course previous European projects, to identify and categorise all the technologies and arrive at a ‘state of the art’. (We also suspect that there is a lot of relevant research ‘out there’ of which industry is quite unaware).

Once we have a taxonomy we can start industry consultation – interviews, webinars, and face to face workshops (if they are allowed). We’ll be working closely with the SPIRE member associations to develop the stakeholder engagement process and identify the industries and contacts most willing and able to drive the work forward. Fortunately, there is already a rich bank of contacts from previous projects such as INSPIRE. We need to achieve comprehensive cross-sector representation but without an unmanageable number of participants.

We want to define ‘maturity levels’ for adoption and use this to explore why, for example, a particular sector may be very advanced in using a technology for one set of situations, but failing to exploit comparable technologies elsewhere. This should suggest methodologies for promoting technology transfer. We also need to assess the impact that AI/BD is having – on the value chain, on plant operations, and perhaps most particularly the impact on human operations since while on the one hand there are understandable fears of job losses, on the other there may be a shortage of the human skills required to work successfully with AI.

By mapping the technologies and implementations, populating the AI-CUBE, we can identify the gaps in Research, Development and Implementation, and begin to develop guidelines for filling those gaps, addressed to industry, to researchers and to policymakers so that all can work together.

Finally, in a work package that ZLC will be leading, we will define a roadmap for creating cross industry/sector synergies that will increase adoption, and develop case studies, value chain configurations and business models (current if possible, new if required).

We will also need to develop tools for Cost Benefit Analysis, bearing in mind that we are potentially talking about very large investments that can’t necessarily be carried out incrementally. But we also need to consider SMEs which are an important component in the process industry landscape – potentially, these can innovate at pace, but not if they are excluded from the process. We’ll be using a Multi-Actor Multi-Criteria analysis, and validating our findings and deliverables through ‘gaming’ workshops and similar approaches.

The deliverables may be deceptively simple, but this is a complex and multi-faceted project. Fortunately, with SPIRE as the ‘customer’ we ae confident that we will have the wholehearted support of the European process industries.

For more information please email [email protected]