Benefits and barriers to adoption of AI in the materials value chain

Materials are fundamental to achieving net zero. All the products and systems we use on a daily basis rely on the production and manufacture of the required materials. The materials value chain plays a fundamental role in the embodied emissions of the systems and developing new innovations to accelerate the transition to net zero.

Opportunities in the materials value chain

Our tools for developing solutions are radically changing. Major advances in robotics, automation, computer science are acting as a catalyst for revolutionising traditional research and development processes and manufacturing practices. Artificial intelligence (AI) connects these areas together to offer new solutions across the materials value chain to accelerate the journey to net zero.

• Discovery: The discovery and development of new functional materials has always been a catalyst for innovation. AI with computational chemistry is being used to rapidly discover new materials which could enable technological advancements in renewable energy and clean power systems, for example.
• Optimisation: With wider availability of analytical techniques and data, materials design and optimisation is being accelerated with support from AI. In some areas, there is a move to autonomous labs, to rapidly iterate and search the parameter space.
• Production: The embodied emissions of the most widely used materials have a significant environmental impact. Continuous improvement approaches provide incremental improvements but recent focus on this impact is encouraging wider adoption of more radical changes, using the use of AI, to achieve step changes in emissions reductions.
• End-of-life: As producers are required to take more responsibility, the design, production and use of materials must consider the end-of-life of the products and how to minimise environmental impact. AI enabled systems are being used to understand material flows and help identify improvements, but there remains an opportunity to connect to this information to materials design, production and procurement.

The application of AI is recognised as potentially game changing but adoption in the materials value chain is low due to the nascent and fragmented capability in the UK. We have been consulting with the industry through seminars, workshops and interviews to understand the opportunities for, and barriers to, adoption of AI in the materials value chain.

Barriers to adoption

In a panel discussion at Materials Research Exchange, experts from industry and academia explained how the barriers to adoption are complex and interconnected. Watch the recording of the session here. It was clear that while the barriers could be grouped into themes, they are correlated and need to be considered together.

This topic was explored in greater detail at a workshop hosted at the Materials Processing Institute in May. The event included three workshop sessions covering opportunities and applications, barriers to implementation and interventions to help overcome these barriers. The participants in the workshop represented a wide range of organisations from different elements of the materials value chain. Development and deployment of AI for materials requires data, skills and investment and generally the barriers are linked to these three elements.

Increasingly more processes in materials development and production are digitised. Consequently, there is more data available to support machine learning and AI. However, this data without context is not very useful. Sector or application specific knowledge is needed to provide the context. But this needs to be combined with digital knowledge and skills to understand what data is valuable and where there may be data gaps for AI applications. Without this, there is the potential to record more data but still miss vital information for successful AI implementation. A combination of sector specific and digital domain skills is needed to find, develop and deliver potential AI applications in the materials value chain.

To address this, some organisations are recruiting digital capabilities to train them on sector knowledge. Others are training sector experts in digital technologies. In this way the intellectual property is retained in-house. One challenge, however, is digital expertise come at a price. The tech industry, e.g. software development, offers higher pay than materials and manufacturing industries. Those trained with digital skills become highly desirable and may leave the industry for higher paid jobs in tech.

In other cases, organisations are working with external solution providers to get the digital technology and expertise. This can raise issues associated with intellectual property, particularly when there is uncertainty around the value of the data. There are many examples of successful collaborations, with several being presented as part of the workshop event. Through these projects, the solution providers are developing expertise in certain areas, so while there may be broad applicability for the AI technology, solution providers are focussing on particular target markets or applications. This helps to build trust which can itself be a barrier to adoption through lack of understanding or transparency of the technology. Improving trust in AI is being addressed in an Innovate UK funded collaborative project led by Intellegens in collaboration with several other partners from materials and chemicals industries, including Johnson Matthey, Domino Printing Sciences, Goodfellow, Welding Alloys Group.

Whether developing capability in-house or working with external solution providers, investment is required. The return on investment is unclear because without testing AI, it is difficult to know what the potential returns may be. Larger organisations have the means to explore the space but it is more difficult for smaller organisations. Some solution providers have recognised this issue and aim to identify particular applications with a clear definition of requirements where they can demonstrate returns within a relatively short timescale, for example, Synbiosis and Reliable Insights. It is important to share the successes, and failures, of these pilots to raise awareness of the opportunities and support the case for more investment in the right application areas.

Conclusion

There are many ways in which AI can help accelerate the transition to net zero across the materials value chain. Various complex and interconnected issues are holding back adoption, which hampers achieving net zero.

At the risk of oversimplifying the situation, the slow implementation of AI in the materials value chain is related to uncertainty. Uncertainty on the availability, suitability and value of data, which needs to be addressed by combining sector specific and digital skills and expertise, which requires investment, which may not be forthcoming due to uncertainty in the returns. In discussions at the workshop, it seems progress is being made to address these issues but more collaborative approaches are required to collectively reduce the uncertainties by sharing case studies and, as far as possible, data to further develop solutions.