Mixed progress for AI in decarbonising manufacturing – report

Treecha / Shutterstock.com

PROGRESS in applying AI to decarbonise manufacturing has been uneven, according to the latest report by net zero research organisation Energy Systems Catapult.

The report, commissioned by the UK department for energy security and net zero (DESNZ), tracked how effectively AI is being applied for decarbonisation across a wide range of use cases – from lowering barriers to heat pump installation to predicting lower emission fuel mixes for heating cement kilns. The report found that amid the “torrent of news and hype” surrounding AI, progress has been slow in some key areas of manufacturing.

In particular, the report identified limited progress in decarbonising manufacturing inputs, which it said would require the “wholesale redesign of processes and products”. AI has been used in early-stage studies to predict material properties for lower-clinker cement and to optimise alloy use in lower-carbon steel. However, the report noted that real-world adoption has been “relatively limited”, adding that the UK is “not particularly a leader” in developing or deploying these applications.

However, the report said the “outlook for the coming year is more positive”, highlighting the development of the Henry Royce Institute’s Digital Materials Foundry which brings together open source libraries of experimental materials data. These datasets could support AI-driven property prediction models and accelerate materials innovation.  

Progress was faster in deploying AI for improving manufacturing process efficiency, where there is greater commercial incentive, partly due to high energy costs.

One example cited was a trial at Heidelberg Materials’ Mokra cement works in Czechia, where an AI tool developed by Carbon Re was used to predict clinker quality in real time. This enabled operators to optimise kiln conditions and adjust fuel use, delivering a 2.2% reduction in specific heat consumption and a 2% reduction in overall emissions.

The report also highlighted Deep.Meta’s machine learning tool that reduced emissions in a steel rolling mill by 5% by predicting when steel had reached sufficient temperature, avoiding unnecessary reheating.

The fastest progress was seen outside manufacturing, particularly in AI-managed tariffs for electric vehicle (EV) charging, which a study last year found reduced electricity demand from EVs by 42%. There was also a “notable step forward” in AI-enabled tools for decarbonising homes, including technologies that have lowered the cost of heat pump installation.

The report’s lead author, Sam Young, urged caution in how AI is deployed for climate action. “Indiscriminate use of AI can increase emissions,” he said. “But smart, targeted uses are already breaking down some of the hardest barriers to a low-carbon economy.”