IIT-Delhi researchers develop AI models for quality check of cement

Researchers at IIT Delhi have developed artificial intelligence (AI) models capable of instantly predicting the quality of cement clinker — an innovation that could drastically cut delays and reduce waste across the cement industry.

Concrete, which makes up over 90 per cent of the world’s built environment, relies on cement as a fundamental ingredient. However, cement production is one of the world’s most carbon-intensive processes, contributing nearly 8 per cent to global carbon emissions. With over 4.1 billion tonnes produced annually, improving the efficiency and sustainability of cement manufacturing has become an urgent priority.

Traditionally, clinker — the partially processed form of cement — is tested using high-energy X-ray analysis to assess quality. This method can take up to four hours, often resulting in reprocessing if problems are detected. In response, a research team led by Sheikh Junaid Fayaz, a PhD scholar at IIT Delhi’s Civil Engineering Department, under the supervision of Prof. NM Anoop Krishnan, has developed AI models that can perform the same assessment in mere fractions of a second.

“X-ray-based clinker checks can take up to four hours, while our AI models deliver predictions in just 1/100 of a second — making quality control a million times faster,” said Prof. NM Anoop Krishnan. “This allows engineers to adjust plant parameters in real time, ensuring target quality is achieved before production, instead of relying on delayed post-production checks.”

Commenting on the broader potential, Sheikh Junaid Fayaz said, “The significance of this work goes well beyond cement. It showcases how AI can modernise traditional industrial practices and support sustainability targets. Given the strong performance of our models, several cement plants globally have already shown interest in adopting this technology. We hope it encourages wider AI integration across the sector.”

The study, titled “Industrial-scale prediction of cement clinker phases using machine learning”, has been published in Communications Engineering.