Glass Futures builds AI-driven digital twin to reinvent glass manufacturing

Overview

Glass Futures, based in St Helens, England, built an AI-driven digital twin of its first-in-the-world multi-fuel pilot glass furnace, developed with NVIDIA and the University of Liverpool's Virtual Engineering Centre as part of the AI-GLASS project (£1.5m Innovate UK/Made Smarter UK funding). Using NVIDIA RTX Pro 6000 GPUs and NVIDIA PhysicsNeMo, the digital twin combines real pilot-line sensor data (16 sensor points) with physics-informed neural networks to run over 11,530 calculations simultaneously, accurately predicting glass output and letting manufacturers test new fuels, electric heating and low-carbon alternatives like hydrogen and biofuels before trying them on the real furnace.

The challenge

Predicting how changes such as temperature, pressure, density, new fuels or heating methods would affect real furnace output required slow, costly physical experimentation, and complex phenomena like flame shape were difficult to describe to a computer.

The solution

Glass Futures, working with NVIDIA and the University of Liverpool's Virtual Engineering Centre on the AI-GLASS project, built a physics-informed AI digital twin of its multi-fuel pilot furnace using NVIDIA RTX Pro 6000 GPUs and NVIDIA PhysicsNeMo, combining real pilot-line sensor data from 16 sensor points with the laws of physics to predict glass output.

Reported business value

The digital twin can process over 11,530 calculations at once to accurately predict glass output, letting manufacturers test new fuels, electric heating, bubbling and other techniques virtually before trying them on a real furnace, removing barriers to experimentation and accelerating adoption of low-carbon alternatives such as hydrogen and biofuels.