Teesside University and Vianet Limited KTP 24_24 R3

A pint of beer that pours flat or foams over the glass is about to get a machine-learning fix. Researchers at Teesside University are building an AI model, paired with a Large Language Model, to automatically calibrate draught beer flow lines and predict when a pump or pipe is about to fail. The problem is that pub cellars are complex mechanical systems. Temperature, pressure, and line length all affect how beer pours, and faults often go unnoticed until a customer gets a bad pint. Currently, calibrating these systems is manual, time-consuming, and reactive. The team aims to replace guesswork with real-time data, using sensors and analytics to spot drift or blockages before they ruin a pour. If successful, the project could cut beer waste, reduce service interruptions, and lower maintenance costs for pubs and breweries. It also creates a Data as a Service business model—selling anonymised, predictive analytics to the hospitality industry. This is applied research with a clear commercial endpoint: fewer wasted kegs, happier customers, and a smarter supply chain for draught beer.