Researchers build a tool to predict how old buildings will crack and crumble

A research team combined IoT sensors, machine learning, and physics simulations to predict how cultural buildings degrade over time. This means conservators can now spot problems before they happen instead of reacting after damage occurs.

Building conservation has always been reactive. You wait until a crack appears, call a specialist, and hope it isn't structural. This framework lets you feed real environmental data into a model that predicts degradation patterns weeks or months ahead. The approach works by embedding actual physics (stress, material behavior, moisture) into neural networks, so the model doesn't just pattern-match on historical data — it reasons about why buildings fail. For institutions managing thousands of aging buildings, this shifts maintenance from expensive emergency repairs to planned interventions. The question is whether heritage organizations have the technical capacity and funding to actually deploy IoT networks and maintain these models in production.