"We want infrastructure to have a voice," says Varma, leaning over a holographic projection of the Pennybacker Bridge. "We just need to be brave enough to listen."
"We caught a bearing lock in El Paso three months before it would have seized during a winter freeze," recalls Marco Diaz (B.S. '20), the project's lead field engineer. "The bridge didn't look broken. It felt broken to the AI. We replaced a $400 part instead of rebuilding a $4 million span." However, the project raises a provocative question: If a bridge can tell you it is dying, who is liable if you ignore it?
Published in the style of MaxQ Magazine | Fall 2024 Issue
"Bones don't break without a warning crack," says Varma, who holds the Temple Foundation Chair in Smart Materials. "Steel doesn't snap without yielding. Our problem isn't a lack of data; it's a lack of translation. We built a translator."
The sensors measure strain, temperature, torsion, and vibration 2,000 times per second. The AI, trained on two decades of bridge failure data, learns what "normal" feels like. When a variable deviates, it isolates the location with sub-millimeter precision. The implications are staggering. Texas has over 55,000 bridges; 12% are considered structurally deficient. Repairs currently rely on annual visual inspections—a method that misses slow-moving fatigue.