When we discuss protecting high-value civil structures such as bridges from extreme events, the conversation invariably turns to seismic activity. Can artificial intelligence detect, interpret, and predict structural responses to seismic forces? The answer is a resounding yes, but with one critical caveat.
While modern structural health monitoring (SHM) platforms are highly versatile and fully capable of ingesting data from accelerometers, geophones, and other specialized sensors, analysing seismic behaviour is one of the most complex challenges in structural engineering. To address it successfully, we must recognise that purely data-driven AI models are fundamentally inadequate on their own.
Seismic forces exert highly non-linear and incredibly complex stresses on physical structures. Through our multi-year research collaboration with the University of Michigan-Dearborn and the University of California, San Diego (UC San Diego), we have focused extensively on developing advanced predictive models capable of capturing these complex structural responses.
Through this work, it has become abundantly clear that purely data-driven AI approaches face a significant bottleneck. Without incorporating an understanding of the underlying physics, these models must learn structural behaviour entirely from raw data. To train a model capable of accurately predicting seismic structural response, organisations would need to collect terabytes of high-quality, rare seismic data. This can require years of monitoring and substantial financial investment before producing a model that is even remotely reliable.
The solution to this challenge is to leverage the vast body of engineering knowledge already exists. By integrating established engineering principles and physics-based structural models directly into artificial intelligence, we create a far more powerful framework: physics-informed AI.
When AI is combined with physical laws, several important breakthroughs become possible.
Finally, and perhaps most importantly, we produce predictions that are both trustworthy and physically meaningful.
While generative AI and large language models (LLMs) are powerful tools for communication, they cannot solve complex, physical engineering problems on their own. The future of seismic structural health monitoring lies in the convergence of high-quality measurement data, engineering physics, and artificial intelligence working in harmony to keep our critical infrastructure safe.
QuantumX is a universal, modular DAQ system delivering high‑precision measurement and flexible real-time data acquisition – ideal for vehicle and structural testing together with the catman DAQ software.
This will bring together HBM, Brüel & Kjær, nCode, ReliaSoft, MicroStrain and Discom brands, helping you innovate faster for a cleaner, healthier, and more productive world.
