Risk-informed Seismic Damage Diagnosis Leveraging Deep Learning

Seyed Omid Sajedi

Deep learning frameworks for rapid structural inspections.

Deep learning frameworks for rapid structural inspections.

Graduate Student Project


Earthquakes are terrible things! Unfortunately, human casualties are just the "tip of the iceberg". After an earthquake, most buildings and bridges cannot be used before they are properly inspected. Therefore, the social and economic consequences could be very severe and potentially escalate beyond the seismic zone. A recent report showed that an earthquake in California, which supplies more than 30% of the food in the United States, could cost over 200 billion dollars. We need to quickly identify if our infrastructures are safe so people can get back to their normal lives as soon as possible. Unfortunately, human inspections are not a good option. They require a lot of time and money. Therefore, my research at UB focuses on developing efficient systems that can identify damage in real-time. Like a human heart-rate monitor, structural vibrations can tell us a lot about a building's condition. With the help of artificial intelligence, we translate these vibrations into damage in just a few seconds. As a result, by just installing several cheap accelerometers here and there on a bridge or a building, you can quickly find out not only the locations of damage but also their severities. My research helps us to train robot doctors for our structures and take the proper actions immediately after earthquakes.


The inspection of civil infrastructures is an integral part of recovery after earthquakes. By taking into account several factors such as time-cost constraints, reliability, and life-safety concerns of human-based inspections, there has been a growing incentive for automation in structural health monitoring (SHM). In this project, we propose a deep learning framework to translate the vibration data from buildings into structural damage in real-time. This information will help decision-makers to allocate the existing resources better through emergency management and accelerate the recovery process. While being very accurate, the framework has a warning system to trigger human interventions in case the automation system has high uncertainty. The findings of this research will pave the way for fast, robust, and reliable inception of buildings and bridges after earthquakes.

See the Full Poster

Click on the file below to see the full poster in your browser. 

Digital Accessibility

The University at Buffalo is committed to ensuring digital accessibility for people with disabilities. We are continually improving the user experience for everyone, and applying the relevant accessibility standards to ensure we provide equal access to all users. If you experience any difficulty in accessing the content or services on this website, or if you have suggestions about improving the user experience, please contact the Experiential Learning Network via email (ubeln@buffalo.edu) or phone (716-645-8177).