VAlue proposition
Structural health monitoring ensures the safety and longevity of structures like buildings and bridges. As the volume and scale of structures and the impact of their failure continue to grow, there is a need for techniques that are scalable, inexpensive, can operate passively without human intervention, and are customized for each mechanical structure without the need for complex baseline models. We have developed a “deploy-and-forget” approach for automated detection and localization of damage in structures. It is a synergistic integration of entirely passive measurements from inexpensive sensors, data compression, and a mechanics-informed autoencoder. Once deployed, the model continuously learns and adapts a bespoke baseline model for each structure, learning from its undamaged state’s response characteristics and can autonomously detect and localize different types of unforeseen damage.
Description of Technology
This technology is Mechanics-Informed Damage Assessment of Structures (MIDAS), a near-real-time framework for automated damage detection and localization. Our technology is based on the premise that sensor data collected from a structure during its regular operation represents its expected behavior, and any deviation from this behavior indicates potential damage. A structure we wish to assess for damage is instrumented with sensors, and data from its undamaged state is collected to establish the reference (baseline) for damage detection through unsupervised learning. The established reference can be employed to detect and localize damage. This technology affords adaptation to known and unknown damage across diverse structures. The key contribution of MIDAS is the seamless integration of inexpensive sensors, data pre-processing in the form of compression, and a customized autoencoder called Mechanics-Informed Autoencoder (MIAE). From a sensor perspective, our solution is agnostic to the sensor technology. Abnormal patterns in the data are indicative of damage. From the neural network perspective, we adopt an autoencoder that learns a compact representation of the data streams from multiple sensors while incorporating the mechanical relations between their strain responses. Such a design significantly enhances the detection and localization of damage in the structure. Damage detection is achieved by comparing the reconstruction error of the instantaneous sensor data in time windows with that of the undamaged baseline. To localize the damage, we further compute the norms of reconstruction errors at each sensor and interpolate them between the sensors. This approach does not require data from damaged structures for training.
Benefits
Applications
IP Status
Patent Pending
LICENSING RIGHTS AVAILABLE
Full licensing rights available
Developer: Vishnu Boddeti, Nizar Lajnef, Li Xuyang and Mahdi Masmoudi
Tech ID: TEC2025-0149
For more information about this technology, contact Raymond DeVito Ph. D. at devitora@msu.edu or 1(517)884-1658