We invite you to join us for a guest lecture titled "Vibration Monitoring as a Tool for Rapid Damage Assessment in Civil Structures" presented by Dr. Raimondo Betti, Professor in the Department of Civil Engineering and Engineering Mechanics at Columbia University.  

Date:  Tuesday, September 19, 2023

Time:  4:30 PM

Location:  STEPS 101, 1 W. Packer Ave., Bethlehem, PA 18015

Lecture will be in person.  Please feel free to share this announcement with others who may be interested.

ABSTRACT:

In recent years, advances in sensors and computer technologies have supported various promising developments in structural health monitoring (SHM) techniques.  Data obtained from sensors installed on a structure can help engineers continuously assess the structural  integrity, reduce the operational costs, and optimize the available resources.  In dealing with buildings and bridges, the most common measurement available for such analyses is represented by the time histories of the structural response, i.e., acceleration and/or displacement, recorded at different locations on the structure in service conditions or during particular single events (e.g., an earthquake or hurricane).  Because of the nature of the data used, these SHM methodologies fall into the category of vibration-based SHM approaches.  In this lecture, methodologies that can be used for a rapid evaluation of the damage conditions in civil structural systems, e.g. bridges and buildings, are presented and validated using simulated data as well as data from real-life applications. Falling into the category of data-based methods, these methodologies only rely on the measurements of the acceleration recorded at some locations on the structure and, when possible, on the recordings of the ground acceleration, without any specific information about the geometrical and material properties of the structure.  Emphasis is given to techniques that can be framed within a statistical pattern recognition framework, ideal for machine learning applications.  These approaches focus on detecting patterns in damage sensitive features that are sensitive to the occurrence of damage and that can be easily extracted from the time-history of the system’s response.  Cepstral Coefficients extracted from the time history of the structural acceleration through simple digital signal processing tools are used as damage sensitive features to assess damage in a structure.  To overcome the limitation imposed by an unbalanced dataset (many features from the undamaged condition vs few features from the damaged condition), an original transfer learning strategy that uses audio datasets is presented.