Speaker: Andres Felipe Calderon Hurtado
School of Civil and Environmental Engineering, University of New South Wales, Sydney, Australia
Title: Unsupervised Computer Vision-Based Approach for Bridge Damage Assessment Applying Drive-by Inspection Technology
Abstract:
Over the last decade, the use of drive-by inspection technology for bridge damage assessment has been widely studied by scholars. It consists of identifying bridge damage from the response of an instrumented sensing vehicle. Most current methods are based on identifying bridge properties and supervised learning techniques. However, these approaches require data from the bridge at its different states (i.e., healthy and damaged conditions), which is not always available. Hence, this study proposes a fully unsupervised computer vision-based methodology for bridge structural health monitoring (SHM) based on the time-frequency domain analysis of the acceleration signal recorded by a two-axle vehicle. A convolutional variational autoencoders (CVAE) algorithm is trained only with the Continuous Wavelet Transform (CWT) of vehicle acceleration responses while passing over a bridge at its benchmark state. The damage index is defined from the measured error between the original and the reconstructed CWT images. During testing, the error between the original and the reconstructed CWT is compared with the damage index from the benchmark state to classify the new samples as healthy or damaged. The methodology is tested on a numerical and experimental vehicle-bridge interaction (VBI) model. Different damage types and severities are considered. The effect of road roughness is also studied.
The Machine Learning Seminar is a regular weekly seminar series aiming to harbour presentations of fundamental and methodological advances in data science and machine learning as well as to discuss application areas presented by domain specialists. The uniqueness of the seminar series lies in its attempt to extract common denominators between domain areas and to challenge existing methodologies. The focus is thus on theory and applications to a wide range of domains, including Computational Physics and Engineering, Computational Biology and Life Sciences, Computational Behavioural and Social Sciences. More information about the ML Seminar, together with video recordings from past meetings you will find here: www.jlengineer.eu/ml-seminar/
10 май 2023
