Premium
An acousto‐ultrasonics approach for probabilistic modelling and inference based on Gaussian processes
Author(s) -
TorresArredondo Miguel Angel,
TibaduizaBurgos Diego Alexander
Publication year - 2018
Publication title -
structural control and health monitoring
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.587
H-Index - 62
eISSN - 1545-2263
pISSN - 1545-2255
DOI - 10.1002/stc.2178
Subject(s) - probabilistic logic , structural health monitoring , context (archaeology) , ultrasonic sensor , computer science , sensor fusion , wavelet , bayesian network , data mining , dynamic bayesian network , inference , statistical model , gaussian , gaussian process , engineering , artificial intelligence , acoustics , structural engineering , physics , quantum mechanics , paleontology , biology
Summary In recent years, there has been an increasing interest in employing and analysing complex ultrasonic waves for the purpose of designing health monitoring systems for guaranteeing the safe structural operation by means of early damage prognosis strategies and reducing the costs of maintenance and downtimes. For this reason, monitoring systems are essential and require highly efficient techniques providing a probabilistic interpretation of their diagnostics. On that account, a Bayesian framework within the context of Gaussian processes is adopted in the present work for the purpose of probabilistic data‐driven modelling and damage detection. A network of permanently installed piezoelectric sensors in a pitch‐catch configuration is used for collecting the propagated ultrasonic waves together with the discrete wavelet transform for feature extraction purposes and matrix unfolding procedures such as the ones used in the monitoring of batch processes for sensor data fusion. The effectiveness of the proposed methodology is experimentally evaluated in three different composite structures. Receiver operating characteristics are used as a statistical measure to assess the damage detection capabilities of the presented method.