Damage identification in concrete using multiscale computational modeling and convolutional neural networks

Concrete is a composite material with heterogeneities across multiple length scales. Degradation of concrete due to external loadings starts diffuse microcracking, followed by damage localization that eventually leads structural failure. Identification at an early stage degradation reduces the costs associated maintenance structure. Weak changes can be detected using ultrasonic waves (so-called Coda waves). In this contribution, virtual testing environment for assessment coda presented. The test combines multiscale computational modeling damage, wave propagation, and supervised learning. At scale mortar material, microcrack growth modelled combination continuum micromechanics linear elastic fracture mechanics. model incorporated into reduced-order Lippmann-Schwinger based mesomodel concrete. Synthetic specimens various levels are generated subsequently these subjected propagation analysis rotated staggered-grid-finite-difference scheme. A convolutional neural network (CNN) learning framework further employed classify given signals.