Approximating Dynamic Elastic Modulus of Concrete for an Old Aqueduct Using Dynamic Tests and BP Neural Network

Monitoring the degradation of dynamic elastic modulus (Ed) concrete is great importance to track durability deterioration for hydraulic structures. For aqueduct under investigation in this study, bent frames and moment frame supports (Ed-frame), arch trusses (Ed-arch) shear stiffnesses asphaltic bearings U-shaped flumes (Kflume) are main parameters define behavior structure, which have direct correlation with its vibrational characteristics thus practicably can be estimated by a BP (back-propagation) neural network using modal frequencies as inputs. Since it impossible obtain sufficient experimental field data train network, full-scale 3D FE model entire created, analyses different combinations Kflume, Ed-arch Ed-frame conducted generate analytical dataset network. After network’s architecture refined cross-validation process modeling accuracy verified test procedure, primary obtained from situ tests put into final approximations Ed-frame, sets an evaluation baseline general Ed status indicates that makeshift basically treated three-directional hinge model. It also found more inputs improve prediction