Development of Artificial Neural Networks for Predicting the Response of Bonded Concrete Overlays of Asphalt for Use in a Faulting Prediction Model

Abstract

Transverse joint faulting is a common distress in bonded concrete overlays of asphalt pavements (BCOAs), also known as whitetopping. However, to date, there is no predictive faulting model available for these structures. To account for conditions unique to BCOA, a computational model was developed using a three-dimensional finite element program, ABAQUS, to predict the response of these structures. The model was validated with falling weight deflectometer (FWD) data from existing field sections at the Minnesota Road Research Facility (MnROAD) as well as at the University of California Pavement Research Center (UCPRC). A large database of analyses was then developed using a fractional factorial design. The database is used to develop predictive models, based on artificial neural networks (ANNs), to rapidly estimate the structural response at the joint in BCOA to environmental and traffic loads. The structural response will be related to damage using the differential energy concept. Future work includes the implementation of the developed ANNs in this study into a faulting prediction model for designing BCOA.

Artificial Neural Networks for Predicting the Response of Unbonded Concrete Overlays in a Faulting Prediction Model

“…The critical response obtained from each model included the deflection basin and corner deflections on the approach and leave slabs. Deflection basins, established using a 2- × 6-ft area, were selected to more accurately represent the difference in energy density on both sides of the joint as compared to just consideration of the corner deflections ( 13 ). However, corner deflections were also obtained for the development of an ANN to be used for defining joint shear capacity in the faulting model.…”

“…The development of ANNs to predict the response of UBOLs as a result of traffic and environmental loadings eliminates the necessity of producing a computational model for all scenarios. These ANNs can produce responses rapidly to facilitate the incorporation of the response of the overlay into a faulting model for UBOLs, similar to that incorporated within Pavement ME for the prediction of JPCP faulting and the prediction of faulting in BCOA ( 2 , 10 – 13 ).…”

“…In the current faulting model in Pavement ME, the critical response to determine differential energy for rigid pavement design is the deflections at the corners on the approach slab (loaded side of the joint) and leave slab (unloaded side of the joint) ( 2 ). In previous studies, corner deflections, full lane width deflection basins, triangular deflection basins and 2 × 6 ft deflection basins were all considered for UBOLs and BCOA ( 7 , 13 ). The deflection basin was selected to characterize slab response, instead of deflections at the corner, because the basin is able to more accurately represent the difference in energy density on both sides of the joint.…”

“…The database is used to develop ANNs, to rapidly estimate the structural response of UBOLs to environmental and traffic loads. These ANNs can then be used in a faulting model for UBOLs, similar to those within Pavement ME for jointed plain concrete pavement (JPCP) faulting, faulting for bonded concrete overlays of asphalt (BCOA), and stress prediction in the BCOA-ME procedure ( 2 , 10 – 13 ).…”

Accelerated Testing of Full-Scale Thin Bonded Concrete Overlay of Asphalt

“…The HVS testing (both dry and wet phases) was conducted with bidirectional traffic which was not expected to reproduce transverse joint faulting, one of the main distress mechanisms for BCOA in past experience ( 6 ). That was regarded as a major limitation of the HVS testing presented in this paper.…”