In this study, Neural Network (NN) model is developed to quantify nano-level adhesion force of moisture damaged asphalt binder using Atomic Force Microscopy (AFM) test data. AFM data contains five point force-distance values determined for some specific asphalt chemical functional groups. Asphalt binder samples contain different types and percentages of polymer modifiers and antistripping agents (ASA). Due to complex and nonlinear interaction between the asphalt properties and adhesion force of asphalt, it is difficult to assess the effects of asphalt binder properties on the adhesion forces using laboratory AFM testing. NN has the ability to recognize and trace the complex relationship trend existing between inputs and outputs; therefore NN is chosen to be used for the development of the model in this study. Two neural network models are developed, one for lime treated and another for chemical antistripping agent treated asphalt samples. To train the network, AFM tip-sample distance data, percentage of lime, type and percentage of polymer and asphalt chemical functional groups are considered as inputs and AFM force as an output. On the basis of performance, 12-9-16-3 and 11-25-25-5 NN are selected as the final structure of models for lime and chemical antistripping agents treated asphalt samples respectively. The models show good agreement with the laboratory data for both models. To this end, the developed models are used to predict adhesion of both lime treated and chemical antistripping agents treated dry and wet asphalt for same inputs. This allows observing the effect of lime and chemical additives in resisting adhesion loss due to moisture conditioning thus moisture damage of bond forces of asphalt. NN induced results show that lime performs better in resisting moisture effect for samples containing 3% SB polymer compared to other polymer modified samples. Also, lime fails to resist the degradation of adhesion force in wet sample determined by silicon nitride tip for all types of modified asphalt samples. Among all the chemical ASAs, Morlife shows best performance in presence of 3% SB and 3 and 5% SBS in reducing moisture effect on adhesion and cohesion bond forces of asphalt at nano-level. In all cases, increase in percentage of additives above 1% does not aid in resistance to moisture damage.
Antistripping Agents, Adhesion, Asphalt, Neural Network
National Science Foundation
Level of Degree
First Committee Member (Chair)
Second Committee Member
Ahsan, Sanjida. "Assessment of Antistripping Agents on Adhesion of Damaged Asphalt by Neural Network." (2014). https://digitalrepository.unm.edu/ce_etds/87