Neural network modelling of asphalt adhesion determined by AFM

Lay Description About 90% roads in US are asphalt pavements, which are made of asphalt binder coated aggregate particles. Asphalt pavements are susceptible to moisture induced-damage, because moisture attacks or weakens the bonds at the interface of asphalt-aggregate or within the asphalt binder. A common practice of controlling moisture damage in the asphalt industry is strengthening these bonds (adhesive strength or bond strength) using additives such as lime. Although lime has been used for over 30 years, moisture-induced of asphalt still remains an unsolved problem. Bond strength or adhesion loss has never been measured due to difficulty of measuring it at a minute scale. Therefore, an attempt was made in this paper to address the effect of lime on the adhesion property of asphalt binder under wet and dry conditions. Adhesion force of lime treated asphalt binder was measured using an Atomic Force Microscope (AFM). In an AFM test, a cantilever tip is brought to asphalt sample and there resulting adhesion force between the tip and the sample is measured. In this study, asphalt film samples were prepared with 0.5-1.5% lime and the adhesion forces between the atoms of the AFM cantilever tip and asphalt molecules of film surface (-COOH, -CH3, -NO, -Si3N4 and -OH) were determined. A neural network (NN) model was constructed to quantify adhesion using the data from AFM testing. It is observed that adhesion force in wet sample is higher than that in dry sample. As adhesion force represents bond strength, which indicates that bond damage or weakening occurs in asphalt film due to water action. The research confirms the damage of moisture on asphalt binder even with the presence of lime. Adhesion force or bond strength of asphalt sample varies with polymer content. The NN model was tested and found to have good predictive capability. The NN model was then applied to determine adhesion of wet and dry asphalt binders samples with further increase in percentage of lime beyond 1.5%. The model induced trends of increasing percentage of lime on the determination of adhesion force are plotted and compared with wet and dry samples to investigate the effectiveness of the increasing percentage of lime in reducing adhesion loss, that is, moisture damage. Summary This study constructs a neural network (NN) model to quantify adhesion from atomic force microscopy (AFM) data. AFM data contain five-point force-distance values. A total of 760 observations are used to build NN model. To train the network, AFM tip-sample distance data, percentage of lime, type and percentage of polymer and asphalt chemical functional groups are given as inputs and AFM force as an output. To select the NN architecture, one and two hidden layers with varying neurons are tried with 10 input nodes in the input layer and 5 output nodes in the output layer. Two hidden layers with 9 and 17 nodes in the first and second layer, respectively, show the best performance. A 10-9-17-5 NN is selected as the final structure of the NN model. Test results for the trained model show good prediction ability. The model is further applied to evaluate the effect of five different percentages of lime on the adhesion of asphalt. Results show that increase in the percentage of lime is very effective at reducing moisture damage in a styrene butadiene polymer modified asphalt sample. However, increase in lime percentage above 1.5% does not help reduce moisture damage in the styrene butadiene styrene polymer modified sample.