SOM-and-GEP-Based Model for the Prediction of Foamed Bitumen Characteristics

Due to significant interaction between properties of bitumen and test conditions, prediction of foamed bitumen characteristics [i.e., half-life (HL) and expansion ratio (ER)] is a challenging exercise. This work presents a novel hybrid clustering-gene expression programming (GEP) approach to predict foamed bitumen characteristics. To develop these predictive models, a database consisting of 190 observations (arising out of different combinations of eight distinct binder types, six water contents, and eight test temperatures) was used. The self-organizing map (SOM)-based clustering of this database helped in obtaining homogeneous groups under highly complex interaction. Further, the C5.0 algorithm was used to decipher underlying patterns among clusters identified by SOM. A GEP approach was used to develop four global models to predict HL and ER. Subsequently, hybrid models were obtained through recalibration of these global models but using data from individual clusters. Statistical analysis indicated that hybrid models outperformed corresponding global models in all cases. Global sensitivity analysis indicated that among various parameters, water content had a significant effect on ER prediction. This was followed by temperature and viscosity. However, for predicting HL, this order was ER (if used), water content, temperature, and viscosity.