Correlative Study of Light-Weight Deflectometer, Plate Loading, and California Bearing Ratio Tests for Unbound Pavement Layers Using Genetic Algorithm

This research aimed to utilize California bearing ratio (CBR), plate loading test (PLT), and light-weight deflectometer (LWD) tests for quality control of unbound pavement layers and correlate the results with new genetic algorithms (GAs). Seven unbound pavements for subgrade, subbase, and base layers, classified as clayey sand, clayey gravel, silty sand, and silty gravel, were constructed with different gradations at a real scale. These layers, along with a field subbase from a road under construction, were tested by CBR, LWD, and PLT methods. Finally, three different GAs—offspring selection genetic programming, grammatical evolution, and genetic programming—were applied for the correlative study, and their results were compared to those from the conventional linear regression (LR) method. The elastic modulus of LWD and PLT tests diminished with increasing the percentage of fine-grained aggregates in the layers. Compared to LR models, the obtained GA models had higher accuracy, appropriate correlation coefficient, and low error; however, the relationships were complicated to analyze. The GA models can obtain PLT and CBR results using LWD data with high precision at the minimum cost and time. More studies should be conducted to determine an applicable relation in various local conditions.