ANN-based evaluation system for erosion resistant highway shoulder rocks

Abstract

Highway shoulder rocks are exposed to continuous erosion force due to extreme rainfall that could be caused by global warming to some extent. However, the logical design method for erosion-resistant highway shoulder is not well-researched yet. This study utilized a large-scale UNLETB (University of Nebraska Lincoln–Erosion Testing Bed) with a 7.6 cm nozzle width and a 4000 cm³/sec flow rate to study the erosion characteristics of highway shoulder rocks. Test results showed that different shoulder materials currently used had vastly diverse erosion resistance. However, the clear criteria between the erosion-resistant gradation and other gradation could not be determined easily. Then, this study trained ANN (Artificial Neural Network) with test results to conveniently distinguish the erosion resistance of rocks from other rocks. The ANN predicted the acceptable/non-acceptable erosion characteristics of shoulder rocks with close to 99% accuracy based on the three gradation parameters (D₁₀, D₃₀, and D₆₀).