Parametric Study of Modified Subgrade Reaction Model Using Artificial Neural Network Approach

摘要

Modulus of subgrade reaction (k) is widely used to evaluate subgrade strength and soil structure interaction, and design the rigid pavements. In the pavement ME design, the modulus of subgrade reaction is characterized by the dense liquid or Winkler model, which has no consideration of interface bond within the supporting media. To overcome this limitation, a modified subgrade reaction model is developed in this study to take into account the shear interaction between the concrete slab and base course. Formulation of the modified subgrade reaction model contains two steps: (1) transform the original slab and base system into an equivalent cross section based on the slab-base interface bond; and (2) develop a formula for the modified subgrade k-value using the falling weight deflectometer (FWD) defection patterns on the equivalent section. An artificial neural network (ANN) approach was employed to quantify the influences of pavement layer modulus, thickness, and bonding condition on the modified k-value. A three-layered ANN model was constructed with 1,296 different combinations of pavement structural properties, which included Portland cement concrete (PCC) slab and base thickness and strength properties, including slab, base, and subgrade moduli, and slab-base interface bonding ratio. The FWD deflection pattern for each combination was calculated using the finite element software ABAQUS. The results showed that the modified k-value increased directly with the increasing degree of bonding. In addition, PCC slab and base moduli individually changed the modified k-value significantly but the subgrade modulus had minimal effects on the k-value.