ELM-Based Approach for Analyzing One-Dimensional Electro-osmotic Consolidation

Electro-osmosis consolidation is an important ground improvement technique for soft clay. It is significant to obtain the analytical or numerical solutions to the partial differential equations describing electro-osmotic consolidation to understand its underlying reinforcement mechanism. Neural network-based solvers have been proven to be effective in solving partial differential equations. Inspired by the basic principles of extreme learning machines, a new method is proposed herein for solving the one-dimensional partial differential equations describing electro-osmotic consolidation in double-layered foundations. In the proposed framework, the solutions to the equations are represented by extreme learning machines, and then the partial differential equations are transformed into linear problems to obtain differentiable closed solutions. The effectiveness of the proposed method is verified in two examples by comparing its results with those of analytical method and finite element method (the relative error is within 4%). In addition to inheriting the benefits of the neural network solver, the proposed method effectively minimizes the computation time required. Therefore, the proposed method can serve as an attractive and effective tool for solving the electro-osmotic consolidation equations.