Based on machine-learning(ML) and analytical methods, a hybrid method is developed herein to predict the ground-displacement field(GDF) caused by tunneling. The extreme learning machine(ELM), as a single hidden layer ...Based on machine-learning(ML) and analytical methods, a hybrid method is developed herein to predict the ground-displacement field(GDF) caused by tunneling. The extreme learning machine(ELM), as a single hidden layer feedforward neural network, is used as an ML model to predict maximum settlement smaxof the ground surface. The particle swarm optimization(PSO) algorithm is applied to optimize the parameters for the ELM method, namely, weight and bias values from the input layer to the hidden layer. The mean square error of the k-fold cross validation sets is considered the fitness function of the PSO algorithm. For 38 data samples from published papers, 30 samples are used as the training set, and 8 samples are used as the test set. For the test samples, the error of five samples ranges between-5 and 5 mm. The error of only one sample is slightly greater than 10 mm. The proposed PSO-ELM method demonstrates good prediction performance of smax. A deformation parameter of the nonuniform displacement mode for the tunnel cross-section is calibrated based on predicted smax. When the determined nonuniform displacement mode is used as the boundary condition of the tunnel cross-section, the GDF of a shallow circular tunnel is analytically predicted based on the complex-variable method prior to tunnel excavation. For a specific engineering case,i.e., the Heathrow Express tunnel, the proposed PSO-ELM-analytical method can well predict the surface-settlement trough curve, horizontal displacements at different depths, and vertical displacements above the tunnel.展开更多
基金supported by the National Natural Science Foundation of China (Grant No. 52025084)。
文摘Based on machine-learning(ML) and analytical methods, a hybrid method is developed herein to predict the ground-displacement field(GDF) caused by tunneling. The extreme learning machine(ELM), as a single hidden layer feedforward neural network, is used as an ML model to predict maximum settlement smaxof the ground surface. The particle swarm optimization(PSO) algorithm is applied to optimize the parameters for the ELM method, namely, weight and bias values from the input layer to the hidden layer. The mean square error of the k-fold cross validation sets is considered the fitness function of the PSO algorithm. For 38 data samples from published papers, 30 samples are used as the training set, and 8 samples are used as the test set. For the test samples, the error of five samples ranges between-5 and 5 mm. The error of only one sample is slightly greater than 10 mm. The proposed PSO-ELM method demonstrates good prediction performance of smax. A deformation parameter of the nonuniform displacement mode for the tunnel cross-section is calibrated based on predicted smax. When the determined nonuniform displacement mode is used as the boundary condition of the tunnel cross-section, the GDF of a shallow circular tunnel is analytically predicted based on the complex-variable method prior to tunnel excavation. For a specific engineering case,i.e., the Heathrow Express tunnel, the proposed PSO-ELM-analytical method can well predict the surface-settlement trough curve, horizontal displacements at different depths, and vertical displacements above the tunnel.
