Evaluating and Predicting the Stability of Roadways in Tunnelling and Underground Space Using Artificial Neural Network-Based Particle Swarm Optimization

Abstract

In this study, a new technique for predicting roadways stability in tunneling and underground space was proposed based on a combination of particle swarm optimization (PSO) algorithm and artificial neural network (ANN), called ANN-PSO model. The dataset from five tunneling and underground mines in the 2006-2019 period was recorded monthly and used for this aim with 145 observations. Accordingly, the stability of roadways in tunneling and underground space was evaluated based on the geomechanical parameters. The uniaxial compressive strength, internal friction angle, rock mass rating, tensile strength, cohesion, density, Young's modulus, shear strength, and slake durability were used as the influence parameters for evaluating and predicting roadway stability. Five other intelligent methods were also developed and compared with the proposed ANN-PSO model in order to have a comprehensive assessment, including support vector machine (SVM), hybrid neural fuzzy inference system (HYFIS), multiple linear regression (MLR), classification and regression tree (CART), and conditional inference tree (CIT). Three model assessment indices, such as MAE, RMSE, and R2 were used to simulate the accuracy of the roadway stability predictive models. Besides, ranking and color intensity techniques were also applied for further assessment. The results showed that the stability of the roadway could be accurately assessed by the proposed ANN-PSO model with an RMSE of 9.708, R2 of 0.972, and MAE of 7.161. They also revealed that the proposed ANN-PSO model yielded the most outperformed over the other models. The sensitivity analysis resulting also indicated that the uniaxial compressive strength, shear strength, quench durability index, density, and rock mass rating were the most important parameters for predicting roadway stability. They should be used in predicting the stability of roadways in tunneling and underground space.