Evaluation of geotechnical parameters affecting the penetration rate of TBM using neural network (case study)

Abstract

Rate of penetration of a tunnel boring machine (TBM) in a rock environment is generally a key parameter for the successful accomplishment of a tunneling project. Tunneling time is one of the factors that determine the tunneling method (blast or mechanical work). In traditional method, tunneling time depends on the progression function of blast cycle, while in mechanized tunneling method the required time for boring is determined by advance rate and consequently penetration rate. The advance rate is a function of penetration rate of TBM, which is the ratio of the length of boring part of the tunnel to the excavated time of the same part. Penetration rate, on the other hand, is a function of lithology and geomechanical properties of the rock and also the boring machine. So far, many efforts have been made to develop a method to predict the penetration rate (TBM) of the machine. Among these methods are empirical models, particularly NTH model, including models that can be used to predict the penetration rate. In this study, statistical analysis was used to obtain importance of parameters involved in penetration rate and compare the performance of neural networks with other mathematical models based on the principles of probability. In addition, the artificial neural network (ANN) was compared with models of Innaurato and NTH, which indicates high performance of neural networks in predicting penetration rate compared with the other two models. As a result, neural network was chosen and then proceeded to build the network optimized.