Predicting the penetration rate of TBM in Cretaceous limestone of the south of Tehran by machine learning method

Much research has been done so far in predicting the performance of tunnel boring machines in rock, but previous studies have often been conducted in conditions where chipping is the dominant mode, and machine performance prediction models have also been developed in these conditions. When, for any reason, the thrust force is not enough to penetrate the rock and the cutting is not complete, the efficiency of the existing models reduces, and new models should be provided for grinding conditions. In this study, the data obtained from the southern extension of Tehran subway Line 6 (SEL6), which part of its route was located in the Cretaceous limestone units and due to the insufficient disc cutter thrust, grinding was the dominant mode, was evaluated. The purpose of implementing different machine learning algorithms is to predict the penetration rate of cutterhead in studied limestone in grinding conditions. So, different linear and non-linear regression algorithms have been implemented and finally, the results have been compared. Since some of these algorithms are among the latest machine learning approaches, the prediction error is much less than in conventional regression methods.