Prediction of hard rock TBM penetration rate using particle swarm optimization

Abstract

The aim of this study is to predict the performance of tunnel boring machines (TBMS) using particle swarm optimization technique (PSO). With this aim, a database including intact rock parameters comprising of strength and brittleness, and rock mass properties such as distance between planes of weakness and orientation of discontinuities, together with field machine performance data, was established using data collected along a 7.5 km long hard rock mechanical tunnel. The particle swarm optimization technique was applied to develop new predictive model for TBM performance. Seven different PSO models were developed using the assortment of datasets having various percentages of rock type in the dataset. Additionally, the PSO model was developed using the entire dataset in random without paying attention to rock type to generalize the model. As a result of the developed models via a variety of generated testing and training datasets, it is concluded that Model 7 and its resultant equation are the most precise among the seven models tested.

Introduction

The prediction of tunnel boring machine (TBM) performance is one of the complex and crucial tasks encountered frequently to excavate the mechanical tunnels. Estimating the machine performance may reduce the risks related to high capital costs typical for excavation operation [1], [2]. Since the first TBM was built, various prognosis models have been developed based on both intact and mass rock properties together with machine specifications. Prediction of TBM performance requires the assessment of the penetration rate (ROP), the ratio of excavated distance to the operating time during continuous excavation phase and advance rate (AR), and the ratio of both mined and supported actual distance to the total time [1], [2], [3]. Developing the prediction models is one of the main tasks and it has been under progress since many years [2], [4], [5], [6], [7], [8], [9], [10], [11], [12]. Besides these theoretical and empirical models, Artificial Neural Networks (ANN) have been used to predict the rate of penetration [13], [14], [15]. Furthermore, fuzzy logic, genetic algorithms and ANNs have been utilized for establishing predictive models in hydrology, mining and civil engineering applications in recent years [15], [16], [17], [18], [19], [20], [21], [22]. A majority of these models are generated based on the experience gained and the data compiled from the past tunneling projects. Although many models have been produced for estimating the ROP with linear, nonlinear multivariable regression, ANN and fuzzy techniques, there has yet been no attempt made to estimate the rate of penetration using the PSO method.

The aim of the current study is develop a TBM performance prognosis using the particle swarm optimization algorithm. To obtain the goal, the database including strength and brittleness of intact rock, orientation and distance between planes of weakness in the rock mass, machine specifications and performance data, have been established by collecting the field and laboratory data from completed hard rock tunnel in the City of New York, USA. Afterward, using the established dataset, the PSO model is developed for predicting the machine performance by means of the rate of penetration in both fractured and mass rock conditions.