This study is to develop a practical empirical model based on a database compiled from 38 tunnel sections along the Karaj water conveyance tunnel for estimating TBM performance with emphasis on Field Penetration Index (FPI). A multiple linear regression correlating FPI with basic RMR input parameters and three additional parameters including tunnel depth, Elastic modulus and the angle between tunnel axis and discontinuity planes was presented. Analysis of the first developed model showed that Alpha angle, tunnel depth and groundwater condition must be excluded from the model because of their poor correlation with FPI. Analyzing the new model by elimination of these parameters resulted in creating a new predictive model with no significant multi-collinearity between independent variables. The comparison between the predicted and actual measured FPIs showed high strength of correlation with coefficient of 0.84. This paper discusses previous works in this area, reviews the available data from Karaj tunnel project, methodology for analysis, and introduces a convenient empirical predictive model for TBM performance.
Due to accelerated development of mechanized tunneling with tunnel boring machines (TBMs) rather than the classical drill and blast method, the attempts for correlating TBM performance with rock mass properties is crucial because of the significant role machine performance has on project scheduling and cost. On the other hand, large variability in rock mass properties even in short intervals and the complex interaction between the rock mass and TBM make performance prediction more difficult. TBM performance and operational characteristics, their relationship with geological conditions, and also the physical and mechanical properties of rock and the rock mass have been the subject of extensive research. However, there is no universally accepted model for TBM performance prediction. This is probably due to the fact that “estimating TBM performance involves understanding the rock fragmentation process in wide range from micro-scale (i.e. the interaction of surface contact of rock material and cutter tip) to macro- scale (i.e. the interaction of rock mass and TBM)” . TBM performance predictive models are categorized into three distinguished groups, namely, theoretical, semi-theoretical and empirical ones. In empirical models researchers have tried to present models for TBM performances which are mainly based on past experience and the statistical interpretation of previously recorded field data. The accuracy and reliability of these models depend on the quality and amount of the data. One example to field data and rock mass properties is the Norwegian hard rock TBM prognosis model developed by Blindheim  at the Norwegian University of Science and Technology (NTNU). Afterwards, the model has been updated several times with more TBM tunneling case histories [3, 4]. Among other empirical models, the efforts of other researchers [5-11], considering several rock and rock mass parameters along with machine design and operational parameters to estimate machine performance is noted. Some authors have also correlated TBM performance to rock mass classification systems [11-17]. In this study, this limitation is mitigated by using multivariate linear regression analysis of RMR input parameters.