Examining the possibility of estimating activity time distributions from recorded downtimes in TBM tunnelling operations using discrete event simulation (DES) models

Abstract

While the tunnel boring machines (TBMs) have been broadly accepted by tunneling industry, the calculation of TBM advance rates has always been a challenge in setting project budgets and schedules. TBM advance rate is a function of penetration rate and machine utilization, which is simply the ratio of boring time to total project time. Utilization is an essential factor that must be taken into consideration in performance predictions since it is the part of the advance rate calculation that defines the efficiency of tunneling operations. Previous researchers have tried to estimate TBM utilization using various models. Some of these models were empirical, some were statistical, and some were numerical. The CSM2020 is a recently proposed model that uses discrete event simulation (DES) to incorporate stochastic nature of tunneling activities in modeling tunneling operation by TBM to estimate machine utilization. The model is based on using recorded downtime data from previous projects as the input to estimate TBM utilization rate as the output. Given that various activity times have not been commonly recorded or studied in many of the tunneling operations, this thesis aims to evaluate the possibility of estimating these tunneling activity durations from the recorded machine downtimes and utilization rate. This allows for expansion of the databases of activity time distributions that can be used for use of DES modeling for estimation of machine utilization in the future TBM tunneling projects. To that end, recorded downtimes from a selected project site were used to create representative time distributions for completion of a section of the selected tunnel where a slurry shield TBM was employed. These distributions were used to simulate the tunneling operations using DES modeling, or CSM 2020 model, and the results compared with the recorded downtimes in the project. A new set of time distributions was subsequently created by varying the mean value for the activity times in the original datasets, and the new distributions were used in sensitivity analyses to verify the relationship between the activity times and the resulting downtimes and overall machine utilizations. The sensitivity analyses facilitated observation of the impact of various activity time distributions on utilization and overall downtime contribution, hence making it possible to evaluate the possibility of estimating the activity times from recorded downtime data in the field. Because the CSM2020 uses a combination of activity time and delay times as the input for modeling, it was concluded that estimation of activity times is not possible with the current version of the model. The except is the activities that are directly linked or listed as downtime, such as segment installation.