Only 14 pages are availabe for public view
This thesis presents an assessment study of the TBM performance in clay soils by selecting the key-influencing parameters pertaining to tunnel, soil and TBM in order to predict and maximize the penetration rates.
6.1.1 Parameters selection and influence
During this study, several parameters were selected to be used for carrying out the computational and statistical-based models.
These parameters were categorized in 4 categories considering the tunnel, soil, TBM and TBM operator experience. The statistical analysis using the coefficient of determination (R2) and Pearson correlation coefficient has shown that all chosen parameters have significant direct or indirect effect on the penetration rate.
These parameters are tunnel depth (m), tunnel slope, face pressure (bar), air bubble pressure (bar), cutterhead speed (rpm), total thrust force (kN), feed flow (m3/hr.), slurry flow (m3/hr.), average grout pressure (bar), clay content (%), silt content (%), plastic limit (%), bulk density (%), liquidity index (%), undrained shear strength (kN/m2) and TBM operator experience.
While comparing the inter-correlation among the TBM Operating parameters and the penetration rate, it was concluded that the first bored (North) tunnel has a slightly different performance compared to the South tunnel that was excavated at later stage. This behavior could be attributed to the fact that the North tunnel was bored through virgin (green-field) ground conditions, while the South tunnel was later bored through
disturbed and overstressed ground media caused by the excavation of the first (North) tunnel.
Furthermore, it was concluded after analysis, that all of these parameters have varying levels of contribution on TBM performance. Where, certain parameters were found to have major contribution than other parameters as follows:
1. The main TBM Operating Parameters were found to be the most important parameters affecting the penetration rate, where the Cutterhead Speed was found to be the most important factor with normalized importance ratio 100%, then Air Bubble Pressure with normalized importance ratio 39%, then Total Thrust Force with 27.2%, then Face Pressure with 23.5%.
2. Soil Parameters came at the second order of their impact on the tunnel boring performance; where the undrained shear strength of the clay layer was found to be the most significant contributing factor, with normalized importance ratio 43.8%. This is followed by the factors pertained to the clay composition and plasticity characteristics with far less impact, such as Liquidity Index, Silt Content, Clay Content and Plastic Limit with normalized importance ratios 20%, 18.4%, 16.1% and 14.8% respectively.
3. Tunnel geometrical parameters including slope and depth were found also to have a considerable influence with normalized importance ratios of 16.7% and 16.3% respectively.
4. The rest of the considered key-parameters, namely; Slurry Flow, Average Grout Pressure, Feed Flow and Bulk Density
were found to have the least contributing influence with normalized importance ratios 14.2%, 13.2%, 9% and 5.2% respectively.
6.1.2 Modeling results
Multi-Linear Regression and Artificial Neural Network modeling were undertaken to assess these parameters, which affect the TBM penetration rates. The developed models in this research can be further used to assist the tunneling TBM operators and designers. Doing so, in return, can lower the overall cost and maximize the efficiency by completing the tunnel construction in shorter time and steadier performance.
Based on previous studies and the obtained results from this research, it was confirmed that such modeling approaches, whether using the multi-linear regression or the artificial neural network, can efficiently predict the TBM performance with acceptable accuracy for tunneling in soft soils. However, the ANN-based models have proven better prediction than the multi linear regression models.
Furthermore, while applying the regression analysis using MLREM and MLRBM, the MLREM have shown better indices than the MLRBM, which enhance the criteria of parameters selection that all selected parameters have significant effect on the penetration rate and shall be used while developing the ANN-based models.
Each developed model using multi-linear regression analysis or artificial neural network technique has been used to predict the penetration rates in mm/rot. Then, the prediction performance indices (R2, MAPE, RMSE and VAF) have been calculated for the obtained models as an
evaluation tool to select the best performance prediction model. The prediction performance indices using the ANN have shown better indices than the MLREM and MLRBM models, which could be attributed to the efficiency of the ANN and ability for adaptation and capturing the potential relation between the independent variables and the non-linearity between the variables.
Following this, and while the obtained indices from the ANN-based Models were varied from model to the other, a rating system was obligatory to be developed. This rating system has simplified the criteria of evaluating and selecting the best model.
Consequently, in this study, the North tunnel ANN-based Model 5 and the South tunnel ANN-based Model 1 yielded the best performance models in North and South tunnels respectively. These models achieved the highest R2 values with minimum errors. Accordingly, the best obtained models in both tunnels has proven the high-efficiency and that such model can successfully predict the penetration rates.
However, the models’ validation has revealed another feedback, as after using the North tunnel models for validation using the South tunnel data measurements, the calculated prediction performance indices showing that the North tunnel ANN-based Model 3 has the best indices with R2 of 0.63 and the lowest RMSE of 3.15 mm/rot. (0.19 m/hr.)
Accordingly, the validation process is very important stage and could be considered as a must to be implemented after developing the prediction models as a further evaluation tool for the prediction models.
In that case, the ANN-based Model 3 in the North tunnel has shown the best prediction model and could be used in the future investigations while estimating the TBM performances in similar soil conditions (clay
soils) in order to provide time estimate and potentially minimizing the downtime and overcome any difficulties during the tunneling stage with the TBM. This could be achieved by changing the values of the input parameters in the developed models to predict and check the impact on the output parameter (penetration rate). Nevertheless, the developed ANN-based prediction models were able to capture such impact in very satisfactory level.
6.2 Thesis Contribution
This thesis mainly contributed in studying performance of tunneling using TBM in soft soils, since most of previous studies have been focusing on tunnel boring machines performance in hard rock ground and rock formation.
The following points are showing the main contributions of this research:
• This research focused on tunneling using bentonite slurry TBM supported by separation treatment plant, while most researchers focused on single and double shield TBM.
• This research focused on tunneling in soft soils (clay soils), while most researchers studied the TBM performance in hard rock soils.
• This thesis provides the methodology to the researchers to investigate, choose and select different parameters affecting the TBM performance while developing the prediction models.
• The developed models are showing good representation of the TBM performance and could be used in the future to predict the performance of other TBM in different projects.
• Having the selected ANN-based Models validated using the data measurements from the other tunnel, confirmed that the selected parameters can successively and efficiently represent the TBM performance in soft soils.
6.3 Future Work
This thesis has studied the TBM performance in soft soils (clay soils), while the used soil parameters are clay content, silt content, plastic limit, bulk density, liquidity index and undrained shear strength. The best-obtained model (North tunnel ANN-based model 3) could be used to predict the penetration rates of the TBM in different projects, and this could be implemented by changing the input parameter values to predict and check the impact on the output parameter (penetration rate). Furthermore, the surface settlement measurements associated with TBM advance can be considered in the future models to control the ground movement.
Finally, it is recommended to use the best obtained model in this study for further investigations and validation in predicting the TBM performance in projects with similar soil and tunneling conditions. So that; the prediction models developed in this research can be further investigated, improved and validated.