A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Predicting Rock Unconfined Compressive Strength Based on Tunnel Face Boreholes Measurement-While-Drilling Data
Quick and accurate acquisition of the uniaxial compressive strength (UCS) of the surrounding rock at the tunnel face effectively ensures the safety of tunnel construction. This paper proposes a model for estimating the USC of the tunnel surrounding rock based on boreholes measurement-while-drilling data and stacking ensemble algorithm. Firstly, four original drilling parameters of hammer pressure (Ph), feed pressure (Pf), rotation pressure (Pr), and feed speed (Vp) as well as the rock UCS are collected from 1489 rock UCS test cases. Then, data cleaning and feature extension are carried out, and a UCS estimation database containing 66 features of the drilling parameters is established. Furthermore, traditional machine learning algorithms (SVM, KNN, RF, ET, GB, Bag), Bayesian optimization, cross-validation, and staking ensemble algorithms are employed to build a rock UCS estimation model. The performance of six traditional and integrated machine learning models is comparatively analyzed. The R2, RMSE and MAE of the prediction set are used as model performance evaluation metrics. The results show that the ensemble model performs best with an R2 of 87.9%. Finally, the reliability of the model is verified by field tests. Compared with the traditional field UCS testing method, this method can accurately and quickly predict the UCS of rocks without additional manpower and material resources, which possesses a greater application prospect.
Predicting Rock Unconfined Compressive Strength Based on Tunnel Face Boreholes Measurement-While-Drilling Data
Quick and accurate acquisition of the uniaxial compressive strength (UCS) of the surrounding rock at the tunnel face effectively ensures the safety of tunnel construction. This paper proposes a model for estimating the USC of the tunnel surrounding rock based on boreholes measurement-while-drilling data and stacking ensemble algorithm. Firstly, four original drilling parameters of hammer pressure (Ph), feed pressure (Pf), rotation pressure (Pr), and feed speed (Vp) as well as the rock UCS are collected from 1489 rock UCS test cases. Then, data cleaning and feature extension are carried out, and a UCS estimation database containing 66 features of the drilling parameters is established. Furthermore, traditional machine learning algorithms (SVM, KNN, RF, ET, GB, Bag), Bayesian optimization, cross-validation, and staking ensemble algorithms are employed to build a rock UCS estimation model. The performance of six traditional and integrated machine learning models is comparatively analyzed. The R2, RMSE and MAE of the prediction set are used as model performance evaluation metrics. The results show that the ensemble model performs best with an R2 of 87.9%. Finally, the reliability of the model is verified by field tests. Compared with the traditional field UCS testing method, this method can accurately and quickly predict the UCS of rocks without additional manpower and material resources, which possesses a greater application prospect.
Predicting Rock Unconfined Compressive Strength Based on Tunnel Face Boreholes Measurement-While-Drilling Data
KSCE J Civ Eng
Ling, Xuepeng (author) / Wang, Mingnian (author) / Yi, Wenhao (author) / Xia, Qinyong (author) / Sun, Hongqiang (author)
KSCE Journal of Civil Engineering ; 28 ; 5946-5962
2024-12-01
17 pages
Article (Journal)
Electronic Resource
English
ANN Model for Predicting Mud Loss Rate from Unconfined Compressive Strength and Drilling Data
| Springer Verlag | 2024
Estimating aggregate properties from the unconfined compressive strength of rock
| British Library Online Contents | 1994
Rock strength characterization using measurement while drilling technique
| Springer Verlag | 2020