터널 심발공 비장약량 예측을 위한 머신러닝 기법의 적용성 평가
Evaluation of Machine Learning Techniques for Predicting Specific Charge of Cut Hole Blasting in Tunnel
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
본 연구는 국내 18개 터널 현장의 데이터를 활용하여 터널 발파 설계에서 심발공의 비장약량을 예측하 기 위한 최적의 머신러닝 모델을 개발하는 것을 목적으로 한다. 입력변수로는 굴진장, 단위면적당 천공 수, 굴착단면계수, 전체 막장면의 단면적, 심발공의 최대 및 최소 공간격, 심발 공법 유형, 사용한 폭약의 종류, RMR, 최대 토피고, 암종이 사용되었다. 선형회귀(LR), Lasso회귀, Ridge회귀, 서포트벡터머신 (SVM), K-최근접이웃(KNN), 랜덤포레스트(RF), LightGBM, 히스토그램 기반 그래디언트부스팅 (HGB), 인공신경망(ANN)의 9가지 머신러닝 모델이 적용되어 비교되었다. 랜덤포레스트 모델이 테스트 데이터에서 R2 0.852, RMSE 0.082로 가장 우수한 성능을 보였다. SHAP (SHapley Additive exPlanations) 분석을 통한 변수 중요도 분석 결과, 심발 공법 유형이 가장 큰 영향을 미치는 것으로 나 타났으며(Mean Absolute SHAP Value: 0.1627), 최소 공간격(0.0393)과 최대 공간격(0.0170)이 그 뒤를 이었다. 반면, 최대 토피고(0.0041), 암반 등급(0.0045), 사용한 폭약 종류(0.0029), 굴진장 (0.0008)은 상대적으로 낮은 중요도를 보였다. 본 연구에서 개발된 모델은 터널 현장의 다양한 지질 조 건에서 최적의 비장약량을 예측하는데 활용될 수 있으며, 이는 터널 건설 프로젝트의 효율성과 안전성을 크게 향상시킬 수 있다.
Abstract
This study aims to develop an optimal machine learning model for predicting the specific charge of cut holes in tunnel blast design using data collected from 18 tunnel sites in South Korea. Input variables included round length, number of charge holes per area, excavation section coefficient, cross-sectional area, maximum and minimum spacing of cut holes, cut method type, explosive type, RMR, maximum overburden depth, and rock type. Nine different machine learning models were applied and compared: Linear Regression (LR), Lasso Regression, Ridge Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), Light Gradient Boosting Machine (LightGBM), Histogram-based Gradient Boosting (HGB), and Artificial Neural Network (ANN). The Random Forest model showed the best performance with an R2 of 0.852 and RMSE of 0.082 on test data. Variable importance analysis using SHAP (SHapley Additive exPlanations) revealed that cut method type was the most influential factor (Mean Absolute SHAP Value: 0.1627), followed by minimum spacing (0.0393) and maximum spacing (0.0170). In contrast, maximum over- burden depth (0.0041), RMR (0.0045), explosive type (0.0029), and round length (0.0008) showed relatively low importance. The developed model can be utilized to predict optimal specific charge under various geological conditions in tunnel sites, significantly enhancing the efficiency and safety of tunnel construction projects.
참고문헌 (23)
- Adesida, P. A., "Powder factor prediction in blasting operation using rock geo-mechanical properties and geometric parameters", International Journal of Mining and Geo-Engineering, 56(1), pp.25~32 (2022)
- Alipour, A., "Application of ANNs and MVLRA for estimation of specific charge in small tunnel", International Journal of Geomechanics, 12(2), pp.189~192 (2012)
- Alipour, A., "Support vector machines for the estimation of specific charge in tunnel blasting", Periodica Polytechnica Civil Engineering, 65(3), pp.967~976 (2021)
- Bieniawski, Z. T., "Engineering classification of jointed rock masses", Civil Engineering= Siviele Ingenieurswese, 1973(12), pp.335~343 (1973)
- Choi, B. H., "Tunnel Blasting Design Suited to Given Specific Charge", Explosives and Blasting, 27(2), pp.33~41 (2009)
- "Home page" (2024)[URL]
- Freund, Y., "Experiments with a new boosting algorithm", icml, 96, pp.148~156 (1996)
- Hanwha Corporation, "Technical Manual of Commercial Explosives Products", Hanwha Corporation, ~ (2017)
- Hong, J. P., "Estimation of Cerchar abrasivity index based on rock strength and petrologicalcharacteristics using linear regression and machine learning", Journal of Korean Tunnelling and Underground Space Association, 26(1), pp.39~58 (2024)
- Hwang, H., "A Study on pilot Test of the Composition Presplitting Angle Cut method in tunnel blasting", Journal of Korean Tunnelling and Underground Space Association, 4(4), pp.287~300 (2002)
- Jong, Y. H., "Influence of geological conditions on the powder factor for tunnel blasting", International Journal of Rock Mechanics and Mining Sciences, 41(3), pp.533~538 (2004)
- Jong, Y. H., "Application of Neural Network Theory for Predicting Specific Charge and Vibration Velocity in Tunnel Blasting", Proceedings of the Korean Society for Rock Mechanics Conference and Seminar, pp.75~84 (2001)
- Jung, D. H., "A Study on the Determination of Suitable Specific Charge in Tunnel Blasting Design", Journal of Korean Tunnelling and Underground Space Association, 3(1), pp.21~26 (2001)
- Kahriman, A., "Estimation of the powder factor in bench blasting from the Bond work index", Mining Technology, 110(2), pp.114~118 (2001)
- Kim, Y. K., "A fundamental study on the automation of tunnel blasting design using a machine learning model", Journal of Korean Tunnelling and Underground Space Association, 24(5), pp.431~449 (2022)
- Kwon, K. B., "A study on EPB shield TBM face pressure prediction using machine learning algorithms", Journal of Korean Tunnelling and Underground Space Association, 24(2), pp.217~230 (2022)
- Lee, J. K., "A study on the optimization of tunnel support patterns using ANN and SVR algorithms", Journal of Korean Tunnelling and Underground Space Association, 24(6), pp.617~628 (2022)
- Min, H. D., "Blast Design Methods Considering Site Specific Characteristics(II)", Journal of Korean Tunnelling and Underground Space Association, 7(4), pp.63~80 (2005)
- Min, H. D., "A Case of Application in Hard Rock Tunnel and Development of High Performance Emulsion Explosives(MegaMEX)", Explosives and Blasting, 23(1), pp.55~64 (2005)
- Son, J. H., "A Numerical Study on the Reduction Effect of Blasting Vibration with Cut Method", Explosives and Blasting, 37(1), pp.1~13 (2019)
- Taiwo, B. O., "Explosive Utilization Efficiency Enhancement: An Application of Machine Learning for Powder Factor Prediction using Critical Rock characteristics", Heliyon, 10(12), pp.e33099~ (2024)