Prediction of mixed-mode I/II fracture toughness of rock-concrete bi-material disc with interface crack: Interpretable NRBO-XGBoost-SHAP model and experimental validation

Abstract

Accurately determining the effective fracture toughness (K_eff) of rock-concrete (R-C) bi-materials, governed by interface inclination and ambient temperature, is a prerequisite for assessing their structural stability. This study developed a hybrid NRBO-XGBoost prediction model using the Newton-Raphson-Based Optimizer (NRBO) to tune the hyperparameters of Extreme Gradient Boosting (XGBoost) model. The established model was developed based on 154 datasets obtained from laboratory tests and numerical simulations with the cracked straight-through Brazilian disc (CSTBD) specimens, including twelve input parameters. The NRBO-XGBoost model for K_eff prediction was investigated and compared with seven more models. Furthermore, the Shapley Additive exPlanations (SHAP) method was employed to quantify the contributions of inputs to K_eff to improve the interpretability of the developed model. Finally, new data were used to validate the model. Evaluation results demonstrate that metaheuristic optimization algorithms significantly enhance the performance of XGBoost, with NRBO-XGBoost performing the best. The models rank from highest to lowest prediction performance as follows: NRBO-XGBoost, WOA-XGBoost, PSO-XGBoost, XGBoost, RF, CatBoost, LightGBM, and AdaBoost. The interpretable analysis shows that the interface inclination angle exerts the dominant influence. The validation results demonstrate that NRBO-XGBoost achieves high predictive accuracy on a new dataset, showing promising implications for practical applications.

Keywords

Effective fracture toughness / R-C bi-material / Extreme Gradient Boosting / Newton-Raphson-Based Optimizer / SHAP

References

[1] Guo TF, Liu KW, Li X, Wu Y, Yang JC. Effects of thermal treatment on the fracture behavior of rock—concrete bi—material specimens containing an interface crack. Theor Appl Fract Mech 2023; 127:104071.

[2] Wu DY, Su HJ, Yu LY, Zhang FL, Qin H, Wei C, Geng ST. Experimental and numerical investigation of dynamic tensile behavior of granite—concrete bimaterials with different rough interfaces. J Build Eng 2024; 84:108565.

[3] Lu JY, Chen XN, Min XH, Wu CS, Rui YC, Guo TF. Numerical investigation of temperature—dependency fracture behavior of early—age concrete—sandstone interfaces. Geomech Geophys Geo—energ Geo—resour 2026; 12(1):24.

[4] Gao RB, Wu F, Li CB, Ye CF, He QC, Xie HP. Damage evolution and failure modes of coal—concrete composites with varying height ratios under cyclic loading. Int J Min Sci Technol 2025; 35(7):1093-106.

[5] Guo TF, Liu KW, Song RT. Crack propagation characteristics and fracture toughness analysis of rock—based layered material with pre—existing crack under semi—circular bending. Theor Appl Fract Mech 2022; 119:103295.

[6] Yu LY, Wu DY, Su HJ, Li SC, Liu RC, Geng ST. Influences of interface roughness and loading direction on tensile behavior of rock—concrete bimaterials under Brazilian test conditions. Rock Mech Rock Eng 2023; 56(8):5861—83.

[7] Guan JF, Zhang YL, Meng JF, Yao XH, Li LL, He SH. A simple method for determining independent fracture toughness and tensile strength of rock. Int J Min Sci Technol 2022; 32(4):707—26.

[8] Chang X, Lu JY, Wang SY, Wang SR. Mechanical performances of rock—concrete bi—material disks under diametrical compression. Int J Rock Mech Min Sci 2018; 104:71—7.

[9] Deng XJ, Shen MX, Zhao Y, Bi J, Wang CL, Zhang YF, Li Y, Ning L. Effect of interface roughness on mode I—II fracture behavior of concrete—sandstone. Theor Appl Fract Mech 2025; 138:104943.

[10] Williams ML. The stresses around a fault or crack in dissimilar media. Bull Seismol Soc Am 1959; 49(2):199-204.

[11] Rice JR. Elastic fracture mechanics concepts for interfacial cracks. J Appl Mech 1988; 55(1):98-103.

[12] Fang K, Fall M. Insight into the mode I and mode II fracture toughness of the cemented backfill—rock interface: effect of time, temperature and sulphate. Constr Build Mater 2020; 262:120860.

[13] Sheity DK, Rosenfield AR, Duckworth WH. Fracture toughness of ceramics measured by a Chevron—Notch diametral—compression test. J Am Ceram Soc 1985; 68(12):C325—27.

[14] Chang X, Guo TF, Zhang S. Cracking behaviours of layered specimen with an interface crack in Brazilian tests. Eng Fract Mech 2020; 228:106904.

[15] Liu KW, Guo TF, Yang JC, Ma SZ. Static and dynamic fracture behavior of rock—concrete bi—material disc with different interface crack inclinations. Theor Appl Fract Mech 2023; 123:103659.

[16] Guo TF, Liu KW, Ma SZ, Yang JC, Li X, Zhou KF, Qiu T. Dynamic fracture behavior and fracture toughness analysis of rock—concrete bi—material with interface crack at different impact angles. Constr Build Mater 2022; 356:129286.

[17] Soares JB, Tang TX. Bimaterial Brazilian specimen for determining interfacial fracture toughness. Eng Fract Mech 1998; 59(1):57-71.

[18] Banks—Sills L. 50th anniversary article: review on interface fracture and delamination of composites. Strain 2014; 50(2):98-110.

[19] Albaijan I, Fakhri D, Hussein Mohammed A, Mahmoodzadeh A, Hashim Ibrahim H, Babeker Elhag A, Rashidi S. Several machine learning models to estimate the effect of an acid environment on the effective fracture toughness of normal and reinforced concrete. Theor Appl Fract Mech 2023; 126:103999.

[20] Afrasiabian B, Eftekhari M. Prediction of mode I fracture toughness of rock using linear multiple regression and gene expression programming. J Rock Mech Geotech Eng 2022; 14(5):1421—32.

[21] Zhou J, Qiu YG, Khandelwal M, Zhu SL, Zhang XL. Developing a hybrid model of Jaya algorithm—based extreme gradient boosting machine to estimate blast—induced ground vibrations. Int J Rock Mech Min Sci 2021; 145:104856.

[22] Fattahi H, Ghaedi H, Armaghani DJ. Optimizing fracture toughness estimation for rock structures: a soft computing approach with GWO and IWO algorithms. Measurement 2024; 238:115306.

[23] Mahmoodzadeh A, Fakhri D, Hussein Mohammed A, Salih Mohammed A, Hashim Ibrahim H, Rashidi S. Estimating the effective fracture toughness of a variety of materials using several machine learning models. Eng Fract Mech 2023; 286:109321.

[24] Ziamiavaghi B, Toufigh V. Fracture toughness of alkali—activated concrete subjected to aggressive environment. Constr Build Mater 2024; 451:138796.

[25] Wang HQ, Liu ZX, Guo TF. Study on mechanical properties and fracture behaviours of rock—mortar specimens with an interface crack under radial compression. Constr Build Mater 2023; 406:133399.

[26] Chen TQ, Guestrin C. XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco: ACM; 2016. p. 785-94.

[27] Sowmya R, Premkumar M, Jangir P. Newton—Raphson—based optimizer: a new population—based metaheuristic algorithm for continuous optimization problems. Eng Appl Artif Intell 2024; 128:107532.

[28] Weerakoon S, Fernando TGI. A variant of Newton’s method with accelerated third—order convergence. Appl Math Lett 2000; 13(8):87-93.

[29] Li Z, Wang M, Xu R, Jiang JJ, Li J, Zhang ZQ, Yang AM. Prediction of viscosity of blast furnace slag based on NRBO—DNN model. Alex Eng J 2025; 119:124-37.

[30] Ahmadianfar I, Bozorg—Haddad O, Chu XF. Gradient—based optimizer: a new metaheuristic optimization algorithm. Inf Sci 2020; 540:131-59.

[31] Mirjalili S, Lewis A. The whale optimization algorithm. Adv Eng Softw 2016; 95:51-67.

[32] Lundberg SM, Lee SI. A unified approach to interpreting model predictions. Adv Neural Inf Process Syst 2017; 30:4765-74.

[33] Hu XJ, Liao D, Ma DD, Xie SL, Xie N, Hu HB, Gong XN. Machine learning models for predicting rock fracture toughness at different temperature conditions. Case Stud Constr Mater 2023; 19:e02622.

[34] Xu L, Wen SM, Huang HF, Tang YF, Wang YF, Pan CF. Corrosion failure prediction in natural gas pipelines using an interpretable XGBoost model: Insights and applications. Energy 2025; 325:136157.

[35] Sheng H, Ren ZG, Wang D, Li Q, Li PP. Estimation and interpretation of interfacial bond in concrete—filled steel tube by using optimized XGBoost and SHAP. Structures 2024; 70:107669.

[36] Breiman L. Random forests. Mach Learn 2001; 45(1):5-32.

[37] Liu WJ, Liu ZX, Xiong S, Wang M. Comparative prediction performance of the strength of a new type of Ti tailings cemented backfilling body using PSO—RF, SSA—RF, and WOA—RF models. Case Stud Constr Mater 2024; 20:e02766.

[38] He DS, Ying Z, Yang BY. Predicting the consumed heating energy at residential buildings using a combination of categorical boosting (CatBoost) and Meta heuristics algorithms. J Build Eng 2023; 71:106584.

[39] Luo M, Wang YF, Xie YH, Zhou L, Qiao JJ, Qiu SY, Sun YJ. Combination of feature selection and CatBoost for prediction: the first application to the estimation of aboveground biomass. Forests 2021; 12(2):216.

[40] Freund Y, Schapire RE. A decision—theoretic generalization of on—line learning and an application to boosting. J Comput Syst Sci 1997; 55(1):119-39.

[41] Xiao CJ, Chen NC, Hu CL, Wang K, Gong JY, Chen ZQ. Short and mid—term sea surface temperature prediction using time—series satellite data and LSTM—AdaBoost combination approach. Remote Sens Environ 2019; 233:111358.

[42] Ke G, Meng Q, Finley T, Wang T, Chen W, Ma W, Ye Q, Liu TY. Lightgbm: a highly efficient gradient boosting decision tree. Adv Neural Inf Process Syst 2017; 30.

[43] Wang D, Thunéll S, Lindberg U, Jiang LL, Trygg J, Tysklind M. Towards better process management in wastewater treatment plants: process analytics based on SHAP values for tree—based machine learning methods. J Environ Manag 2022; 301:113941.

[44] Kuang W, Guo TF, Chang X, Rui YC. Fracture behavior of heterogeneous layered sandstone under mixed loading after thermal treatment: a thermal—mechanical grain—based FDEM model. Comput Geotech 2025; 188:107599.

[45] Jia CJ, Wang L, Zhang S, Zheng YN, Shi CH, Peng Z. Effects of temperature gradient and curing time on shear properties of concrete—rock interfaces in geothermal tunnels: experimental investigations. Eng Fract Mech 2025; 325:111323.

[46] Taylor KE. Summarizing multiple aspects of model performance in a single diagram. J Geophys Res 2001; 106(D7):7183-92.