Prediction of shear strength parameters of Zhoushan soft soils using probabilistic analysis and XGBoost

Abstract

Thick soft muddy clay layers in coastal regions pose significant engineering challenges due to their poor mechanical properties, threatening projects’ stability and safety. Accurate determination of soil parameters is essential for reliable geotechnical design. This study utilizes 178 soil samples from multiple projects in Zhoushan, China, to analyze the statistical distributions and probabilistic characteristics of key parameters, including water content, unit weight, void ratio, degree of saturation, and Atterberg limits. Focusing on shear strength as a critical indicator, the research investigates its relationship with various physical properties. By employing readily available physical parameters as inputs, a machine learning model combining Bayesian optimization and XGBoost (BOA-XGBoost) is developed to predict shear strength parameters. The BOA-XGBoost model achieved R² values of 0.859 for cohesion and 0.814 for the internal friction angle, with corresponding low error metrics, demonstrating superior performance and generalization capability compared to other models. This approach enables efficient and timely prediction of mechanical properties, offering practical guidance for the design and construction in soft clay areas.

Keywords

Soft soils / Machine learning / Parameter prediction / XGBoost / Shear strength parameters

References

[1] Khanlari GR, Heidari M, Momeni AA, Abdilor Y. Prediction of shear strength parameters of soils using artificial neural networks and multivariate regression methods. Eng Geol, 2012, 131–132: 11-18.

[2] Ching J, Phoon KK, Li KH, Weng MC. Multivariate probability distribution for some intact rock properties. Can Geotech J, 2019, 56(8): 1080-1097.

[3] Ma X, et al. . Characteristics of physical parameters and predictive modeling of mechanical properties in loess-like silty clay for engineering geology. Eng Geol, 2024, 339: 107672.

[4] Li Y, Rahardjo H, Satyanaga A, Rangarajan S, Lee DTT (2022) Soil database development with the application of machine learning methods in soil properties prediction. Eng Geol 306:106769

[5] Niyogi A, Ansari TA, Sathapathy SK, Sarkar K, Singh TN (2023) Machine learning algorithm for the shear strength prediction of basalt-driven lateritic soil. Earth Sci Inform 16:899–917

[6] Han D, Xue, et al (2024) Machine Learning-Based Prediction of Shear Strength Parameters of Rock Materials. Rock Mech Rock Eng. 57(10):1–25

[7] Reddy NDK, Gupta AK, Sahu AK (2022) A novel soil liquefaction prediction model with intellectual feature extraction and classification. Adv Eng Softw 173:103233

[8] Pham BT, Son LH, Hoang TA, Nguyen DM, Bui DT. Prediction of shear strength of soft soil using machine learning methods. CATENA, 2018, 166: 181-191.

[9] Diksha, Dev N, Goyal PK (2023) Prediction of compressive strength of alccofine-based geopolymer concrete. Iran J Sci Technol Trans Civil Eng 48:2077–2093

[10] Spearman C. The proof and measurement of association between two things. Am J Psychol, 1987, 100(3/4): 441-471.

[11] Chen T, Guestrin C (2016) XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp 785–794

[12] Pelikán M, Goldberg D, Cantú-Paz E (1999) BOA: the Bayesian optimization algorithm. In: Proc Genetic & Evolutionary Computation Conference, vol 1, p 525–532

[13] Snoek J, Larochelle H, Adams RP (2012) Practical Bayesian Optimization of Machine Learning Algorithms. Statistics 2:2951–2959.

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

[15] Rumelhart DE, Hinton GE, Williams RJ. Learning representations by back-propagating errors. Nature, 1986, 323(6088): 533-536.

[16] Cortes C, Vapnik V. Support-vector networks. Mach Learn, 1995, 20(3): 273-297.

Funding

Research Funds for the Central Universities(22120230077)

RIGHTS & PERMISSIONS

The Author(s)