Prediction of crack coalescence stress in rock-like specimens with non-persistent joints under direct shear test based upon machine learning algorithms

Message:
Article Type:
Research/Original Article (دارای رتبه معتبر)
Abstract:
Concretes frequently contain joints and microcrack fractures, and the failure mechanism of these fractures is highly dependent on the pattern of crack coalescence between pre-existing flaws. Determining the non-persistent joints' failure behavior is an engineering challenge that incorporates several factors, including the ratio of the joint surface to the total shear surface, normal stress, and the mechanical characteristics of the concrete. This paper aims to utilize grey wolf optimizer (GWO) and gene expression programming (GEP) algorithms for the prediction of the crack coalescence stress (CCS). For this purpose, 8 input parameters affecting the CCS including jointing coefficient (JC), normal stress (σn), uniaxial compressive strength (σc), tensile strength (σt), Poisson's ratio (υ), modulus of elasticity (E), cohesion strength (C) and internal friction angle (φ) were selected based on the results of 450 direct shear tests conducted on specimens including 2 sets of non-persistent joints made of gypsum, cement, and water. The GWO and GEP techniques were then implemented. Three performance indicators of determination coefficient (R2), root mean square error (RMSE), and mean absolute error (MAE), were employed for the training and testing phases to evaluate the efficiency of the suggested models. The R2 values for GWO and GEP for the training phase were 0.962 and 0.938, respectively, while for the testing phase were 0.996 and 0.981, indicating that the GWO algorithm is more efficient than GEP. Moreover, the findings reveal that the GWO algorithm exhibits lower RMSE and MAE values in both the training and testing phases compared to the GEP method. However, it can be professed that the two methods used have high reliability and accuracy. Also, based on the GEP method, a formula was derived and presented for prediction of CCS. At last, according to the sensitivity analysis, it was found that the normal stress (σn) and jointing coefficient (Cu) have the greatest and least influence on CCS, respectively...
Language:
Persian
Published:
Journal of Aalytical and Numerical Methods in Mining Engineering, Volume:14 Issue: 40, 2024
Pages:
35 to 47
https://www.magiran.com/p2816292  
سامانه نویسندگان
  • Shadman Mohammadi Bolbanabad
    Author (3)
    Phd Student Mining Engineering, Mining, Materials and Mining Engineering, Tarbiat Modares University, Tehran, Iran
    Mohammadi Bolbanabad، Shadman
اطلاعات نویسنده(گان) توسط ایشان ثبت و تکمیل شده‌است. برای مشاهده مشخصات و فهرست همه مطالب، صفحه رزومه را ببینید.
مقالات دیگری از این نویسنده (گان)