To study the vertical compressive bearing characteristics of large-diameter rock-socketed cast-in-place piles, eight manually-excavated rock-socketed cast-in-place piles were subjected to vertical compressive on-site load and pile stress tests. The test results showed that the load–displacement (Q-s) curves of the eight test piles were all slow-varying, and the settlement of the piles was less than 11 mm, which met the minimum engineering requirements. The unloading rebound rate was between 55 and 75%, and the elastic working properties of the piles were apparent. The pile axial force gradually decreased with depth, and the slope of the axial force distribution curve reached a minimum in the moderately weathered muddy siltstone layer while the pile side friction resistance reached its maximum value. Pile end friction increases with the increase of load. But the pile end resistance was inversely proportional to the single pile length-to-diameter (L/D) ratio and the depth of rock embedment for the pile. The percentage of pile side friction resistance under maximum load was 86%, indicating that these were characteristic friction piles. Based on the test results and the current Chinese code, the friction coefficient of the pile side soil layer η and the total resistance coefficient of the rock-socketed section ζ were introduced. A revision to the calculation equation for the vertical bearing capacity of the rock-socketed cast-in-place pile in the code was proposed, together with an optimization design method for large-diameter rock-socketed cast-in-place piles.

During deep foundation pit construction, slight improper operations may lead to excessive deformation, resulting in engineering accidents. Therefore, how to accurately predict the deformation of the deep foundation pit is of significant importance. With advancements in artificial intelligence technology, machine learning has been utilized to learn and simulate complex nonlinear relationships among various factors influencing foundation pit deformation. Prediction accuracy is significantly improved, and the dynamic trend of foundation pit deformation is accurately grasped to curb the risk of safety accidents. This paper systematically reviews the current applications of machine learning in deep foundation pit deformation prediction. The fundamental principles of machine learning models, including neural networks, support vector machines, and Bayesian networks, are elaborated in the context of their application to deep foundation pit deformation prediction. The application effects of various machine learning models in predicting deep foundation pit supporting structure deformation, surrounding surface settlement, and assessing foundation pit risks are summarized. The limitations and future development prospects of current machine learning models for deformation prediction in deep foundation pit construction are discussed. The research results offer valuable insights for the application and advancement of machine learning in the deep foundation pit deformation prediction field.

We conducted anchoring performance, stress distribution, and full-scale indoor pulling tests on glass-fiber-reinforced polymer (GFRP) bolts. The tests were conducted using finite element software while considering the multi-interface contact and BK criterion by using the cohesive element to simulate the contact relations between the anchor rod body and concrete and building an axial symmetry calculation model of the GRFP bolt and concrete. The results indicated that the finite element model based on cohesive element accurately represents the load–displacement relationship of the GFRP bolt and the distribution law of axial stress along the anchoring length. In addition, the simulation outcomes of the load–displacement relationship were in good agreement with the measured test values. Under the same load, the axial-force-transferred depth of the bolt body was identical regardless of the anchorage length. As anchoring length increases, the pull load on the bolt and the decay rate of axial stress along the anchoring length rises gradually. There is a critical value for the anchorage length of the bolt.