Bearing Capacity Prediction of Lightweight Steel Double L-Joints Based on Experimental Investigation and an Improved Gaussian Process Regression Algorithm

Abstract The vertical load-bearing capacity of earthquake-damaged prefabricated frame structures determines their ability to resist collapse during aftershocks. In particular, the response performance of the beam-column joint area of prefabricated frame structures to axial compression after earthquake damage plays a key role in determining the vertical load-bearing capacity of the structure. To study the bearing capacity of beam-column joints in post-earthquake prefabricated light steel frames, firstly, based on nine existing low-cycle repeated loading tests of beam-column joints in different structural prefabricated light steel frames, axial compression performance tests were conducted on nine joints with different degrees of damage and different structures by applying axial forces to the frame columns. The failure modes and bearing capacities of joints with different structures were compared. Secondly, the experimental structures were verified through numerical simulation, and a database was constructed based on experimental and finite element results. Finally, an improved particle swarm optimization algorithm was used to optimize GPR, and the IPSO-GPR algorithm was proposed to form a prediction method for the bearing capacity of light steel frame structures. The results show that the proposed bearing capacity prediction method can better reflect the stress characteristics of such specimens.