| Nowadays, a lot of research has been done on the mechanical properties anddeformation performance of concrete-filled rectangular steel tubular columns by theresearchers at home and abroad, but the methods they have chosen mostly are testresearch, analytical method and numerical analysis method, all of which have somelimitations. However, neural network algorithm is an efficient method for solvingnonlinear problem. Therefore, a neural network model is created to study themechanical properties of concrete-filled rectangular steel tubular columns, whichmakes contribution for further research and application of the concrete-filledrectangular steel tubular members.In the paper, several calculation theories and methods of the bearing capacity ofconcrete-filled rectangular steel tubular columns were presented, and their defects andlimitations were pointed out. The brief introduction to the finite element analysismethod of concrete-filled rectangular steel tubular columns was given. The necessityof finding an effective and uniform calculation method of bearing capacity was putforward, and the neural network method was applied in the calculation of bearingcapacity of concrete-filled rectangular steel tubular columns.ANSYS, which involves large geometric deformation, material nonlinearity andcontact condition, was applied in the analysis of concrete-filled rectangular steeltubular columns under axial compression and eccentric compression. The distributionstress, mechanical property and deformation performance were gained. The resultsdemonstrated that the axial compressive concrete-filled rectangular steel tubularcolumns mostly occurred local buckling failure caused by the material strengthdamage, including the drum-shaped damage and shear damage, and shear failure wasmore and more obvious with the increase of length-width ratio. While eventuallydestroyed, each column all had one continuous buckling loop. The angle between thebuckling loop and the horizontal plane was approximately0-45。. When the columnhad a larger slenderness, overall bending failure would happen. The eccentriccompressive concrete-filled rectangular steel tubular columns mostly occurred localbuckling failure caused by the material strength damage as well. Eventually, the failure mode of steel tube outer convex deformation occurred on the compressiveshort edge and adjacent two long edges close to a cross section of concrete-filledrectangular steel tubular columns.Neural network algorithm was used to respectively calculate the bearing capacityof concrete-filled rectangular steel tubular columns under axial compression andeccentric compression. The results demonstrated that the calculation values of neuralnetwork were closer to the experimental values than the formula calculated values. Inthe paper, the applicable scope of the BP neural network model for the bearingcapacity of concrete-filled rectangular steel tubular columns under axial compressionwas: the length of the section60.2~300mm, length-width ratio1.0~2.02, width-thickness ratio7.5~90.57, slenderness ratio2.31~103.92,steel yield strength190~550MPa, concrete compressive strength12.53~79.85MPa; eccentric compression:the length of the section76.2~300mm, length-width ratio1.0~2.0, width-thicknessratio7.45~73.58, slenderness ratio9.01~92.66,steel yield strength194~495MPa,concrete compressive strength18.6~89.4MPa, eccentricity ratio0.1067~1.9580. |
