Flexural buckling behaviour of 690 MPa high strength steel H-section columns

The paper describes a numerical study on columns fabricated from high strength steel (HSS) plates with nominal yield stress of 690 MPa. The study comprised 4 benchmark models validated by testing data and 144H-section parametric study models. The benchmark models were firstly created and validated against testing data for the accuracy. A full-scale parametric study was carried out to investigate effect of flange width/thickness ratio (alpha), web height/thickness ratio (beta), geometrical imperfection (D) and residual stress (R) on overall buckling behaviour of the columns. The purpose of the study was to understand the relationship between section geometry, imperfections and column strength fabricated by welding from 690 MPa high strength steel plates. The H-section HSS columns were studied in this paper for the buckling behaviour along the minor axis in which direction those columns have smaller stiffnesses. It is shown that all those factors can be classified into three groups according to their impact on column strength (expressed with strength reduction factor chi): favourable factor (alpha), adverse factor (beta and D) and uncertain factor (R). alpha and Chi are positively correlated: Chi increases at different rates as a increases. It is an effective way to increase a to improve column strength when a is a small value. Chi decreases with the increase of beta at very low rate and the effect of beta on chi is not very obvious. D could produce more pronounced deterioration effect on ultimate strength of columns with higher alpha. The effect of residual stress (R) on H-section column strength is related with alpha: for the columns with higher a value, higher residual stress could produce more serious strength reduction effect. Residual stress does not always have negative effect on chi. It could produce beneficial effect on the ultimate strength of high strength steel H-section columns as long as alpha is reasonably small. The paper also shows a comparison of analysis results with existing standards including Eurocode 3 and GB50017-2017. For Eurocode 3, curve a0 can accurately predict chi with (lambda) over bar when D = 0.03%. When D = 0.20%, curve b gives the best-fit prediction.