Fig。 5(f) shows the stress-strain relationship of the concrete infill

for the test specimens。 In the cases of E4 and E5, the effect of stiffeners was addressed by replacing tt=bt in Eq。 (3e) with tt=w (Sakino and Sun 1994)。 Strength degradation due to the size effect

was neglected in the  model。

For the uniaxial stress-strain relationship of steel, a trilinear model was used [Fig。 5(f)]。 For the slender plates of E1 and E2,

the elastic local buckling stress  Fcr  ¼ 9Es=ðbt=tt      2

(AISC 2010) was used as the peak stress。 In the case of E3,  since

the tube section is compact, it was assumed that the yield strength can be developed。 In the analysis of E4 and E5, considering the

strength-reduction factor that accounts for the size effect might be necessary for the modeling of  concrete。

In the numerical analysis for all specimens, the peak loads oc- curred when the extreme compressive strain exceeded 0。003。 In E4 and E5 with stiffeners, the peak loads occurred at the extreme com- pressive strain = 0。00464 (midheight deflection = 4。60 mm) and 0。0037 (midheight deflection = 7。13 mm), respectively。 The corre- sponding midheight deflections were smaller than the measured deflections, 4。64 and 8。96 mm of the   specimens。 Initial local buckling  strains

effect of stiffeners, the tube plates were regarded as compact section that can develop the yield strength。 The descending branch repre- senting the inelastic buckling mechanism was described based on the analytical model of Fujimoto et al。   (2004)。

Using the material constitutive models and linear strain distri- bution, fiber model sectional analysis was performed for the mid- height section of the specimens。 To address the column slenderness effect, the relationship ϕm ¼ Δmðπ=LeÞ2 between the midheight curvature ϕm and deflection Δm was assumed, relating the local response to the global response。 Iterative calculations were per- formed to satisfy the force equilibrium at the midheight section, considering the second-order effect。

Fig。 5 compares the results of the test and fiber model analysis。 The figure also shows the reference points of the axial strains of the compressive flange that represent the main events: the crushing strain (0。003) of the unconfined concrete; yield strain (fy;t=Es ¼ 0。0036 for Grade-800 steel and 0。0015 for Grade-400 steel); and crushing strain (0。005) of the confined concrete for CFTs (Tomii and Sakino 1979)。 Generally, the numerical analysis results showed good agreement with the test results in terms of initial stiffness and strength。 Except for E3, the test results were

98% ∼ 101% of the predictions。 In E3, on the other hand, the prediction overestimated the stiffness and strength of the test speci-

men。 This indicates that in the case of mild steel, even considering the postbuckling deterioration of the tube plates [Fig。 5(f)], the analysis results still significantly overestimate the test results。 It seems that the performance degradation of concrete infill also oc- curred due to the combined effect of local buckling and yielding of the mild steel。 For conservative design, as mentioned in previous studies (Fujimoto et al。 2004; Sakino et al。 2004; Liang 2009), a

Strength Prediction of Current Design Codes

Fig。 10 compares the axial-flexural capacity of the test specimens and the predictions of ANSI/AISC 360-10。 Considering the second-order effect, the midheight flexural moment was calculated as M ¼ Pðe þ ΔmÞ。 In the specification, the plastic stress distribu- tion method (Method 2) is permitted only for compact sections, and the equivalent steel column method (Method 1) is recommended for noncompact and slender  sections。