Jerome F. Hajjar - Department of Civil and Environmental Engineering

O'Shea and Bridge 1997

In a series of papers and reports (O’Shea and Bridge, 1994, 1997a, 1997b, 1997c, 1997d, 1997e), the authors presented a series of experiments on the behavior of circular and square thin-walled concrete-filled steel tubes. In addition, the authors compared the experimental results to the design methods presented in the ACI 318 (1989), AISC LRFD (1993), AISI LRFD (1991), AS4100 (1990), EC4 (1992), and AS 3600 (1988) specifications for concrete-filled steel tube members.

Experimental Study, Results and Discussions

The experimental study consisted of four series of tests. For the first two series, the authors investigated local buckling of the steel tube wall. The main purpose of the third and fourth series was to investigate the confinement of the concrete. Eight different loading types were applied to the specimens. They were labeled as BS, BSU, BSC, CS, CL, CF, E1 or E2. Tests labeled BS included axial loading of the bare steel tube alone. Experiments BSU and BSC had axial loading of the steel tube alone while the steel tube was filled with unbonded concrete. Experiments labeled CS and CF included tests in which the CFTs were loaded axially through both the steel and concrete together. Experiments labeled CL included axial loading of the unbonded concrete only. Tests E1 and E2 had eccentric axial loading of both the concrete and steel simultaneously. For type E1 loading, the eccentricity was D/10, while for type E2 loading, the eccentricity was D/20, where D was the outer nominal diameter of the circular steel tube or the nominal depth of the rectangular steel tube. The eccentricity was provided through thick endplates connected to offset hemispherical bearings.

In the first series of the experimental study, the local buckling behavior of short thin-walled circular tube members was investigated. Both unfilled and concrete-filled specimens were tested. Type BS, BSC, E1 and E2 loadings were utilized. Type E1 and E2 loadings were applied to the bare steel tubes. Specimens with nominal D/t ratios ranging from 55 to 200 were prepared. The yield stresses for the steel tubes ranged from 26.83 ksi to 52.69 ksi according to the coupon tests. The out-of-plane imperfections of the steel tube walls were determined using a fixed measuring ring and were found to increase near the weld. The membrane residual stresses were also high near the weld. The bending residual stresses for the specimens were generally small. In these tests, the concrete was serving as a lateral restraint with an average strength of 6.89 ksi. For each axially loaded specimen, the end conditions were fixed and displacement controlled loading was applied. The BS and BSC specimens showed similar responses and it was found that the ductility improved as the D/t ratio decreased. The failure of each of the axially loaded specimen was due to outward local buckling at one end (with the exception of specimen S16BSC, which had local buckling at its mid-height). This showed that internal concrete as a lateral restraint did not improved local buckling strength for circular CFTs. Pin-pin support conditions were provided for the eccentrically loaded specimens and displacement controlled loading was applied. The peak capacities of the eccentrically loaded specimens were less than those of the axially loaded specimens and the strength of the tubes decreased with an increase in eccentricity. The failure for all of the eccentrically loaded specimens, except two of them, was due to outward local buckling at one end. For specimens S30BSE2 and S30BSE1, local buckling formed at the mid-height. The capacity of S16BSE2 specimen was higher than expected and that of S30BSE1 specimen was less than expected. This was attributed to the differences in the eccentricities. The authors presented the AISC LRFD (1993), AISI LRFD (1991), AS4100 (1990), and Grimault and Janss (1977) procedures to calculate the local buckling capacity of axially loaded specimens. These methods were applied to the axially loaded specimens and the calculated capacities were compared with the experimental results. The moment capacities according to the aforementioned standards were also computed and compared with the experimental results.

In series 2, two groups of thin-walled square box specimens were tested. For the first group, an L/D ratio of 3.45 was used and six different D/t ratios ranging from 37.4 to 130.7 were selected. In the second group, a D/t ratio of 130.7 was used and six different L/D ratios ranging from 0.77 to 3.45 were selected. The specimens in both groups consisted of bare steel tubes and steel tubes filled with either unbonded or bonded concrete. The first group of specimens was tested by loading types BS, BSU and CS. The second group was tested by loading types BS and BSU. The average yield stress for the steel was obtained as 40.90 ksi from the coupon tests. The concrete used for the specimens had a nominal capacity of 2.90 ksi. The out-of-plane geometric imperfections of the steel tube walls were investigated for each specimen using an automatic level with a micrometer. The residual stresses on representative tubes for each D/t ratio were also examined. Displacement controlled loading was applied and the ends of the specimens were fixed. In addition to conducting the experiments, the authors introduced a finite strip method to calculate the elastic buckling stress of a plate and then presented modified von Karman and modified Winter formulations for determining the plate strength. The test results were also compared with the predictions according to the AISC LRFD (1993), AISI LRFD (1991), AS4100 (1990) procedures. The collective results showed that the local buckling strengths of the bare steel tube specimens were found to increase by up to 50% when they were filled with unbonded concrete. This was attributed to the outward buckling mode. The results of the bare steel tube specimens and the unbonded concrete filled specimens with a D/t ratio of 130.7 and an L/D ratio greater than 1.15 indicated that the strength remained almost constant for increasing L/D ratios. The authors presented some modifications to the existing Winter formula and the new formula matched with the experimental results better.

The third series of experiments were performed to investigate the effect of confinement on the cross section strength. Short thin-walled steel tubes filled with medium and high strength concrete were tested by type CL, CS, E1 and E2 loadings. Five different D/t ratios of the specimens were selected ranging from 63 to 190. According to the coupon test results, the yield strengths of the steel tubes were found to range from 26.93 to 52.69 ksi. The membrane residual stresses in the tubes were high near the welds. The bending residual stresses were generally small except for the S30 type specimens. The steel tubes were filled with two concrete mixes having nominal strengths of 7.25 ksi and 11.60 ksi. The geometric imperfections of the steel tube walls were obtained for the tubes tested by type CS, E1 and E2 loadings. It was again found that the imperfections were highest near the seam welds. The type CS and CL tests of the specimens filled with 7.25 ksi concrete were carried out by displacement controlled loading. For all the other tests, force controlled loading was utilized. In the type CS test, two different failure patterns were observed. In the first one, the specimens maintained the bond between steel and concrete until failure. For the other specimens, the bond between steel and concrete was not maintained and local buckling took place. The vertical principal strains on the steel tube were observed to decrease when local buckling occurred. The principal strains increased rapidly and diverged from each other after concrete crushing. In the type CS tests, only one specimen experienced local buckling. This confirmed that the bond between steel and concrete prevented local buckling. For the type CS and CL loadings of the specimens with 7.25 ksi concrete, the ductility and strength of the thicker specimens were higher, as the thicker specimens provided better confinement. When concentric axial load was applied to the 11.60 ksi specimens, the ductility, strength and confinement were also found to improve with an increase in tube thickness. The principle strains for the thickest 11.60 ksi specimens increased in a linear fashion up to 0.006. This showed that the lateral deformation of the high strength concrete was less compared to moderate strength concrete. In the eccentric axial load tests, the 7.25 ksi specimens showed higher strength and ductility with increasing thickness. The divergence in the vertical strains was observed in the post peak region. This was attributed either to local buckling or concrete crushing. Increasing the eccentricity caused the ultimate strength to decrease but it improved the ductility. The eccentrically loaded 11.60 ksi specimens showed the same trend as the eccentrically loaded 7.25 ksi specimens in terms of strength and ductility. For 11.60 ksi specimens, the principle strains started to increase rapidly after the peak load and this was again the effect of concrete crushing. The authors calculated the confining pressures of the specimens in the CL tests and found that the confining pressure decreased for the specimens having high strength concrete and greater D/t ratio. Using the maximum energy distortion theory, they concluded that the steel in all the CS specimens yielded. The moment capacities of the eccentrically loaded specimens were calculated with a fiber analysis method by Wheeler and Bridge (1993). This method gave conservative results for specimens loaded with an eccentricity of D/20 if a confined concrete model was used. However, an unconfined concrete model gave conservative results for the specimens with D/10 eccentricity.

The fourth experimental study was again performed to investigate the effect of confinement on the cross section strength. Short thin-walled very high strength concrete- filled steel tubes were tested by type CF, CL, CS, E1 and E2 loadings. Force controlled loading was utilized in all the cases. The specimen sizes and steel properties were the same as in the third experimental study. The steel tubes were filled with three different concrete mixes having nominal strengths of 11.60, 14.50 and 17.41 ksi. The lateral imperfections of the steel tube walls were obtained for the specimens tested by CS, E1 and E2 type loadings and similar trends were observed with the previous experimental study. The concentrically loaded specimens showed brittle post peak behavior. They failed suddenly and their post peak responses could not be obtained completely. In type CS tests, two different failure patterns were observed. In the first one, the specimens maintained the bond between steel and concrete until failure. For the other specimens, the bond between steel and concrete was not maintained and local buckling took place. The principal strain values of the steel tubes were found to diverge from each other after local concrete crushing. The formation of local buckling was observed with a reduction in the vertical strains. From the available data for CS and CL tests, it was found that increasing the wall thickness also increased the ductility. According to the results of the CL tests, strength enhancement of the concrete due to confinement was also greater for the thick walled tubes. In addition, this enhancement was greater for the concrete batches having smaller strength. For the specimens tested by CS type loading, strength enhancement of concrete did not occur. In the E1 and E2 type tests, the eccentricity was provided using very thick and stiff endplates with an offset half-round. According to the results, the increase in eccentricity caused a decrease in capacity. Some specimens behaved in a more ductile manner when loaded at greater eccentricity. The capacity and ductility of thicker specimens were higher. Local concrete crushing was observed with deviations in vertical strains measured at equal rotational distances from the axis of bending. The authors calculated the capacity of the eccentrically loaded specimens using a fiber analysis method by Wheeler and Bridge (1993). The results were compared with experiments and they concluded that the increase in concrete strength due to confinement was small and did not occur except in the S30 type specimens.

Analytical Study

In the last report (O’Shea and Bridge, 1997e), the authors developed design methodologies for thin-walled concrete filled steel tubes. The capacity of a short, thick-walled circular steel tube filled with medium strength concrete was determined using ACI 318 (1989), EC4 (1992) and modified AS 3600 (1990) procedures. It was also calculated with the fiber method by Wheeler and Bridge (1993). The outputs were compared and EC4 (1992) gave the closest capacity to the one obtained by the fiber method. From the past experiments of several researchers, the specimens, which satisfied the provisions of EC4 (1992) were selected for comparison purposes. The capacity of each specimen was computed according to the three aforementioned codes. When the results were compared, the capacities of EC4 (1992) showed lower scatter than the others. This was attributed to the fact that the confinement effect was taken into account in EC4 (1992). The results for the other two procedures were very conservative because the confinement was neglected. The column curves were utilized for slenderness effect in EC4 (1992). However, minimum eccentricity approach was used both in ACI 318 (1989) and modified AS 3600. It was concluded that for slenderness, EC4 (1992) approach was more rational compared to the others. When local buckling was involved, the authors proposed a modified Winter Formula to calculate the local buckling strength of square steel tubes using the results of the experimental study (O’Shea and Bridge, 1997b). From tests of thin-walled circular steel tubes (O’Shea and Bridge, 1997a), the authors decided that the current methods in the codes could be used to calculate local buckling strength. To investigate confinement of concrete, the concrete models by Martinez, Nilson, and Slate (1984), Mandler, Priestly, and Park (1988) and Attard and Settunge (1996) were presented. The models by Mander, Priestly, Park (1988) and Attard andSettunge (1996) were calibrated according to the results of the type CL experiments conducted by the authors (O’Shea and Bridge, 1997c, 1997d). These modified models showed good correlation with experimental results and the authors concluded that they could be used to estimate the response of confined concrete in circular steel tubes. In the same test series, the experimental results were used to propose design methods for type CS loading of circular concrete-filled steel tubes. The EC4 (1992) provisions predicted accurately the response of the type CS specimens, which had 11.60 ksi concrete and experienced no local buckling. For concrete strengths greater than 11.60 ksi, EC4 (1992) predictions were close to the experimental values when confinement, steel reduction and local buckling were neglected. The formation of local buckling might result higher cross section strengths for concrete strengths up to 11.60 ksi. However, there were not enough tests for this effect. The authors concluded that design without local buckling was a conservative approach for type CS loadings. The authors also examined the experimental results from type E1 and E2 loadings (O’Shea and Bridge, 1997c, 1997d) to propose design procedures. The capacities of the eccentrically loaded specimens were calculated by the EC4 (1992) provisions and no local buckling was assumed. To evaluate the EC4 (1992) procedures, the fiber method by Wheeler and Bridge (1993) was also applied to the specimens. Using the fiber model, two analyses were performed. For the first analysis, an unconfined concrete model was used and the second one was performed by a confined concrete model. The results were compared with the experimental values. For thin-walled circular steel tubes filled with concrete up to a strength of 7.25 ksi, EC4 (1992) gave conservative results and it was recommended to be used. However, for concrete strength greater than 7.25 ksi, the authors decided that a rigorous fiber analysis was required.

Reference

O'Shea, M. D. and Bridge, R. Q. (1994). “Tests of Thin-Walled Concrete-Filled Steel Tubes,” Preliminary Report, Center for Advanced Structural Engineering, University of Sydney, Australia.
O'Shea, M. D. and Bridge, R. Q. (1997a). “Behaviour of Thin-Walled Box Sections with Lateral Restraint,” Research Report No. R739, School of Civil Engineering, University of Sydney, Sydney, Australia, March.
O'Shea, M. D. and Bridge, R. Q. (1997b). “Local Buckling of Thin-Walled Circular Steel Sections with or without Lateral Restraint,” Research Report No. R740, School of Civil Engineering, University of Sydney, Sydney, Australia, April.
O'Shea, M. D. and Bridge, R. Q. (1997c). “Tests on Circular Thin-Walled Steel Tubes Filled with Very High Strength Concrete,” Research Report No. R754, School of Civil Engineering, University of Sydney, Sydney, Australia, November.
O'Shea, M. D. and Bridge, R. Q. (1997d). “Tests on Circular Thin-Walled Steel Tubes Filled with Medium and High Strength Concrete,” Research Report No. R755, School of Civil Engineering, University of Sydney, Sydney, Australia, November.
O'Shea, M. D. and Bridge, R. Q. (1997e). “Design of Thin-Walled Concrete Filled Steel Tubes,” Research Report No. R758, School of Civil Engineering, University of Sydney, Sydney, Australia, November.