0143-974X(94)00020-4

ELSEVIER

J. Construct.Steel Research34 (1995)1-25 1995ElsevierScienceLimited Printed in Malta. All rights reserved 0143-974X/95/$9-50

Flexural Tests of Longitudinally Stiffened Fabricated Steel Cylinders

Q. Chen, G. L. Kulak & A. E. Elwi
Department of Civil Engineering, University of Alberta, Edmonton, Canada T6G 2G7 (Received 5 May 1994)

ABSTRACT Fabricated steel cylinders with large R/t ratios are often used in conveyor galleries, offshore platforms, storage tanks, towers, vessels, and the like. The use of longitudinal stiffeners in these large cylinders has proven to be an eJ~icient way to promote structural stability. The increase of load capacity is significant. However, the attad)ment of the stiffeners complicates the problem by introducing additional buckliing modes and new design parameters. Four large-scale specimens were tested. The collected data included the load versu:~ deformation behavior, as well as the geometric imperfections and welding resid~lal stresses of the specimens. The objective of the research program was to investigate the local buckling behavior and ultimate load capacity of longitudinally stiffened steel cylinders subjected to bending. The results showed two failure modes, a sheJl buckling mode and a general buckling mode. The occurrence of both modes and the corresponding capacity depend on the shell properties, as well as on the spacing and angle of arc covered by the stiffeners. The general buckling mode has strength and ductility much higher than those predicted by existing design guidelines.

INTRODUCTION
Longitudinally stiffened large diameter cylinders subjected to flexural moment have many industrial applications. For example, they are used as conveyor galleries for handling materials. These structures have typical R/t ratios of between 100 and 400 and the outside diameter is usually in the order of 2"5 to 4 meters. Usual practice is to weld circumferential stiffeners at intervals of 0.5 to 2-0 diameters, primarily to help maintain the round shape of the tubes. The span can be as long as 60 meters. An example of

Q. Chen, G. L. Kulak, A. E. Elwi

such a long span is the main tubular conveyor at the Genesee Power Plant in Alberta, Canada. The cylinders are fabricated by first cold-rolling steel plates into curved cylindrical shells. The tube formed in this way is completed by performing longitudinal seam welding. The relatively short 'cans' that are produced in this way are then joined together by circumferential girth welds to form long spans. In addition to the circumferential stiffeners, longitudinal stiffeners are often welded to the shell surface. Inevitably, initial geometric imperfections and residual stresses are introduced as a consequence of the fabrication procedure. The flexural capacity of a longitudinally stiffened cylinder is usually governed by the instability of a local area that is highly stressed in compression. One possible mode of failure is so-called shell buckling, in which shell plating buckles like an unstiffened cylinder. The longitudinal stiffeners essentially remain straight. A second possible failure mode is that of general buckling, in which case the shell and longitudinal stiffeners buckle simultaneously. This happens when the stiffeners are closely spaced. Lastly, failure may be initiated by stiffener buckling, in which the stiffeners are twisted and distorted before shell buckling occurs. This mode may occur when stiffeners with low torsional stiffness, such as thinwalled open cross-sections, are used. Of course, the three buckling modes may interact with one another.

LITERATURE REVIEW Testing of large-scale specimens is the most direct and reliable way to investigate the buckling behavior of large diameter steel cylinders. In order that the material, initial imperfections, and residual stresses be representative of the situation in practical structures, the specimens in such tests should be fabricated by industrial procedures. Green and Nelson 1 reported the testing of three large diameter longitudinally stiffened cylinders subjected to axial compression. The radius of the specimens was nearly 600 mm and the R/t ratio varied from 95 to 285 as different shell thickness were selected. The stiffeners and their arrangement were typical for offshore platforms: plate stiffeners were welded to the inside surface of the cylinders. Residual stresses arising from the welding of stiffeners were measured, and it was observed that the presence of a significant compressive residual stress caused local yielding to precede shell buckling. A review of the Green and Nelson test results indicates that the level of residual stresses probably affected the load-carrying capacity

Flexural tests of fabricated steel cylinders

in a signJificant way. Specimen GU1, which had a high level of compressive residual stress because of its small shell thickness (t = 2 mm), failed by local shell buckling and reached a PulPy ratio of only 69%. Meanwhile, Specimen GU2, which had a relatively low level of residual stress, reached a PulPy ratio of 97%. Unstiffened cylinders with relatively small R/t ratios have also been tested under axial compression and flexure. 2-4 However, the stiffness of these shells is higher because of their large curvature. Consequently, local shell buc:kling behavior and strength are different from those with large R/t ratios. A single test done by Bailey and Kulak, s designated hereafter as SB1, was the first one reported in its category--a large diameter, longitudinally stiffened, fabricated steel cylinder subjected to pure bending. The test set-up and the arrangement of the test specimen were similar to that used in the study reported herein. Cylinder SB1 was made of steel meeting CSA G40.21M 300 W (which is a hot-rolled steel with specified minimum yield and ultimate strengths of 300 MPa and 450 MPa, respectively) and it had an R/t ~ratio of 128. Five longitudinal stiffeners were welded to the compression region of the 5-05-mm thick cylinder. They were rectangular hollow structural shapes (HSS 50.8 x 25-4 x 3"18) and were arranged so as to cover an angle of 76 of the circumference of the cylinder. Bailey and Kulak also tested one unstiffened cylinder in bending. It was made of steel having a specified yield strength of 700 MPa and it had an R/t ratio of 174. The failure mode of the longitudinally stiffened specimen SB1 was general buckling. After yielding of the shell between stiffeners started, buckling was first observed at a bending moment of 2200 kN m. The development of buckles forced the stiffeners to bend together with the shell. The load continued to rise, but at a decreasing rate, as the buckling waves increased in size. As the test was approaching the ultimate load, the deflection of the cylinder became so large that the specimen was approaching the test floor. Photographs of the specimen showed that the buckles were appreciably deep and wide at the time when testing had to be terminated. The maximum moment achieved during the test was 3068 k N m , which is about 89% of the theoretical plastic moment capacity. Stephens et al. ~ tested two large diameter unstiffened steel cylinders made of grade 300 W hot-roiled steel plates. The R/t ratios were 149 and 222. The results of these pure bending tests showed that both shells failed by local buckling in the extreme compression zone. The geometry and material properties of one of the Stephens et al. cylinders matched those of the Bailey and Kulak stiffened cylinder SB1. The unstiffened cylinder

Q. Chen, G. L. Kulak, A. E. Elwi

achieved an ultimate moment to plastic moment ratio Mu/Mp, of 0.48. This is only 53% of that of the stiffened cylinder SB1 (Mu/Mp = 0.89). In this case, the use of longitudinal stiffeners nearly doubled the flexural capacity. Based on test data of fabricated steel cylinders subjected to axial compression, Stephens et al. 6 proposed a set of design formulae to predict the buckling stress of tubular members subjected to uniform compression. As concluded by both Stephens et al. 6 and by Bailey and Kulak, s the flexural buckling stresses of their unstiffened cylinders (a total of three) could also be reasonably predicted by these formulae. The flexural behavior of a longitudinally stiffened cylinder differs from axial compression behavior because the bending stress is not uniform and can be redistributed after local failure. Nonetheless, the assumption that buckling stress in bending is equal to that in uniform compression is widely employed in design guidelines. Although large diameter unstiffened cylinders tend to fail at comparable stress levels in bending or in compression, Bailey and Kulak 5 found that the ultimate moment in general instability failure was about 90% of the fully plastic moment of the stiffened cylinder (or 6% higher than the yield moment). The design criteria based on uniform compression could not predict such a high capacity. These design criteria need to be examined so that the effectiveness of longitudinal stiffeners can be realistically defined.

SCOPE The work presented here forms the experimental part of a wider investigation into the buckling behavior of longitudinally stiffened cylinders subjected to bending. The authors conducted four tests on large-scale specimens. All were fabricated from steel cylinders longitudinally stiffened over part of the circumference using hollow structural sections (HSS). The outside diameter of the cylinders was 1270mm, which is about one-third to one-half of that encountered in practice. The test parameters included the radius-to-thickness ratio (R/t) of the cylinder, the spacing between the stiffeners, and the total angle of the cross-section covered by the stiffeners. The R/t ratio used was between 130 and 200. The number of stiffeners varied from three to eight. This paper gives a detailed description of the experimental program, including the fabrication of the test specimens, material properties, initial imperfections, and residual stresses. The experimental behavior of the stiffened cylinders is presented along with observational comments. Numerical modeling of the physical tests has been done, but that work is not included in the present paper.

Flexural tests of fabricated steel cylinders

TEST S P E C I M E N S A segment of cylinder under pure bending can be used as a representative model for the study of flexural buckling behavior of longitudinally stiffened cylinders. Since the wavelength of a local buckle can be expected to be very short compared with the radius of the cylinder, a length equal to the cylinder diameter is sufficient for the segment. Practical boundary effects, such as imperfections due to girth welds and the variation of shell thickness, should also be reflected in the model. With this in mind, all specimens were chosen to have the general configuration shown in Fig. 1. Each was longitudinally stiffened on the outside of the shell in the compression zone. The cylinder consisted of three segments: a 1300-mm long tesl: portion at the center, and two 768-mm long thicker segments at the ends. The end segments were added to preclude local buckling near the end boundary because of non-uniform stress distribution at those locations. All three segments had the same outside diameter, and they were joined together by a circumferential groove weld. The dimensions of the cylinders and the stiffeners were measured prior to the test and the mean values of these quantities are listed in Table 1. The cylinders were fabricated from hot-rolled steel plates (CSA G40.21 300 W or 260 G) by a local steel fabricator using standard procedures. The steel pl~Lte was passed through rollers several times until the desired curvature was reached. The two edges of the plate were then joined and welded together by a complete penetration groove weld. During this process, there was no restraint against springback (the stretching of the plate to release the elastic deformation). Thus, any so-called locked-in stress, which might exist when springback is not free, is likely to be minimal. After each specimen was delivered to the laboratory, light interior fixture frames were installed into both ends to maintain the circular shape of the cross-section.

Thick end segment

Test segment

Thick end segment

]~

1270 O.D.

768

=l"

13002836

__ vl-

768

Fig. 1. Schematic of the test specimens (dimensions in mm).

Q. Chen, G. L. Kulak, A. E. Elwi

TABLE 1 Dimensions of Test Specimens

Shell Specimen Rang (mm)

629.8 631.5 631.5 631.2

Stiffener Ravg/t l hx bx t (ram)

50.7 x 25.4 x 25.4 x 25"5 x 25.3 x 3-04 25.4 x 2.38 25.4 x 2.38 25-5 x 2.33

t~ (ram)
4-72 3"07 3"07 3.34

tb2 (mm)
7.94 6.23 6.23 6.47

No.

Spacing (mm)
180 175 175 140

SB2 SB3a SB3b SB4

133 206 206 189

3 5 7 8

Thickness of the test segment. b Thickness of the end segments.

Stiffeners with open cross-sections may buckle torsionally under compression. Such a buckling mode could interact with other failure modes and complicate the investigation. For this reason, rectangular HSS sections were chosen for stiffeners. They were attached to the cylinders along their short dimension by means of 50-mm long intermittent 5-mm fillet welds separated by 100-mm intervals. The location of the intermittent fillet welds on a given stiffener alternated from side to side. In order to determine the material properties of the cylindrical shells, coupons were cut from the same plate that had been used to fabricate the test specimens. These coupons were taken in an orientation parallel to the longitudinal axis of the cylinder. They were prepared as specified in ASTM E8-897 and tested in a 1000 kN capacity universal testing machine. The material properties from the coupon tests are summarized in Table 2 and the coupon test curves are shown in Fig. 2. All showed the sharp yield point that is typical of hot-rolled structural steels. A long yield plateau can be observed from the stress versus strain relationship of the materials used in specimens SB2 and SB4. However, the response of the coupon corresponding to specimens SB3a and SB3b does not have the same characteristic. Instead, there is a moderate hardening slope shortly after the yield stress is reached. The yield stress values reported in Table 2 are static yield stress values. The cylinder tests were carried out using a very slow rate of loading. A further complication is that the tension coupons were taken from flat source plate used to roll the tubes. Cold forming the cylinders would result in a change in the yield stress both in the circumferential direction and through the thickness. However, for cylinders with such large R/t ratios, it is likely that changes in the static yield stress arising from the cold forming are negligible, particularly in the longitudinal direction.

Flexural tests offabricated steel cylinders

TABLE 2 Material Properties

Shell material Test segment Specimen

E ~y

Stiffeners

End segments
E ~y E ay

(MPa)
SB2 SB3a; SB3b SB4 210x 103 204x 10a 207 x 103

(MPa)
303 244 298

(MPa)
199 x l03 207 x l03 210x 103

(MPa)
248 290 305

(MPa)
209x 103 214 x 103 193 x 103

(MPa)
483 512 492

500

M O.

300

200

100

........
..........

SB2 SB3a; SB3b

SB4
I . . . . I

50000

100000

150000

200000

250000

Strain (microstrain)

Fig. 2. Tensile coupon test results for the cylinder materials.

The stiffeners were made from steel meeting CSA G40.21 350W (specified minimum values of yield and tensile strengths are 350 MPa and 450 MPa, respectively). Because of the way in which HSS shapes are manufactured, residual stresses can be quite significant. The yield stress is raised due to cold-forming and it may not be uniform along the crosssection. In order to assess the average value of material properties under compression, stub column specimens were prepared and tested in accordance with the guidelines of the Structural Stability Research Council. '~ A typical stress versus strain behavior from the stub column tests is plotted in Fig. 3. A nominal yield strength can be obtained using the 0.2% strain offset as specified in various standards. 4'7 However, the average stress versus strain curve of small size HSS stiffeners is so severely affected by the residual stresses and the nonhomogeneous material properties that the

8
600 _

Q. Chen, G. L. Kulak, A. E. Elwi

400 I~ & <~ 300 200 100

i i i . ! . . . . | . . . . ! . . . . | . . . . I

SO00

10000

15000

20000

25000

Strain (microstraln)

Fig. 3. Stub column test results of the HSS stiffeners of specimen SB2.

nominal yield strength corresponding to the 0.2% strain offset is not particularly meaningful. As an alternative, the yield point shown in Fig. 3 is defined as the intersection between the extended elastic and strainhardening lines. The stub column test curve can then be interpreted as the response of a uniform, bi-linear elastic-plastic hardening material with axial residual stress present that causes yielding at an early stage. Test results from the HSS stub column tests are also included in Table 2.

INITIAL IMPERFECTIONS Analytical studies of axially loaded cylinders have indicated that initial geometric imperfections of stiffened cylinders are important in evaluating the capacity, s In this study, specimens SB2 and SB3a were chosen as representative of the 5-mm and 3-mm thick specimens, respectively, and their initial imperfections were examined in detail. The measuring device used for recording imperfections was similar to that used by Obaia e t al. 9 It consisted of an aluminum bar that was placed along the longitudinal axis of the test specimen, a rod arranged at fight angles to the aluminum bar, and a linear variable differential transformer (LVDT). The aluminum bar could rotate around its axis, and its ends were supported at the centers of the interior frames located at each end of the test specimen. An LVDT was mounted on the end of the rod adjacent to the tube, while a trolley was attached to the other end. The trolley allowed the rod to move longitudinally on the track of the aluminum bar. The

Flexural tests of fabricated steel cylinders

LVDT output gave the difference between the inside shell wall and the referenc~ circle drawn by the rotating LVDT head. The imperfections were measured after the welding of stiffeners but before the tube was welded to the end plates that would attach it to the test set-up. A total of 192 measurements (12 cross-sections with 16 circumferential points at each section) were taken for each specimen. To ensure that the measured configuration coincided with that which would be present during the test, the specimen was suspended horizontally at its ends by a crane while the imperfections were being measured. This position is similar to that used subsequently in the test set-up. Chen et al. 1 provide full details of the measurements. The data obtained from the measurements described above must be interpreted properly so that the three-dimensional profile of the imperfect cylinder can be recreated. As a first step, different forms of imperfections need to be classified. A perfect cylinder can be characterized by a straight longitudinal axis, uniform radii along the length, and a perfectly round shape for every cross-section. The imperfections in a cylinder are thus classified into three categories: out-of-straightness, variation of the radii, and out-of-roundness. The measurements located points relative to a reference cylindrical surface that is determined by the measuring device. However, it is difficult to make the axis of the measuring device coincident with the axis of the specimen. In fact, neither of these axes is a straight line because both deflect under gravity load. Therefore, the reference cylindrical surface cannot be used as a perfect cylinder and the raw data do not represent the real imperfections. Such errors must be eliminated during data analysis. In the,, procedure established during this study, a best fit circle is found by the least squares method at each cross-section. The centers of these circles tiarm the axis of the imperfect cylinder. The perfect reference cylinder, which is used to defined imperfections, has the average radius and a straight axis connecting the two end cross-sections. A summary of the results of the data analysis for specimens SB2 and SB3a is presented in Table 3. As can be seen from the table, the
TABLE 3 Maximum Magnitude of Initial Imperfections

Specimen
SB2 SB3a

Out-of straightness (mm)

-0"8 -0"6

Variation of r a d i i (mm)
-0-6, +0-4 -0"9, +0"5

Out-of-roundness (ram)
-3"7, +4.9 -5.2, +6.7

10

Q. Chen, G. L. Kulak, A. E. Elwi

initial deviation of the central axis and the change of the radius is not significant for such short cylinders, and the initial shell deviation is in the order of wall thickness. The results also showed that the two specimens had similar imperfection patterns, with the thinner one (SB3a) having the more severe imperfections. The minus sign in the table denotes downward deflection of the centroidal axis for the out-ofstraightness imperfections, shrinkage of the radius for variation of radii imperfections, and inward shell deflection for the out-of-roundness imperfections. The recreated initial profiles of specimen SB2 are plotted in Fig. 4. These plots are scaled by a factor of 30. The deflection of the axis and the shrinkage due to the girth weld can be seen from the side view plot. The peter shape of the cross-section could be attributed to the residual deformation due to the seam weld as well as the result of the coldforming process, in which the curvature applied might not be exactly uniform throughout the plate. The part away from the edges was overrolled and, consequently, had a relatively large curvature. The interior fixture frames located at the cylinder ends helped to overcome the ovalization due to the self weight of the tube. Dwight 11 examined three offshore cylinders (R=7300-9200mm, t = 16--40mm) with similar results. Lastly, and in addition to the pear shape, the longitudinal weld underneath the stiffeners introduced circumferential waves in the local area.

Seam weld

Fig. 4. Measured initial imperfections of specimen SB2 (magnification factor = 30).

Flexural tests of fabricated steel cylinders

11

RESIDUAL STRESSES Cold rolling of a steel plate into a tube is a repetitive process of applying successive inelastic bending deformation to the plate and then unloading elastically. Circumferential residual stresses are induced in such a process. However, if springback is not restrained, which is considered to be the case for specimens fabricated for these tests, the level of these residual stresses is reduced considerably. Longitudinal residual stresses caused by seam welding are concentrated in the vicinity of the weld. 2 In this test program, the cylinders were arranged such that the seam was located at the mid-depth, near the neutral axis. Consequently, the influence of seam welding on either the compression or the tension zone was minimized. Four circumferential welds were required in each specimen, two performed by the fabricator to join the segments, and the other two attaching the ends of the specimen to the test set-up. The residual strains resulting from the latter were measured using strain gages and results showed that a strain of about 40 x 10 -6 existed in the test portion of cylinder SB2. However, information on residual strains resulting from the two shop welds was not available. The residual stresses that arise from this particular source are present mainly in the circumferential direction, and, of course, the distribution is localized near the girth weld. Welding of the longitudinal stiffeners to the shell produces longitudinal compressive residual stress near the welds. This is considered to be a major source of residual stresses in the compression zone of a stiffened cylinder, where local shell buckling may be promoted. As reported by other investigators, 1 the distribution of compressive residual stresses in the panels between the stiffeners is nearly uniform. Its average value varies mainly according to the shell thickness and the size of welds. The tensile residual stress in the immediate vicinity of the welds is much higher than the equilibrium compressive residual stress. The combined residual stresses from the above sources are not necessarily equal to the summation of all distributions. In fact, some later processes, such as the welding of stiffeners, may eliminate part of the earlier distribution because of the large amount of heat input. For various practical reasons it was only possible to measure directly the residual strains arising from welding of the stiffeners and such measurements were taken only for specimens SB2 and SB3a, as representative of the 5-mm and 3-mm thick cylinders, respectively. The intensive measurement of SB2, by Demec gages as well as strain gages, was aimed at collecting data for the residual stress distribution: The strain gages were

12

Q. Chen, G. L. Kulak, A. E. Elwi

mounted on the inner surface of the cylinder wall, while the Demec gage marks were placed on the outer surface at a gage length of 51 mm. The residual strain measurements from SB2 (Fig. 5) show that the distribution pattern of compressive residual stresses generally agrees with the description of Green and Nelson. ~ The average compressive residual strain was about 35% of the yield strain (~y) in the internal panels between stiffeners and about 15% of ey in the shell outside the stiffened area. The strain measurements taken on SB3a were concerned merely with the magnitude and only strain gages were used. Compared with SB2, a higher compressive residual stress level was observed in this thinner shell, even though the welding was very similar to that of SB2. The average compressive residual strain was about 55% in the panels between the stiffeners and about 30% of ey in the shell outside the stiffened area. Since residual stresses came from many sources, the measured level reflected only the change of residual stresses due to the welding of stiffeners, rather than the total residual stresses in the specimen. However, it is considered that the measured data represent the dominant residual stresses in the compression zone of these test specimens. Using the measured strains, the corresponding longitudinal residual stresses can be calculated by assuming that the shell is in a state of plane stress and is free to expand and shrink circumferentially. Thus, a0=0; az=Eez (1)

The subscripts z and 0 refer to the longitudinal and circumferential directions, respectively and ez is the measured longitudinal residual strain.
1000 Strain gages Demec gages

E g 2
e

5O0 (shell between stiffeners)

0

(shell outside the stiffened area)

8

_.R
t Q E

a
e-

-500

Centerot the stiffener

-1000 50 1 O0

w o, l
150

lWe"
200 250 300 350

Distance trom apex (mm)

Fig. 5. Measured residual strains of specimen SB2.

Flexural tests of fabricated steel cylinders

13

THE TEST SET-UP The test set-up used in these tests was similar to that used in earlier work in the program, s'6 As shown in Fig. 6, it consists of two major parts: a large composite beam and a loading frame. The composite beam consisted of two side trusses and a test specimen in the center. The test region was subjected to pure bending under a two-point loading system. The ends of the specimen were expected to remain plane during bending because they were welded to heavily stiffened base plates at the ends of the side trusses. Each side truss consisted of two parallel plane truss elements interconnected by .~ihort I-shapes. The chords of the plane trusses were welded wide-flange shapes and the web members were channel sections. They were assembled using A325 bolts. The composite beam described above was simply supported at both ends, with horizontal translation allowed only at one end. There were two separate supports at each end, one for each plane truss. Every support consisted of a knife edge and a set of rollers. After the trusses had been positioned, the rollers at one end were restrained while those at the other end rem,-Lined movable. The load was applied by means of two 2220-kN hydraulic jacks that were controlled by an air pump. By balancing the pressure valves of two
Load Endplate ~Load l~dfdiaphragm/ Specimen ~ 1 ~ /~

(a) Elevation view

Thickendsegment
=11 1

/
~'~'
12440
(b) Plan view

X~

|L

I ~ I
~._

I ,q

r,I p

IS
I

Fig. 6. Test setup (dimensionsin mm).

14

Q. Chen, G. L. Kulak, A. E. Elwi

outlets in the pump, the same load level could be maintained in the two jacks. The jacks were simply supported between the loading frame and the composite beam. During the test, horizontal rollers above the jacks enabled them to move with the loading points. A spherical device below the jack was present so as to allow for possible rotations. The jack load was transferred to the plane trusses through a distributing beam.

INSTRUMENTATION Since the load was applied through a temporarily assembled system, not a standard testing machine, accurate measurement of loads and reactions was emphasized in the design of instrumentation. The applied load was measured using 2200-kN load cells under each loading jack. To check the uniformity of the reactions, the supporting force underneath each plane truss was measured individually. Two 900-kN and two 1300-kN load cells were distributed at the four supports under the composite beam. All the load cells were calibrated prior to the test. The output of the six load cells was zeroed before the specimens were loaded. During the tests, the agreement between the loads and the reactions was within 3%. The applied m o m e n t was calculated according to the average of the load cell readings. However, the total moment must also include that caused by the gravity load, which in this case contributes a large percentage. In order to determine this portion of load, the end trusses were weighed by load cells and their centers of mass were located. It was found that each end truss had a weight of 62.3 kN and the center of mass was 2-99 m away from the end support. This resulted in a bending moment of 186.3 kN m at the test specimen. When the weight of the specimen itself was included, the 5-mm thick cylinder (SB2) had an initial m o m e n t of 204 kN m before testing, and the 3-mm thick cylinders (SB3a, SB3b and SB4) had an initial moment of 202 kN m. The displacements at various locations were measured using LVDTs, cable transducers, or dial gages. All of the LVDTs and cable transducers were calibrated before the installation. Deflections at three cross-sections were measured by cable transducers above and beneath the specimen, as shown in Fig. 7(b). At the same cross-sections, LVDTs on both sides recorded the horizontal displacements. The rotation of the test segment was obtained in two ways. Rotation meters were mounted at both ends of the test segment to measure the change of the angle (Fig. 7(a)). In addition, LVDTs were horizontally mounted at the top and bottom of the specimen. These measured the

Flexural tests of fabricated steel cylinders

15

,voT

Lv-Tvw,
I Straingages ~LVDT SectionA-A
(b) Strain gagesand LVDTs

LVDT

(a) Rotationmeters and LVDTs

Fil~ 7. Instrumentation for displacements and strains (dimensions in mm).

relative change in length of the top and bottom fibers of the test span. The average curvature was then calculated from these measurements. Eight dial gages were used to check that the end plates between the specimen and the trusses remained plane during deformation. All dial gages we:re installed in the same plane so that any deviation from a plane would be detected. Several other dial gages on the loading frame monitored the lateral movement of the trusses. Fifty strain gages were mounted on the specimen. They formed three rings at the same cross-sections as deflections were measured (Fig. 7(a)). Most of l~hese gages were in the longitudinal direction except five that were mounted circumferentially on the compression side of the central crosssection. Because of different arrangements of the stiffeners, the locations of the gage,; varied slightly among the specimens.

TEST PROCEDURE After the trusses had been assembled, their base plates were carefully shimmed into a plane before the truss bolts were finally tightened. Then the specimen was positioned and welded to the heavy plates present at the end of each side truss (see Fig. 6). The alignments of the side trusses, the specimen, and the loading jacks were checked both before and after the welding. In order to observe yielding and the buckling pattern, the specimen was whitewashed before testing. A proliminary run up to 30% of the estimated failure load was carried out before each test in order to check for any defects in the set-up and the instrumentation. Loads, displacements, and strains were measured throughout the duration of every test. These measurements were taken by a data acquisition

16

Q. Chert, G. L. Kulak, A. E. Elwi

system driven by a microcomputer. The load increment started at 5% of the estimated failure load, but this was decreased as the moment versus curvature behavior changed. All the data were scanned and recorded after each load increment. Dial gage readings were taken at longer intervals. During the test, the behavior of the composite beam was monitored by dial gages. Lateral translation of the trusses was negligible and ceased completely after a few load steps. The out-of-plane displacement of the heavy plate that formed the boundary between the specimen and the side trusses was a maximum of 1.6mm in the tension area of SB2. After reinforcing the connection by adding four additional bolts, this amount dec "eased to less than 0"4 mm in subsequent tests.

TEST RESULTS

Specimen SB2
The test segment of this specimen was a 4.72-mm thick cylinder with three HSS 50.8 x 25.4 x 3.18 stiffeners spaced at 180 mm covering an angle of 32 in the compression zone. Including the influence of compressive residual stresses, yield was anticipated at a moment of 141.4 kN m. The calculated plastic moment was 2659 kN m, based on the measured dimensions and yield stress. In the test, specimen SB2 failed by local buckling in the shell outside the stiffened area. The moment versus curvature response of specimen SB2 is plotted in Fig. 8. The linear portion agrees well with prediction based on elastic
2200 |

i !

E"

1800

i 1400
1000 600 Rotation metem Elastic beam theory
2 0 0 - "

0.0

. . . . . . . .

0.5

'

. . . . . . . . .

1.0

' . . . . . . . . . . .

1.5

' . . . . . . .

2.0

'

Curvature (xlO" s )

Fig. 8. Moment versus curvature response of specimen SB2.

Flexural tests of fabricated steel cylinders

17

beam theory. The yield in the shell between stiffeners, which only accounted for a small area in the compression zone, did not have any significant nonlinear effects on the measured rotations. The change of the behavior near the ultimate point was sudden. At a bending moment of 1888kNm, yield lines at 45 appeared on the whitewashed shell surface in the compression zone. These occurred in the shell outside the stiffened area and near the circumferential weld joining the central and end segments. During the load increment leading to 1947 kN m, flaking of the whitewash signaled local shell buckling near the location of previous yielding. At this time, the strain gage readings indicated a large increment of deformation throughout the whole compression zone, which was confirmed by the observation of several shell buckles immediately after the first buckle. The ultimate moment of 1947 kN m is about 73% of the theoretical plastic moment. The stiffeners remained straight until the ultimate point was reached, but then bent with the shell at the post-peak stage. The behavior after the ultimate load point was stable. Figure 8 shows that large rotations took place with only a small decrease in moment. The strain distribution of the central cross-section at various load stages is presented in Fig. 9. The elastic strains at M = 773 kN m show that the cross-section remained planar. The neutral axis at this moment was close to the position predicted by elastic beam theory (41 mm above the center of the circle). At higher load levels, however, nonlinear inelastic shell deformation meant that the assumption of plane sections was no longer valid. After the ultimate moment was reached, the neutral axis shifted downwards as the buckled area extended towards the mid depth of the cross-secdon.
Z (mm)

~
. . . . . . . . . . . .

200
. . . .

Strain !mlcrostrain) 1500

-3000

-1500 Moment (kN-m) m~

-20 -40

ot\\
~

. ~.

773 1947

of
b i,

\\
iS

o---

188s

-soo[.

Fig. 9. Strain distribution in specimen SB2.

18

Q. Chen, G. L. Kulak, A. E. Elwi

Specimen SB3a
Specimen SB3a had five HSS 25-4 x 25.4 x 2.54 stiffeners on its 3.07-mm thick steel cylinder. These stiffeners were spaced at 175 m m and covered an angle of 64 . Using measured material properties and residual stresses, the predicted yield m o m e n t is 5 1 8 k N m and the fully plastic moment is 1475 kN m. The moment versus curvature behavior of this specimen is presented in Fig. 10. The curvature was computed from both the rotation meters and the horizontal LVDT measurements. The response was linear up to about 660 kN m, at which time yielding spread over the compression area. The departure of the LVDT curve from the rotation meter curve indicated that the cross-section was no longer a plane. The failure mode of SB3a was similar to that of SB2: the ultimate load capacity was controlled by shell buckling outside the stiffened region. Local shell buckling was first identified by strain reversal in the strain gages at various locations outside the stiffened region when the m o m e n t was about 1110 kN m. The buckling waves spread over the whole length of the test segment, as indicated by strain gages at different cross-sections. At a m o m e n t of l l 2 0 k N m , several shell buckles and yield lines were observed. These buckles were close to the circumferential welds. Those outside the stiffened area were much longer and deeper than those between stiffeners. With increasing deformation, the load started to drop after reaching an ultimate moment of 1138 k N m , which is about 77% of the plastic moment.
12oo ! !

1000 z 800

=E

600

-----a~ 400

Rotation m e t e r s LVDT's Elastic b e a m t h e o r y

200; . . . . 0.00

t . . . . 0.25

, . . . . 0.50 s)

, . 0.75

I 1.00

Curvature ( x l O "

Fig. 10. Moment versus curvature response of specimen SB3a.

Flexural tests of fabricated steel cylinders

19

The buckling deformation was not severe at the time the ultimate load was reached. On the basis of the observations of the previous test, continuing the experiment would only have provided a stable post-peak behavior similar to that of SB2. It was therefore decided to stop the test and unload the specimen so that the cylinder could be modified and tested again. Specimen SB3b The two relatively shallow shell buckles that were present at the end of the SB3a teslL were pulled out and two additional stiffeners were welded to the shell. The size and spacing of the new stiffeners were the same as those of existing stiffeners. Altogether, there were now seven stiffeners, and they covered !95 of the circumference. This constituted specimen SB3b. The specimen was initialized in the same way as the previous specimen had been and was tested using the same procedure. This specimen failed at a moment of 1333 kN m, having reached 87% of the fully plastic moment (1524 kNm). The observed failure mode was of the so-called general buckling type, in which the stiffeners buckle together with the ;adjacent portion of the shell. Yield lines first appeared at l 1 2 9 k N m and the first new buckle occurred at 1234 k N m , in the panel next to the central stiffener. As the load was increased further, shell buckles in the top panels extended rapidly from the circumferential welds towards the center, while the old buckles from SB3a outside the stiffened area remained unchanged. The difference
1400 1200 1000 800 600 400
200=

If
r ,, I,., I, ,, I , . . 1 * , ,

o ....
I , , , 1 . ,

P (2) Deflection at bottom (3) - (2)- (1)

, 1 , , , 1 , , , I , , , I

10

20

30

40

50

60

70

80

90

100

Deflection (rnm)

Fig. l l . Moment versus deflection response of specimen SB3b.

20

Q. Chen, G. L. Kulak, A. E. Elwi

Fig. 12. Deformed shell and stiffeners of specimen SB3b.

between the top and the bottom deflection, as shown in Fig. 11, indicated that buckling of the central stiffener started at about 1200 kN m. Bending and yielding of five central stiffeners were observed near the ultimate load point (1333 kN m). The strain distribution of the central cross-section, not shown here, indicated that a substantial inelastic tensile deformation developed before the failure. After passing the peak point, shell buckles extended to the unstiffened area. Figure 12 shows that the whole compression zone was buckled by the time the test was stopped. On the other side of the cylinder, tensile yield lines could be seen at the bottom of the specimen. The moment versus curvature graph, shown in Fig. 13, indicates that a large increase in the applied load took place after the initiation of buckling. This behavior is significantly different from the shell buckling failure of specimens SB2 and SB3a, in which shell buckling outside the stiffened area led to an abrupt failure.

Flexural tests of fabricated steel cylinders

1400 .t

21

'If
o Rotation meters LVDT's

*
....... 200. , , , , I i 0.5 0.0
i , I . , I , , , I * ,

Elastic beam theory

, , I , , I

1.0

1.5

2.0

2.5

3.0

Curvature ( x l O " 5 )

Fig. 13. Moment versus curvature response of specimen SB3b.

Specimen SB4 Specimen SB4 used eight HSS 25.4 x 25.4 x 2.54 sections to stiffen its 3.34-mm thick cylinder. The stiffeners were spaced at 140 m m and covered a total a:agle of 89 . On the basis of the assumption that the compressive residual stress in this specimen was equal to that measured in SB3a, the predicted yield m o m e n t was 730 kN m. The fully plastic moment for SB4 was 1970 kN m. Readings of the strain gages located between the stiffeners indicated that compressive yield occurred at a moment of about 750 kN m. It took place throughout the length of the test segment. Such a large volume of yielded shell reduced the stiffness considerably, as can be seen from the change of slopes of the curves in Fig. 14. The yield m o m e n t observed in the test agrees well with the predicted value. Specimen SB4 also failed in general buckling, similarly to SB3b. The first loca]L buckle was observed at the panel of extreme compressive area when the: m o m e n t was 1308 kN m. The deformations developed slowly and new yield lines and buckles appeared at about 1 5 0 0 k N m . The stiffeners bent slowly with the shell surface until the ultimate m o m e n t was reached. 'fielding of the tension zone was indicated by strain gage readings at a m o m e n t of about 1 6 0 0 k N m . At 1800kNm, 45 yield lines were observed at the bottom of the cylinder. As the load approached and passed the ultimate point, buckling at the top and yielding at the bottom proceeded rapidly. The ultimate moment reached was 1857 kN m, or 94% of the plastic moment.

22

Q. Chen, G. L. Kulak, A. E. Elwi

ooof .-,
1700 1
1400 F

:.

41~

BB

t**

*%
o []

~"

1100

800 [] 500 4, o 200 . -10000 , Central

North
South = -8000 = , -6000 . . i . -4000

o to * ~ Q ~ " o Shell yielding a o

, . i . -2000 , II (I . ,h 0

Strain (mlcroatratn)

Fig. 14. Moment versus compressive strain measurements at different cross-sections in specimen SB4.

The moment versus curvature graph for SB4, shown in Fig. 15, illustrates how the behavior of this specimen differed from the shell buckling failures observed in SB2 and SB3a. The development of deformation was slow and gradual, even though nonlinearity started early because of the yielding around the welding of stiffeners. The load was still able to increase significantly after the shell and the stiffeners buckled. The measured strain distribution at the central cross-section showed an upward migration of the neutral axis. This indicates that there was a
2000 /

1700

400
1100

8oo!
500 ~'

LVDT's

Elastic beam theory

' . 0.2 . . i . 0.4 . , 0.6 , 0.8 , , i , 1.0 , , i 1.2

200. 0.0

Curvature ( x l O "

s )

Fig. 15. Moment versus curvature response of specimen SB4.

Flexural tests of fabricated steel cylinders

23

substantial stress redistribution after the initial buckling in the compression zone. A~t the time of failure, nearly all material below the center of the cylinder had yielded. Meanwhile, the buckling waves covered the compression zone. The capacity of the full cross-section was nearly exhausted.

SUMMARY Four fabricated large diameter steel cylinders, reinforced by longitudinal stiffeners, were tested under pure bending moment. The test results are summariized in Table 4, wherein the ultimate moment is compared with the plastic moment of the stiffened cross-section as well as design moments of unstiflimed cylinders. The buckling stresses of unstiffened cylinders were obtained using the Stephens e t al. 6 formulae. The experimental study has led to a number of observations. The initial imperfections of the specimens, exclusive of the ovalization, were moderate. The 3-mm thick cylinder had more severe imperfections as compared with the 5-mm thick shell. In addition, significant compressive residual stresses were introduced by the process of welding the stiffeners to the shell. Detailed measurement of two specimens showed that a higher residual stress level existed in the thinner shell. The flexural behavior in the elastic stage can be predicted adequately by beam theory. However, substantial nonlinear deformation was observed after yielding. Failure was always by buckling, whether before or after yield took place.
TABLE 4 Test Results

Specimen

M. (kN m)
1947 1138 1333 1857

Mpa (kN m)
2659 1475 1524 1970

Mu/Mp

Failure mode

M../Mp ~

SB2 SB3a SB3b SB4

0.73 0-77 0-87 0.94

Shell buckling outside stiffened area Shell buckling outside stiffened area General buckling General buckling

0.62 0.48 0.48 0.50

Plastic moment (stiffeners included) is calculated using the measured dimensions and material properties. Unstiffened moment capacity is calculated from the Stephens et aL6 formula. Plastic moment (stiffeners excluded) is calculated using the measured dimensions and material properties.

24

Q. Chen, G. L. Kulak, A. E. Elwi

The failure of specimens SB2 and SB3a occurred suddenly as a result of shell buckling outside the stiffened region. The first specimen was relatively thick (R/t = 133), but it had only three stiffeners covering an arc of 30 in the compression zone. The second specimen was relatively thin (R/t = 206), but its stiffeners covered an arc of 60 in the compression zone. However, the post-ultimate behavior was stable. The moment capacity was lower than that if a general buckling failure had occurred, but it was still much higher than that of unstiffened cylinders of otherwise identical dimensions. Specimens SB3b and SB4 failed in general buckling when the stiffeners buckled gradually with the shell. The ultimate moment was close to the plastic moment. Both specimens were thin (R/t = 206). Both were stiffened over approximately 90 . Specimen SB4, however, had stiffeners more closely spaced than any of the others. The last column of Table 4 shows the ratio of the ultimate moment of equivalent unstiffened cylinders to the corresponding plastic moment. The capacities of the unstiffened cylinders were obtained using the Stephens et al. 6 formulae. All specimens have failed at substantially higher ratios of moment to plastic moment than those predicted or permitted by any existing design code. Full investigation of the behavior of longitudinally stiffened fabricate steel tubes is therefore warranted. The authors have already carried out a numerical simulation of the test results. On the basis of the success of the finite element analysis, a parametric study was also carried out. The results of the numerical simulation and the parametric study have been used to predict the failure mode regimes and to suggest empirical design rules for the stiffened tubes under flexure. 1

ACKNOWLEDGEMENT This work presented herein constitutes a part of research leading to the degree of Doctor of Philosophy for the first author. Funding was provided in part by the Natural Sciences and Engineering Research Council of Canada.

REFERENCES 1. Green, D. R. & Nelson, H. M., Compression tests on large-scale, stringerstiffened tubes. In Buckling of SheUs in Offshore Structures, Offshore Structures Engineering III, ed. J. E., Harding, P. J. Dowling & N. Agelidis. Gulf Pub. Co., London, UK, 1982.

Flexural tests offabricated steel cylinders

25

2. Chert, W. F. & H a n , D. J., Tubular Members in Offshore Structures. Pitman Adw~nced Publishing Program, Boston, 1985. 3. Prion, H. G. L. & Birkemoe, P. C. Beam-Column Behavior of Unstiffened Fabricated Steel Tubes. Proc. SSRC Annual Technical Session, Houston, Tex~s, 1987. 4. SSRC, Guide to Stability Design Criteria for Metal Structures, 4th cdn, ed. T. V. Galambos. Structural Stability Research Council, Wiley, New York, 1988. 5. Bailey, R. W. & Kulak, G. L., Flexural and Shear Behaviour of Large Diameter Steet Tubes. Struct. Eng. Rep. No. 119, Department of Civil Engineering, University of Alberta, Edmonton, Canada, 1984. 6. Step]hens, M. J., Kulak, G. L. & Montgomery, C. J., Local Buckling of Thin-Walled Tubular Steel Members. Struct. Eng. Rep. No. 103, University of Alberta, Edmonton, Canada, 1982. 7. ASTM. Standard methods of tension testing of metallic materials. ASTM Standard, 01.01E 8M-89, American Society for Testing and Materials, Philadelphia, 1989. 8. Hutchinson, J. W. & Koiter, W. T., Postbuckling theory. Applied Mechanics Review, 23 (1970) 1353-66. 9. Obaia, K. H., Elwi, A. E. & Kulak, G. L., Inelastic Transverse Shear Capacity of l_~rge Fabricated Steel Tubes. Struct. Eng. Rep. No. 171, Department of Civil Engineering, University of Alberta, Edmonton, Canada, 1991. 10. Chcn, Q., Elwi, A. E. & Kulak, G. L., Bending Strength of Longitudinally Stiffened Steel Cylinders. Struct. Eng. Rep. No. 192, Department of Civil Engineering, University of Alberta, Edmonton, Canada, 1993. 11. Dwight, J. B., Imperfection levels in large stiffened tubes. In Buckling of Shells in Offshore Structures, Offshore Structures Engineering III, ed. J. E. Harding, P. J. Dowling & N. Agelidis. Gulf Pub. Co., London, UK, 1982.