Thin-Walled Structures 92 (2015) 115–129

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Practical strain-hardening material properties for use

in deformation-based structural steel design
A.S.J. Foster a,n, L. Gardner b, Y. Wang c
a
Det Norske Veritas, London, UK
b
Imperial College London, London, UK
c
Ove Arup and Partners, Hong Kong

art ic l e i nf o a b s t r a c t

Article history: Through the development of an innovative full cross-section tensile testing method, a programme of
Received 25 October 2014 experiments was conducted to investigate the influence of average cross-section properties on the
Received in revised form constitutive relationships for carbon steel, to validate the use of an elastic linear hardening model in
3 February 2015
practical design, and to assess the resulting accuracy enhancements to the new deformation-based
Accepted 3 February 2015
continuous strength method (CSM) of structural steel design. A total of seventeen full cross-section
Available online 19 March 2015
tensile tests on hot-rolled I-sections, hollow sections and cold-formed hollow sections were performed
Keywords: and these were compared with coupon test data obtained from a supplementary programme of 14
Full cross-section tensile tests tensile coupon tests and data carefully obtained from the literature. The overall behavioural response of
Residual stress
the cross-section tensile tests demonstrated that assuming an elastic, linear hardening material model
Strain-hardening
for the CSM is a reasonable assumption and the previous assumption concerning the magnitude of the
Material modelling
Section properties strain-hardening modulus, based upon the recommendations of EN 1993-1-5, is overly conservative. A
Plastic design revised suite of material models was presented and was shown to furnish the CSM capacity equations
with a higher degree of accuracy when compared against experimental data.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction and key design aspects basic assumptions are made: (1) the underlying material model is
elastic, linear hardening and (2) in the elastic range the relationship
Current structural design codes generally represent the material between stress and strain is defined by Young's modulus E and beyond
stress–strain characteristics of structural steel by means of an elastic, the yield stress fy this relationship is defined by a strain-hardening
perfectly plastic model. This leads naturally to the concept of elastic modulus, taken as Esh ¼ E=100 as recommended by EN 1993-1-5. An
and plastic moment capacities and the process of cross-section elastic, linear hardening model is able to represent strain-hardening
classification. Although simple, this treatment can lead to overly effects, and the slope can be adjusted to suit the grade, section type
conservative designs. A newly proposed, deformation-based approach and forming method. Tensile coupon test data typically exhibit a
to structural steel design, referred to herein as the continuous strength prolonged Lüders or yield plateau with the implication that prior to
method (CSM) [1], represents an alternative treatment to cross-section the onset of any strain-hardening, significant strains must develop;
classification that is based upon a continuous relationship between however in this paper it will be demonstrated that this plateau is
cross-section slenderness and deformation capacity, as well as a substantially eroded when considering the stress–strain response of
rational exploitation of strain-hardening. Strain-hardening can be the full cross-section, due to it encompassing variable plate thickness,
broadly defined as the additional strength beyond yield arising as a residual stresses and localised strain-hardening due to cold-forming,
result of plastic deformation; its importance in the design of steel as well as variations in the yield stress throughout the cross-section.
structures has been previously recognised, notably by [2,3]. The Basing structural design equations on full-cross section tests in
continuous strength method has been shown to offer increases in compression are well documented (see [4]); the purpose of this
member resistance of up to 15% over current European standards, as paper is to take advantage of a modification of this approach,
well as a reduction in scatter when compared with test data. whereby the whole cross-section is tested in tension with the aim to
Amongst the key parameters required to develop and use the CSM,
material properties are of fundamental importance. For the CSM, two (i) Examine the strain-hardening behaviour of various hot-rolled
and cold-formed carbon steel sections to determine section
dependent values of Esh.
n
Corresponding author (ii) Propose a suite of material models suitable for the CSM as an
E-mail address: [email protected] (A.S.J. Foster). enhancement to the general provisions made by EN 1993-1-5.

http://dx.doi.org/10.1016/j.tws.2015.02.002
0263-8231/& 2015 Elsevier Ltd. All rights reserved.
116 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129

(iii) Validate the assumption of using an elastic, linear hardening

stress strain curve with immediate strain-hardening in the
post-yield range.
(iv) Quantify the accuracy enhancements to the CSM resulting
from the improved material models.

A wide range of parameters (steel grade, cross-section shape,

forming process, loading conditions and local plate thickness)
affecting the stress–strain response of structural steel have been
identified and their relevance to the CSM has been discussed by
Wang [5] whose conclusions were drawn from the analysis of
Fig. 1. Typical stress–strain curve for hot-rolled carbon steel.
tensile and compressive coupon test data.
Assuming the same general properties (steel grade, cross-
section shape, forming process and loading conditions), the
stress–strain response of any given specimen will encounter 2. Overview of material modelling approaches
location specific variations in material properties (i.e. the material
properties will vary depending on the location from which the 2.1. General
coupon is extracted), which are determined by factors such as
plate thickness, work hardening due to forming and the distribu- The typical mechanical properties of hot-finished structural
tion of residual stresses due to differential cooling rates through steel subjected to static uniaxial load are illustrated Fig. 1. In the
the cross-section. A coupon test will only provide a representative elastic range the slope is linear and defined by the modulus of
stress–strain response for the area from which it has been cut; elasticity (Young's modulus) E, which is valued at 210,000 N/mm2
coupons taken from multiple locations will provide a family and in EN-1993-1-1. The elastic range is limited by the yield stress, fy,
hence a range of stress–strain responses, but these will still neglect and corresponding yield strain ϵy. Beyond ϵy a plateau forms with
any interactions that develop when the full cross-section is no increases in stress until ϵsh is reached, which is the strain at
stressed. which strain-hardening initiates. At this point, stress accumulation
recommences at a reduced rate Esh which is the tangent modulus
of the slope at the onset of strain-hardening.
Various idealisations of this relationship exist and can be grouped
1.1. The continuous strength method as (1) elastic or rigid, perfectly plastic; (2) elastic, linear hardening; or
(3) elastic, multi-linear hardening or non-linear hardening. The rigid
The continuous strength method is a deformation-based design plastic model is illustrated Fig. 2a and forms the basis of current
approach for steel elements that allows for the beneficial influence plastic design methods that neglect strain-hardening. Linear hard-
of strain-hardening. To date, design equations for the CSM have ening models have at least two distinct phases of stress accumulation
been developed for cross-section resistance in bending and characterised by the initial slopes at each transition strain. For the
compression [6]. The CSM bending resistance function Mcsm,Rd, elastic, linear hardening model illustrated in Fig. 2b there is an initial
which applies for λ p r 0:68 is defined in (1) as elastic phase where stress and strain are related by E, followed by a



 ! strain-hardening phase whose rate of stress accumulation is reduced
W pl f y Esh W el ϵcsm W ϵcsm  2 by some proportion of E to give Esh.
M csm;Rd ¼ 1þ  1  1  el
γ M0 E W pl ϵy W pl ϵy A comprehensive review of the form of material model to be
ð1Þ adopted by the CSM is given in Wang [5] where it was identified
that the following criteria should be satisfied:
where E is the modulus of elasticity, Esh is the strain-hardening
slope taken equal to E=100 for structural steel, Wel and Wpl are the (i) A minimal number of parameters.
elastic and plastic section moduli and ϵcsm =ϵy is the strain ratio, (ii) Overall accuracy of the stress–strain description for mechan-
defining the limiting strain in the cross-section ϵcsm as a multiple ical behaviour.
of the yield strain ϵy, and given by (2) (iii) Consideration of strain-hardening.
(iv) Stress can be solved for explicitly.
ϵcsm 0:25 (v) Consistency with the current design code (Eurocode 3).
¼ 3:6 but r 15 ð2Þ
ϵy λ p
It was concluded that the elastic, linear hardening model best
in which λ p is the local cross-section slenderness, given by (3) as satisfied the criteria and has since been used throughout the
development of the CSM elsewhere [6].
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
λp ¼ f y =σ cr ð3Þ 2.2. Current modelling approaches adopted by the CSM

with σcr being the elastic buckling stress of the cross-section, or Early work on the CSM concentrated on applications to stain-
conservatively its most slender constituent plate element. less steel [7–9], and as such employed the Ramberg–Osgood
This research will present (1) a brief summary of the most material model. Subsequent extensions to carbon steel [1] have
widely adopted material modelling approaches; (2) a summary of led to the general application of the elastic, linear hardening
previous studies and proposals for material models based on local material model; this model has recently been applied to stainless
coupon test data; (3) the results of an experimental programme steel design [10], motivated by its simplicity and consistency with
carried out at Imperial College London; (4) an updated proposal current design codes.
for the material models to be used in the CSM, taking into account EN 1995-1-1-5 suggests a value of Esh ¼ E=100 for all types and
the average tensile cross-section stress–strain properties deter- grades of steel section. Previous work conducted by Wang [5] on
mined in the experimental investigation. carbon steel properties demonstrated that such a generalised
 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129 117

Fig. 2. Material models with and without strain-hardening. (a) Elastic, perfectly plastic. (b) Elastic, linear hardening.

Table 1
Summary of existing material models for the CSM [5].

Hot-rolled I-sections
Esh =E ¼ 0:015
ðf u =f y  1:0Þ for f u =f y r 1:7
1:7  1:0
Esh =E ¼ 0:015 for f u =f y 4 1:7
Hot-rolled hollow sections
Esh =E ¼ 0:003
ðf u =f y  1:0Þ for f u =f y r 1:3
1:3  1:0

Esh =E ¼
ðf =f  1:0Þ
0:003 þ 0:007 u1:6y 1:3 for 1:3 o f u =f y r 1:6
Esh =E ¼ 0:01 for f u =f y 4 1:6
Cold-formed hollow sections
ðf =f  1:0Þ
Esh =E ¼ 0:01 1:25
u y for f u =f y r 1:25
 1:0
Esh =E ¼ 0:015 for f u =f y 4 1:25

be representative of overall material behaviour. Broadly, Esh can be

determined according to the following methods:

(i) Basing Esh on the initial slope at the onset of strain-hardening.

(ii) The equal energy dissipation method.
(iii) Direct linear method.

The initial slope method is generally only used for tri-linear

representations of material behaviour [11,12] and assumes a
constant value of Esh based on the initial post-yield tangent slope
taken at ϵsh (Fig. 1). This model is unsuitable for elastic, linear
Fig. 3. Elastic linear-hardening model based upon equal energy dissipation.
hardening models as it often provides a value of Esh that is too high
measure of strain-hardening behaviour can be refined. In particular, to be used for the entire post yield range of the stress–strain curve.
following a survey of experimental data collected from the litera- The concept of equal energy dissipation is illustrated schema-
ture, it was shown that tically in Fig. 3. Here, the post-yield portion of the stress–strain
curve is represented by a straight line connecting ϵy to ϵu via a
(i) The degree of strain-hardening is a function of f u =f y . value of Esh that ensures the areas of the shaded portions between
(ii) The yield plateau for hollow sections exceeds that of the model and actual curves are the same [13]. This method can be
I-sections. applied generally, but requires some effort to compute.
(iii) For a given value of f u =f y , I-sections exhibit a higher degree of The direct method involves fitting a straight line through two
strain-hardening than hollow sections. values obtained from the experimental stress–strain curve. In a
(iv) The ultimate non-dimensional strength f u =f y of hot-rolled survey of 50 mill tests taken from hot-rolled I- and H-sections,
sections is higher than it is for cold-formed sections, but hot- Byfield et al. [14] examined strain-hardening behaviour by calcu-
rolled sections exhibit a lower initial degree of strain-hardening. lating the slope of the lines connecting 1% and 4% strains (or 6ϵy).
(v) Stub column tests exhibit shorter yield plateaus and a higher The approach developed by Wang [5] and that will be adopted
initial degree of strain-hardening than coupon tests. in this paper draws upon methods used for material properties
analysis of stainless steel [10], as well as the elements of the direct
Expressing the degree of strain-hardening in non-dimensional approach outlined above; this method will then be applicable to
form (Esh =E), the proposed models for the CSM suggested by both hot-rolled and cold-formed carbon steel sections. The parti-
Wang [5] are presented in Table 1. cular elements of the method are as follows:

2.3. Methods to calculate the strain-hardening modulus (i) For hot-rolled I- and hollow sections, a line is constructed
between 0.5% and 3% strain (Fig. 4a).
For all considered section types and materials, the value of the (ii) For cold-formed sections, a line is constructed between the
strain-hardening modulus varies with progression along the 0.2% offset strain and the 3% strain (Fig. 4b).
stress–strain curve. Selecting a representative value of Esh for an (iii) If the ultimate stress is reached at a strain less than 3%, then
elastic, linear hardening material model needs some care if it is to the ultimate strain ϵu is used.
118 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129

Fig. 4. Methods to determine Esh. (a) Hot-rolled. (b) Cold-formed.

2.4. Previous experimental procedures 3.2. Local tensile material properties

Although the tensile coupon test is the most prevalent method for The tensile coupons were cut and milled from the web and
extracting fundamental information on material behaviour, it faces a flanges of the I-sections in the longitudinal (rolling) direction only
number of limitations. Firstly, residual stresses present in the parent (Fig. 5a). Testing was carried out in accordance with the provisions
material due to non-uniform cooling and plastic deformations are of EN 10002-1 (2001). The nominal dimensions of the tensile
released upon cutting the coupons [15]. Secondly, the stress–strain coupons were 320  30 mm with the parallel length region having
curve will vary according to the location on the section from which a nominal width of 20 mm (Fig. 5b).
the coupon is cut [16,5]. Compression coupon tests face similar Prior to testing, half gauge lengths were marked onto the
limitations and because of the need to prevent buckling, the range of tensile coupons to allow the final strain at fracture, ϵf, to be
strain over which data can be collected is limited. calculated,
pffiffiffiffiffi based on the elongation over the standard gauge length
Average cross-section behaviour can partially be taken into account 5:65 Ac where Ac is the cross-sectional area of the coupon. A
using stub column tests – these encompass both residual stresses and summary of the measured coupon dimensions is provided in
the interactions between varying local stress–strain relationships. Table 2, in which F denotes material extracted from the flange
However, their primary role is for evaluating local buckling capacity and W denotes material extracted from the web, and bc and tc are
[17] and not Esh as often local buckling occurs at low strain levels. the measured width and thickness of the coupons, respectively.
Hence, performing a tensile full cross-section test will confer the
benefits of stub column tests, but they will be observed over the
ranges of strain typically associated with a tensile coupon test, as well 3.2.1. Testing conditions
as permitting a more meaningful comparison between the two. This The tensile coupon tests were carried out in an INSTRON
type of test will be discussed in detail in Section 3.3. 500 kN hydraulic loading machine with an initial strain rate of
0.001%/s for ϵ o 0:5 %. Between 0:5% and 4% strain, the strain rate
3. Experimental programme was 0.002%/s; between 4% and 17% strain, the strain rate was
0.04%/s. These test rates are summarised in Table 3.
3.1. Introduction Once the coupon reached ϵ ¼ 17% testing switched to displace-
ment control at a constant rate of 0.1 mm/s until failure. Static
A testing programme comprising tensile material coupon tests yield and ultimate strengths were determined by holding the
and full cross-section tension tests was carried out on hot-rolled strain constant for 2 min in the yield plateau and at four points
and cold-formed steel I - and hollow sections in the structures near the ultimate stress.
laboratory at Imperial College London. Full cross-section tensile Pausing the tests permits the stress relaxation associated with
tests were carried out on the following sections: plastic straining to take place. In the plastic range, total strain can be
decomposed into elastic and plastic components, of which the latter
(i) Hot-rolled grade S355JR I-sections in seven sizes – 305  is time dependent and typically takes several minutes to develop.
127  48 UB, 305  165  40 UB, 305  102  28 UB, 254  When pausing a test performed in displacement control, the total
102  28 UB, 203  133  25 UB, 152  152  23 UC, and strain is maintained at a constant level and hence, as the plastic
152  152  30 UC. strain increases with time, the elastic strain decreases at the same
(ii) Hot-rolled grade S355J2H square and rectangular hollow rate accordingly [18]. Reductions in the elastic strain component
sections in four sizes – 40  40  3 SHS, 40  40  4 SHS, result in a reduction in the stress under constant total strain; this is
60  40  4 RHS, and 60  60  3 SHS. known as the stress relaxation. Underpinning this process is the
(iii) Cold-formed grade S235JRH square and rectangular hollow observation by Lay [19] that when the shear stresses at the
sections in four sizes – 40  40  3 SHS, 40  40  4 SHS, structural level of the material reach a critical value, slip planes
60  40  4 RHS, and 60  60  3 SHS. form in the direction of the shear stresses; these are unstable and
will progress until a change of surface is encountered. It was also
Tensile material coupon tests were carried out for the I-sections as shown by Lay [19] that the stress required to initiate a slip plane is
part of this programme to determine the local engineering stress– greater than that required to maintain it. Thus, when slip planes are
strain material response of the full cross-section tensile specimens; forming under a critical stress, stopping any further strain incre-
complementary tensile coupons for the hollow sections were tested ments will cause the stress to fall to a static level, the level at which
and reported by Wang [5]. a change in surface is encountered. It is shown by Krapf [20] that
 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129 119

Web
coupon

Flange
coupon

Fig. 5. Details of tensile coupon specimens. (a) Typical locations of tensile coupons. (b) Tensile coupon geometry and dimensions.

Table 2
Basic tensile coupon specimen geometry.

Tensile coupon designation bc tc Ac

(mm) (mm) (mm2)

305  127  48 F 19.5 14.4 280.5

305  127  48 W 19.7 8.5 167.2
305  165  40 F 19.5 10.0 194.5
305  165  40 W 20.2 6.4 128.3
305  102  28 F 20.0 8.1 163.1
305  102  28 W 19.9 6.3 124.5
254  102  28 F 20.1 9.7 194.8
254  102  28 W 20.1 6.3 126.4
203  133  25 F 20.0 7.1 141.5
203  133  25 W 20.1 6.4 128.7
152  152  23 F 20.0 6.8 136.6
152  152  23 W 20.0 6.2 123.3
152  152  30 F 19.9 9.1 180.6
152  152  30 W 20.1 6.7 133.9

Fig. 6. Tensile coupon test set-up.

Table 3 A summary of the results of these tests is provided in Table 4 together

Summary of tensile coupon testing rate.
with the test results of Wang [5] in Table 5; in the tables, the strain-
Strain Test rate hardening modulus Esh was determined using the direct method, as
set out in Section 2.3. In the coupon designation system HR denotes
0–0.5% 0.001 %/s hot-rolled, CF denotes cold-formed and C denotes material taken from
0.5–4% 0.002 %/s the corners of the cross-section.
4–17% 0.04 %/s
417% 0.1 mm/s
Summary stress–strain curves for all the coupon tests conducted
in this programme are shown in Fig. 7a, with a second plot (Fig. 7b)
focusing on the region up until the onset of strain-hardening.
The tensile coupons for the hot-rolled sections exhibited the
stress decreases logarithmically with time during pausing, suggest- anticipated response of a well-defined yield point, followed by a
ing that relatively brief pauses will be sufficient to obtain reliable plateau before the initiation of strain-hardening. Fig. 8a and b it is
static values during tensile coupon tests. noted that material taken from hot rolled hollow sections exhibit a
Tensile strain was measured using a clip gauge extensometer and longer yield plateau than material taken from I-sections. The cold-
strain gauges affixed to the face of the specimen (Fig. 6). The clip gauge formed material does not exhibit a well defined yield point, with a
was used to control the testing machine and the strain gauges were progressive and rounded transition from elastic to inelastic behaviour.
used to provide accurate readings in the elastic range, against which Examination of Fig. 8a also illustrates the differences in response of
any necessary adjustments to the clip gauge data can be made. material taken from the flanges and webs, with the flange material
Preliminary analysis in [21] demonstrated that in the elastic range, generally experiencing a shorter yield plateau than the web.
strain-gauges are more accurate and so are used for determining the Examination of the results presented in Table 4 indicates that
modulus of elasticity, but the full stress–strain response is based upon the value of Esh is generally higher for thicker material. Studies
the corrected readings from the clip-gauge extensometer. have shown that whilst this pattern holds true in some cases, it is
not a general relationship [14], but it has been argued elsewhere
that the ratio f u =f y decreases with decreasing thickness and that
3.2.2. Test results and discussion Esh increases with increases in f u =f y ; thus thickness is implicitly
All test data, including load, displacement, strain and input voltage related to the strain-hardening modulus via the ratio of ultimate
were recorded at 1-s intervals using the DATASCAN acquisition system. strength to yield stress [5].
120 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129

Table 4
Measured tensile coupon test results.

Tensile coupon designation fy fu ϵsh ϵu ϵf Area reduction E Esh

(N/mm2) (N/mm2) (%) (%) (%) (%) (N/mm2) (N/mm2)

UB 305  127  48 F 381 (392) 499 (534) 1.6 17.1 30.8 62.8 191,700 2700
UB 305  127  48 W 396 (403) 494 (528) 2.0 19.1 24.2 56.1 198,700 2100
UB 305  165  40 F 428 (438) 579 (585) 1.6 15.7 25.1 67.0 204,200 2900
UB 305  165  40 W 452 (459) 569 (599) 2.0 13.7 21.6 47.9 201,340 2900
UB 305  102  28 F 363 (375) 488 (518) 1.0 16.0 33.9 63.8 207,400 2600
UB 305  102  28 W 357 (366) 492 (523) 1.1 17.5 32.1 59.4 188,500 2600
UB 254  102  28 F 382 (386) 501 (530) 1.1 12.0 26.7 65.0 198,800 2700
UB 254  102  28 W 386 (401) 513 (549) 1.2 16.3 32.5 58.0 212,500 2400
UB 203  133  25 F 354 (364) 492 (522) 0.7 14.4 28.1 61.4 204,100 3200
UB 203  133  25 W 379 (385) 490 (519) 1.6 16.2 32.9 61.9 203,100 1900
UC 152  152  23 F 345 (355) 471 (500) 0.8 15.5 30.0 60.9 201,300 2900
UC 152  152  23 W 362 (371) 476 (507) 1.2 16.0 32.9 54.0 211,400 2300
UC 152  152  30 F 374 (382) 498 (528) 0.7 14.3 28.3 57.5 210,000 3000
UC 152  152  30 W 376 (383) 490 (521) 1.4 15.5 30.8 57.0 205,300 2300

Table 5
Collated tensile coupon test data [5].

Tensile coupon designation fy fu ϵsh ϵu ϵf E Esh

(N/mm2) (N/mm2) (%) (%) (%) (N/mm2) (N/mm2)

SHS 40  40  3 HR 504 581 3.9 9.1 36.0 219,600 0

SHS 40  40  3 CF 451 502 1.2 5.9 24.0 212,900 1280
SHS 40  40  3 CF-C 534 589 0.4 3.3 16.0 196,700 2380
SHS 40  40  4 HR 496 572 4.3 14.2 34.0 212,300 0
SHS 40  40  4 HR-C 499 578 3.9 8.5 37.0 215,500 0
SHS 40  40  4 CF 410 430 0.4 5.1 38.0 201,600 730
SHS 40  40  4 CF-C 479 507 0.4 1.5 17.0 210,900 2870
RHS 60  40  4 HR 468 554 4.7 9.8 37.0 213,800 0
RHS 60  40  4 CF 400 452 0.4 5.0 21.0 212,000 1880
RHS 60  40  4 CF-C 480 570 0.4 2.4 15.0 202,400 3910
SHS 60  60  3 HR 449 555 3.8 9.9 31.0 215,200 0
SHS 60  60  3 CF 361 402 0.4 12.0 49.0 207,400 1030
SHS 60  60  3 CF-C 442 471 0.4 1.4 21.0 208,000 2550

600 600

500 500
Stress σ (N/mm2)

Stress σ (N/mm2)

400 400

300 300

200 200

100 Flange coupon 100 Flange coupon

Web coupon
Web coupon
0 0
0 5 10 15 20 25 30 0 0.5 1 1.5 2 2.5 3
Strain ε (%) Strain ε (%)
Fig. 7. Summary of tensile stress strain curves for I-sections. (a) Complete stress–strain curves. (b) Initial portion of stress–strain curve.

3.3. Full cross-section tensile tests section types. In this section, the details of the test procedure will
be discussed, along with a presentation of the basic results and how
In order to account for the interactions between varying material they compare to data obtained from local tensile coupon tests.
and geometric properties of the different section types that are
otherwise overlooked by tensile coupon tests, a new, full cross-
section tensile test is proposed. As previously discussed, taking into 3.3.1. Test specimens
account average cross-section properties will give a more accurate Full cross-section tensile tests were performed on hot-rolled
representation of the uniaxial tensile behaviour of the different (S355JR) I-sections, hot-rolled (S355J2H) hollow sections and cold-
 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129 121

600

Stress σ (N/mm2) 500

400

300

200

100 305x102x28 UB Flange

305x102x28 UB Web

0
0 2 4 6 8 10 12
Strain ε (%)
Fig. 8. Comparative stress–strain response for different cross-section types. (a) I-sections. (b) Hot-rolled hollow sections.

Table 6 distribution of bolts at sufficiently close intervals to spread the

Basic geometric properties of tensile I-sections. load evenly.
(iii) Sufficiently strong welds between the end plates and the
Cross-section tensile test H B tf tw r A
specimen; vertical stiffener fins were used with the primary
designation (mm) (mm) (mm) (mm) (mm) (mm2)
function of increasing the welded area between the specimen
305  127  48 UB S1 311.48 127.39 13.93 8.60 15.00 6183 and the end plates.
305  127  48 UB S2 311.48 127.39 13.93 8.60 15.00 6183
305  165  40 UB S1 305.88 166.78 10.01 6.54 12.00 5331 A summary of these design features is provided in Fig. 10.
305  165  40 UB S2 306.15 166.89 9.96 6.30 12.00 5249
Prior to testing, gauge lengths were scribed along the length of the
305  102  28 UB 307.67 105.08 8.44 6.35 9.40 3697
254  102  28 UB 260.25 102.74 9.36 6.39 12.40 3599 specimens to allow the final strain at fracture, ϵf, to be calculated. The
pffiffiffiffiffi
203  133  25 UB 201.49 133.95 6.96 6.34 12.00 3178 standard gauge length L0 ¼ 5:65 Ac (where Ac is the pre-test cross-
152  152  23 UC 153.77 152.38 6.53 5.94 14.50 3006 sectional area of the specimen) suggested by EN 10002-1 (2001) for
152  152  30 UC 155.91 154.28 8.75 6.57 14.40 3788
tensile coupons is too long for this type of test. The following
modifications to the gauge length were therefore introduced:

Table 7 (i) For the hollow sections, the local plate cross-sectional area
Basic geometric properties of tensile hollow sections. was p ffiffiffiffiffi for Ac and substituted into the expression L0 ¼
used
5:65 Ac to determine L0.
Cross-section tensile test H B t ro A
(ii) For the I-sections, a non-proportional gauge length L0 ¼ 80 mm
designation (mm) (mm) (mm) (mm) (mm2)
was used. Elongation values based on this gauge length are
pffiffiffiffiffi
SHS 40  40  3 HR 40.00 40.44 3.13 3.80 453 comparable to those based upon L0 ¼ 5:65 Ac by means of a
SHS 40  40  3 CF 40.24 40.34 2.83 5.50 404 conversion provided in Tables 4 of EN 2566-1 (1999).
SHS 40  40  4 HR 40.13 40.08 3.99 4.00 563
SHS 40  40  4 CF 40.69 40.45 3.82 7.50 525
The full cross-section tensile tests were carried out using an INSTRON
RHS 60  40  4 HR 40.50 60.47 3.92 5.30 708
RHS 60  40  4 CF 40.27 60.33 4.01 5.60 718 3500 kN hydraulic universal testing machine. The end plates of the
SHS 60  60  3 HR 60.40 60.59 3.29 5.50 731 specimen were securely bolted to the end platens of the testing
SHS 60  60  3 CF 60.11 59.86 2.80 7.50 611 machine prior to testing. Displacements were recorded by 8 L.V.D.T.s
positioned at uniform locations on both ends of the specimen
(Fig. 11a). The applied load was recorded with a load cell and strain
formed (S235JRH) hollow sections. All of the test specimens were gauges were attached to each face of the specimen at the mid-height
cut to a nominal length of 600 mm. The basic measured geometric (six for I-sections and four for hollow sections) – Fig. 11b.
properties of the specimens are reported in Tables 6 and 7. For the Testing was carried out under displacement control with similar
305  127  48 UB and 305  165  40 UB specimens, two tests relative variations in the testing rate to the coupon tests, which are
of each section were carried out for initial assessment purposes summarised in Table 8. Unlike the coupon tests, displacement
and these are denoted S1 and S2. Notation for the geometric holding points were not introduced into the test profile as firstly,
properties is defined Fig. 9. Tolerance compliance was confirmed the testing machine was not capable of reliably introducing such
according to EN 10310-2 (2006) and EN 10319-1 (2006). holds and secondly, by having a very slow test rate, the difference
between static and dynamic stresses is likely to be negligible [22]. All
test data, including load, displacement, strain and input voltage, were
3.3.2. Testing conditions recorded at 1-s intervals using the DATASCAN acquisition system.
The following conditions were sought from the testing arrange-
ment:

(i) An even distribution of strain across the cross-section; this 3.3.3. Test results
was achieved by welding end plates to the specimen. All specimens failed by ductile fracture, which was the antici-
(ii) Minimal deformation of the end plates under load; this is pated mode of failure. In most, but not all cases, fracture occurred at
achieved by using thick end plates (25 mm) with a symmetrical the mid-height of the specimens. Examples of this mode of failure
122 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129

Fig. 9. Notation for full section tensile test specimens. (a) I-sections. (b) Hollow sections.

Stiffener

Specimen

Bolted
Weld
end plate

Fig. 10. Typical details of fabricated specimens for full cross-section tension tests. (a) I-sections. (b) Hollow sections.

are presented in Fig. 12, in which Fig. 12a and b shows this mode for resulted in small discrepancies between L.V.D.T. and strain gauge
I-sections and Fig. 12c shows this mode for square hollow sections. initial stiffness readings.
The procedure for accounting for non-uniform strain along the
length of the test member was developed by Wang [21] and the
3.3.3.1. Data processing. During the tests, measurements based key steps involved are presented in this section. The first step is to
upon L.V.D.T., strain gauge, load cell and machine displacement construct the basic stress–strain (σ ; ϵ) curve directly from the
readings were collected. Prior to analysis, certain considerations load-displacement readings obtained from the tests, where
need to be taken into account. Firstly, unlike a tensile coupon test, σ ¼ F=A and ϵ ¼ δ=L ð4Þ
inclusion of the additional stiffening material at the ends of the
specimens resulted in non-uniform strains developing along the in which δ is the axial displacement recorded by the L.V.D.T.s, L is
length of the test member. Secondly, like the stub column tests, the specimen length, F is the recorded load and A is the specimen
the end plates experienced small elastic deformations which cross-section area. Using this data, the initial tangent modulus E0 is
 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129 123

Fig. 11. Test set-up for full cross-section tension tests.(a) Schematic instrumentation. (b) In situ.

(iii) The welded edges of each stiffener provide a rigid body

Table 8
connection with the specimen.
Test rate for full tensile cross-section tests.
(iv) The free edges of the stiffener and the triangular region that
Displacement Test rate they enclose have no contribution to the stiffness of the
(mm) (mm/s) segment Lend.
(v) The effective material area can be simplified into equivalent
0–3 0.004
3–14.5 0.008
rectangular regions (Fig. 13b).
14.5–20 0.016
20–24 0.08 For equilibrium, the corrected initial modulus ELVDT must satisfy
4 24 0.16
ELVDT ¼ kLVDT E0 ð7Þ
Correspondingly

calculated. This value is an average for the whole specimen and σ end σ mid F=ðA þ Astiff Þ F=A
¼ ⟺ ¼ ð8Þ
includes the additional rigidity of the stiffeners. ϵend ϵmid δend =Lend δmid =Lmid
The second step is to calculate a corrected modulus of elasticity To derive an expression for kLVDT it is necessary to solve for ϵmid in
from the L.V.D.T. data, ELVDT, allowing for the influence of the Eq. (8). Solving initially for ϵend
stiffeners, which is achieved using (5)
ϵmid A 1
ϵend ¼ ¼ ϵ ð9Þ
ðA þAstiff Þ α mid
ELVDT ¼ kLVDT E0 ð5Þ
Axial displacement δ can be expressed as
in which
δ ¼ δmid þ 2δend ¼ ϵmid Lmid þ2ϵend Lend ð10Þ
β  
kLVDT ¼ þ 1β ð6Þ Noting, with reference to Fig. 14, Lmid ¼ L  2Lend and by definition,
α
ϵ ¼ δ=L, Eq. (10) becomes (upon substituting Eq. (9))


where α ¼ ðA þ Astiff Þ=A and β ¼ 2Lend =L and Astiff is the area of the L  2Lend 1 2L
ϵ ¼ ϵmid þ ϵmid end ð11Þ
stiffeners. With reference to Fig. 13a this area is defined by the L α L
shaded triangle formed by the welded edges (the triangle defined by
the free edges is assumed to have a zero contribution), which in turn
From the earlier definition of β, Eq. (11) becomes
provides a rigid body connection between the stiffeners and the   β   β
ϵ ¼ ϵmid 1  β þ ϵmid or ϵ ¼ 1  β þ ϵmid ð12Þ
specimen. With reference to Fig. 13b, the triangular area is simplified α α
to an equivalent rectangular region. This region, in conjunction with
Rearranging
Fig. 14, defines the length of the specimen, Lend, over which the
stiffeners influence the elastic rigidity of the specimen.
ϵ
In deriving the stiffness correction factor kLVDT the following ϵmid ¼ h  i ð13Þ
assumptions are made: 1  β þ αβ
With reference to Eqs. (7) and (8)
(i) The whole specimen consists of three regions: two identical
σ mid F=A ð1  βÞ þ β=α
end regions defined over the lengths Lend and a single middle ELVDT ¼ ¼ ð14Þ
ϵmid ϵ
region Lmid (Fig. 14).
(ii) The material properties of the stiffener and the test specimen For Eq. (7) to hold, E0 ¼ ðF=AÞ=ϵ, which is the initial slope of the
are homogeneous. stress–strain curve derived using Eq. (4), and kLVDT ¼ ð1  β Þ þ
124 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129

Fig. 12. Typical failure modes for all cross-section tensile tests.

Fig. 13. (a) Area contributions of stiffeners and (b) simplification of effective stiffener geometry.

Fig. 14. Nomenclature for stiffness reduction factor calculations.

 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129 125

Table 9 The final corrected value for strain in the region Lmid is then
Summary of stress calculations for stiffeners (I-sections).
δmid;true
ϵmid;true ¼ ð20Þ
Cross-section tensile test designation A Fu Astiff f stiff Lmid
(mm2) (kN) (mm2) (N/mm2)
It is acknowledged by the authors that certain aspects of this
305  127  48 UB S1 6183 3071 4000 302 procedure may be omitted by adopting a different L.V.D.T. arrange-
305  127  48 UB S2 6183 3063 4000 301 ment, as exemplified by [24] for tests in compression.
305  165  40 UB S1 5331 2772 4000 297
305  165  40 UB S2 5249 3059 4000 331
305  102  28 UB 3697 1983 4000 258 3.3.3.2. Summary of test results. Tables 11 and 12 present the key
254  102  28 UB 3599 1957 4000 257 test results for all of the tests; all notation is as previously defined
203  133  25 UB 3178 1697 4000 236 for the tensile coupon tests. Fig. 15 shows the stress–strain curves
152  152  23 UC 3006 1535 4000 219
derived from the full cross-section tensile tests for all specimens
152  152  30 UC 3788 1946 4000 250
tested. The final strain at fracture, ϵf, is based upon the weighted
average area contributions of each component of the specimen
cross-sections. Attempts to quantify area reduction for some of the
Table 10 I-sections are also presented, though in general this proved difficult
Summary of stress calculations for stiffeners (hollow sections). to calculate with any reliability. The modulus of elasticity was
obtained from strain gauge readings and the strain-hardening
Cross-section tensile test designation A Fu Astiff f stiff
modulus was estimated using the method described in the
(mm2) (kN) (mm2) (N/mm2)
preceding section. For cold-formed sections, fy and ϵy are taken as
SHS 40  40  3 HR 453 242 2000 99 the 0.2% proof stress and 0.2% offset strain.
SHS 40  40  3 CF 404 212 2000 88 The key observations from the test results are:
SHS 40  40  4 HR 563 300 2000 117
SHS 40  40  4 CF 525 247 2000 98
RHS 60  40  4 HR 708 372 2000 137
(i) All hot-rolled cross-sections still exhibit a yield plateau,
RHS 60  40  4 CF 718 350 2000 129 though this is significantly reduced when compared to the
SHS 60  60  3 HR 731 406 2000 149 results from the coupon tests. For I-sections, the average
SHS 60  60  3 CF 611 256 2000 98 ϵsh ¼ 1:28% for the coupons, whilst for the full section tensile
tests the average ϵsh ¼ 0:72% – a 43% reduction in the length
of the yield plateau. For hot-rolled hollow sections, the
average ϵsh ¼ 4:16% for the coupons, whilst for the full tensile
β=α. The values α and β can be readily evaluated from measure- tests the average ϵsh ¼ 2:91% – a 30% reduction in the length
ments obtained from the test specimens. of the yield plateau. For cold-formed sections, the average
By introducing the correction in Eq. (6) it is assumed that over ϵsh ¼ 0:59% for the flat coupons (average ϵsh ¼ 0:43% for
the length Lend, the material remains in the elastic range throughout corner coupons), whilst for the full section tensile tests the
the test. Defining the stress in this region as f stiff ¼ F u =ðA þ Astiff Þ, average ϵsh ¼ 0:42%.
Tables 9 and 10 show that all stresses in this region are indeed (ii) In all cases there is an absence of a well-defined yield point,
below the respective material yield stress values (assuming that the reflecting the interactions of varying material properties
material properties of the stiffeners are the same as those of the around the cross-section and residual stresses.
specimen), indicating that this is a reasonable assumption. (iii) Comparing values of Esh to those derived from the coupon
Assuming that the region Lend deforms elastically confines all tests, higher values for Esh are observed in all cases, with the
the inelastic deformations to the region Lmid (Fig. 14). The third exception of the hot-rolled hollow sections which typically
step is then to construct the appropriate stress–strain relationship exhibit zero strain-hardening (up to 3% strain).
(σmid, ϵmid). Noting that
δmid
ϵmid ¼ ð15Þ
Lmid
4. Analysis of test data
it is necessary to determine δmid. With reference to Fig. 14, the
total axial displacement δ can be decomposed as In this section, the results of the full cross-section tensile tests
δ ¼ δmid þ 2δend ⟺ δmid ¼ δ  2δend ð16Þ for each section type are compared and discussed. Thereafter, the
relative attributes of the full cross-section tensile tests are exam-
Evaluating the modulus of elasticity from the strain gauges, ESG, and ined alongside those of the tensile coupon tests. Summary non-
invoking the previous assumption of elastic behaviour in the region dimensional Esh =E values derived from the tensile full cross-
Lend, the quantity δend ¼ ϵend Lend can be readily evaluated using section tests and the coupon tests are plotted against f u =f y in
σ end F=ðA þ Astiff Þ Fig. 16. These will be used in conjunction with the results
ϵend ¼ ¼ ð17Þ
ESG ESG presented in Tables 11 and 12.
The fourth step compensates for the small elastic deformations of the
end plates. Using a procedure developed by [23] for stub columns, the 4.1. Full cross-section tensile tests: hot-rolled I-sections vs. hot-rolled
recorded L.V.D.T. displacements δmid (adjusted from δ using Eq. (17)) hollow sections
can be modified to obtain true axial displacement δmid,true, where
Comparing Esh =E values obtained for the hot-rolled I-sections with
δmid;true ¼ δmid  2Δend ð18Þ those of the hot-rolled hollow sections in Fig. 16, it is clear that the
The deformation of the end plates Δend can be calculated using Eq. (19) hot-rolled I-sections exhibit a higher degree of strain-hardening.
Furthermore, they also show a higher f u =f y ratio, which is in contrast

 to a similar comparison made by Wang [5] for equivalent coupon test
Lmid 1 1 data where the range of f u =f y differed for the two cross-section types.
Δend ¼ σ  ð19Þ
2 E0;LVDT ESG This discrepancy is most likely to be explained by the comparatively
126 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129

Table 11
Key cross-section tensile test results (hot-rolled I-sections).

Cross-section tensile A fy fu ϵsh ϵu ϵf Area reduction E Esh

test designation (mm2) (N/mm2) (N/mm2) (%) (%) (%) at fracture (%) (N/mm2) (N/mm2)

305  127  48 UB S1 6183 392 – 0.94 – – – 218,000 3100

305  127  48 UB S2 6183 392 – 0.92 – – – 214,900 3100
305  165  40 UB S1 5331 427 – 0.95 – – – 211,100 –
305  165  40 UB S2 5249 451 580 0.92 9.70 – – 217,300 3400
305  102  28 UB 3697 375 536 0.61 10.00 24.50 48.10 196,000 4000
254  102  28 UB 3599 392 544 0.66 9.50 26.20 41.70 202,700 3800
203  133  25 UB 3178 372 534 0.40 9.20 22.90 46.70 212,000 3900
152  152  23 UC 3006 373 511 0.44 8.50 28.90 44.60 205,200 3700
152  152  30 UC 3788 403 – 0.68 – – – 209,700 3400

Table 12
Key cross-section tensile test results (hot-rolled and cold-formed hollow sections).

Cross-section tensile A fy fu ϵsh ϵu ϵf E Esh

test designation (mm2) (N/mm2) (N/mm2) (%) (%) (%) (N/mm2) (N/mm2)

SHS 40  40  3 HR 453 471 535 3.46 10 45 208,200 0

SHS 40  40  3 CF 404 474 521 0.43 4.2 21.9 211,900 1600
SHS 40  40  4 HR 563 450 532 2.57 10.4 45.9 207,100 0
SHS 40  40  4 CF 525 450 467 0.43 1.2 26.8 193,700 2200
RHS 60  40  4 HR 708 443 525 2.98 11.5 45.2 206,600 0
RHS 60  40  4 CF 718 433 486 0.41 2.7 20 204,700 2300
SHS 60  60  3 HR 731 456 555 2.61 12.3 41.3 212,300 300
SHS 60  60  3 CF 611 375 418 0.39 4.6 38.5 197,800 1600

600 600

500 500
Stress σ (N/mm2)

Stress σ (N/mm2)

400 400

300 300

200 200
Hot rolled
100 100
Cold formed

0 0
0 5 10 15 20 0 5 10 15 20
Strain ε (%) Strain ε (%)
Fig. 15. Summary of full cross-section tensile stress–strain curves. (a) I-sections. (b) Hot-rolled and cold-formed hollow sections.

small sample size of the current experimental programme rather than 4.3. Full cross-section tensile test vs. tensile coupon test
any specific deterministic relationship.
4.3.1. Yield plateau
With reference to Tables 11 and 12 and Fig. 17, it can be
4.2. Full cross-section tensile tests: hot-rolled vs. cold-formed hollow observed that for hot-rolled specimens, reductions in the length of
sections the yield plateau by up to 43% can be attained, which supports the
assumption of the elastic, linear hardening material model cur-
For the cross-section tensile tests, the hot-rolled hollow sec- rently employed by the continuous strength method.
tions have higher f u =f y ratios than their cold-formed counterparts.
This is in line with the observations made in Wang [5] for tensile
coupon tests. As to be expected, cold-formed hollow sections 4.3.2. Strain-hardening modulus
exhibit a far higher degree of strain-hardening in the initial stage With reference to Tables 11 and 12 and Fig. 16 the most significant
than their hot-rolled equivalents, due to the rounding effect of the observation is that for I-sections, the ratio Esh =E as defined by the
cold-forming process on the shape of the stress–strain curve. tensile coupon tests is somewhat conservative for I-sections and
 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129 127

suggests the type of compromise between corner and flat material the three sets of full cross-section tensile tests are plotted. Based
behaviour expected of cold-formed hollow sections. upon earlier observations, the following revisions are proposed:

(i) For I-sections: enhancements to the degree of strain-

5. Revisions to the CSM material model for carbon steel hardening from Esh =E ¼ 0:01 to Esh =E ¼ 0:015 and a reduction
in the limiting value of f u =f y over which the maximum value
Fig. 18a is a graphical representation of the cross-section of Esh =E applies from f u =f y ¼ 1:7 to f u =f y ¼ 1:3.
dependent suite of material models based upon coupon test data (ii) For cold-formed hollow sections: the degree of strain-
as proposed by Wang [5], against which the data obtained from hardening shall remain unchanged from Wang [5], but the
range over which Esh =E ¼ 0:015 applies shall commence from
0.02 f u =f y ¼ 1:15 rather than f u =f y ¼ 1:25.
(iii) For hot-rolled hollow sections the strain-hardening modulus
proposed by Wang [5] shall be replaced with Esh =E ¼ 0.
0.015
A summary of the revised material models is provided in Table 13
and plotted in Fig. 18.
Esh/E

EN 1993-1-5 HR I cross section

0.01
HR I coupon
HR hollow cross section
HR hollow coupon 5.1. Associated improvements to the CSM
0.005 CF hollow cross section
CF hollow coupon
To assess the benefits of an updated material model to the CSM,
CF hollow coupon (corner)
the predictive capacity of the CSM design equations will be
0
1 1.1 1.2 1.3 1.4 1.5 compared with data obtained from a collection of laboratory tests
in the literature. Expressions for compression and bending resis-
fu/fy
tances (see [6] for details) shall be evaluated. For compression,
Fig. 16. Summary relationship between Esh =E and f u =f y for all full cross-section and stub-column tests data was obtained for hot-rolled and cold-
tensile coupon tests. formed hollow sections [17,25–30]; for bending resistance, test

1.6 1.6

1.4 1.4
305x102x28 UB
1.2 305x102x28 UB 1.2 full cross-section
coupon test
1 1
305x102x28 UB
f / fy
f / fy

0.8 0.8 coupon test

0.6 0.6
305x102x28 UB
0.4 full cross-section 0.4

0.2 0.2

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.005 0.01 0.015 0.02
δ/L δ/L

Fig. 17. Comparative responses of full cross-section and tensile coupon tests. (a) Non-dimensional full stress–strain response. (b) Relative length of yield plateaus and
properties of the yield point.

0.025 HR I-section model

HR hollow section model 0.025 HR I-section model
CF hollow section model
HR I-section CF hollow section model
0.02 HR hollow section 0.02
CF hollow section

0.015 0.015
Esh/E

Esh/E

0.01 0.01
EN 1993-1-5 (2006)
EN 1993-1-5 (2006)

0.005 0.005

0 0
1 1.2 1.4 1.6 1.8 2 1 1.2 1.4 1.6 1.8 2
fu/fy fu/fy

Fig. 18. Previous and proposed material models for the CSM. (a) Wang (2011), based on coupon tests. (b) Proposed, based on full cross-section tests.
128 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129

Table 13
Summary of revised material models for the CSM.

Hot-rolled I-sections
Esh =E ¼ 0:015
ðf u =f y  1:0Þ for f u =f y r 1:3
1:3  1:0
Esh =E ¼ 0:015 for f u =f y 4 1:3
Hot-rolled hollow sections
Esh =E ¼ 0
Cold-formed hollow sections
Esh =E ¼ 0:015
ðf u =f y  1:0Þ for f u =f y r 1:15
1:15  1:0
Esh =E ¼ 0:015 for f u =f y 4 1:15

Table 14 for Esh for each section type, derived from the values reported in
Comparison between previous and updated CSM predictions in compression (test/ Tables 11 and 12. For axially loaded members, there is an improve-
predicted). ment in the mean predictive capacity and C.O.V. for cold-formed
  hollow sections, but on average for all section types, the mean
Esh ¼ f f u Constant Esh
predictive capacity is marginally worse, albeit with an improved C.O.
Wang [5] Updated Wang [5] Updated V. In bending the most significant improvement in mean predictive
capacity is associated with the hot-rolled I-sections, although these
Hot-rolled hollow sections demonstrate a slight increase in C.O.V. values. On average, for all
Mean 1.067 1.082 1.009 1.082 section types, the mean predictive capacity remains unchanged. Such
C.O.V. 0.076 0.081 0.054 0.081
Cold-formed hollow sections
inconsistencies in the results at this stage are not entirely surprising
Mean 1.091 1.074 1.087 1.056 given the relatively small sample sizes of (i) the test results used to
C.O.V. 0.085 0.080 0.085 0.081 derive new Esh values and (ii) the highly variable sample sizes of the
All section types bending and axial test data, with considerably more relevant data
Mean 1.083 1.077 1.061 1.065
being available for cold-formed hollow sections than for other types
C.O.V. 0.084 0.080 0.087 0.080
of material or cross-section type.

6. Conclusions and design recommendations

Table 15
Comparison between previous and updated CSM predictions in bending (test/
predicted). A programme of seventeen full cross-section tensile tests on hot-
 
rolled I-sections, hollow sections and cold-formed hollow sections
Esh ¼ f f u Constant Esh has been conducted in order to investigate the influence of average
cross-section properties on the constitutive relationships for carbon
Wang [5] Updated Wang [5] Updated
steel. Using data obtained from the literature, as well as a supple-
Hot-rolled hollow sections mentary programme of 14 tensile coupon tests, the overall beha-
Mean 1.111 1.131 1.017 1.131 vioural response of the cross-section tensile tests demonstrated that
C.O.V. 0.026 0.026 0.025 0.026 assuming an elastic, linear hardening material model for the
Cold-formed hollow sections
continuous strength method is a reasonable assumption. Further-
Mean 1.180 1.151 1.101 1.106
C.O.V. 0.071 0.070 0.074 0.074 more, it was shown that the values obtained for the strain-
Hot-rolled I-sections hardening modulus based upon the tensile coupon tests are overly
Mean 1.072 1.045 1.090 1.045 conservative, except in the case of hot-rolled hollow sections, where
C.O.V. 0.021 0.026 0.024 0.026 for the range of validity, it is zero. Revising the suite of material
All section types
Mean 1.140 1.112 1.095 1.096
models originally proposed by Wang [5] in the light of these
C.O.V. 0.070 0.070 0.068 0.071 findings has been shown to have the potential to furnish the CSM
capacity equations with a higher degree of accuracy when com-
pared with experimental data. With additional full cross-section
data was obtained for hot-rolled I-sections, as well as hot-rolled tensile tests, there is also potential for a reduction in scatter in the
and cold-formed hollow sections [17,31–33]. data. These improvements apply to all cross-section types investi-
Tables 14 and 15 present comparisons between ultimate test gated, both in compression and in bending.
values and predicted capacity values using the CSM equations for
compression and bending. The first comparisons are between the
Acknowledgements
multi-linear material models reported in Tables 1 and 13. There is
a small improvement in the accuracy of the CSM when the
This project is funded by the Building Research Establishment
updated material model (Table 13) is used; this holds both within
Trust, whose sponsorship is gratefully acknowledged. The authors
groupings for each cross-section type and over the entire sample
would like to thank Julie Bregulla and David Brooke from the Building
for axially loaded specimens (Table 14), with the exception of hot-
Research Establishment, as well as Gordon Herbert from Imperial
rolled tubes; the coefficient of variation (C.O.V.) values for the
College London for their assistance in the experimental programme.
updated material models show very slight decreases in scatter. In
bending (Table 15), similar observations can be made, with all
sections (with the exception of hot-rolled tubes) showing References
improvements in mean predictive capacity.
The second comparison assumes a single value of Esh that is [1] Gardner L. The continuous strength method. Proc Inst Civil Eng—Struct Build
2008;161(3):127–33.
independent of f u =f y . Here the comparisons are between the [2] Horne MR, Medland IC. Collapse loads of steel frameworks allowing for the
previously assumed Esh ¼ E=100 by Wang [5] and the average values effect of strain-hardening, Proc Inst Civil Eng—Struct Build 1966;35(4):681–7.
 A.S.J. Foster et al. / Thin-Walled Structures 92 (2015) 115–129 129

[3] Davies JM. Strain hardening, local buckling and lateral-torsional buckling in [19] Lay MG. Yielding of uniformly loaded steel members. J Struct Div (ASCE) ST
plastic hinges. J Construct Steel Res 2006;62:27–34. 1965;6:49–66.
[4] Rasmussen K, Hancock G. Design of cold-formed stainless steel tubular [20] Krapf C. Stress relaxation of stainless steel rods for bridge pier cap shear
members. II: beams. J Struct Eng (ASCE) 1993;119:;2368–86. strengthening [Master's thesis]. Georgia Institute of Technology; 2010.
[5] Wang F. A deformation based approach to structural steel design [PhD thesis]. [21] Wang Y. Strain-hardening properties of structural steel [Master's thesis].
Imperial College of Science, Technology and Medicine; 2011. Imperial College of Science, Technology and Medicine; 2013.
[6] Gardner L, Wang F, Liew A. The influence of strain hardening on the behaviour [22] Krempl E, Nakamura T. The influence of the equilibrium stress growth law
and design of steel structures. Int J Struct Stab Dyn 2011;11(05):855–75. formulation on the modelling of recently observed relaxation behavior. JSME
[7] Gardner L. A new approach to structural stainless steel design [PhD thesis]. Int J, Ser A (Solid Mech Mater Eng) 1998;41(1):103–11.
Imperial College of Science, Technology and Medicine; 2002. [23] CASE. Compression tests of stainless steel tubular columns. Technical report.
[8] Gardner L, Nethercot D. Experiments on stainless steel hollow sections—Part Investigation Report S770, University of Sydney Centre for Advanced Struc-
1: material and cross-sectional behavior. J Construct Steel Res tural Engineering; 1990.
2004;60:1291–318. [24] Lecce M, Rasmussen K. Distortional buckling of cold-formed stainless steel
[9] Gardner L, Theofanous M. Discrete and continuous treatment of local buckling sections: experimental investigation. J Struct Eng (ASCE) 2006;132(4):497–
in stainless steel elements. J Construct Steel Res 2008;64(11):1207–16. 504.
[10] Afshan S, Gardner L. The continuous strength method for structural stainless [25] Gao L, Gun H, Jin F, Fan H. Load carrying capacity of high-strength steel box-
sections I: stub columns. J Construct Steel Res 2009;65:918–24.
steel design. Thin-Walled Struct 2013;68(0):42–9.
[26] Greiner R, Kettler M, Lechner A, Freytag B, Linder J, Jaspart JP, et al. Plastic
[11] Galambos TV. Guide to stability design criteria for metal structures. New York:
member capacity of semi-compact steel sections—a more economic design.
John Wiley and Sons; 1998.
Technical report. European Research Fund for Coal and Steel; 2008.
[12] Kato B. Deformation capacity of steel structures. J Construct Steel Res 1990;17
[27] Hu SD, Ye B, Li LX. Material properties of thick-wall cold-rolled welded tubes
(1–2):33–94.
with a rectangular or square hollow section. Construct Build Mater
[13] Bruneau M, Uang C, Whittaker A. Ductile design of steel structures. New York:
2011;25:2683–9.
McGraw-Hill; 1998. [28] Rasmussen KJR, Hancock GJ. Plate slenderness limits for high strength steel
[14] Byfield MP, Davies JM, Dhanalakshmi M. Calculation of the strain hardening
sections. J Construct Steel Res 1992;23:73–96.
behaviour of steel structures based on mill tests. J Construct Steel Res 2005;61 [29] Sakino K, Nakahara H, Morino S, Nishiyama I. Behaviour of centrally loaded
(2):133–50. concrete-filled steel tube short columns. J Struct Eng (ASCE) 2004;130(2):180–88.
[15] Alpsten G. Thermal residual stresses in hot-rolled steel members. Technical [30] Elchalakani M, Zhao XL, Grzebieta R. Tests on concrete filled double-skin (CHS
report 337.3. Lehigh University; 1968. outer and SHS inner) composite short columns under axial compression. Thin-
[16] Roderick JW. Behaviour of rolled-steel joists in the plastic range. Br Weld J Walled Struct 2002;40:415–41.
1954;1:261–75. [31] Wilkinson T, Hancock G. Tests to examine the compact web slenderness of
[17] Gardner L, Saari N, Wang F. Comparative experimental study of hot-rolled and cold-formed RHS. J Struct Eng (ASCE) 1998;124(10):1166–74.
cold-formed rectangular hollow sections. Thin-Walled Struct 2010;60 [32] Zhao X, Hancock G. Tests to determine plate slenderness limits for cold-
(7):495–507. formed rectangular hollow sections of grade C450. J Austral Steel Inst 1991;25
[18] Jeong CY, Nam SW, Ginsztler J. Activation processes of stress relaxation during (4):2–16.
hold time in 1Cr–Mo–V steel. Mater Sci Eng A (Struct Mater: Prop Microstruct [33] Byfield MP, Nethercot DA. An analysis of the true bending strength of steel
Process) A 1999;264(12):188–93. beams. Proc Inst Civil Eng—Struct Build 1998;128(2):188–97.