Professional Documents
Culture Documents
ABSTRACT
Heat curving is commonly used in the fabrication of curved structural steel bridge
used for numerical modeling in the development of the AASHTO code provisions. This
iterative analysis can take into consideration multiple heat /cool cycles, initial residual
This paper describes a simplified analysis based on the Duhamel Analogy that can
be carried out using “hand calculations”. Results obtained from this method are within
element solution. The background, basis and steps required for the proposed analysis are
described and an illustrative numerical example presented. The proposed analysis may be
heating width for a single heat/cool cycle for steels such as high performance steel (HPS)
Key Words: Duhamel analogy, bridges, heat curving, HPS, thermal analysis, curvature,
stress, simplified
1
Assistant Professor, University of Balamand, Lebanon, formerly graduate student, Department of Civil
and Environmental Engineering, University of South Florida, Tampa, FL 33620
2
Samuel and Julia Flom Professor, Department of Civil and Environmental Engineering, University of
South Florida, Tampa, FL 33620
1
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
INTRODUCTION
Heat curving is widely used for fabricating curved, bridge steel girders. In this
process, flange tips along one side of the girder are subjected to repeated heat/cool
(natural cooling to ambient temperatures) cycles to achieve the desired curvature. The
width of the heated flange (ha) can vary and three types (Fig. 1) are recognized in
2c
ha Heating Cooling
Longitudinal
axis
(a) I-Girder Prior to Heating (b) Convex Curvature (c) Concave (Reverse) Curvature
(Duhamel Analogy steps 1–5) (Duhamel Analogy steps 6–7)
Heat curving is subject to the AASHTO provisions (AASHTO 1993, 1996) that
are based on full-scale test results (Brockenbrough 1970). In the testing, six distinct
heat/cool cycles (see Table 1) were used to obtain a terminal radius of curvature of 121 m
The complexity of the heat curving operation is reflected in its numerical analysis
2
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
heating) and elastic (during cooling) response of the heat curved girder must be taken into
consideration.
1966) was used to model the US Steel test girder (Brockenbrough 1970a). It was
subsequently used to develop heat curving design charts (Brockenbrough 1972, 1973) for
steel fabricators to determine the heating regime, i.e. the heating temperatures needed to
obtain desired curvatures under a specific heat /cool cycle. More recently, three-
dimensional finite element analyses were carried out to analyze heat curving (Gergess
2001). In the course of this study, a simplified analysis was developed to check the
intermediate results for a specific heat/cool cycle This analysis is suitable for “hand
This paper describes the basis and limitations of the proposed simplified analysis.
numerical example presented in a step-by-step format. The accuracy of the results from
the simplified analysis is assessed by comparisons with solutions from the Duhamel
3
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
DUHAMEL ANALOGY
analysis that can predict residual curvature in a heat curved girder. A brief review is
presented to provide the necessary background for the proposed simplified analysis. More
The basic steps in the Duhamel Analogy are illustrated in Fig. 2. In essence, a
heat curved girder is assumed to be restrained under thermal loading and the final stress
stresses, e.g. σ = F/A + My/I due to heating and cooling. The corresponding strains are
variation over the width of the flange. This variation is non-uniform since only one side
of the flange is heated (Fig. 1). Brockenbrough obtained the temperature distribution
from heat flow analysis in a semi-infinite thin plate for a moving-point heat source along
its edge (Myers & Vyehars, 1967). This led to a distribution (hx) that was constant over
the heated width and transitioned linearly over a width (c/6) to ambient temperature (T0)
The steps required for analyzing a simply supported girder subjected to a single
heat /cool cycle are first outlined followed by the more general case involving multiple
heat /cool cycles (Table 1). Temperature distribution across the heated width (Fig. 2) (not
the flange width) is assumed to be uniform for illustrative purposes. The effect of heating
4
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
Step 1: Parameters
x ∆T= (T – T0) = n∆Ti
L
T
2c ha
NA
tf
Simply Supported Girder Flange Section Temperature Profile (rectangular
shape shown for illustration)
Step 2:
Thermal Stress c σt ha
σt = Eα∆Ti where E & α are calculated at Tn = T0 + Σn∆Ti
c
Step 3: F
End Reactions (applied in opposite direction) c ha
e
F F x
M M c
c c
-F/A
x
c c -Mx/I
Mc/I
Axial Stress Bending Stress
Step 5A:Self Equilibrating Stresses - Elastic Step 5B: Inelastic
Fy
σt - F/A - Mc/I
c c x2
- F/A + Mc/I
σ2
x1, x2, σ1, σ2 determined from ∫σtdA & ∫σtxdA = 0
∆Ti)
Step 6: Cooling - Steps 1 – 5A repeated for (-∆
σt = -Eα∆Ti, E & α at Tn = T - Σn∆Ti Mc/I -σt + F/A + Mc/I
5
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
Single Cycle : Heating
sub-dividing the heating temperature increase (∆T) into n increments (∆Ti, ∆T = n∆Ti).
As a result, a multi-step analysis (steps 2-5 are repeated n times) is required. Fig. 2 and
A = 4tfc + (d – 2tf)tw
Step 3 Calculate end reactions F (axial force due to longitudinal constraint) and
F= ∫ σ t dA and M =
(over h a )
∫ σ t xdA , x is the offset from the neutral axis.
(over h a )
Step 5A Elastic:
step 2) and negative end reaction stresses (from step 3), Fig. 2, step 5A.
6
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
Table 2. Duhamel Analogy versus Proposed Simplified Method
MULTI-STEP ANALYSIS FOR SINGLE-STEP ANALYSIS FOR Single step analysis does not
incorporate residual stresses
EACH TEMPERATURE INCREMENT (∆Ti) ONE TEMPERATURE INCREMENT
ANALYSIS but provides curvature values
(∆Tmax) within 15% of results from
sophisticated analyses.
7
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
Step 5B Inelastic:
∫ σ edA and
(over 2c )
∫ σ exdA = 0 . Refer to Fig. 2, step 5B.
(over 2c )
At the end of this step, the girder develops a concave curvature (Fig. 1b) due to the temperature
increase (∆T). Self-equilibrating stresses (Fig. 2, Step 5) constitute the initial conditions for the
cooling cycle.
∆Tmax
∆T
hx
0.1∆T
0.03∆T c
c c/6 0 -c 0 -c
h’a=2ha
tf tf
2c 2c
ha 2ha
d d
x x
tw tw
tf tf
8
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
Single Cycle: Cooling
The results from step 5B for the last heating increment, i.e. total temperature
increase ∆T, constitute the initial conditions for the 1st cooling decrement.
However, the cooling operation is elastic since reductions in stresses and strains
during unloading are reversible (Smith 1996). As a result, only steps 2 – 5A (not
5B) as defined for heating are repeated but for a temperature decrease (-∆T = T0 –
T). Temperature-dependent material properties (Fy, E and α) are calculated for the
Step 7 After cooling to ambient temperature (T0, last decrement in step 6), there will be
permanent residual stresses, strains, and curvatures that define the girder
conditions after one heat/cool cycle. The residual curvature is of primary interest
Summary
During heating, steps 1 – 5 are repeated for each successive temperature increment ∆Ti,
with the results from step 5 for a specific increment constituting the initial conditions for the
following increment. For example, for a girder heated to a temperature T of 299C (570F) and
ambient temperature 21C (70F), the temperature increase (∆T = 299C – 21C = 278C (570F –70F
= 500F)) may be subdivided into say five intervals (∆Ti = 278/5 = 55.6C (100F)) and thermal
analysis (steps 1 – 5) carried out for each increment. Results from the fifth increment (last
interval), constitute the initial conditions for cooling (steps 6 – 7 performed for successive
9
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
cooling temperatures (-∆Ti = -55.6C (-100F)) until ambient temperature (T0 = 21C (70F) is
reached).
For multiple heat/cool cycles (Table 1), steps 1 - 7 are repeated for each heat/cool cycle
separately, with the results from step 7 for a specific cycle constituting the initial conditions for
the following cycle. If for example the girder from the previous paragraph (heated to 299C
(570F)) was cooled to ambient conditions, then heated to 354C (670F), the temperature increase
in the 2nd heating cycle (∆T = 354C – 21C = 333C (670F – 70F = 600F)) would be subdivided
into six intervals (∆Ti = 55.6C (100F)) and thermal analysis (steps 1 – 7) performed with the
results from the final step 7 (after cooling, previous paragraph) being the initial value. Such an
results obtained from the US Steel study. However, was this not the case for fabrication aids
(Brockenbrough 1972, 1973) where only one heat/cool cycle was considered.
It should be noted that step 5B requires a trial and error numerical solution due to the
SIMPLIFIED ANALYSIS
for the dependence of material properties on temperature. In the proposed analysis, the essential
assumptions were made so that the analysis can be completed in a single step without unduly
compromising the accuracy of the numerical solution. These assumptions are listed below and
10
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
1. Replace Brockenbrough’s non-uniform temperature distribution (Fig. 3a) by an
The idealized triangular distribution uses a heated width h’a = 2ha and a maximum
for details). The heated width (2ha) is kept the same for all three heat types (Fig. 3b) but the
and strains for the total temperature variation (∆T = T – T0) thereby eliminating the need for a
multi-step solution.
CLOSED-FORM EQUATIONS
The steps outlined for the Duhamel Analogy (Fig. 2) are also used in the simplified
analysis. Closed-form equations required for calculation of thermal stresses using the proposed
triangular temperature distribution are presented. Average temperatures for the three different
∆T= T-T0)
Table 3. Average Temperature for Single Step Analysis - (∆
11
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
End Reactions
End reaction forces and moments are calculated by integrating the thermal stresses that
develop for the assumed triangular heat profile, as given by Eqs. (1) and (2):
(∆Tmax (2h a − c + x ) )
σt = Eα (see Fig. 4 for definition of x).
2h a
2h a
M = 2αh a t f E∆Tmax (c − ) (2)
3
Yield initiates along the heated girder flange section (point of maximum stress at the tip
of the flange width, x = c), for temperature increase (∆Ty) calculated by equating Eq. 4 to (Fy) at
(x = c):
Fy
∆Ty = (5)
2h
c c - a
3
Eα − 2αEh a t f 1 +
A I c.g
The stress varies linearly with the heated zone (tension (+) at the heated flange tip (x = c),
compression (-) at the tip of the heated area (x = c – 2ha)) as shown in Fig. 4a.
The point of zero stress (x0) in the heated area is calculated by setting the self-
12
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
- c + 2h a 2h a t f
−
2h a A
x0 = (6)
2h a
c - 2h a t f
3 1
−
I cg 2h a
The distances x1, x2, x3 shown in Fig. 5a define the shape of the inelastic region and are
c c
determined to satisfy equilibrium ( ∫ σ ex dA = 0, ∫ σ ex xdA = 0, ). From these equations, the
-c -c
γ = 3+
[ ]
β 3 − 3β - α 4 (α 3 + α 5 ) − a α 3 (6 − 3α 3 ) + α 5 (3 − 3α 3 − α 5 )
≤ 2 − α4 (7)
4 − (1 + a )(α 3 + α 4 )
γ2
α2 = (8)
4 − (1 + a )(α 3 + α 4 )
1 εx
= (9)
R x - x NA
where εx is the strain (σx/E) at offset (x), and (xNA), is the offset of the centroid of the flange to
Strain variations in the elastic and inelastic ranges are shown in Figs. 4b and 5b
respectively:
13
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
( F + Fe x)
A I c.g
εx = Elastic Analysis (10)
E
F' y x + c − x1
εx = Inelastic Analysis (11)
E' x2
(1) During cooling, elastic analysis is performed for (-∆Tmax), e.g. Eqs. 1-4, 9 & 10 above are
(2) Residual stresses (σrx, Fig.6a) after unloading are listed in Appendix II. Residual
2h a
F' y x + c − x1 1 x (c - )
ε rx = − 2α∆Tmax h a t f + 3 (12)
E' x2 A I c.g
2h
x(c - a )
1 2α∆Tmax h a t f 1 3
=- + (13)
Rc (x - x NA ) A I c.g
where Rc is the radius of curvature calculated at the Neutral Axis (NA) and XNA is the distance to
I c.g
the Neutral Axis, x NA = - .
2h a
A(c - )
3
1 1 1
= + (14)
Rr R Rc
14
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
Axial Force: F
Bending Moment: M=Fe
Thermal Stress: σt
-F/A - Mc/Icg + σt
Heated Width
(F/A + Mc/Icg)/E c
c
- Compression
2ha
+ Tension
NA NA
-F/A + Mc/Icg
-c -c (F/A - Mc/I )/E
cg
σex)
Self-Equilibrating Stress (σ Strain
(a) Stress Distribution due to Heating (b) Strain distribution due to Heating
Axial Force: F
Bending Moment: M=Fe
Thermal Stress: σt
- Compression
Heated Width
Heated Width
c Fy + Tension c
2ha
2ha
x3
-F’y Flange Width (2c)
Flange Width (2c)
xo
x F’y/E x
0 0 x2
x2 CG CG
(1/R) = εy/x2
xNA
NA NA
x1 x1
F’y(x1/x2)
-c -c -F’ (x /x )/E
y 1 2
σex)
Self-Equilibrating Stress (σ Strain
(a) Stress distribution due to Heating (b) Strain distribution due to Heating
15
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
Axial Force: F
Bending Moment: M=Fe
Thermal Stress: σt
- Compression
F/A + Mc/Icg - σt + Tension
Heated Width
Heated Width
c c -(F/A + Mc/Icg)/E
2ha
2ha
NA NA
xNA
(a) Stress distribution due to Heating (b) Strain distribution due to Heating
Fig. 6. Residual Stress and Strain Distribution –Inelastic Range (Steps 6 –7)
APPLICATION
Heat curving induces lateral deformations and curvatures. In the proposed analysis, a heat
curved girder subjected to a simplified triangular temperature profile (Fig. 3b) develops end
For elastic conditions, (∆Tmax ≤ ∆Ty, ∆Ty being the temperature at which yielding
initiates, Eq. 5), the resulting self-equilibrating stresses in the heated and non-heated zone are
determined from Eqs. 3 and 4 respectively. The corresponding curvature may be obtained from
Eq. 9. For inelastic conditions (∆Tmax > ∆Ty), stresses during heating inside and outside the
heated area are calculated using coefficients listed in Appendix II that are functions of the plastic
zone depth. The corresponding strains and curvatures are given by Eqs. 9 & 11.
16
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
During cooling, the curvature direction is reversed and elastic stress-strain-curvature
relationships are used in the analysis. Residual stresses may be calculated using functions listed
in Appendix II. The curvature during cooling (1/Rc) is obtained from Eq.13 and corresponding
NUMERICAL EXAMPLE
The numerical example selected to illustrate the simplified method is taken from
experimental data obtained from US Steel’s pioneering full-scale test conducted in 1968. As this
example was also analyzed by Brockenbrough (1970a) using the Duhamel Analogy and by
Gergess (2001) finite elements, its selection is convenient from the standpoint of assessing the
Description
The symmetrical Grade 250 (Fy = 36 ksi) heat-curved I-girder was 14.02 m (46 ft) long
with 61 cm × 5.1 cm (24 in. × 2 in.) flanges and a 116.8 cm × 1.27 cm (46 in. × ½ in.) web. At its
ends, the bottom flanges were placed on mobile platforms that permitted longitudinal and lateral
movement. At the middle, the web was bolted to a rigid platform (Fig. 7). Curving was achieved
Of the six heat/cool cycles, the 3rd and 6th cycles (referred to subsequently as runs 3 and
6) were of particular importance since: (1) run 3 had the largest heating temperature (544C
(1011F)) over a heated width of 8.9 cm (31/2 in.) and (2) run 6 had the largest heating width of
13.3 cm (51/4 in.) for a heating temperature of 409C (768F)). For an ambient temperature of 21C
(70F), temperature increases were 523C (941F) and 388C (698F) respectively. The measured
radius of curvature was 200 m (655 ft) after run 3, i.e. following runs 1 and 2 and 121 m (397 ft)
17
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
after run 6, i.e. following runs 3, 4 and 5. A summary of the results from the intermediate steps is
presented in Table 4.
tf = 5.1 cm L = 14.02 m
d = 116.8 cm
2c = 61 cm
End Support Middle Support
Mobile Platform Rigid Platform
CROSS-SECTION ELEVATION
Solution: Steps 1–7 are followed to calculate the residual curvature for run 3 (run 6 results are
presented in Table 4). Note that the measured heated width (Table 1) did not accurately reflect
the 1/6 flange width for run 3 (Type II heat) and ¼ flange width for run 6 (Type III heat).
Brockenbrough 1970b, used actual values in the theoretical analysis e.g. 10.16 cm (4 in.) instead
of 8.9 cm (31/2 in.) for run 3 and 15.24 cm (6 in.) instead of 13.3 cm (51/4 in.) for run 6
Step 1: Ic.g = 191,800cm4 (4608in4), A = 770.54cm2 (120in2), ∆Tmax = 1.09×(523C) = 570C (1026F)
Step 2: At T0: F’y=250MPa (36 ksi), E’=200GPa (29,000 ksi), α’ = 1.1E-5/C (6.29E-6/F)
At Tave = 2/3∆Tmax + T0 = 2/3(570C) + 21C = 401C (754F), Fy=218MPa (31.7 ksi), E=176GPa (25,550 ksi), α=1.39E-
18
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
Step 3:
Step 5A (elastic):
13,970 331,030
σex = − − x = - (18.1 + 1.73x)kN/cm 2 , x (cm), (-(26.2 + 6.36x)ksi, x(in.))
770.54 191,800
σex = - (18.1 + 1.73x) + (1.39E − 5 × 176E 2(29.1x − 295.5 − 21)) = (5.39x - 95.5) kN/cm 2 , x (cm), ((19.8x − 138.5)ksi, x (in.))
218,000
∆Ty =
2(10.16)
30.48 30.48 -
3
1.39E − 5 × 176 E6 − 2 × 1.39E − 5 × 176E 6 × 10.16 × 5.08 1 +
770.54 191,800
∆Ty = 187C (348F) < ∆Tmax = 570C (1026F), inelastic analysis required
Eq.7: γ = 3+
[
β 3 − 3β - α 4 (α 3 + α 5 ) − a α 3 (6 − 3α 3 ) + α 5 (3 − 3α 3 − α 5 ) ] Eq.8: α = γ2
2
4 − (1 + a )(α 3 + α 4 ) 4 − (1 + a )(α 3 + α 4 )
19
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
From Appendix II:
α3 = (1 – 0.57)(1+0.88) – (0.87)(0.67) = 0.22, α5 = (0.67 + 0.57 – 1)0.88 =0.21. β= 0.21 + 0.67 + 0.22 = 1.1
γ = 1.42 > (2 - α4) = 1.33, then use 1.33. α = 1.33 2 = 0.75, α1 = 1.33 – 0.75 = 0.58
2
4 − (1 + 0.88)(0.22 + 0.67)
@ x = (x1 + x2 –c) = (17.7 + 22.9 - 30.48) = 10.12 cm (3.98 in.), εx2 = 250,000/200,000,000 = 0.00124
2h
x (c - a )
Steps 6 and 7: Eq.13: 1 = 2α∆Tmax h a t f 1 3
R c - (x - x ) A + I
NA c.g
2(10.16)
10.12(30.48 - )
(1/Rc) = 2(0.0000139)(570)(10.16)(5.08) 1 3 = -0.0000973
- +
(10.12 - (-10.5)) 770.54 191,800
Eq.14: 1 = 1 + 1
Rr R Rc
1 1 1 , Rr = -224.8m (-742 ft), ∆ = -10.9cm (-4.3 in.) compared to Rr = -200m (-655 ft), ∆ = -
= +
Rr 184.6 - 101.4
20
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
Table 4. Proposed Method, Steps 1 – 7 (Numerical Example)
Fy T − 100
= 1 − ave 100F < Tave < 800F
F' Fy = 218MPa (31.7 ksi) Fy = 228MPa (33.2 ksi)
2 y 5833
cont’d E=176GPa (25,550 ksi) E=180GPa (26,354 ksi)
( 2
)
= − 7.2E5 + 4200Tave − 2.75Tave 10 −6 800F<Tave<1200F
α=1.39E-5/C(7.49E-6/F) α=1.36E-5/C(7.16E-6/F)
2
cont’d Temperature (Tx) = (28.1x – 264), x (cm) (13.9x + 21), x (cm)
((128.3x – 443)), x (in.) ((63.7x + 70)), x (in.)
Thermal Stress: σt = Eα∆Tx (68,744x – 645,850), (34,027x + 51,408)
21
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
3 Equivalent forces: F = 2 α h a t f E ∆ Tmax (Eq.1) 13,970 kN (3140 kips) 15,390 kN (3460 kips
Fy
∆Ty = (Eq.5)
2h a
c c - 3
5B
in- 187C (348F) < ∆Tmax = 312C (590F) < ∆Tmax =
Eα − 2 αEh a t f +
1
elastic 570C (1026F) 425C (764F)
A I c.g
- c + 2 h a 2h a t f
− (Eq.6)
2h a A
x0 =
5B 2h a
cont’d c - 2h a t f x0 = 17.4 cm (6.85 in.) x0 = 12.3 cm (4.84 in.)
3 1
−
I cg 2h a
γ = 3+
[
β 3 − 3β - α 4 (α 3 + α 5 ) − a α 3 (6 − 3α 3 ) + α 5 (3 − 3α 3 − α 5 ) ]
4 − (1 + a )(α 3 + α 4 )
γ = 1.42>1.33 use 1.33 γ = 1.123>1.0 use 1.0
= α0 , 1 = α 1
5B (Eq.7) x0 x
cont’d c c α0 = 0.57, α4 = 0.67 α0 = 0.4, α4 = 1.0
= α2 3 = α3, , x5 ,
x2 x 2h Fy a = 0.88, α3 = 0.22 a = 0.92, α3 = 0.225
a
= α4 = α5 = a
c c c c F' y
α5 = 0.21, β = 1.1 α5 = 0.372, β = 1.6
α 3 + α 4 + α 5 = β , α 3 = (1 − α 0 )(1 + a) − aα 4
5B γ2 (Eq.8)
α2 = 0.75 α2 =0.61
α2 =
cont’d 4 − (1 + a )(α 3 + α 4 ) x2 = 22.9cm (9 in.) x2 = 18.6 cm (7.32 in.)
5B α1 + α 2 = γ α1 = 0.58 α1 = 0.39
cont’d x1 = 17.7cm (7 in.) x1 = 11.9 cm (4.68 in.)
5B 1 ε x Eq.9, F' y x + c − x 1 (Eq.11)
= εx=
cont’d R x - xNA E' x2
R = 184.6 m (606 ft) R = 149 m (489 ft)
2h a
x (c - ) (Eq.13)
6 1 2 α ∆ T max h a t f 1 3 Rc = -101.4 m (-333 ft) Rc = -112 m (-366 ft)
=- +
Rc (x - x NA ) A I c.g
7 1
=
1
+
1 (Eq.14) Rr = -224.8 m (-742 ft) Rr = -450 m (-1446 ft)
Rr R Rc
22
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
DISCUSSION
The numerical example illustrated the applicability of the proposed simplified method for
heat curving. Stresses, strains and curvatures during heating and cooling were calculated for the
US Steel test girder corresponding to specific heat /cool cycles (see Table 1) referred to as run 3
(Type II heat, ha = 10.16 cm (4 in.)) and run 6 (Type III heat, ha = 15.24 cm (6 in.)).
Table 5 compares results for residual curvature obtained from the simplified analysis with
those available results using the Duhamel Analogy and three-dimensional finite element analysis.
For Type II heat (see Fig. 3) the simplified method provides curvature values that are within 13%
of the measured values. The agreement is similar for Type III heat. In both cases, results are in
Heat curving is an art and regardless of the method of analysis, a final corrective heating
operation is required during fabrication to bring the girder to the desired curvature. Thus, the
accuracy of the numerical solution is not of paramount importance. Under the circumstances, the
23
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
CONCLUSIONS
This paper presents a single-step analytical solution for heat curving based on the two-
Fig. 3 and Table 2). These simplifying assumptions allowed the development of closed-form
equations for thermal stresses and strains from which curvature can be directly calculated.
example that utilizes actual data from US Steel’s full-scale test conducted in 1968. Results show
good agreement and indicate that the proposed method may be conveniently used to predict
curvatures for specified conditions. Alternatively, the analysis may be used to determine the
maximum temperature required for heat curving an unstiffened symmetrical steel plate girder
section of various grades, including newly developed high performance steel (HPS, AASHTO
2000).
Although the closed form equations are complex, they can be readily incorporated in a
spreadsheet format with the main variables being the yield stress (Fy, function of the steel grade),
heating temperature (T), heated width (ha), and girder properties (flange thickness and width).
Such manual analyses serve to de-mystify heat curving and may be considered as a catalyst for
ACKNOWLEDGEMENTS
This study is part of a doctoral dissertation that was funded by a Flom Fellowship,
Department of Civil and Environmental Engineering, University of South Florida, Tampa. The
24
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
SYMBOLS
a ratio of Fy to F’y
A total cross-sectional area of I-girder, cm2 (in2)
Ah heated area of the girder, cm2 (in2)
c half flange-width of I-girder, cm (in.)
d web depth of I-girder, cm (in.)
e eccentricity, cm (in.)
E’ modulus of elasticity at ambient temperature (T0), 200 GPa (29,000 ksi)
E modulus of elasticity at heating temperature (T)
F equivalent prestress force, kN (Kips)
Fy yield stress at temperature T, MPa (ksi)
F’y yield stress at ambient temperature, MPa (ksi)
ha heated width in Brockenbrough’s analysis cm (in.)
h’a heated width (=2ha) in proposed simplified analysis cm (in.)
Icg moment of inertia of flange sections, cm4 (in4)
L total length of I-girder, m (ft)
M equivalent moment (Fe), KN-m (kips-ft)
R radius of curvature during heating, meter (ft)
Rc radius of curvature during cooling, meter (ft)
Rr residual radius of curvature after cooling, meter (ft)
T heating temperature, C (F)
Tave average temperature from proposed profile (Tmax), C (F)
Tmax equivalent temperature for proposed profile, C (F)
Tx heating temperature at any point along flange width, C (F)
Ty heating temperature at which yielding occurs in un-heated area, C (F)
T0 ambient temperature 21C (70F)
tf flange thickness of I-girder, cm (in.)
tw web thickness of I-girder, cm (in.)
25
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
x offset from flange neutral axis, cm (in.)
xo distance from centerline to point of zero stress in heated area, cm (in.)
x1 depth of compressive stress elastic core in un-heated area, cm (in.)
x2 depth of tensile stress elastic core in un-heated area, cm (in.)
x3 depth of compressive stress plastic core in heated area, cm (in.)
xy distance from centerline to point of first yield in heated area, cm (in.)
α coefficient of thermal expansion at heating temperature , /C (/F)
∆ lateral offset (L2/8R), mm (in.)
∆T difference between heating temperature and ambient temperature, C (F)
∆Tmax difference between proposed temperature and ambient temperature, C (F)
∆Ty difference between yield temperature and ambient temperature, C (F)
εrx residual strain after cooling, mm/mm (in./in.)
εx strain during heating, cm/cm (in./in.)
εy yield strain, cm/cm (in./in.)
σ axial + flexural stress, kN/m2 (ksi)
σt thermal stress, MPa (ksi)
σe self-equilibrating stress, MPa (ksi)
26
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
APPENDIX I. EQUIVALENT TEMPERATURE PROFILE
∆T ∆Tmax
h*a = ha/2 for Type I Heat ΣA = ∆Tmax h’a/2
= 3ha/4 for Type II Heat
ΣAx = ∆Tmax (h’a/2)(c – h’a/3)
= 5ha/6 for Type III Heat
1 2
0.1∆T 4
0.03∆T 3 5 6
c -c c -c
0 0
h*a h’a
c/6
x x
Σ A;Σ
Σ Ax 0.188∆Tc; 0.167∆Tc2 0.355∆Tc; 0.288∆Tc2 0.522∆Tc; 0.382∆Tc2
*∆Tmax based on:
*two ∆Tmax values are calculated: (1) based on Σ A = ∆1Tmax h’a/2, (2) based on Σ Ax = ∆2Tmax
(h’a/2)(c – h’a/3). Since h’a is selected as 2ha, the average between the two values is used.
27
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
APPENDIX II. COEFFICIENTS FOR HEAT CURVING
x0 x1 x2 x3 2h a x5
= α0 = α1 = α2 = α3 = α4 = α5
c c c c c c
Fy
=a α1 + α 2 = γ α3 + α4 + α5 = β α 3 = (1 − α 0 )(1 + a) − aα 4
F' y
α 5 = (α 4 + α 0 − 1)a
( x - x1 + c) –c ≤ x ≤ (x1+x2-c)
σ x = -F'y
x2
2h
x (c - a )
σ rx = Fy + 2α∆T Eh t 1 + 3 - α∆Tmax E ( x-c+2h a ) (c-x3) < x ≤ c
max a f 2h a
A I c.g
2h
( x + 2h a − c) x (c - a ) ( x-c+ 2h a ) (c-2ha) < x ≤ (c-x3)
σ rx = ( Fy + F' y ) − F' y + 2α∆T Eh t 1 + 3
− α∆Tmax E
2h a
(2h a − x3 ) max a f
A I c.g
2h
x (c - a )
σ r x = -F'y + 2α∆T Eh t 1 + 3 (x1+x2–c)<x≤(c-2ha)
max a f
A I c.g
2h a
( x - x1 + c) 1 x (c - 3 ) –c ≤ x ≤ (x1+x2-c)
σ rx = -F' y + 2α∆Tmax Eh a t f +
A I c.g
x2
28
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.
APPENDIX III. REFERENCES
AASHTO (1999). “Guide for Highway Bridge Fabrication With HPS 70W Steel” AASHTO
Subcommittee on Bridges and Structures, San Diego, CA, May 20.
AASHTO (1996). “Standard Specifications for Highway Bridges”, 16th Edition, Washington,
DC.
AASHTO (1993). “Guide Specifications for Horizontally Curved Highway Bridges, 1980:
As Revised by Interim Specifications for Bridges 1981, 1982, 1984, 1985, 1986 and
1993”. Washington, DC.
Brockenbrough, R.L. (1970a). “Theoretical Stresses and Strains from Heat-Curving”. ASCE,
Journal of Structural Division, July, vol. 96, no. ST7, pp. 1421-1444.
Brockenbrough, R.L. (1973). “Fabrication Aids for Girders Curved with V-Heats”. United
States Steel Corporation, Pittsburgh, PA, January.
Brockenbrough R.L., Merrit F.S. (1999). “Structural Steel Designer’s Handbook”, 3rd
Edition, McGraw-Hill, NY, NY.
England G.L. and Tsang C.M. (1996). “Thermally Induced Problems in Civil Engineering
Structures”, Elsevier Science B.V., NY, NY.
Gergess A. (2001). “Cold Bending and Heat Curving of Structural Steel I-Girders”, Ph.D
Dissertation, Department of Civil and Environmental Engineering, University of South
Florida, Tampa, FL, August 2001.
Manson, S.S. (1966). “Thermal Stress and Low Cycle Fatigue”, McGraw-Hill Book Co.,
New York, N.Y.
Myers P.S., Vyehars O.A., and Borman G.L. (1967). “Fundamentals of Heat Flow in
Welding”, Welding Research Council Bulletin 123, July 1967.
Smith J.C. (1996). “Structural Steel Design”. LRFD Approach, 2nd Edition, John Wiley &
Sons, NY, NY.
29
TRB 2003 Annual Meeting CD-ROM Paper revised from original submittal.