Strut-and-Tie Models Using Mul
Strut-and-Tie Models Using Mul
Strut-and-Tie Models Using Mul
The development of strut-and-tie models (STMs) for the design C is independent of the structural layout. As a result, mini-
of reinforced concrete (RC) deep beams considering a general mizing the load path Z for a given design problem is equiv-
multi-material and multi-volume topology optimization framework alent to minimizing the volume V if the structure is fully
is presented. The general framework provides flexibility to control stressed.
the location/inclination/length scale of the ties according to prac-
A pioneering work by Kumar (1978) applies the load path
tical design requirements. Optimality conditions are applied to
theory of truss frameworks to design reinforced concrete
evaluate the performance of the optimized STM layouts. Specifi-
cally, the Michell number Z (or load path) is used as a simple and (RC) deep beams by navigating optimal load transmission.
effective criterion to quantify the STMs. The experimental results Following and building upon Kumar’s study, this work
confirm that the layout with the lowest load path Z achieves the extends Michell’s optimality conditions to understand the
highest ultimate load. Moreover, significantly reduced cracking optimal load path for STMs and uses the load path Z (or
is observed in the optimized layouts compared to the traditional the Michell number) as a criterion to quantify the efficiency
layout. This observation implies that the optimized layouts may of the STM. Compared to existing criteria (Schlaich et al.
require less crack-control reinforcement, which would lower the 1987; Xia et al. 2020; He et al. 2020), the present criterion
total volume of steel required for the deep beams. is simpler. The experimental results in the “Load-deflection
curves” section in this paper verify that the STM layout with
Keywords: load path; Michell number; multi-material topology optimiza-
tion; reinforced concrete (RC) deep beam; strut and tie. the lowest load path Z (or Michell number) achieves the
highest ultimate load.
INTRODUCTION The STMs are powerful tools for analyzing and designing
In 1904, Michell wrote the revolutionary paper “The RC structures. However, traditional STMs dramatically
Limits of Economy of Material in Frame-Structures,” which simplify the complex stress state found in deep concrete
is a landmark in the field of optimization in general and elements in compression, which greatly limits their effi-
topology optimization in particular. He derived the well- ciency in many practical design applications. More recently,
known Michell’s optimality conditions (Michell 1904), that topology optimization has been used to automatically
provide analytical ways to find optimal truss structures. The generate STMs, including the works of Liang et al. (2000,
definition of the optimal structure is the least-weight truss 2001), Leu et al. (2006), Bruggi (2010), Mozaffari et al.
with given allowable stresses, which is also known as the (2020), and Zhou and Wan (2021), which is just a small
minimal total load path theory. The load path has been quan- sample of references in the field. The optimized STM layouts
tified using the Michell number, Z, defined as follows provide deeper insight into the load paths in RC members and,
ultimately, aid in more efficient structural designs. However,
Z = ∑e|Fe|Le = ∑ T|Fe|Le+ ∑ C|Fe|Le
(1) most topology optimization formulations for STMs use
e∈G e∈G only a single material, assuming the struts and ties have the
same linear behavior. Victoria et al. (2011) extend the single
where Le and Fe denote the length and internal axial force of material optimization using a bilinear material model with
the e-th truss member in the structure, respectively; and GT different behaviors in compression and tension to represent
and GC are the sets of tension and compression members, the struts and ties, respectively. Gaynor et al. (2013) and
respectively. For any statically determinate truss that is fully Jewett and Carstensen (2019) consider different materials
stressed (to the tensile stress limit σT and compressive stress for the struts and ties, but most are typically restricted to
limit σC), the volume of the truss can be calculated as follows a single volume constraint for both materials (that is, each
material volume cannot be constrained separately). Thus,
e∈GT|Fe|Le ∑
∑_ e∈GC|Fe|Le
_ (σC+ σT)Z + (σC− σT)C
____________________ these models are limited in practical application. In many
V = + = real-world RC structure design cases, restricting the location
σ T
σ C
2σCσT
(2) of reinforcement (ties) to certain regions while controlling
[ ]
et al. [1964]). In this method, the design domain is discret- 2 _ ∂J
T
ized using a set of nodes that are interconnected by truss Jk(x1, x2) = J(x1k, x2k)+ ∑ (x1k, x2k) [yi(xi)− yi(xik)]
i=1 ∂ yi
members to form an initial ground (that is, reference) struc- (4)
ture (GS). Based on a tailored design update scheme, unnec-
essary members are gradually removed from the initial GS; where x1k, x2kare the values of the design variables at optimi-
the optimal STM design is then obtained. The topology opti- zation step k; and ∂J/∂yi is the gradient of J with respect to
mization formulation for STMs using the GSM is given as the intervening variable yi, which depends on the gradient of
J with respect to xi. In the following, to simplify the notation,
m inJ x , x2) = m
x1,x2 ( 1
in − Π(x1, x2, u(x1, x2))
x1,x2 bi denotes this gradient ∂J/∂yi. Having defined the approxi-
mated objective Jk, a subproblem (by neglecting the constant
terms in Jk) is formulated as
s.t.
∑ jLiTxi− Vmax
j ≤ 0, j = 1,…, n, and i = 1, 2 (3)
i∈G
2
m in∑ [bi( x1k, x2k)]Tyi(xi)
x1,x2 i=1
where x1 and x2 are the vectors of design variables (cross-sec- s.t. ∑ jLiTxi− Vmax
j ≤ 0, j = 1, …, nc (5)
i∈G
tional areas of the truss members) for struts (concrete) and
ties (reinforcement), respectively, which can be constrained
separately, and s.t. is subject to. The objective function J
xi,L ≤ xi(e) ≤ xi,U
(e),k
, Ɐi and e
(e),k
which is obtained as the minimizer of the potential energy design variable xi determined through a user-prescribed
(e)
П; thus, general nonlinear constitutive behavior can be move limit. Introducing a set of Lagrange multipliers λj
incorporated (Sanders et al. 2020). The formulation (Eq. (3)) for each volume constraint, the Lagrangian of the previous
considers a total of n independent volume constraints and subproblem is expressed as
denotes Gj as the set of material indexes for the j-th volume
constraint. The term L iTx1 indicates the total volume asso- L(x1, x2, λ1,…, λnc) =
ciated with the design variable x1, with Li being the length nc
{ [
j=1 i∈G
T
]
∑ ∑ j [bi(x1k, x2k)] yi(xi)+ λjLiTxi − λjVmax }
j (6)
Fig. 3—Deep beam STM design library. Load path Z (that is, Eq. (1)) indicates efficiency of alternative designs in this library.
As load path Z decreases through (a) to (l), efficiency of corresponding design improves. Three highlighted layouts (that is, (a),
(d), and (k)) are selected for experimental validation in “Experimental Evaluation of Optimized Strut-and-Tie Models” section.
Experimental setup and procedure was checked according to ACI 318-19; refer to Fig. 5 for
The beams were cast with concrete having a compressive visualizations of beam and support locations and Appendix
strength of 8.5 ksi (58.6 MPa) at the time of testing and rein- A for reinforcement layouts. The steel reinforcing bars of the
forced with Grade 60 steel reinforcing bars (nominal yield five RC deep beams were instrumented with 350 Ω strain
strength of 60 ksi [414 MPa]). Reinforcing bar development gauges prior to concrete casting. The locations of the strain
Fig. 5—Deep beam specimen geometry for evaluating optimized strut-and-tie models. (Note: 1 in. = 25.4 mm.)
Fig. 6—Schematics of STM reinforcement layouts: (a) to (c) three selected STM layouts designated as standard ACI layout,
optimized layout I, and optimized layout II, respectively; (d) to (f) front view of reinforcement designs; and (g) to (i) perspective
view of reinforcement designs.
gauges were selected to monitor the role of key reinforcing The program of testing consisted of applying an increasing
bars in the optimized layouts; these locations are labeled in load while monitoring crack initiation on the beam. Once a
the three figures of Appendix A. Figures 8(a) to (c) show crack was visually observed, the hydraulic jack valve was
the steel reinforcement used for the tested beams of the closed to hold the load constant, and the cracks and their
ACI layout, optimized layout I, and optimized layout II, corresponding load values were highlighted on the beam.
respectively. Also, the lead wires for the strain gauges are Additional load was then applied; this process was repeated
presented in Fig. 8. Moreover, two linear variable differen- until extensive cracks were observed, which prevented
tial transformers (LVDTs) were used to estimate the effec- further safe monitoring. At that point, continuous loading to
tive strain in the concrete struts. For the midspan deflection, complete failure was carried out, and the failure load was
a string potentiometer was used. A 200 kip (890 kN) load recorded.
cell connected to a hydraulic load ram was used to record
load values. The test setup and instrumentation are shown Load-deflection curves
in Fig. 9. An increasing load was applied to the specimen until
it eventually failed. The load-deflection behavior of the
Fig. 8—Steel reinforcement cages used for: (a) ACI layout; (b) optimized layout I; and (c) optimized layout II.
Fig. 10—Test results of deep beams considering different STM layouts. (Note: 1.0 in. = 25.4 mm; 1.0 kip = 4.45 kN.)
crushing. However, the ACI layout had larger crack widths attempt to avoid an undesirable experimental failure mode
compared to optimized layout I, which can be attributed to in the optimized layout II specimens, the bearing plate width
the more efficient load path Z introduced by more inclined was increased from 6 to 8 in. (0.15 to 0.20 m) for the load
steel reinforcing bars to mitigate large crack widths. On the test of Specimen No. 5. A higher load capacity was observed
other hand, the specimens with optimized STM layout II using the 8 in. (0.20 m) bearing plates, even though the
(Specimens No. 4 and 5) had a different failure mode charac- section still failed in bearing. This shift of the controlling
terized by bearing failure instead of strut failure, as shown in failure mode (for optimized layout II) demonstrates the
Fig. 11(f) and (h). When optimized layout II was tested with effectiveness of the novel STM layout. Future testing can
a 6 in. (0.15 m) bearing support plate (Specimen No. 4), the incorporate more robust bearing layouts, which are expected
failure mode was characterized by bearing failure, showing to result in further optimization of the proposed STM.
the improved design obtained with more inclined steel rein- Figure 11 shows how the total number of observed
forcing bars and a more optimal load path Z, which mitigates tension cracks (cracks in the midspan region of the beam)
the inclined strut failure mode of the tested specimen. To increased for optimized layouts I (Specimen No. 3) and II
Fig. 11—Failure modes of deep beams considering standard ACI layout with 6 in. (0.15 m) bearing plate, optimized layout I
with 6 in. bearing plate, optimized layout II with 6 in. bearing plate, and optimized layout II with 8 in. (0.20 m) bearing plate,
respectively.
Optimized layout I
For optimized layout I, the analysis procedure is the same
as that given in the previous section, except that the resultant
of the forces (CR) acting at Node 4 (refer to Fig. 13(b)) and
the corresponding αR are calculated with Eq. (15) and (16),
respectively. This resolution of forces is intended to simplify
the analytical procedure.
______________________________________ CR =
√ (
C2 sinα2+ C3 sinα3)2+ (C2 cosα2+ C3 cosα3)2 = 1.118P
(15)
(
C2 sinα2+ C3 sinα3
_________________
αR = tan−1
C cosα + C cosα = 0.46
2 2 3 3
)
4 (16)
wc1= b1sinα1 + wT1cosα1 = 7.16 in. (0.18 m) (17)
wc4= (b2/2)sinαR + wT2cosαR = 3.13 in. (0.08 m) (18)
Fig. 13—Deep beam strut-and-tie models: (a) standard ACI As wc4 < wc1, then w
c4is used for the calculation of the
STM layout; (b) optimized STM layout I; and (c) optimized effective compressive stress in the strut at Node 4. Given
STM layout II. the ultimate load P = 114 kip (507 kN) reached during the
Table 2—Relative magnitudes of internal forces (as fraction of applied load P) and angles (in radians) for
ACI layout, optimized layout I, and optimized layout II, as shown in Fig. 13(a) to (c), respectively
T C α1
ACI 1 1.118 tan (1/2)
–1
T1 T2 T3 C1 C2 C3
1 0.2500 0.4714 0.5590 0.6067 0.5336
Optimized I
α1 α2 α3
tan–1(2) tan–1(2/7) tan–1(4/5)
T1 C1 C2 C3 C4 C5
1 0.5590 0.6622 0.1169 0.3030 0.0958
Optimized II
α1 α2 α3 α7 α11
tan–1(2) tan–1(1) tan–1(2) tan–1(3/4) tan–1(1/5)
( )
_
1.118P (25)
_ CR _
1.118P 13 /(0.85 βcfc ′) = 0.719
fce
s = 4 = 4 = 4.52 ksi (31.2 MPa) (19) wc bw
wcbw wcbw
In turn, the value of βs is calculated, assuming that the
strut and node confinement modification factor is equal to For the optimized layout II with an 8 in. bearing plate (that
1.0 (βc = 1.0) is, b1 = 8 in. [0.2 m]), the width of the strut at Nodes 1 and
13 is obtained as
fce
s
__________
βs = = 0.625 (20)
0.85βcfc ′
wc1= b1sinα1 + wT1cosα1 = 8.94 in. (0.23 m) (26)
The larger value of βs = 0.625 calculated for optimized
layout I compared to that calculated for the standard ACI
layout indicates a more efficient load path in the optimized
wc13 = (b2/2)sinαR + wT2cosαR = 3.13 in. (0.08 m) (27)
layout. It should be noted that the βs value calculated for
optimized layout I is less than the value βs = 0.75, which Given the ultimate load P = 189 kip (841 kN) from the
assumes the inclusion of minimum distributed reinforce- experiment (refer to Table 1), the value of βs is obtained,
ment for crack control. assuming that the strut and node confinement modification
factor is equal to 1.0 (βc = 1.0)
Optimized layout II
Similar to the procedure shown in the previous section, the
resultant of the forces acting at Node 13 (refer to Fig. 13(c))
_
1.118P
( )
βs = 13 /(0.85βcfc ′) = 1.038
wc bw
(28)
Note: 1.0 in. = 25.4 mm; 1.0 kip = 4.45 kN; 1.0 ksi = 6.89 MPa.
APPENDIX A
This Appendix details the reinforcement design including
the location of strain gauges for the standard ACI STM
layout (Fig. A1), optimized layout I (Fig. A2), and optimized
layout II (Fig. A3), respectively.
APPENDIX B
Appendix B includes an example to show the calculation
of the Michell number Z for the STM layout in Fig. 3(d).
Because this truss system is statically determinate, the
internal axial force of the members can be calculated using
the equilibrium conditions—that is, N12 = N67 = –0.56P, N24 =
N46 = –0.61P, N34 = N45 = –0.53P, N13 = N57 = 0.25P, N23 =
N56 = 0.47P, and N35 = P (refer to the labeled node numbers
in Fig. B1). Moreover, the length of each truss member is
given as L12 = L67 = 0.56H, L24 = L46 = 1.82H, L34 = L45 =
1.6H, L13 = L57 = 0.75H, L23 = L56 = 0.71H, and L35 = 2.5H.
Therefore, the Michell number Z can be obtained using
Eq. (1) as Z = Σe|Fe|Le = 2|F12|L12 + 2|F24|L24 + 2|F34|L34 +
2|F13|L13 + 2|F23|L23 + F35|L35 = 8.08PH.