Professional Documents
Culture Documents
ABSTRACT: Masonry walls are very sensitive to flexural effects due to low tensile strength which, in
turn, greatly influences the load bearing capacity under compression. The main source of flexural effects
may result from the eccentric loading at the ends of the wall or from any lateral loading like the wind
action, the earth pressure, or the second order effect of the applied actions. Several analytical solutions
were proposed in the literature to solve the differential equation of the problem, but those solutions were
limited to special conditions. In the current contribution, a general formulation for the non-linear stability
problem has been formulated numerically based on the transfer-matrix method. Despite the method is out
of professional use today and don’t possess the potential and flexibility of the finite elements but for the
currently addressed problem, it is still the most efficient. A relative form description has been introduced
to formulate the stability theory of masonry walls. This form of description has been used to minimize
the dimensions of matrixes in transfer-matrix method and to produce the equations in a compact form.
The algorithms of the method have been derived for general boundary and loading conditions with a
user-defined non-linear material model. Algorithms and solution procedures have been explained and
implemented into a computer code. The convergence of the iterative solution has been studied with a clear
definition for the cases at which the stability or material failures occur. The results of the developed solu-
tion procedure have been validated by comparing them with the existing solutions and the experimental
results. The developed solution procedure provides a powerful tool to solve a wide range of problems
related to the stability of masonry walls and to check the existing empirical methods.
71
rBj
m=
rh2j
(
⋅ ω 2j ⋅ C1 ⋅ cos ω j ξ + C2 ⋅ sin ω j ξ
) (5)
or m = Φ ⋅ re ;
rBj
v=
rh2j
(
⋅ ω 3j ⋅ −C1 ⋅ sin ω j ξ + C2 ⋅ cos ω j ξ
) (6)
or v = Φ ⋅ rϑ .
This yields:
Figure 1. A discrete model for masonry wall under rh2j ⋅ rq j rϑi
compression and lateral load and the transfer-matrix C1 = rei + ; C2 = . (8)
method derived for one element. Φ ⋅ ω 2j ωj
72
73
rne = R ne ⋅ r0 ;
ne R11 R12 R13
(23)
R ne = ∏Ωne +1− i = R21 R22 R23 .
i =1 0 0 1
(24)
rϑ n = R21re0 + R22rϑ 0 + R23 .
74
The relative eccentricity at the mid of element j The relative eccentricity at the mid of the
can be calculated using the following equation: element can be updated as follows after getting
the value of ε j from step 5-c:
ωj
Ye j = cos
ωj
⋅ re + 2 ⋅r
sin
ϑi re j =
1
⋅
( ) ( )
I m ε1 j − I m ε 2 j 1 ε1 + ε 2 j
− j ,
2
2 i
rh ⋅ rq
ωj
ωj
( ) ( )
ε1 j − ε 2 j I n ε1 − I n ε 2
j j
2 ε1 j − ε 2 j
− j 2j 1 − cos . (29) (33)
Φ ⋅ω j 2
where
b. Calculation the relative strain difference
The relative strain difference ∆ ∈j can be deter- mε (ε ) , ε ≥ εt
I m (ε ) = ;
mε (εt ) , ε < εt
mined at the mid of element j by:
ε (34)
1 Φ
∆ ∈j = ⋅ ⋅ Ye j . (30) mε (ε ) = ∫ε ⋅ σ ⋅ d ε
ε c rBj 0
where εc is the strain corresponding to com- The relative strains ε1 j and ε 2 j can be calculated
pressive strength of masonry, rBj the relative as follows:
stiffness of the element.
εj , ∆ εi ≥ 0
c. Solving the equilibrium equation ε1 j = ;
This requires calculating the maximum normal- ε j − ∆ ε j , ∆ εi < 0
(35)
ized compression strain ε j at the mid of element j ε j − ∆ ε j , ∆ε j ≥ 0
which leads to the equilibrium state. Based on the ε2 j = .
defined material model and the calculated value εj , ∆ε j < 0
of | ∆ε j |, the maximum normalized strain ε j can
be found as a solution of the following equation:
When the updated value gets bigger than half,
i.e. | re j | > 1/ 2, the calculation must be ended as
Φ=
( ) (
I n ε j − I n ε j − ∆ε j ). (31)
this indicates a stability failure in the system.
∆ε j b. Updating the relative stiffness
By using the updated relative stiffness at the mid
where I n (ε ) depends on the material model of the element j, the relative flexural stiffness
and can be calculated as follows: can be also updated as follows:
Φ
nε (ε ) , ε ≥ εt rBj = ⋅ re . (36)
I n (ε ) = ; εc ⋅ ∆ε j j
nε (εt ) , ε < εt (32)
ε
nε (ε ) = ∫ σ ⋅ d ε . Step 7: Calculation the damage state within the
0
cross section
where εt the maximum relative tensile strain To get an idea about the state of damage
of the material. The value of ∆ε j always takes occurred due to cracking of the cross section, it is
75
rt1 =
(
ε1 j − max ε 2 j , εt ); Figure 3.
The function Φ is non-smooth at εt + ∆ε and its
j
∆∫j derivative is discontinuous at this point. The func-
(38) tion Φ reaches its peak at the relative strain εm ,
rt2 =
ε 2 j − max ε1 j , εt( ). which is the root of the derivative. It is useful to
determine εm at the peak point to check whether
Downloaded by [California State University, Fresno] at 13:00 05 December 2016
j
∆ε j a solution exists or not. Φ > Φ m means there is a
material failure and thus no solution exists. Φ ≤ Φ m
Step 8: Checking the convergence means there is a solution and it may be a single solu-
The convergence of the solution can be checked tion or double solutions. This is mainly dependent
at each iteration using the following convergence from the characteristics of the post-peak branch of
index: the stress-strain relationship. If εu ≤ εm , only one
solution exists and can be found in the range [ 0, εm ]
ne . if εu > εm , two solutions are possible and the sec-
∑ (re )
2
Conv = − Yei ≤ err. (39) ond solution should be found in the range [εm , εu ]
i =1
i
. The maximum values of Φ m and its correspond-
ing εm have been determined numerically using the
Since the assumption made in step 2 might be golden section search method, by successively nar-
incorrect, the solution state must be checked at rowing the searching domain. The initial search-
this step. If the convergence index is bigger than ing domain is taken between [ 0, εu ] . The probe
the permissible error, the steps from 3 to 8 must be point within the searching domain is chosen with
repeated, otherwise, the iteration must be ended. a proportional spacing ratio equal to the golden
ratio g = 1+2 5 . The secant method has been applied
to find the solution numerically within the range
4 Solving the equilibrium [ε1, ε2 ]. The following recurrence relation were used
equation to find better approximation of the solution:
In step 5-c from the iterative procedure described
above, it is required to find the relative maximum
compressive strain ε at the mid of the element
by solving Eq. (31). The solution of Eq. (31) is
dependent from the material integral I n and the
relative strain difference ∆ε . The solution of eq.
(31) can be found from the stress-strain relation-
ship at normalized stress level equal to Φ. When
an inflection point on the curve I n (ε ) exists, two
secants of slope equal to Φ can be determined at
two positions; one secant is located to the right and
the second one is located to the left of the inflec-
tion point. The function Φ can be described in a
stepwise form as follows:
Φ (ε, ∆ ε, εt ) =
nε (ε ) − nε (ε − ∆ ε )
, ε ∈[εt + ∆ ε, εu ]
∆ε (40)
.
nε (ε ) − nε (εt )
, ε ∈[εt , εt + ∆ ε ]
∆ε
The derivative of the function Φ with respect to Figure 3. Graphical representation of Φ as a stepwise
ε can be also written in a stepwise form as follows: function with transition point at εt + ∆ε.
76
the analytical solution of the differential equation mines if a solution exists or if a material failure
of buckling without lateral load (Angervo, 1954, occurs. The outcome of this routine provides an
Angervo, 1961, Angervo et al., 1957), (Chapman important indication to terminate the iterative
and Slatford, 1957), and (Yokel, 1971) and with solution when a material failure exists. All algo-
lateral load (Hellers, 1967). rithms have been developed to use a wide variation
The values of the capacity reduction factor of nonlinear material models. For the practical
obtained by the transfer-matrix method were iden- use of the developed TMM solution, the capac-
tical or very close to the values obtained by the ity reduction factor has been determined based
analytical solution. When masonry wall divided on bisection method. The solution procedure has
into 20 elements in the transfer-matrix method, the been checked for several examples and different
difference in the values of the capacity reduction failure modes. This provides an important tool to
between both solutions becomes less than 10−3. study the influence of different parameters on the
Since masonry has a wide variation in the non- capacity reduction factor and to help to verify the
linearity of the material model, the proposed solu- existing empirical methods.
tion procedure is further verified with experimental
data of masonry walls of different material com-
binations (Pflücke, 2006). The capacity reduction References
factor of the experimental and numerical results
have been represented in Figure 4. A comparison Angervo, K. (1954) Über die Knickung und Tragfähigkeit
between the experimental and numerical values of eines exzentrisch gedrückten Pfeilers ohne Zugfestigkeit
the capacity reduction factors demonstrates very unter- und oberhalb der Proportionalitätsgrenze mit
besonderer Berücksichtigung des rechteckigen Quer-
good conformity of both results. schnitts, Helsinki.
Angervo, K. (1961) Erweiterung der Theorie der Biegung
eines Pfeilers ohne Zugfestigkeit, Helsinki.
6 Concluding remarks Angervo, K., Putkonen, A.I. & Attila, U. (1957) Erweiter-
ung der Theorie der Biegung eines Pfeilers ohne
A numerical solution has been proposed to solve Zugfestigkeit und ihre Anwendung zur Berechnung von
the differential equation of masonry buckling Rahmentragwerken mit unbewehrten Stielen, Helsinki.
under combined vertical and lateral actions using Chapman, J.C. & Slatford, J. (1957) The elastic buckling
of brittle columns. Institution of Civil Engineers Pro-
ceedings, 6, 107–125.
Falk, S. (1955) Biegen, Knicken und Schwingen des
mehrfeldrigen geraden Balkens. Abhandlungen der
Braunschweigischen Wissenschaftlichen Gesellschaft,
7, 74–92.
Falk, S. (1956) Die Berechnung des beliebig gestützten
Durchlaufträgers nach dem Reduktionsverfahren.
Ingenieur-Archiv, 24, 216–232.
Falk, S. (1986) Das direkte (natürliche) Reduktionsver-
fahren, Teil II. Acta Mechanica, 62, 5–18.
Hellers, B.G. (1967) Eccentrically compressed columns
without tensile strength subjected to uniformly dis-
tributed lateral loads. National Swedish Institute for
Building Research. Stockholm.
Marguerre, K. (1960) Matrices of Transmission in Beam
Problems. IN Sneddon, I.N. & Hill, R. (Eds.) Progress
Figure 4. Comparison between experimental and in Solid Mechanics, Vol I. Amsterdam, North-Holland
numerical results of the capacity reduction factor. Publishing Company.
77
rialverhaltens. IN Jäger, W. & Vassilev, T. (Eds.) Bau- with no tensile strength. Journal of the Structural
forschung und Baupraxis—From Research to Practice in Division, 97, 1913–1926.
Construction. Dresden, Lehrstuhl Tragwerksplanung,
Fakultät Architektur, Technische Universität Dresden.
78