Professional Documents
Culture Documents
RESEARCH ARTICLE
© Higher Education Press and Springer-Verlag GmbH Germany, part of Springer Nature 2018
ABSTRACT The present study goes into the search for the safety domain of civil engineering structures. The objective
is to show how a reliability-evaluation brought by a mechanical sizing can be obtained. For that purpose, it is necessary to
have a mechanical model and a reliability model representing correctly the behavior of this type of structure.
It is a question on one hand, to propose a formulation for the nonlinear calculation (mechanical nonlinearity) of the
spatial structures in trusses, and on the other hand, to propose or to adapt a formulation and a modeling of the reliability.
The principle of Hasofer-Lind can be applied, in first approach, for the reliability index estimation, scenarios and the
probability of failure.
The made check concerned metallic in truss structures. Finally, some structures are calculated using the method adapted
by Hasofer-Lind to validate the probability approach of the reliability analysis.
KEYWORDS modeling, nonlinearity mechanical, truss, probability, reliability, response surface, probability of failure
main influential factors on the strength and behavior of key-input parameters, which used to evaluate the Young
concrete-filled multi-planar CHS Inverse-Triangular tubu- modulus of the polymer (epoxy) clay nanocomposite
lar truss. (PCNs). They compared the computation of sensitivity
Truss structures are commonly used in construction indices and to the simulation time between the kriging
because they can lighten the weight of a building while regression (KR) model and the quadratic regression (QDR)
ensuring greater stability. In the first part of this article, we model.
propose a formulation for the nonlinear material calcula- N. Vu-Bac et al. [17] carried out a sensitivity analysis
tion of truss structures (bars elements) in static behavior consisting in quantifying the influence of uncertain input
under an increasing loading until fracture. Our formulation parameters on uncertain model outputs. The results are
will be implemented in a calculation program in based on a probability density function (PDF) provided for
FORTRAN language processing the beams elements, the input parameters.
which was developed in references [10,11]. It has been demonstrated that the spline regression
The safety of a mechanical system is ensured by a safety model is more robust than polynomial regression model. It
coefficient: the ratio between a variable loading and a is necessary to take penalized spline regression models
variable strength determined by a deterministic approach. using global separate penalties or separate global penalties
For a complex structure these efforts are poorly known, into account in order to approximate the observed data.
and its strength is uncertain, so there is always a risk of the N. Vu-Bac et al. [18] used a hierarchical multiscale
structure fracture. To do this, the probabilistic approach model brindging four (nano, micro, meso and macro)
allows evaluating the risk by methods of analysis of scales to study the effect of uncertain model inputs on the
reliability of the mechanical systems developed during the macroscopic Young’s modulus and the Poisson’s ratio.
last years. They used sensitivity analysis (SA) methods to estimate
In structure reliability, the FORM and SORM methods the effect of the uncertain correlated (dependent) inputs on
are an approximation methods allowing to determine a the Young's modulus and the Poisson's ratio for the
particular structures design point and therefore to estimate multiscale model in the context of a global SA. Estimates
the failure probability. These methods are therefore for correlated parameters are performed for both first-order
intrinsically linked to this famous design point it permit and total sensitivity indices.
to define the distance between the design point and the It is proposed in this article in the second part, first a
failure surface point in the original space. This distance is classical technique for the calculation of the reliability
called Reliability: Reliability or Safety Index. (method Hasofer Lind level 2) which uses the probability
The so-called safety index method was also used by and the statistic and makes it possible to check the
Ravindra et al. [12] to design reinforced concrete beams reliability of the metallic structures and ensuring dialogue
and structural steel members. between mechanics and reliability model with indirect
Karamchandani and Cornell [13] developed a method coupling method. The reliability model and its coupling
that approximates the parameter sensitivity with respect to with the mechanical model are writing in FORTRAN
distribution parameters that can take second order effect language.
into account, based on SORM and the finite difference Finally, in this paper we propose a non linear
method. computational calculation method for the truss structures.
N. Vu-Bac et al. [14] performed a sensitivity analysis We have modeled the reliability problem by the Hasofer-
(SA) based on their MD results to quantify the influence of Lind principle for the reliability index estimation the
uncertain input parameters on the predicted yield stress and failure probability. Once the mechanical model and the
elastic modulus. The sensitivity analysis (SA) is based on reliability model have been available, they have to be
response surface (RS) models (polynomial regression and coupled by a response surface, and then the probability
moving least squares). They used partial derivatives (local laws of the random variables retained to approximate the
SA) and variance-based methods (global SA) in this study statistical law (real law).
an coefficient of determination (COD) is computed for
allowing an estimation of the quality of the approximation.
N. Vu-Bac et al. [15] used simulations of molecular 2 Methodology of nonlinear analysis of
dynamics (MD) or studied the effect of single-walled truss structures
carbon nanotube (SWCNT) radius, the temperature and the
pulling velocity on interfacial shear stress (ISS). For 2.1 Introduction
computational efficiency, the sensitivity analysis (SA) is
based on surrogate models (polynomial regression, mov- The elements of a truss work only in tension or
ing least squares (MLS) and hybrid of quadratic poly- compression, thus they are modeled by finite elements
nomial and MLS regressions). such bars. This is usually an element with two nodes,
N. Vu-Bac et al. [16] proposed a stochastic framework which includes three degrees of freedom. Each node
based on sensitivity analysis (SA) methods to quantify the represents the components of its movement in space. The
Karim BENYAHI et al. Nonlinear analysis and reliability of metallic truss structures 579
nodes have no rotations freedom (thus not running); This condition, taking into account the Eq. (3), can be
because they have no physical meaning. Indeed, the writing as:
existence of a rotational freedom degree mean the presence ( ) " #( )
within the bar element (pin-ended element) of bending ΔF sn ½k mn O Δεn
¼ (6)
moment or torsion, which is excluded. 0 O O 0
2.4 Flowchart of calculation 2.5.1 Formulation of the bar element stiffness matrix in its
intrinsic system
The flowchart of the equilibrium state search, in the metal
profile section is described below (Fig. 2): The loads on the extremity of the bar element is (Fig. 3):
fF N g ¼ fN j ,0,0,0,0,0gT (10)
!0 ½LðxÞTfδðxÞgdx
L The element nodes displacements in the intermediate
fS N g ¼ (17) system coordinate can be written as:
Where : fS U g ¼ ðu,v,w,0,0,0,0,0,0ÞT (27)
fδðxÞg ¼ fεðxÞ,0,0,0,0,0gT (18) Where:
8
Using the virtual work theorem and neglecting the < u ¼ uj – ui
>
element length variation in computing of the nodes v ¼ v j – vi (28)
displacements permits to write: >
:w ¼ w –w
j i
!0 ½LðxÞTfΔδðxÞgdx
L
fΔS N g ¼ (19) The relationship between the displacements and loads in
these two systems are given by:
By replacing the relations (8) and (17) in Eq. (19), one
gets the flexibility matrix of the element K N such that: fS U g ¼ ½T 0 ½S L (29)
2.5.2 Formulation of the bar elements stiffness matrix in its fΔF U g ¼ ½BT fΔF N g þ ½ΔBT fF N g (33)
local system coordinate
When:
2 v w 3
The loads at the element nodes in the local system 1 0 0 0 0 0 0
coordinates x0y0z0, are: 6 L0 L0 7
60 0 0 0 0 0 0 0 07
6 7
fF L g ¼ ðF ix0 ,F iy0 ,F iz0 ,0,0,0,F jx0 ,F jy0 ,F jz0 ,0,0,0ÞT (24) 6 7
60 0 0 0 0 0 0 0 07
½B ¼ 6
6
7
7 (34)
The displacements at the element nodes in the local 60 0 0 0 0 0 0 0 07
system coordinates x0y0z0, are: 6 7
6 7
40 0 0 0 0 0 0 0 05
fS L g ¼ ðui ,vi ,wi ,0,0,0,uj ,vj ,wj ,0,0,0ÞT (25)
0 0 0 0 0 0 0 0 0
The element nodes loads in the intermediate system
coordinate can be written as: And ½ΔBT fF N g ¼ ½DfΔS U g (35)
fF U g ¼ ðF jx0 ,F jy0 ,F jz0 ,0,0,0,0,0,0Þ T
(26) With
582 Front. Struct. Civ. Eng. 2018, 12(4): 577–593
2 3 2 3
– vN j – wN j ½R0 0 0 0
6 0 2 0 0 0 0 0 07 6 7
6 L0 2 L0 7 6 0 0 0 07
6 7 ½RT ¼ 6 7 (43)
6 7 6 7
6 – vN j Nj 7 4 0 0 ½R0 0 5
6 0 0 0 0 0 0 07
6 L0 2 7
6 L0 7 0 0 0 0
6 7
6 7
6 – wN j Nj 7 The geometric transformation matrix [R0] of a three-
6 0 0 0 0 0 0 07
6 L0 2 7 dimensional bar element; it given by:
6 L0 7 2 3
½D ¼ 6
6
7
7 cx cy cz
6 0 0 0 0 0 0 0 0 07 6 7
6 7 6 – c x Y S – cy Z 7
6 7 6 7
6 0 0 0 0 0 0 0 0 07 ½R0 ¼ 6 S
6
6
7
7 6 L S 77
(44)
6 0 0 0 0 0 0 0 0 07 4 –Z –X 5
6 7 0
6 07 S S
6 0 0 0 0 0 0 0 0 7
6 7
6 07 8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4 0 0 0 0 0 0 0 0 5 >
>
> L ¼ ðxj – xi Þ2 þ ðyj – yi Þ2 þ ðzj – zi Þ2
0 0 0 0 0 0 0 0 0 >
> qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
>
>
>
> S ¼ ðxj – xi Þ2 þ ðzj – zi Þ2
(36) >
>
>
>
Finally, the relationship between the nodes loads < xj – xi X
With cosðx,xÞ ¼ ¼ (45)
increase and the nodes displacements increase in the >
> L L
>
> yj – yi Y
local system coordinate is given by: >
> cosðy,yÞ ¼ ¼
>
>
fΔF L g ¼ ½K L fΔS L g >
> L L
(37) >
> zj – zi Z
>
: cosðz,zÞ ¼ ¼
The stiffness matrix [KL], of the bar element in the local L L
system coordinate is defined by:
Finally, the relationship between the loads increase and
½K L ¼ ½T 0 T ð½BT ½K N ½B þ ½DÞ½T 0 (38) the displacements increase of the bar element nodes in the
absolute system coordinate OXYZ, is:
2.5.3 Formulation of the bar element stiffness matrix in the fΔF X g ¼ ½K X fΔS X g (46)
absolute system
The element stiffness matrix in the absolute system
The nodes loads increase in the absolute system coordinate coordinate can be written as:
are defined as: ½K X ¼ ½RT T ½K L ½RT (47)
fΔF X g ¼ ðΔF ix ,ΔF iy ,ΔF iz ,0,0,0,ΔF jx ,ΔF jy ,ΔF jz ,0,0,0ÞT
The stiffness matrix [KS] and [KN] are singular and
(39) cannot remain so, because in the resolution of the general
stiffness matrix in the global system coordinate, the null
The nodes displacements increase in the absolute system pivots will appear. For this purpose, disrupts the terms of
coordinate are defined as: the diagonal corresponding to the shear forces, instead of
being zero, are multiplied by a real coefficient between 0
fΔS X g ¼ ðΔX i ,ΔY i ,ΔZ i ,0,0,0,ΔX j ,ΔY j ,ΔZ j ,0,0,0ÞT (40)
and 1.
The relationship linking the nodes displacements After implementing the procedure described above, the
increase in the local system coordinate and the nodes other null pivots to appear due to the singularity of the
displacements increase, in the absolute system coordinate matrix [L(x)] and [B]then to solve this problem we proceed
can be written as: in the same way as for the stiffness matrix (instead of
canceled the terms of the diagonal of the matrix, they are
fΔS L g ¼ ½RT fΔS X g (41) taken equal to 10–4).
The relationship between the loads increase in the local
system coordinate and the loads increase in the absolute 2.5.4 Node displacement calculation
system coordinate is:
The nonlinear problem is solved by using an iterative
fΔF X g ¼ ½RT T fΔF L g (42)
method [19,20,21], based on the displacements method, in
With which we recalculate, for each step, the matrix structure’s
Karim BENYAHI et al. Nonlinear analysis and reliability of metallic truss structures 583
reliability analysis methods [25,26,27], developed in Dividing the equation by norm krHðuÞkuðkÞ and by
recent years. introducing the direction cosines of H in P k we get :
HðuðkÞ Þ
3.2 Research design point þ hαiðkÞ fugðkþ1Þ – fugðkÞ ¼ 0 (57)
krHðuÞkuðkÞ
The design point (or the most probable failure) is the point
of the limit state surface where U probability density is a It comes:
maximum; it is also defined as the point of the state limit HðuðkÞ Þ
surface as close to the origin: fugðkþ1Þ fαgðkÞ ¼ fugðkÞ fαgðkÞ – (58)
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi krHðuÞkuðkÞ
βHL ¼ min fugT fug (53)
gfxi ðuj Þg£0 rHðuðkÞ Þ
With αðkÞ ¼ is the vector cosine directors
Under constraint HðuÞ£0. krHðuÞkuðkÞ
In this study the constrained minimization problem is (or the vector of the normalized gradient) of H in P(k).
solved by using the algorithm of Hasofer-Lind-Rackwitz- In the limit when k↕ ↓1,dðuðkÞ Þ ¼ β and fug ¼ – βfαg
Fiessler which is an adaptation of a first order optimization if the algorithm is convergent. At iteration (k), let:
algorithm to the problem of research design point.
fugðkþ1Þ ¼ – βðkÞ fαgðkÞ ) βðkÞ
3.3 Optimization algorithms rackwitz-fiessler
¼ – huiðkþ1Þ fαgðkÞ (59)
There are many algorithms allow solving the optimization
problem [28,29]. In practice, Rackwitz-Fiessler algorithm This leads to the iterative relationship giving the
is the most used because of its simplicity and good results. reliability index:
The algorithm Hasofer-Lind-Rackwitz-Fiessler (HL-RF) is
an adaptation of a first order optimization algorithm to the HðuðkÞ Þ
βðkÞ ¼ – huiðkÞ fαgðkÞ þ (60)
problem of research design point. The algorithm (HL-RF) krHðuÞkuðkÞ
is an improvement of the Hasofer Lind algorithm, which
The search algorithm of reliability index stops when the
assumes that the basic variables are Gaussian and
norm
uncorrelated. As for him (HL-RF) is more general. It has
no restrictions on basic variables. The algorithm to be used kfugðkþ1Þ – fugðkÞ k£ε:
in the following is that found in the book of Lemaire [30].
The assumptions considered in the algorithm are: And fugðkþ1Þ is deducted by substituting Eq. (60) into
- The limit state function H has a gradient at the point of Eq. (59):
coordinate u. HðuðkÞ Þ
- The gradient rHðuÞ is not null at any point in the uðkþ1Þ ¼ huiðkÞ fαgðkÞ fαgðkÞ – fαgðkÞ (61)
hyper limit state surface. krHðuÞkuðkÞ
To determine the design point, we consider a normalized
space at a point P(k) to coordinate {u}(k), the origin point of
the iteration (k). This point does not necessarily belong to 3.3.1 Summary of the algorithm
constraint and H(u) can be different from zero.
Taylor series development of the state of the limit The algorithm of Hasofer Lind-Rackwitz-Fiessler (HL-RF)
function H(u) about the point P(k) gives: is summarized by the following steps:
1- Choose a starting point {u}(0);
HðuÞ ¼ HðuðkÞ Þ þ hrHðuÞiuðkÞ fug – fugðkÞ 2- Evaluate the limit state function H(u(K));
2 3- Calculate the gradient of the limit state frHðuÞgðkÞ
þO fug – fugðkÞ (54)
and norm krHðuÞkðkÞ , to deduce fαgðkÞ by: αðkÞ ¼
The equation of the tangent has hyper plan H(u) as
{u}(k): rHðuðkÞ Þ
;
hrHðuÞiuðkÞ fug þ c ¼ 0 (55) krHðuÞkuðkÞ
4- Calculate the reliability index βðkÞ ;
rHðuðkÞ Þ is the gradient H(u) at the point P(k). Then, we 5- Calculate the coordinate of the next iteration {u}(k + 1);
define P(k + 1) by: 6- Convergence tests:
If kfugðkþ1Þ – fugðkÞ k£ε, stop calculating;
Hðuðkþ1Þ Þ ¼ HðuðkÞ Þ þ hrHðuÞiuðkÞ fugðkþ1Þ – uðkÞ ¼ 0 Else put k = k + 1 and go 2.
(56)
Karim BENYAHI et al. Nonlinear analysis and reliability of metallic truss structures 585
3.3.2 Flowchart Hasofer-Lind-Rackwitz-Fiessler (HL-RF) Where: X is the vector n basic variables, and these ci ,cij
are the coefficients of the polynomial sought.
The flowchart Hasofer-Lind-Rackwitz-Fiessler (HL-RF) is
described below (Fig. 5): 3.4.2 Flowchart of coupling reliability-mechanical by
response surface
To realize this coupling, there are three methods of control 4 Comparisons with numerical and
[30]: The direct coupling, the coupling response surface, experimental results
the coupling optimization.
In this present study, the mechanical-reliability coupling 4.1 Validation of the non-linear calculation of the truss
will be directed by response surface. structures
The analytical response surface method is to replace the A truss plane in three equal spans has been tested by
function of unknown performance g(x) by approximated LOVEGROVE and analyzed by X. SUN and S.L CHAN
function g^ðxÞ. The choice of a high order polynomial is to [31] using the finite element method.
better represent the model. Very often, a quadratic shape is The layout and dimensions of the truss are shown in
chosen for writing the substitute g^ðxÞ of the performance Fig. 7, and sections of the frames. The values of modulus
function g(x). In other words, the substitute can be written of elasticity and the elastic limit data are respectively equal
as follows: to 214 GPa, and 285.4 MPa.
Xn Xn
gðxÞ g^ðxÞ ¼ c0 þ cX þ c X2 (62) The element passing through the point (A) is designed
i¼1 i i i¼1 ii i
such that the load applied in the vicinity makes it as the
Xn Xn – 1 Xn critical point.
gðxÞ g^ðxÞ ¼ c0 þ cX þ
i¼1 i i i¼1
c XX
j¼iþ1 ij i j The evolution of the load versus lateral displacement at
(63) mid point (A) is shown in Fig. 8.
586 Front. Struct. Civ. Eng. 2018, 12(4): 577–593
Fig. 8 Plane truss beam- load curve – arrow The geometrical and material characteristics of the planar
truss studied (scope section, loading, boundary condition,
The experimental ultimate load is given equal to 88 kN. elastic modulus, elastic limit) are described in section
The calculated ultimate load is 90 kN, a difference of 4.1.1.
2.27% (in the sense of an overstatement). By against it is The procedure to estimate random variables probability
found that the shape of the calculated curve approximates distributions selected for this study in order to approximate
the experimental curve. The agreement between the the statistical law (real). Most often the mean and standard
experiment and our calculation is pretty good. deviation of a random variable are known, however, this
Fig. 12 (a) Probability distributions of the random variable P; (b) Probability distributions of the random variable δ
588 Front. Struct. Civ. Eng. 2018, 12(4): 577–593
the aim of as close as possible real probability distributions 4, 5, 6, 7). Consequently, there is a significant difference
of random variables used in this example, it is retained for between the probability of failure of the random variables
seven case of combining the distribution laws. The seven in the same distribution law (cases 1, 2, 3) and the
cases selected random variables are shown in Table 3 and probability of failure of random variables act differently
limit state (response surface) approached in the reduced distribution (cases 4, 5, 6, 7) it is because in reality the
^
space centered HðP,δÞ are shown in Fig. 16. random variables may not follow all the same law.
We also note that if four gives the greatest reliability
index β ¼ 1:29306, where reliability of 90.20%, which
confirms the validity of the first method for estimating
probability distributions that can approximate the statistics
law real of the random variable.
In our study, we considered that continuous and
independent random variables, when in reality it is still
not the case for all systems.
Table 4 Results of the mechanical reliability analysis for the various cases treated
case reliability index probability of failure direction cosine (α1 ,α2 ) design point (U1, U2) design point (X1, X2)
case 1 0.20974 0.4169 ( – 0.1094, 0.9939) (0.0229, – 0.2084) (51.755, 0.0045)
case 2 0.11476 0.4542 ( – 0.1157, 0.9932) (0.0132, – 0.1139) (36.724, 0.0032)
case 3 0.09772 0.4611 (0.2409, – 0.9705) ( – 0.0235, 0.0948) (11.84, 0.00086)
case 4 1.29306 0.09800 ( – 0.3047, 0.9524) (0.3940, – 1.2315) (28.831, 0.0018)
case 5 0.70959 0.2390 ( – 0.2901, 0.9569) (0.2058, – 0.6790) (69.763, 0.0074)
case 6 0.82593 0.2044 ( – 0.3148, 0.9491) (0.2600, – 0.7839) (71.391, 0.0079)
case 7 0.4054 0.3426 (04997, – 0.8661) ( – 0.2026, 0.3511) (33.498, 0.0022)
590 Front. Struct. Civ. Eng. 2018, 12(4): 577–593
Table 8 Results of the mechanical reliability analysis for the various cases treated
case reliability index probability of failure direction cosine (α1 ,α2 ) design point (U1, U2) design point (X1, X2)
case 1 0.3385 0.3674 ( – 0.4174, 0.9086) (0.1413, – 0.3076) (0.6942, 0.0014)
case 2 0.2506 0.4010 ( – 0.5398, 0.8417) (0.1353, – 0.2109) (0.6406, 0.0012)
case 3 0.2360 0.4067 (0.2460, – 0.9692) ( – 0.0580, 0.2287) (0.226, 0.00033)
case 4 1.0300 0.1515 ( – 0.2794, 0.9601) (0.2878, – 0.9889) (0.362, 0.00063)
case 5 0.7535 0.2255 ( – 0.2558, 0.9667) (0.1928, – 0.7283) (0.6811, 0.0013)
case 6 0.4710 0.3188 ( – 0.2491, 0.9684) (0.1173, – 0.4561) (0.7426, 0.0014)
case 7 0.4364 0.3312 (0.4279, – 0.9037) ( – 0.1867, 0.3944) (0.476, 0.00077)
592 Front. Struct. Civ. Eng. 2018, 12(4): 577–593
5 Conclusion ui, vi, wi : Components of the displacement vector in the local coordinate
system,
Steel real behavior laws are used in this present study to ½S S i – 1 : Sections flexibility matrix of the iteration (i-1),
treat the structures non-linear calculation. This calculation εs : Strains balanced in the previous step,
gives best estimates the real displacement of structures. fΔF s gr : Forces increase in the step r,
The formulation proposed for a bar element, in the case fΔεg0 : Initial strains increase,
of steel truss, and its implementation allowed to treat the
[K]i: Structure stiffness matrix at the iteration (i),
nonlinear material analysis of truss structures under
increasing monotonic loading until failure. {Us}: Node displacement vector at the latest stable step,
Finally, we addressed the reliability problems on some fΔPgr : Applied load increase in the r step,
examples to estimate their reliability index and determine {P}: External structures applied loads,
their failure probability. The results obtained by combina- {Pint}: Internal structures applied loads,
tion of the continuous random variables laws show that the
Φ: The normal law distribution function reduced centered (mean 0 and
method for estimating probability distributions used in the
standard deviation 1),
first approach approximate correctly the real statistical
mR: Means strength,
distribution of the random variable.
The reliability model coupled to the mechanical model mS : Means loads,
applied to the truss structures; assess the reliability index R : Standard deviations of the strength,
relating to the ruins of mechanical systems with nonlinear S : Standard deviations of the loads,
behavior and not normal variables.
P*: Point of the most probable failure,
αðkÞ : Vector cosine directors.
Notations
Ea: Young modulus of steel, References
εe : Steel yielding strain,
1. Yaghmai S. Incremental analysis of large deformations in mechanics
e : Steel yielding stress,
of solids with applications to axisymmetric sheus of revolution.
εu : Steel ultimate strain, Technical Report SESM 68–17, Univ. Califorma, Berkeley, 1968
εx : Strain in the gravity center of the total area caused by the normal force N, 2. Grelat A. Nonlinear analysis of reinforced concrete hyperstatics
{Fmn}: Contribution caused by the concrete and / or metal profile, frames, Doctoral thesis Engineer. University Paris VI, 1978
Sm: The metal profile’s cross-section. 3. Grelat A. Non-linear behavior and stability of reinforced concrete
frames, Annals of I.T.B.T.P., N° 234, 1978, France
Em (y,z): The longitudinal elastic modulus at a current point of the metal
profile cross-section.
4. Sargin M. Stress-Strain relationship for concrete and the analysis of
structural concrete sections. Solid Mechanics division, University
Δm ðy,zÞ: Normal stress in a current point of the metal profile,
of waterloo, Canada, 1971
[kmn]: Section stiffness matrix, 5. Rabah O N. Numerical simulation of nonlinear behavior of frames
{Fsn}: Vector of the sections normal forces, Space, Doctoral Thesis. Central School of Paris, France, 1990
e: Element length increase, 6. Robert F. Contribution to the geometric and material nonlinear
analysis of space frames in civil engineering, application to
L0: Element initial length,
structures, Doctoral Thesis. National Applied Sciences Institute,
L: Element length after deformation,
Lyon, France, 1999
[B]: Geometric transformation matrix, 7. Espion B. Contribution to the nonlinear analysis of plane frames.
[KL]: Element stiffness matrix in the local coordinate, Application to reinforced concrete structures, Doctoral Thesis in
[R0]: Geometric transformation matrix, Applied Science, vol. 1 and 2. Free University of Brussels, Belgium,
1986
[KX]: Element stiffness matrix in the absolute coordinates,
8. Machacek J, Charvat M. Design of Shear Connection between Steel
[KN]: Bar element’s stiffness matrix in the intrinsic system coordinate.
Truss and Concrete Slab. Structures and Techniques, 2013, 57:
[KU]: Element Stiffness matrix in the intermediate local coordinate system, 722–729
{FX}: The nodes load vector in the absolute system coordinate, 9. Feng R, Chen Y, Gao S, Zhang W. Numerical investigation of
{SX}: The nodes displacements vector in the absolute coordinate system, concrete-filled multi-planar CHS Inverse-Triangular tubular truss.
Thin-walled Structures, 2015, 94: 23–37
[FL]: The nodes load in the local coordinate system.
10. Adjrad A, Kachi M S, Bouafia Y, Iguetoulène F. Nonlinear
[SL]: The nodes displacements vector in the local system coordinate.
modeling structures on 3D, In: Proc. 4th Annu.icsaam 2011.
[FU]: The nodes loads vector in the intermediate system coordinate. Structural Analysis of Advanced Materials, Romania, pp. 1-9, 2011
[SU]: The nodes displacements vector in the intermediate system coordinate. 11. Kachi M S. Modeling the behavior until rupture beams with external
Karim BENYAHI et al. Nonlinear analysis and reliability of metallic truss structures 593