Professional Documents
Culture Documents
Multiphysics Modelling
and Simulation for
Systems Design and
Monitoring
Proceedings of the Multiphysics Modelling
and Simulation for Systems Design Conference,
MMSSD 2014, 17–19 December, Sousse, Tunisia
Applied Condition Monitoring
Volume 2
Series editors
Mohamed Haddar, National School of Engineers of Sfax, Tunisia
Walter Bartelmus, Wroclaw University of Technology, Poland
Fakher Chaari, National School of Engineers of Sfax, Tunisia
e-mail: fakher.chaari@gmail.com
Radoslaw Zimroz, Wroclaw University of Technology, Poland
About this Series
The book series Applied Condition Monitoring publishes the latest research and
developments in the field of condition monitoring, with a special focus on indus-
trial applications. It covers both theoretical and experimental approaches, as well
as a range of monitoring conditioning techniques and new trends and challenges in
the field. Topics of interest include, but are not limited to: vibration measurement
and analysis; infrared thermography; oil analysis and tribology; acoustic emissions
and ultrasonics; and motor current analysis. Books published in the series deal with
root cause analysis, failure and degradation scenarios, proactive and predictive tech-
niques, and many other aspects related to condition monitoring. Applications con-
cern different industrial sectors: automotive engineering, power engineering, civil
engineering, geoengineering, bioengineering, etc. The series publishes monographs,
edited books, and selected conference proceedings, as well as textbooks for ad-
vanced students.
Multiphysics Modelling
and Simulation for Systems
Design and Monitoring
Proceedings of the Multiphysics Modelling
and Simulation for Systems Design
Conference, MMSSD 2014, 17–19 December,
Sousse, Tunisia
ABC
Editors
Mohamed Haddar Tamer Elnady
National School of Engineers Faculty of Engineering
of Sfax ASU Sound and Vibration Lab
Sfax Ain Shams University
Tunisia Cairo
Egypt
Mohamed Slim Abbes
National School of Engineers Andrei Kanaev
of Sfax Laboratory of Materials Sci. & Process
Sfax University of Paris 13
Tunisia Villetaneuse
France
Jean-Yves Choley
Superior Engineering Institute Mounir Ben Amar
of Paris University of Paris 13
Saint-Ouen Laboratory of Materials Sci & Process
France Villetaneuse
France
Taoufik Boukharouba
Laboratory for Advanced Mechanics Fakher Chaari
University of Sciences and Technology National School of Engineers
Houari Boumediene (USTHB) of Sfax
Alger Sfax
Algeria Tunisia
But, what is the future of multiphysics systems? After three days of lively dis-
cussion all the participants concurred on the necessity of increasing the level of
collaboration between researchers and professionals with different expertise. This
synergy is fundamental to deal with the increasing complexity of future systems.
1 Introduction
The mechanical properties of a laminated structure are related to the design pa-
rameters, including the number of layers, thickness, orientation, and the material
of each fold and the stacking sequence of these folds. This is very advantageous
compared to conventional materials because it gives the possibility of designing
structures according to the properties that meet the needs and requirements. Typi-
cal problems of optimal design concern particularly a laminate, the rigidity,
strength, critical load, the natural frequencies and the minimization of thermal ex-
pansion coefficients. However, mechanical properties of a laminate cannot gener-
ally be predicted trivially, even less optimized. Pedersen (1987) showed that in
term of fibers orientation, the composite optimization problems tend to have mul-
tiple local optima which make them difficult to approach by using local optimition
methods based on gradients. Luersen et al. (2004) used an optimization method
called Globalized and Bounded Nelder-Mead. As for Girard (2006), he has used a
genetic algorithm for maximizing the longitudinal rigidity module with con-
straints on the shear modulus and Poisson's ratio. The design variables are the ply
orientations which vary in a continuous or discrete manner.
The purpose of this study is to determine the optimal fibers orientations, for each
ply of a laminated structure, which give the modulus of longitudinal stiffness with
maximum constraints on the shear modulus and Poisson's ratio. The study will use the
Monte Carlo method as a method of stochastic optimization under constraints
2 Modelling
For a laminate consisting of n layers numbered from the bottom to the top, aver-
age surface is chosen as the reference plane. The constitutive equation of a lami-
nated plate expressing the resultants and moments based on the membrane
deformations and bends is given by:
(1)
…. …. ….. …. (2)
The relationships of the off-axis rigidity constants and the main axes for
orthotropic, symmetrical and balanced material can be written in the following forms:
Influence of the Plies Orientation on the Rigidity of the Laminated Composites 3
1 0 0
1 0 0
0 0 1 0 (4)
0 0 0 1
0 0 0
0 0 0
1 0 0
1 0 0
0 0 1 0
(5)
0 0 0 1
0 0 0
0 0 0
3 Method of Resolution
The optimization method used to solve this problem is that of Monte-Carlo, but
with reduction of interval research. The method involves introducing a reduction
factor of the interval research (the best solution found neighbourhood) in which
are carried out random tests. At the beginning of the process, the size of the
neighbourhood is made wide enough but when the number of iterations increases,
there is a narrowing of this neighborhood. This allows well explore the field of re-
search in the initial phase of the calculations. The reduction comes to allow re-
finements of the solution. On the other hand, one introduces a number of successive
chess (E) which will be counted since the last improvement. E will remain small as
long as the process of optimization improves the solution, but when E begins to take
4 A. Belkadi, A.A. Benyahia, and N. Ouali
large values, this means that the process brings no improvement, then, it will be more
profitable to reduce the research interval to allow refinements by introducing a pa-
rameter _ which is a threshold of decision-making. The effectiveness of the re-
duction technique of research interval depends on essentially the wise choice of the
reduction factor of research interval and of the failures threshold parameter.
Consider the example treated by Girard (2006). It is a symmetrical and balanced
laminate with 16 glass/epoxy unidirectional plies (UD). The laminate must be de-
signed to have a modulus of elasticity longitudinal (maximum) while ensuring that
its module of shear ( ) higher or equal to 12 GPa, and that its poisson’s ratio ( )
is smaller than 0,5. The elastic properties of the glass-epoxy UD plies are: E1 = 45
GPa, E2 = 10 GPa, G12 = 4.5 GPa and = 0.31. Since the laminate is symmetrical
and balanced, so there are 4 possible orientations of the UD plies limited between 0°
and 90°. Thus, the optimization problem can be posed as follows:
Maximize the modulus of rigidity by changing the orientation of the 16
plies under the constraints:
- symmetrical and balanced Stacking.
- With : 12 and ν 0.5
The idea of solving this problem is to work in the search space , as shown
in figure 1 and then back to the directions of the plies.
If one takes ξ and ξ as design variables, the field of feasibility is a parable
and each point in this area (point lamination) corresponds to a tensor A. Indeed,
the shear modulus and the poisson’s ratio of the optimal solution were re-
spectively 12 GPa and 0.5, either the lower and upper limits of these constraints.
In this case, a solution violating one of the constraints will probably have a longi-
tudinal Young modulus greater than the optimal solution. The mode of evo-
lution of the Young module based on the change in the orientation of the
plies is illustrated by figure 2. That is why it must penalise enough this solution,
this violates one of the constraints, in order to make it less attractive compared to
solutions that do not violate any constraints. Otherwise, the algorithm will focus
on a no-workable solution and, therefore, will generate a solution violated one or
more constraints.
In the case of this study, a programming our algorithm was made under
MATLAB, where we have considered 0.85 as value of reduction factor, with a
maximum number of chess factor _ 250. The program was rolled out 100
times to ensure the optimum values.
Optimal solutions obtained by the proposed algorithm are given in table 1.
Influence of the Plies Orientation on the Rigidity of the Laminated Composites 5
plies num-
Strafication sequence optimized Ex max.
ber
[ ± 48.0047 ] 12.8259
04 [ ± 48.0048 ] 12.8258
[ ± 48.0056 ] 12.8256
A comparison between the results, obtained for a 16 plies, by our algorithm and
those obtained by Girard are grouped in table 2
6 A. Belkadi, A.A. Benyahia, and N. Ouali
Table 2 Comparison of the results obtained with those reported by Girard [4]
5 Conclusion
In this work a stratified structure optimization technique based on the parameters
of stratification (lamination parameters) is presented, this technique is based on
the decomposition of the parameters that influence the behavior of an elastic lami-
nate invariant parts and dependent parts of the stacking (the parameters of stratifi-
cation). Thus, everything that depends on the material is separated of what is es-
sentially geometric (sequence of stacking and orientations). For this purpose,
the base material is preselected, then the parameters that influence the change in
elastic behavior are precisely the lamination parameters, which are the only para-
meters which describe the design of the structure, because the stacking sequence,
the thickness and the number of layers are already selected. To this end, the Monte
Carlo method is applied as a constrained stochastic optimization, based on stratifi-
cation parameters for the determination of optimal fiber orientations laminate
Influence of the Plies Orientation on the Rigidity of the Laminated Composites 7
References
Pedersen, P.: On sensitivity analysis and optimal design of specially orthotropic laminates.
Engineering Optimization 11, 305–316 (1987)
Vanucci, P.: Méthodes Avancées en Optimisation des Structures, Conception optimale des
composites. Université de Versaille (2007)
Luersen, M.A., Le Riche, R., Guyon, F.: A constrained, globalized, and bounded Nelder-
Mead method for engineering optimization. Thèse de doctorat, Ecole Doctorale SPMI,
pp. 52–59 (Décembre 2004)
Girard, F.: Optimisation de stratifiés en utilisant un algorithme génétique. Thèse de
maîtrise, Université LAVAL, pp. 9–20, 49–69 (2006)
Voyiadjis, G.Z., Kattan, P.I.: Mechanics of Composite Materials with MATLAB. Louisiana
State University, pp. 9–190 (2005)
Irisari, F.-X.: Stratégies de calcul pour l’optimisation multiobjectif des structures compo-
sites. Thèse de doctorat, l’Université Toulouse III (2009)
Allaire, G.: Conception optimale de structures. Springer (2007)
Vannucci, P.: Matériaux composites structuraux: Méthode d’analyse et de conception. Un-
iversité de Versailles et Saint-Quentin-en-Yvelines (2007)
Simplified Calculation Methods on Smoke
and Temperature Stratification in Ventilated
Compartments
1
LCEMSM Laboratory, Hassiba Benbouali University, Po Box 151, UHB Chlef, Algeria
agrsousou@yahoo.fr
2
Mechanical Department, Faculty of Technology, UHB Chlef, Algeria
a.benarous@univ-chlef.dz
3
ISAE-ENSMA, Po Box 40109, Futuroscope Chasseneuil, France
Djamel.karmed@isae.ensma.fr
4
LCEMSM Laboratory, UHB Chlef, Algeria
l.loukarfi@yahoo.fr
Abstract. Using different calculation methods with respect to smoke and heat
evacuation systems (SHEVS), a comparative study is presented with regards to
enclosure fire in large single storey compartments. A focus is made on European
and American manual methods for which a set of graphical Matlab routines are
developed. The results obtained with several modeling approaches are discussed
in this paper.
1 Introduction
A fire can be defined as undesirable burning of materials with release of heat and
toxic gases, causing hazards to people and structures. In this context, it has been
shown that human causalities are much more due to smoke exposure than burns by
flames (Stec et al., 2010). Consequently, some efforts are needed to explore fire
safety engineering issues, more particularly those related with heat and smoke dy-
namics. In this direction, there is a specific need for fire simulation softwares, as
well as predictive tools for the design of smoke and heat evacuating systems
(SHEVS). These devices are supposed to allow for thermal comfort under safe (no
fire) conditions while providing optimal circumstances for survival and egress of
occupants. Design procedures for SHEVS use manual-based methods, which obey
to regionally standardizations and safety regulations.
*
Corresponding author.
2 Calculation Methodology
If this value induces an absolute thermal rise ΔT , the mass flow rate of the
smoke can be expressed by:
Q c
M s = (2)
cΔ T
where c is the specific heat capacity for the air-smoke mixture.
Owing to buoyancy forces, a stratification of the hot layer is created and the
smoke lies at a vertical distance Ys from the floor:
(
Ys = M s / C e Pf )
2/3
(3)
δ s = H ceiling − Ys (4)
As the static pressure is supposed to be constant throughout the hot and cold
layers, the smoke volumetric flow rate can be written as:
where Ts = ΔT + T∞ stands for the static temperature within the smoke layer.
t
In the case of natural ventilation, a set of N NV vents whose total area ANV are
expected in the compartment, so that:
t M s Ts
ANV C v = (6)
M s Ts .T∞
2
2
2 ρ ∞ gδ s T∞ ΔT −
( Ai C i )2
where C v , C i denote loss coefficients of ventilation holes and air openings (gates,
windows) respectively. The cross sectional area of each vent is easily obtained
from the following relation :
n t
ANV = ANV / N NV (7)
⎧1.3 δ s2
M MV = ⎨
crit
g .δ s T∞ ΔT if X VW < Dv
⎩ Ts
(8)
⎧ 2.05 δ s2 ρ ∞
M MV =⎨
crit
g .T∞ ΔT D v if X VW ≥ Dv
⎩ Ts
Dv , X VW are respectively the hydraulic diameter and the distance to the wall of
each device. The number of elements (Fig.3) that should be set to extract the
smoke is given by:
N MV ≥ M s / M MV
crit
(9)
Simplified Calculation Methods on Smoke and Temperature Stratification 13
The same procedure is used for the calculation of the smoke layer thickness,
according to a slight different value for the constant C e = 0.188 .
The total area required for a natural ventilation case is obtained by solving the
nonlinear equation:
2
⎛ At C ⎞
Ts2 +
⎜ NV v ⎟ T T
⎜ ⎟ ∞ s
t M s ⎝ Ai C i ⎠
ANV C v = (11)
ρ∞ 2 g δ s T s ΔT
N MV ≥ Vs / VMV
crit
(13)
14 S. Agred et al.
The first test case concerns a polyvalent hall which dimensions are 95 m length,
66m width and 11m height. The hall features ten gates whose total opening area is
109 m2.
The fire source is taken to be a square 9m×9m pool (category 2) with a total
heat release rate of HRR = 2.02 Megawatts. A natural ventilation configuration is
considered here. Basing on 20% radiative loss, the temperature rise is estimated at
ΔT = 150 °C for both methods, with a corresponding smoke flow rate of
M = 108 kg/s. In this case, the average smoke temperature is almost 170 °C
s
which shows an overestimation of 42% from that of the hot layer, as predicted by
the CFAST solver (Fig.5). The important discrepancy is certainly, due to the fact
that the “manual” formulations do not account for thermal losses to the ceiling
wall.
Furthermore, the CFAST solver makes use of a well-known entrainment model
that accurately predicts the air-smoke mixing (Mc Caffrey, 1983). A model that
justifies the moderate temperature, in the vicinity of the ceiling.
Simplified Calculation Methods on Smoke and Temperature Stratification 15
180
160
European + American method
CFAST calculation
140
Hot layer temperature (C°)
120
100
80
60
40
20
0 50 100 150 200 250 300 350 400 450 500 550
Time (s)
Consequently, the steady state temperature (120°C) of the hot layer seems to be
most revealing of the thermal behaviour. Regarding the smoke stratification, the
European as well as the American methods recover approximately the same value
Ys ≈ 6.3 m against Ys = 8.0 m for the CFAST solver (Fig.6).
Fig. 6 Hot layer height for various calculation methods (natural ventilation)
16 S. Agred et al.
If one assumes that the ventilation components are sufficiently remote from the
walls, a minimum of N MV = 7 ventilators having 1m diameter, are required within
the compartment.
The American method considers the existence of a threshold for the volume
flow rate. The calculated value VMV = 3 m3/s corresponds to a minimum
crit
180
160
European + American method
CFAST calculation
140
Hot layer temperature (C°)
120
100
80
60
40
20
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750
Time (s)
Although the compartment has a large surface where flow of air and smoke
may develop in a three dimensional pattern, the mechanical ventilation contributes
to favour a spreading (blowing) direction. Consequently, the ceiling flow (at least
at the blowing level) can be considered as quasi one-dimensional. The present
configuration certainly enhances the applicability of the zone modelling calcula-
tion while allowing for a reduction of fire simulations cost.
18 S. Agred et al.
4 Conclusion
Different calculation methods have been applied for two test cases of large single
storey compartments. Both European and American method recover almost the
same values for SHEVS characteristics. Indeed, these “manual” procedures con-
tain quasi-identical formulations while basing upon a similar design philosophy. It
has been shown that the American method seems more conservative with respect
to critical volume flow rate, which is a more appropriate yardstick of a ventilator
performance. Mixing of air and smoke in the plume region were accurately quanti-
fied by means of an entrainment sub-model that exists in the zone modelling
solver. Smoke and temperature stratification were predicted fairly good using the
CFAST solver. This was not the case of the manual methods for which a common
drawback is the steady state assumption. Several fire issues can be investigated by
means of zone/field modelling formulations where much pertinent informations
can be provided. Nevertheless, field/zone calculations are much time and resource
consuming and require more skills from the user to deal with realistic applications.
References
CR 12010-5: Guidelines on functional recommendations and calculation methods for
smoke and heat evacuating systems. CEN Technical Report No.191, Brussels, Belgium
(2005)
Klote, J.H., Milke, J.A., Turnbull, P.G., Kashef, A., Ferreira, M.J.: The handbook of smoke
control engineering, 1st edn. American Society of Heating, Refrigerating and Air Condi-
tioning Engineers (ASHRAE), Washington (2012)
Mc Caffrey, B.J.: Momentum implications for buoyant diffusion flames. Combustion and
Flame (1983), doi: 0010-2180(83) 90129-3
Paranthoen, P., Gonzalez, M.: Mixed convection in a ventilated enclosure. International
Journal of Heat and Fluid Flow (2010), doi:10.1016/2010.01.006
Peacock, R.D., Forney, G.P., Reneke, P.A.: Consolidated model for fire growth and smoke
transport. NIST Technical Report No.1026 R1, USA (2011)
Stec, A.A., Hull, T.R.: Fire toxicity, Part 2, 1st edn. Woodhead Publishing, Cambridge
(2010)
Analysis of the Notched Specimens on the
Fracture Behavior by the Volumetric Method
Keywords: Weight function, Notch stress intensity factor, Notch effect, Effective
stress.
1 Introduction
1999). The result of this works and others indicate that this approach gives a good
description in relation with the notch effect.
Under a plane stress mode, two methods are used especially volumetric ap-
proach and ones of Irwin for analysis of the stresses field using FEM methods.
This present study focus on the effect of the variation of dimensional quantities of
the notch on the fracture behaviour of notched specimens by the calculation of
notch stress intensity factor and analysis the stresses field that reigns at the notch
root.
Considering that the material has perfect elastic plastic behavior and having the
following mechanical properties in Table 1.
Irwin’s approach has set a goal: the determination of the stress intensity factor in
mode I, taking into account the presence of a plastic zone that is assumed to be
small compared to the crack length. The resulting value of this factor, KI , will
establish a critical “intrinsic” value corresponding to the elastic–plastic fracture
(Irwin 1964).
Irwin therefore assumes an extension of the fracture of ‘re << a’ and considers
that the stress, obtained from this zone does not exceed the elastic limit, (the law
of perfect elastic–plastic behavior, without extension). It is therefore required to
estimate the value re of the stress applied at a distance (Naman. R 2012)
From the stress field, which is determined in linear elasticity, for θ = 0 and in
mode I, we can write:
Κ Ι = σ π (a + re ) (1)
1 ∂σ yy ( x )
χ( x ) = (2)
σ yy (x ) ∂x
Where Κ Ιρ , σ eff and Xeff are respectively the notch stress intensity factor, the
effective stress, and the effective distance. The effective distance represents the
upper limit of this zone. The latter starts at the inflection point of the curve, which
is given by the minimum of relative stress gradient.
Fig. 6 shows an elastic stress distribution analysis of an open and thin notch hav-
ing semi-elliptical shape with dimension b equal to 0.5 and 4 [mm] respectively, a
notch having a depth equal to 6 [mm]. It shows a comparison of analytically ob-
tained results with those obtained by the finite element method. Various formalisms
representing the stress distribution can be found in literature (Pluvinage1998).
The results are similar near the end of notch where the stress concentration is
high. However at remote distances from notch bottom, i.e. where the stress gradi-
ent is lower, the results obtained by finite element method decrease rapidly
towards lower values.
Analysis of the Notched Specimens on the Fracture Behavior 23
Fig. 4 Evolution of the notch stress intensity factor for thinned and opened semi elliptical
notch
The results obtained are compared with those calculated according to the two
Irwin models (Nui et al 1994, Chaboche1988). Fig. 4, show the notch stress inten-
sity factor evolution versus the ratio b/a, calculated using the two Irwin models,
and the three weight functions: unit, delta and gradient functions and considering
semi-elliptical notch.
24 M. Moussaoui and S. Meziani
The obtained results of the stress intensity factor for the thin semi-elliptical
notch (b = 0.5 [mm]) show , The Irwin corrected model gives converging results
with the results of unit weight function and the results obtained from delta weight
function represented the upper limit and gradient weight function is lower limit
(Fig.4a). At increased value of ration (b/a), the results obtained from Irwin without
corrected model tend to delta function.
Near to the notch root, the results obtained by the three weight functions move
away from those obtained by the Irwin corrected model and are close to those of
Irwin without correction model (Fig.4b).
The increase of the major axis "a" and the reduction of the minor axis "b" lead
to a flattened semi-elliptical notch, corresponding to a deep notch, create a region
where the stress concentration is high, and generate an increased stress gradient as
a result of the increase of the stress intensity factor values.
The thin and deep semi-elliptical notches are much more dangerous than the
short semi-elliptical and semi-circular notches
The effective stress is the average value of the stress distribution over the effective
distance, weighted by the relative stress gradient inside the elaboration volume of
fracture, in fracture elaboration volume is expressed according of a weighting
function, in order to take into account the stress gradient due to geometry and
loading mode (Minor et al 2003, Adib and Jeong.2007) by the following relation:
1 x eff
σ eff =
x eff ∫
0
σ ij ϕ ( x , χ ) dx (4)
Three weight functions are used in this study, the unit weight function is
ϕ( ) ( )
x , χ = 1 , delta weight function is δ x − X eff and gradient is (1− x.χ ) .
Fig. 5 show the evolution of effective stress calculated using the three weight
functions (unit, delta and gradient) with the maximum elastic-plastic stress, ac-
cording to the ratio (b/a), relating to the thinned and opened semi-elliptical
notches. In the case of an opened notch with a lower ratio (b/a), the elastic-plastic
stresses value (Fig. 5b) is greater than the effective stresses value calculated by the
volumetric approach. And for thinned notch where the gradient is strong, the val-
ues of effective stresses calculated using the three weight functions; unit, delta and
gradient are identical (Fig. 5a).
The more the minor axis 'b' of semi-elliptical notch increases the more the
results obtained by the three weight functions converge between them (Fig. 5).
Analysis of the Notched Specimens on the Fracture Behavior 25
For both thin and open notch configurations, all the results obtained using
weight delta function are situated in a superior position and are close to the values
of the maximum elastic-plastic stress. The increase of the minor axis 'b' reduces
the error gap between the results obtained using delta weight function and those
obtained using the maximum elastic-plastic stress.
Fig. 5 Evolution of effective stress & maximum elastic-plastic stress for thinned and
opened semi elliptical notch
7 Conclusion
References
Shi, S.Q., Puls, M.P.: A simple method of estimating the maximum normal stress and plas-
tic zone size at a shallow notch. International Journal of Pressure Vessels and Pip-
ing 64(1), 67–71 (1995)
Pluvinage, G.: Fatigue and fracture emanating from notch; the use of the notch stress inten-
sity factor. Nuclear Engineering and Design 185, 173–184 (1998)
Pluvinage, G.: Fracture and Fatigue Emanating from Stress Concentrators, Université de
Metz, France. Kluwer Academic Publishers (2003)
Nui, L.S., Chehimi, C., Pluvinage, G.: Stress filed near a large blunted tip V-notch and
application of the concept of the critical notch stress intensity factor (NSIF) to the frac-
ture toughness of very brittle materials, Engineering Fracture Mechanics. Perga-
mos 49(3), 325–335 (1994)
Lemaitre, J., Chaboche, J.-L.: Mécanique des matériaux solides, © Bordas, Paris (1988)
Hadj Meliani, M., Matvienko, Y.G., Pluvinage, G.: Corrosion defect assessment on pipes
using limit analysis and notch fracture mechanics. Engineering Failure Analysis 18,
271–283 (2011)
El Minor, H., Kifaini, A., Louah, M., Azari, Z., Pluvinage, G.: Fracture toughness of high
strength steel-using the notch stress intensity factor and volumetric approach. Structural
Safety 25, 35–45 (2003)
Adib-Rammezani, H., Jeong, J.: Advanced volumetric method for fatigue life prediction
using stress gradient effects at notch roots. Computational Materials Science 39,
649–663 (2007)
Pluvinage, G., Azari, Z., Kadi, N., Dlouhy, I., Kozak, V.: Effect of ferritic micro-structure
on local damage zone distance associated with fracture near notch. Theoretical and Ap-
plied Fracture Mechanics 31, 149–156 (1999)
Irwin, G.R.: Structural aspects of brittle fracture. Applied Material Research 3, 65–81
(1964)
Shin, C.S., Man, K.C., Wang, C.M.: A practical method to estimate the stress concentration
of notches. Fatigue 16 (June 1994)
Recho, N.: Fracture Mechanics and Crack Growth. ISTE Ltd and John Wiley & Sons, Inc.
(first published, 2012)
Temperature Effect on Shear Flow and Time
Dependant Modelling of Cutting Oil Emulsion
Laboratory of Rheology, Transport and Treatment of the Complex Fluids, University of Science
and Technology, Mohamed Boudiaf, B.P. 1505, El M’Naour, Oran 31000, Algeria
Hammadi7280@yahoo.fr
1 Introduction
(1)
Where τ0 is the yield stress in Pa, K the consistency index in Pa sn and n is the
flow index.
The fitted parameters of the model of Herschel-Bulkley as a function of tem-
perature are presented in Figs 2 and 3. It appears that yield stress and consistency
index decreases with increasing temperature but the flow index increases.
12
11 20°C
40°C
10 60°C
80°C
9 100°C
8 Model
Shear stress (Pa)
7
0,5
6 20°C
0,4
40°C
5 60°C
Viscosity (Pa.s)
0,3
80°C
4 0,2
100°C
3 0,1
2 0,0
0 50 100 150 200
0
0 50 100 150 200 250 3 00 350
-1
She ar rate (s )
3,5
3,0
2,5
Yield stress (Pa)
2,0
1,5
1,0
20 30 40 50 60 70 80 90 100 110
Temperature (°C)
Fig. 2 Variation of yield stress as a function of temperature (60% of cutting oil emul-
sion)
30 L. Hammadi, N-E. Boudjenane, and M. Belhadri
0,85 0,40
0,80
0,75
0,35
0,70 n
0,65
Temperature (°C)
(2)
0,070
0,065 Experimental
Arrhenius equation
0,060
Limit viscosity (Pa.s)
0,055
0,050
0,045
0,040
0,035
(1/T)*10 3 (1/°K)
Fig. 4 Limit viscosity as a function of 1/T at volume concentration 60% of cutting oil emul-
sion. The solid lines correspond to the curve fitting to Eq.2.
0 ,09 5
0 ,09 0
0 ,08 5
. -1
0 ,08 0 γ = 50 S
. -1
γ = 100 S
0 ,07 5 . -1
γ = 150 S
Viscosity (Pa.s)
0 ,07 0 .
γ = 200 S
-1
. -1
0 ,06 5 γ = 250 S
Equation (10)
0 ,06 0
0 ,05 5
0 ,05 0
0 ,04 5
0 ,04 0
0 ,03 5
0 ,03 0
0 20 40 60 80 100 120 140
Fig. 5 Viscosity of cutting oil emulsion as a function of time shears at 20°C for constant
shear. The solid line corresponds to curve fitting to eq.10.
For their rate equation Tiu and Boger (1974) employed a second-order kinetic
equation:
for (4)
model of Herschel-Bulkley and the apparent yield stress value obtained by extra-
polation to 0. The resulting equation was:
.
0.547 0.6 , 0.994 (5)
10
60% Vol
9
Herschel-Bulkley model
4
50 100 1 50 200 25 0
-1
Shear ra te (s )
Fig. 6 Zero time shear stress as a function of shear rate for 60% of cutting oil emulsion and
comparison with the Hershel Bulkley model (temperature 20°C)
The structural parameters and are expressed in terms of the apparent vis-
cosity in order to find k2. Equating the rheological model of Eq. (3) to the
definition of the apparent viscosity which, in this case, is a function of both shear
and time, it is found:
(6)
, (7)
Where
(8)
, ,
Integrating equation (7) at constant shear rate from at 0
yields:
(9)
The equilibrium viscosity was found by fitting the time dependence of the
viscosity (Fig. 5) by an exponential decay curve (Hammadi et al. 2014) of the
form:
+ exp (10)
0,09
0,06
0,05
0,04
0,03
50 1 00 150 200 25 0
-1
shear rate (s )
Fig. 7 Initial viscosity and equilibrium viscosity determined by equation (10) as a function
of shear rate (60% of cutting oil emulsion)
Using Eq. (9), a plot of versus t’, at constant shear rate, should yield a
straight line with slope equal to . The same procedure is repeated at different
shear rate. Fig.8 shows the data for 60% cutting oil emulsion.
10000
-1
250 s
-1
200 s
8000 -1
150 s
-1
100 s
1/(η-ηe) [1/(Pa.s)]
-1
6000 50 s
Fitting
4000
2000
0 5 10 15 20 25 30
Time(s)
Fig. 8 Linear relationship between and time for different shear rates (60% cutting oil
emulsion) at temperature 20 °C. The solid line corresponds to a fitting of data by linear
regression.
Fig.9 shows the evolution of the parameters a1 and k2 as a function of shear rate
for 60 % of cutting oil emulsion. It was found that both a1 and k2 can be described
by a power function of shear rate [7]:
(11)
(12)
With
1.567 0.334; 0.961 0.0390, 0.996
0.965 0.190; 0.460 0.039, 0.982
The value of the structural parameter was determined at each shear rate by
using Eq. (13).
(13)
34 L. Hammadi, N-E. Boudjenane, and M. Belhadri
Where are the initial viscosity and the equilibrium viscosity deter-
mined by equation (10).
350
12
K2
300
Fitting
10
250
8
a1 [1/(Pa.s2)]
K2 (s )
-1
200
6
150
4
100
a1
2
50
0
50 100 150 2 00 250
-1
shear rate (s )
Fig. 9 Parameters a1 and k2 of Tiu and Boger model as a function of shear rate for 60 % cut-
ting oil emulsion
(15)
0,9 95
Equilibrium structural λ e (-)
0,9 90
0,9 85
60% Vol
Herschel-Bulkley model
0,9 80
0,9 75
50 1 00 150 200 25 0
-1
Shear rate (s )
Fig. 10 Equilibrium structural parameter versus shear rate at 20°C together with Herschel-
Bulkley model (60% of cutting oil emulsion)
Temperature Effect on Shear Flow and Time Dependant Modelling 35
1,0 00
. -1
γ =250 s
0,9 95 . -1
γ =200 s
. -1
γ =100 s
0,9 85
0,9 80 . -1
γ =50 s
0,9 75
0 20 40 60 80 100 12 0 140
Time (s)
Fig. 11 Structural parameter as a function of time at 20° (60% of cutting oil emulsion)
4 Conclusion
In this study the effect of temperature on the rheological behavior of cutting oil
emulsion and the effect of shear rate on structure buildup and breakdown of cut-
ting oil emulsion have been performed. The model of Herschel-Bulkley, whose
parameters are temperature dependent, was found to be well correlated with the
behavior of cutting oil emulsion. With increased temperature, the yield stress and
the consistency index decreased, while the flow index increased. The limit viscosi-
ty of the cutting oil emulsion decreased with increasing temperature, and their
relationship could be modeled by the Arrhenius equation.
The time dependant viscosity decreased rapidly with shearing time and reached
an equilibrium stage. In order to take into account the structural evolution of the
cutting oil emulsion, the model of Herschel-Bulkley was modified by the introduc-
tion of a structural parameter which allows a good description of the thixotropic
behavior of the cutting oil emulsion.
References
Bensadok, K., Benammar, S., Lapicque, F., Nezzal, G.: Electrocoagulation of cutting oil
emulsions using aluminium plate electrodes. Journal of Hazardous Materials 152,
423–430 (2008)
Butler, F., McNulty, P.: Time dependent rheological characterization of buttermilk at 5°C.
Journal of Food Engineering 25, 559–580 (1995)
Cheng, D.H., Evans, F.: Phenomenological characterization of the rheological be-havoiur
of inelastic reversible thixotropic and antithixotropic fluids. Br. J. Appl. Phys. 16,
1599–1617 (1965)
Dullaert, K., Mewis, J.: Thixotropy: Build-up and breakdown curves during flow. Journal
of Rheology 49, 1213–1230 (2005)
Hammadi, L., Ponton, A., Belhadri, M.: Temperature effect on shear flow and thixo-tropic
behavior of residual sludge from wastewater treatment plant. Mechanics of Time-
Dependent Materials 17, 401–412 (2013)
Hammadi, L., Boudjnane, N., Belhadri, M.: Effect of polyethylene oxide (PEO) and shear
rate on rheological properties of bentonite clay. Applied Clay Science 99, 306–311
(2014)
36 L. Hammadi, N-E. Boudjenane, and M. Belhadri
Liu, J., Han, R., Zhang, L., Guo, H.: Study on lubricating characteristic and tool wear with
water vapor as coolant and lubricant in green cutting. Wear 262, 442–452 (2007)
Moore, F.: The rheology of ceramic and slip bodies. Trans. Br. Ceram. Soc. 58, 470–494
(1959)
Mujumdar, A., Beris, A.N., Metzner, A.B.: Transient phenomena in thixotropic systems.
Journal of Non-Newtonian Fluid Mech. 102, 157–178 (2002)
Pottirayil, A., Kailas, S.V., Biswas, S.K.: Lubricity of an oil in water emulsion in metal cut-
ting: The effect of hydrophilic/lypophilic balance of emulsifiers. Colloids and Surfaces
A: Physicochem. Eng. Aspects 384, 323–330 (2011)
Stachowiak, G.W., Batchelor, A.W.: Physical Properties of Lubricants. Engineering Tribol-
ogy, 3rd edn., pp. 11–50 (2006)
Tiu, C., Boger, D.V.: Complete rheological characterisation of time dependent food prod-
ucts. J. Texture Studies 5, 329–338 (1974)
A Comparison between Two Reliability Based
Optimization Strategies of Tuned Mass Damper
Parameters under Stochastic Loading
1
Unité de Recherche en Dynamique des Structures, Modélisation et Ingénierie des Systèmes
Multi-physiques, Preparatory Engineering Institute of Nabeul (IPEIN), 8000 M’rezgua,
Nabeul, Tunisia
2
Laboratory of applied mechanics and engineering, ENIT. Mechanical engineering department,
École Nationale Supérieure des Ingénieurs de Tunis, University of Tunis, 5, Av.
Taha Hussein Montfleury 1008 Bab Menara TUNIS Tunisia
3
LTDS UMR5513 Ecole Centrale de Lyon, 36 Avenue Guy de Collongue,
69134 Ecully, France
1 Introduction
The use of the TMD devices in engineering to mitigate vibrations in structures sub-
ject to stochastic loading has taken wide interests in last years. One of the main ques-
tions asked to the designer is how to choose the optimum parameters of the TMD so
that a certain criterion is met. In the literature (Yu et al. 2013) and (Marano et al.
2010), several strategies can be used such as the RBO (Chakraborty and Roy 2011)
where the objective function is a failure probability related to a certain failure mode.
A failure mode can be the exceedence, for the first time, of the displacement response
(Chakraborty and Roy 2011) or the Von Mises stress (Gupta and Manohar 2005)
through a certain given thresholds during a time interval.
mT
TMD
kT cT yT
ms
Primary
ks cs ys
structure
yb
The system is subjected to a base acceleration yb due to seismic motion. Let
y(t ) = ( yT , ys )T be the relative displacement vector of the TMD device and the
A Comparison between Two Reliability Based Optimization Strategies 39
primary structure, respectively, the governing equation (Greco and Marano 2013)
of the dynamic system can be expressed as follows:
⎛ mT 0 ⎞
where the overdot represents the time derivative, M = ⎜⎜ ⎟,
⎝ 0 ms ⎟⎠
⎛ cT − cT ⎞ ⎛ k − kT ⎞
C = ⎜⎜ ⎟ , K = ⎜⎜ T ⎟ and r = [1 1]T .
⎝ − cT cT + c s ⎟⎠ ⎝ − kT kT + k s ⎟⎠
⎛ 0 0 1 0 ⎞
⎜ ⎟
⎜ 0 0 0 1 ⎟
A0 = ⎜ ⎟
− ωT2 ωT2 − 2ξT ωT 2ξT ωT
⎜ ⎟
⎜ μω 2 − ( μωT2 + ωs2 ) 2μξT ωT − (2μξT ωT + 2ξ sω s ) ⎟⎠
⎝ T
where w(t ) is a stationary zero mean white noise process applied at the bed rock.
It represents the excitation of the dynamic system and it has intensity S 0 . ξ f , ω f
and y f are the damping ratio, the natural frequency and the relative response
(with respect to the ground) of the elastic filter (Chakraborty and Roy 2011). Con-
sidering the Kanai-Tajimi model, the global space state vector of the mechanical
( )
system is given by Z = yT , y s , y f , yT , y s , y f T and the equation of motion can be
40 E. Mrabet et al.
⎛ 0 0 0 1 0 0 ⎞
⎜ ⎟
⎜ 0 0 0 0 1 0 ⎟
⎜ 0 0 0 0 0 1 ⎟
A=⎜ ⎟
⎜ − ωT 2 ωT 2 ωf 2 − 2ξT ωT 2ξ T ωT 2ξ f ω f ⎟
⎜ ⎟
⎜ μωT − ( μωT + ω s ) ωf 2 2μξ T ωT − 2(ξ sω s + μξT ωT ) 2ξ f ω f ⎟
2 2 2
⎜ 0 0 −ωf
2
0 0 − 2ξ f ω f ⎟⎠
⎝
AR + RA T + B = 0 (3)
where B has all zero elements, except the last one [B]( 6, 6 ) = 2πS0 . The root mean
square displacement (rmsd) of the primary system can be obtained as
σ y = R (2,2) . The root mean square velocity (rmsv) is expressed as
s
σ y = R (5,5) . In stationary condition, the DPO in the primary structure and its
s
Hence, the root mean square σ y of the relative acceleration ys can be imme-
s
In the present work, two optimization strategies are considered in order to get op-
timum design parameters of the TMD. The first strategy is the classical approach
for the RBO problems (Chakraborty and Roy 2011) where the failure probability
is related to the primary structure displacement ys , representing the performance
quantity. This RBO strategy consists to find the optimum design parameters
d = (ωT , ξT ) of the TMD device that minimizes the failure probability Pf , D cha-
racterized by the exceedence, for the first time, of the displacement ys of a certain
threshold value β during some interval time [0, T]. The second strategy is a new
RBO problem. It's related to the DPO in the primary structure e(t ) . This RBO
problem consists to find d that minimizes a failure probability Pf , E characterized
by the up-crossing, for the first time, of e(t ) across a certain threshold value
β E during some interval time [0, T]. Since the system described in Fig.1 is subject
to stochastic loading, the evaluation of both quantities Pf , D and Pf , E constitute a
time dependant problem of reliability analysis that can be transformed into time
invariant problem (Gupta and Manohar 2005) by introducing the random variables
y s , max = max y s (t ) for the displacement approach and emax = max e(t ) for the
0≤ t ≤T 0≤ t ≤T
DPO approach, respectively. Considering the Poisson assumptions (Li and Chen
2009) and adopting the classical Rice's formula, the failure probabilities can be
approximated by:
(
Pf , D ≈ 1 − exp − υ β ( β )T ) (
and Pf , E ≈ 1 − exp − υ β ( β E )TE
) (5)
and the DPO processes, respectively. These mean out-crossing rates can be
obtained using the Rice's formula as follows:
∞ ∞
υ β ( β ) = ∫ y s . pY Y (β , y s )dy s
s s
and υ β ( β E ) = ∫ e. p EE (β E , e)de
E
(6)
0 0
where pY Y and pEE are the joint probabilities functions (PDF) of the considered
s s
processes and their time derivatives. The main difficulty when handling reliability
analysis using the Rice's formula resides in the computation of mean out-crossing
rates which are function of the joint PDF that cannot be always easy to obtain ana-
lytically (Gupta and Manohar 2005).
42 E. Mrabet et al.
s s
(
by υ β = (σ y πσ y ) exp − β 2 2σ y2 s
) and the failure probability is expressed as
{ (
Pf , D = 1 − exp − T (σ y s πσ ys ) exp − β 2 2σ y2s )} . Hence, the RBO problem with
displacement approach can be formulated as follows:
because both processes E and E , describing the DPO in the primary structure
and its time derivative, are non-Gaussian processes.
After manipulations, the mean out-crossing rate can be written as
E s s
{ s
}
υ β ( β E ) = (σ y πσ y ) exp − β E 4ξ sω s msσ y2 and the failure probability is im-
mediately given by:
{ s s
(
Pf , E = 1 − exp − T (σ y πσ y ) exp − β E 4ξ sω s msσ y2
s
)} (8)
4 Numerical Study
The purpose of the numerical study is to analysis the correlation between the pro-
posed RBO problems. The correlation analysis allows establishing relationships
and possible equivalences between the proposed optimization approaches. To this
end, a dimensionless quantity ψ = ω f ω s is introduced to make comparison be-
tween the obtained results of the proposed strategies. The structural parameters are
assumed to be ms = 1 , ω s = 8π , ξ f = 0.4 , ξ s = 0.03 , S 0 = 0.03 , μ = 2% ,
A Comparison between Two Reliability Based Optimization Strategies 43
β = 0.045 , β E = 2.6 and T = 10s . For the optimization, the genetic algorithm
routine available in Matlab is used.
Figure 2 gives the optimum design parameters of TMD when both approaches
are applied. From Fig. 2(a) one can observe that, for both strategies, the optimum
natural frequencies are earthquake dependant for ψ < 2 . Indeed, the optimized
natural frequencies are increasing functions until roughly ψ ≈ 1.5 and then
26
25.8
25.6
ωT(rd/s)
25.4
25.2
25
( a)
24.8
0.5 1 1.5 2 2.5 3 3.5
ψ
Displacement approach Diss. power approach
0.074
(b)
0.072
ξT
0.07
0.068
0.5 1 1.5 2 2.5 3 3.5
ψ
Fig. 2 Optimum natural frequencies (a) and damping ratios (b), obtained using the
displacement and the DPO approaches
Figure 2 (b) shows that the optimized damping ratios obtained using the
proposed strategies are also earthquake dependent for ψ < 2 and they become in-
dependent for ψ > 2 . In effect, the optimized damping ratios are roughly linear
decreasing function until approximately ψ ≈ 1 where a minimum is reached, then
increasing function until ψ ≈ 2 . In addition, the observed curve slope of the ob-
tained results shown in Fig. 2(b) for ψ < 1 also prove that the TMD performances
are strongly dependant on earthquake. The optimized failure probabilities Pf*, E ob-
tained using the DPO approach and the failure probabilities Pf , E evaluated using
the optimum values when the displacement approach is applied as well as their
44 E. Mrabet et al.
corresponding relative error are shown in Fig.3. The obtained results shown in
Fig. 3(a) confirm the remarks made previously in Fig.2. Indeed, the obtained val-
ues of Pf*, E present large variability vs. ψ and the TMD performance is strongly
earthquake dependant especially for the values neighboring ψ ≈ 1.2 where a max-
imum is reached. It is also observed that the optimized failure probabilities are less
sensitive, to earthquake, for ψ > 2 . Besides, one can see that the curve corres-
ponding to the optimized values are below the curve corresponding to the failure
probability evaluated with the optimum parameters when the displacement ap-
proach is applied.
Figure 3(b) shows the relative error between the optimized failure probabilities
obtained when the DPO is applied and the failure probabilities evaluated using the
optimum TMD parameters obtained when the displacement approach is applied.
The examination of this figure demonstrates that a minimum is reached for
ψ ≈ 1.25 with an error of 3.5%. Besides it has been also found that the maximum
error can attain 9.5% for ψ ≈ 0.6 and it is roughly constant for ψ > 2 .
-3
x 10
1.2
1
Failure probability
0.4
( a)
0.2
0
0.5 1 1.5 2 2.5 3 3.5
ψ
10
Relative error (%)
8
(b)
6
2
0.5 1 1.5 2 2.5 3 3.5
ψ
Fig. 3 (a) Pf*, E obtained using the DPO approach compared with those evaluated using the
optimum values of the displacement approach; (b) the relative error.
Figure 4(a) shows the optimized failure probabilities Pf*, D when the displace-
ment approach is applied compared with the failure probabilities Pf , D evaluated
using the optimum TMD parameters if the DPO approach is applied. The same
observations made previously for Fig. 3(a) can be done for Fig. 4(a). Indeed the
A Comparison between Two Reliability Based Optimization Strategies 45
-3
x 10
8
2
( a)
0
0.5 1 1.5 2 2.5 3 3.5
ψ
6
Relative error (%)
5
(b)
4
3
0.5 1 1.5 2 2.5 3 3.5
ψ
Fig. 4 (a) Pf*, D obtained using the displacement approach compared with those evaluated
using the optimum values obtained with the DPO approach; (b) the relative error.
5 Concluding Remarks
In this paper two RBO problem of TMD parameters under stochastic loading are
presented and compared. The selection of one of these strategies totally depends
on the designer aim. The first strategy is based on a displacement approach so that
the optimal TMD parameters are obtained by minimizing the failure probability
46 E. Mrabet et al.
explicitly related to the covariance responses of the primary structure. The second
RBO strategy is a new approach in RBO problems. The approach uses the DPO in
the primary structure. The introduction of this strategy is motivated by the strictly
relationship between the DPO in the primary structure and the damage level of the
vibrating structure.
The expression of the failure probability, for the DPO approach, is presented
and a comparison between the presented RBO strategies is made. The obtained re-
sults show that both strategies are strongly earthquake dependent on values of
earthquake frequencies close to main system frequency and they are less sensitive
outside of this frequency.
The correlation analysis made between the presented strategies showed that
the RBO with DPO approach is more robust in the sense that it can provide the
optimized TMD parameters if the RBO displacement is applied with a maximum
relative error less than 6%. Besides, both RBO strategies are equivalent, from ro-
bustness point of view, for frequencies close to the main system one because the
relative error is roughly the same.
References
Yu, H., Gillot, F., Ichchou, M.N.: Reliability based robust design optimization for tuned
mass damper in passive vibration control of deterministic/uncertain structures. J. Sound
Vib. 332, 2222–2238 (2013)
Marano, G.C., Greco, R., Sgobba, S.: A comparison between different robust optimum de-
sign approaches: Application to tuned mass dampers. Probab. Eng. Mech. 25, 108–118
(2010)
Chakraborty, S., Roy, B.K.: Reliability based optimum design of tuned mass damper in
seismic vibration control of structures with bounded uncertain parameters. Probab. Eng.
Mech. 26, 215–221 (2011)
Gupta, S., Manohar, C.S.: Probability Distribution of Extremes of Von Mises Stress in ran-
domly vibrating structures. J. Vib. Acoust. 127, 547–555 (2005)
Crandall, S.H.: First crossing probabilities of the linear oscillator. J. Sound Vib. 12(3),
285–299 (1970)
Li, J., Chen, J.: Stochastic dynamics of structures. John Wiley, Asia (2009)
Greco, R., Marano, G.C.: Optimum design of Tuned Mass Dampers by displacement and
energy perspectives. Soil Dyn. Earthq. Eng. 49, 243–253 (2013)
Modeling and Simulation for Vertical Rail
Vehicle Dynamic Vibration with Comfort
Evaluation
1
Laboratoire de Génie Mécanique(LGM),
Ecole Nationale d’Ingénieurs de Monastir, Université de Monastir, Tunisie
2
Laboratoire de Mécanique de Sousse (LMS),
Ecole Nationale d’Ingénieurs de Sousse, Université de Sousse, Tunisie
3
On leave at the College of Engineering, The American University of Sharjah, UAE
Keywords: Sperling index, ride quality, rail vehicle, dynamic behavior, ride com-
fort, vertical dynamic vibration.
1 Introduction
Rail transport is one of the major modes of transportation, so it must offer a high
comfort level for passengers and crew. However, the comfort that passengers ex-
perience is usually perceived differently from one individual to another. In several
research works, noise and vibration have been identified as the most important
factors for high comfort. The main sources of vibration in a train are track defects
in welding or rolling defects, rail joints, etc. The nature of vibration itself is ran-
dom and covers a wide frequency range [1]. The improvement of the passenger
comfort while travelling has been the subject of intense interest for many train
manufacturers, researchers and companies all over the world. Although new tech-
niques in manufacturing and design ensure better ride quality in railway carriages,
it is sometimes impossible to completely eliminate track defects or various ground
irregularities. The dynamic behavior of a train also depends on the load and and its
mechanical components, such as springs, dampers, etc., which interact with the
wheels, the train body and bogies. The dynamic performance of a rail road vehicle
as related to safety and comfort is evaluated in terms of specific performance in-
dices. The quantitative measurement of the ride quality is one of such performance
indices. Ride quality is interpreted as the capability of the rail road suspension to
maintain the motion within the range of human comfort. Sperling’s ride index is a
measure of the ride quality and ride comfort used by ISO 2631[2]. Due to the
complex dynamics that exists between the rail and wheel, rail vehicle dynamics
are often difficult to model accurately. This velocity-dependent dynamics justifies
the importance of the track input to railcar modeling. In the physical system, the
input comes from the actual track. In a model, a user-defined input is used to pre-
dict the actual track characteristics. The user-defined input can be created analyti-
cally or can be based on actual measurements. Measured track data are obtained
by running a specialized railcar down the track. Analytic track data are created us-
ing mathematical shapes, such as cusps, bends, and harmonic functions, to
represent the track geometry [3].
There have been several studies, which dealt with the dynamic analysis of rail
vehicles in order to enhance the ride comfort while travelling. Nejlaoui et al [4]
optimized the structural design of passive suspensions in order to ensure simulta-
neously passenger safety and comfort. Abood et al [5] investigated the Railway
carriage simulation model to study the influence of vertical secondary suspension
stiffness on ride comfort of railway car body. Kumar and Sujata [2] presented the
numerical simulation of the vertical dynamic behavior of a railway vehicle and
calculated Sperling ride index for comfort evaluation. Nielsen and Igeland [6] in-
vestigated the vertical dynamic behavior for a railway bogie moving on a rail
which is discretely supported by sleepers resting on an elastic foundation. Effects
of imperfections on the running surfaces of wheel and rail were studied by assign-
ing irregularity functions to these surfaces.
2.1 Assumptions
The assumptions made in formulating the model are as follows:
Some parameters regarding the rigid bodies are already given in Table 1; how-
ever, the other parameters, which are essential for the simulation of the vehicle,
are presented in Table 2.
2Lb
2b2
2Ld Y
h3
Scondary
Ks Cs h2 Suspension
h1 Kp Cp Primary
R Suspension
Z
2b1 K Hertizian
y(lateral) 2a
(pitching)
Z
Speed V
x (logitudinal)
(rolling)
(yawing)
z(vertical)
A typical rail road vehicle system is composed of various components such as car
body, springs, dampers, Bogies, Wheel-set, and so forth. When such dynamic sys-
tems are put together from these components, one must interconnect rotating and
translating inertial elements with axial and rotational springs and dampers, and al-
so appropriately account for the kinematics of the system structure.
X
H
The bump excitations of the left wheels (Figure. 3) of leading bogies are as
follows:
⎧H ⎡ ⎛ V ⎞⎤
⎢1 − cos ⎜⎝ 2π L ( t − tdi ) ⎟⎠⎥
L
⎪ for tdi ≤ t ≤ + tdi
Z ri = ⎨ 2 ⎣ ⎦ V (i = 1..4) (1)
⎪0
⎩ otherwise
⎡ 2 Ld 2 Lb 2 Lb + 2 Ld ⎤
Where [t d 1 , t d 2 , td 3 , td 4 ] = 0, (2)
⎢⎣ V , V , V ⎥⎦
In the present study, H is taken as 0.03m and L is taken as 1m.
4 Simulation Study
The models were built in the MATLAB/ Simulink® environment. The fixed step
solver ODE-45 (Dormand-Prince) was utilized, with the sampling time
Ts=0.0001. Ts is smaller than the fastest half-car Active Vehicle Suspension Sys-
tems (AVSS) model dynamics, enabling observation of all model dynamics [7, 8].
52 M. Graa et al.
Dynamic analysis was carried out for the vehicle at different speeds: 15m/s,
30m/s, 45m/s and 60m/s.
-3
0.03 2.5
x 10
Zr1 15 m/s²
Bum p e xcita tion (m )
Zr2 30 m/s²
0.025 2 45 m/s²
Zr3 60 m/s²
0.5
0.01
0
0.005
-0.5
0
0 0.2 0.4 0.6 0.8 1
-1
Time(s) 0 0.5 1 1.5 2 2.5
Time(s)
Fig. 3 The bump excitations of the left Fig. 4 The vertical Car body displacement
wheels vehicle speed of 15m/s for differents vehicle speeds
Vertical acceleration at the front and the rear bogie center pivot.
The displacement and acceleration responses of the carbody at speeds of 15m/s,
30m/s, 45m/s, 60m/s are shown, respectively, in figure.4 and figure.5. Plots show
that initially, the value of acceleration is nearly equal to 0 m/s2, which is mainly
the acceleration without gravity. Finally it goes to zero, when the vibration of the
car body ceases and stabilizes. The acceleration is generally within an acceptable
range and does not show any instability.
The acceleration response of front and the rear bogie with time is presented at
different velocities of vehicle in figure.6.a) and figure.6.b), respectively. It is clear
from the plots that initially the wheels of the front bogies come in contact with the
track irregularity and the vibration starts in the front bogie and later these vibra-
tions are shifted to the rear bogies. The amplitude of the vehicle vibration also
increased with the vehicle speed.
Where ai denotes the amplitude of the peak acceleration response (in cm/s²)
measured on the floor of the itch frequency identified by the FFT and B(fi) a
weighting factor, given by:
1.911 f 2 + ( 0.25 f 2 )
2
B( f ) = k (5)
(1 − 0.277 f 2 )2 + (1.563 f − 0.0368 f 3 )2
Where k = 0.737 for horizontal vibration and 0.588 for vertical vibration.
1000
1 15 m/s² 15 m/s²
30 m/s² 30 m/s²
45 m/s² 45 m/s²
Car body acceleration (cm/s²)
60 m/s² 60 m/s²
Car body Acceleration (m/s²)
0.5 500
0 0
-0.5
-500 0 1 2 3
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 10 10 10 10
Time(s) Frequency (Hz)
(a) (b)
Fig. 5 The vertical Car body acceleration in time domain (a) and frequency domain (b) for
vehicle speed of 15m/s, 30m/s, 45m/s and 60m/s
54 M. Graa et al.
Table 3 Sperling’s ride comfort index an evaluation for different vehicle velocities
200 200
15 m/s²
15 m/s² 150 30 m/s²
150
30 m/s² 45 m/s²
Front bogie acceleration (m/s²)
50 50
0 0
-50 -50
-100 -100
-150 -150
-200 -200
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.5 1 1.5
Time(s) Time(s)
(a) (b)
Fig. 6 The vertical Front (a) and Rear (b) bogie acceleration for vehicle speed of 15m/s,
30m/s, 45m/s and 60m/s
The maximum and minimum ISO Sperling Index values are respectively 2.53
and 1.62 for the rail vehicle speed respectively 15 m/s and 60 m/s. These values
respectively indicate “The more pronounced but not unpleasant” and “just notice-
able” zones. Figure 7 show that the comfort index decreases as the vehicle speed
increases while maintaining an acceptable level of comfort. This means that the
passengers are not much affected by the vibration as they are exposed to low level
of vibrations.
2.6
2.2
2
Clearly noticeable
1.8
1.6
10 20 30 40 50 60
Speed(m/s)
2.8 2.8
C arb o d y Sp erlin g In d ex W z
2.6 2.6
Carbody Sperling Index Wz
1.4 1.4
0 500 1000 1500 2000 2500 1.65 1.7 1.75 1.8 1.85 1.9
Primary stiffness Kp (N/mm) Secondary stiffness Ks (N/mm) x 10
4
a) b)
2.6 3
Carbody Sperling index Wz 2.4
1 0.5
500 600 700 800 900 1000 6 8 10 12 14 16 18 20 22
Primary damping Cp (Ns/mm) Secondary damping (Ns/mm)
a) b)
Fig. 9 The influence of Primary a) and secondary b) damping
5 Conclusion
Vertical dynamic analysis has been carried out for a Railway Vehicle. A 17 degree
of freedom model is used for the analysis. Velocity input at all the wheelset is giv-
en by considering similar bump irregularities at both right and left rail. A vertical
acceleration response at the car body has been calculated in the frequency domain.
Sperling Ride index has been calculated for the above vehicle. The Sperling Ride
Index values at different speeds are presented. The calculated values of the Sperl-
ing index are found well in the satisfactory limits defined by the ISO 2631 stan-
dard which means that the passengers are not much affected by the vibration as
they are exposed to low level of vibrations. A parametric study has been carried
out with emphasis on better ride index. The parametric study has brought out poss-
ible design changes required in different parameters to deliver better Sperling
comfort index. It should be noticed that the parametric study was carried out to
suggest design modifications to improve Wz index, but others dynamic and con-
trol behaviors like stability, control of secondary suspension were to be considered
when implementing the design modifications.
References
[1] Skarlatos, D., et al.: Railway fault diagnosis using a fuzzy logic method. Applied
Acoustic 65(10), 951–966 (2004)
[2] Kumar, H., Sujata: Vertical dynamic analysis of a typical indian rail road vehicle. In:
Proceedings of Computational Mechanics and Simulation, IIT, India, pp. 8–10 (2006)
[3] Dukkipati, V., Amyot, J.: Computer Aided Simulation in Railway Dynamics. Marcel
Dekker, New York (1988)
[4] Nejlaoui, M., et al.: Multiobjective robust design optimization of rail vehicle moving
in short radius curved tracks based on the safety and comfort criteria. Simulation
Modelling Practice and Theory 30, 21–34 (2013)
Modeling and Simulation for Vertical Rail Vehicle Dynamic Vibration 57
[5] Abood, K.H.A., Khan, R.A.: The Railway carriage simulation model to study the in-
fluence of vertical secondary suspension stiffness on ride comfort of railway carboy.
Journal of Mechanical Engineering Science 225, 1349–1359 (2011)
[6] Nielsen, J., Igeland, A.: Vertical dynamic interaction between train and track-
influence of wheel and track imperfections. Journal of Sound and Vibration 187,
25–26 (1995)
[7] Chen, H., et al.: Application of Constrained H∞ Control to Active Suspension Sys-
tems on Half-Car Models. Journal of Dynamic Systems, Measurement and Control,
Transactions of the ASME 127(3), 345–354 (2005)
[8] Pedro, J., Dahunsi, O.: Neural Network-based Feedback Linearization Control of a
Servo-Hydraulic Vehicle Suspension System. International Journal of Applied Ma-
thematics and Computer Science 21(1), 137–147 (2011)
[9] Forstberg, J.: Ride comfort and motion sickness in tilting trains (2000) - Human res-
ponses to motion environments in train and simulator experiments. Master’s Thesis,
KTH/ FKT/D-00/28-SE, Division of Railway Technology, KTH (2000)
[10] Carg, V.K.: Dynamics of Railway Vehicle Systems. Academic Press, NY (1984)
[11] B153/RP21 Application of ISO Standard to Railway Vehicles: Comfort Index Nmv
Comparison with the ISO/SNCF Comfort Note and with the Wz (European Rail Re-
search Institute) (1993)
Localization of Impact on a Beam by Time
Reversal Method
1
Laboratory for Advanced Mechanics (LMA), USTHB, BP32 El Alia,
Bab Ezzouar, Algiers, Algeria
ochaterbache@usthb.dz
2
Laboratory for Advanced Mechanics (LMA), USTHB, BP32 El Alia,
Bab Ezzouar, Algiers, Algeria
amiloudi@usthb.dz
Abstract. This paper deals with a numerical simulation of time reversal (TR)
process applied to a cantilever beam. To conduct this process we assume a finite
difference scheme based on Euler-Bernoulli theory of transversal vibration of the
beam. The TR process is then used to localize the position of an impact on
the beam. To complete the TR process we need one or more measurements of the
field on different positions. We had study the effect of the measurements number
and their positions on the localization with the TR process. The objective of this
study is to reduce the number of measurements and hopefully use it in an experi-
ment with few measuring instruments.
1 Introduction
The time reversal idea is based on the time invariance of theory formulation of a
physical phenomenon (Sachs, 1987). Robert G. Sachs (Sachs, 1987) has explained
in his book ”The Physics of Time Reversal” that the time invariance concept is
purely theoretical and only applicable to a physical governing ”equations”. Also
the time invariance can’t have a physical sense only if it’s tested experimentally
(Sachs, 1987). Mathias Fink was interested on ultrasonic field time invariance
(Fink et al., 1989) and he used it to ultrasound field focusing on an object through
an inhomogeneous (Fink, 1992). First he supposes the existence of substance act-
ing like an acoustic source when it receives an ultrasonic field. So, the reflected
pressure from that is measured at different positions and then remitted in the in-
verted sense (i.e. time reversed) at the same measurement positions simultaneous-
ly. This last step should focus all the pressure on the body (Fink, 1992). Our work
was inspired from the study by G. Ribay (Ribay, 2006) who compares the time
2 Numerical Model
⎧ ∂ 2 w( x, t )
⎪⎪M x = − E I ∂x 2
⎨ (1)
⎪ ρ A ∂ w( x, t ) = ∂ Mx( x, t ) + δ ( x − x ) δ (t )
2 2
⎪⎩ ∂t 2 ∂x 2
0
The system (1) contains only second order derivatives. It will be discretized
with fourth order finite difference in space and second order in time. The discrete-
zation detail is reported to the appendix.
The impact force is of the form shown in figure 1. As needed in our work we
construct the function (2) of the impact force to be as close as possible to a real
impact (caused by a hammer impactor for example (Fig. 2)) :
2 ⎛ −8t ⎞
⎛ 4 t ⎞ ⎜⎝ τ ⎠
⎟
δ (t ) = e 2 ⎜ ⎟ e (2)
⎝τ ⎠
τ is the impact duration.
Localization of Impact on a Beam by Time Reversal Method 61
The figure 3 shows the response to an impact at the middle of the beam in
terms of transverse displacement. The impact duration was τ = 0.2 ms
* represent the convolution product. Equation (3) shows that the time reversal
gives a maximum at the initial impact position.
4 Impact Localization
The trace of the responses’ maxima after retransmission versus the position allows
us to determine the initial impact position. First we define a set of measurement
positions as the “measuring network” with m positions on the beam and an “exci-
tation network” with n positions. For both defined networks the positions are
Localization of Impact on a Beam by Time Reversal Method 63
λ EI
d= =4 πτ (3)
2 ρA
The measuring network (Fig. 5) was formed by a measuring position every 0.01
m (i.e. m = 100) which can be considered as an ideal case for localization. But for
further experiments consideration it was in our interest to reduce the number of
measuring points. (See sections 5 and 6)
Figure 6 shows the emergence of other maxima; parasites. Given the nature of
the problem, the impact spreads in both directions giving rise to two types of
waves; progressive and evanescent. When the measurement is made on either
sides of the impact position (second case with two measurement points, for exam-
ple), the parasites are less apparent. In the case of a single measurement point the
information on the impact position is not lost but the parasites are more important.
Here we are interested on the third case with only one measuring point. We study
the influence of the measuring position from the initial impact position; left or
right side, and the effect of the single measuring position itself.
Figure 7 shows two cases for a single measuring point; left (x = 0.1 m) and
right (x = 0.8 m) of the impact position (x0 = 0.2 m). In both cases we distinguish
the position of impact. The maxima parasites probably are related to the
interaction between the wave caused by the impact and boundary conditions. The
reflections from the boundaries act as virtual sources.
Using a single measuring point doesn’t affect the time reversal localization but
we need to select carefully its position. In the case of transverse waves in a beam
and because of the reciprocity principle (hij = hji) the impact localization fails if the
measurement is made at x = 0.5 m (the middle of the beam). Figure 8 shows that
for this particular point localization is completely disrupted; two maxima of the
same level have emerged.
7 Conclusion
This study allowed us to analyze the time reversal process for the impact localiza-
tion on a cantilever beam. We demonstrate the effect of the number of measure-
ment points as well as the choice of their positions and allow us to capitalize on
the efficiency of time reversal. The time reversal method shows its robustness for
impact localization even with partial information about the behavior of the beam
(one or two measuring points). The ideal case is to measure the response of the
beam along its entire length which needs advanced experimental technique of
measuring. So by reducing the number of measurement points to one we can con-
sider an experiment with a single accelerometer.
66 O. Chaterbache and A. Miloudi
References
Ernst, R., Dual, J.: Acoustic emission localization in beams based on time reversed disper-
sion. Ultrasonics 54(6), 1522–1533 (2014)
Fink, M.: Time reversal of ultrasonic fields. i. basic principles. IEEE Transactions on Ultra-
sonics, Ferroelectrics and Frequency Control (1992)
Fink, M., Prada, C., Wu, F., Cassereau, D.: Self focusing in inhomogeneous media with
time reversal acoustic mirrors. In: Proceedings of the IEEE Ultrasonics Symposium
(1989)
Francoeur, D., Berry, A.: Time reversal of flexural waves in a beam at audible frequency.
The Journal of the Acoustical Society of America 124(2), 1006–1017 (2008)
Graff, K.: Wave Motion in Elastic Solids. Dover Publications (1975)
Prada, C., Manneville, S., Spoliansky, D., Fink, M.: Decomposition of the time reversal op-
erator: Detection and selective focusing on two scatterers. The Journal of the Acoustical
Society of America 99(4), 2067–2076 (1996)
Ribay, G.: Localisation de source en milieu réverbérant par Retournement Temporel. PhD
thesis (2006)
Sachs, R.: The Physics of Time Reversal. University of Chicago Press (1987)
Appendix
Numeric Scheme
Let’s differentiate the system of equations (1) using a centered scheme with a
constant time step Δt and space step Δx . We write then in second order and using
the notation wi = w(k Δt , i Δx ) :
k
⎪ 12 Δx 2
⎪ k
⎩+ f i
Localization of Impact on a Beam by Time Reversal Method 67
Stability Criteria
The stability condition for this process is determined using the Fourier method by
k
replacing wi by ξ k e I kx i Δx . The condition ξ <1 should be checked by the
process (8) to make it do not diverge over time. This imposes a condition on the
choice of the time step according to the space step:
3 ρA 2
Δt < Δx (9)
8 EI
Boundary Conditions
The boundary conditions we imposed are Fixed - Free and it results:
Fixed: zero deflection and zero rotation.
⎧w(0, t ) = 0
⎪
⎨ ∂w (10)
⎪⎩ ∂x (0, t ) = 0
⎧w0k = 0
⎪ k
⎨ w+1 − w−k1 (11)
⎪ =0
⎩ Δx
The second equation from (11), w−k1 = w+k1 , states that the fixed boundary condi-
tion is a line of symmetry between w+k1 , which is a point from the beam, and a
fictive point w−k1 . So, we can extend this property to the fictive point w−k 2 by
writing:
⎧ ∂2w
⎪⎪Mx(L, t ) = 2 (L, t ) = 0
∂x
⎨ (13)
⎪Qx(L, t ) = ∂Mx (L, t ) = ∂ w (L, t ) = 0
3
⎪⎩ ∂x ∂x 3
Or;
⎧MxNk = 0
⎪ k
⎨ MxN +1 − MxNk −1 (14)
⎪ =0
⎩ Δx
For this condition we differentiate the second derivative with a second order
scheme instead of fourth order. So we get only one fictive point from the second
equation of (14)
The process (8) is used for all points i = 0..N to evaluate the momentum Mx
and the deflection w. For the boundaries we use the second order instead of the
fourth to evaluate the deflection at instant k+1. The bending momentum is always
calculated at the instant k.
Early Detection of Gear Failure by Vibration
Analysis
1
Advanced Mechanics Laboratory, USTHB, Algiers, Algeria
moustapha_merzoug@yahoo.fr, amiloudi@usthb.dz
2
Advanced Mechanics Laboratory, Champagne Ardène University, Reims, France
{khalid.ait-sghir,fabrice.bolaers,jp.dron}@univ-reims.fr
1 Introduction
Gears are commonly used in power transmission achines for many applications,
such as industrial machinery and automotive. Health monitoring of rotating ma-
chines is important to avoid failure of the system in advance. A detailed examina-
tion of models published in 1988 by Ozguven (Ozguven 1988) allowed classified
these models into several groups.
• Simple dynamic factor models. These models include most of the early studies
in which dynamic factors that can be used in gear root stress formula are de-
termined,
• Models with tooth compliance. These models include the tooth stiffness as the
only potential energy storing element in the system,
• Global models of gear transmissions.
These models are currently heavily used, and can include shaft torsional vibra-
tion, shaft bending stiffness, gear tooth bending stiffness, bearings stiffness. In the
defect presence, the mesh stiffness variations give important information about the
transmission. It also reflects the damage severity and can be obtained by calcula-
ting the variation of the tooth mesh stiffness so that preventive action can be envi-
saged. Weber [2], Cornell [3] computed the gear mesh stiffness analytically, and
Kasuba [4] calculated it with a digitization approach. Yang [5] used the potential
energy principle to calculate the total mesh stiffness of meshing gear pairs as a
function of the rotation angle of the gear. Chaari [6] developed an analytical mo-
del to calculate the gear mesh stiffness based on the Weber's tooth bending equa-
tion. In addition, some researchers attempted to extend the gear mesh stiffness
model to involve the presence of the some gear faults, such as tooth breakage and
spalling, gear tooth crack. A comprehensive review of the methods of signal pro-
cessing published in 2006 can be found in [7]. Time-domain analysis is based on
the time waveform itself. Traditional time-domain analysis calculates characteris-
tic features from time waveform signals as descriptive statistics such as mean,
peak, peak-to-peak interval, standard deviation, crest factor, high-order statistics:
root mean square, skewness, kurtosis, etc. These features are usually called time-
domain features, which have proved to be effective in fault diagnosis and are now
well established [8] and [9]. The advantage of frequency domain analysis over
time domain analysis is its ability to easily identify and isolate certain frequency
components of interest. The most widely used conventional analysis is the spec-
trum analysis by means of fast Fourier transform (FFT). One limitation of fre-
quency-domain analysis is its inability to handle non-stationary waveform signals,
which are very common when machinery faults occur. Thus, time-frequency ana-
lysis, which investigates waveform signals in both time and frequency domain,
has been developed for non-stationary waveform signals. Traditional time-
frequency analysis uses time-frequency distributions, which represent the energy
or power of waveform signals in two-dimensional functions of both time and fre-
quency to better reveal fault patterns for more accurate diagnosis. Short Time Fou-
rier Transform (STFT) or spectrogram (the power of STFT) [10] and Wigner-Ville
distribution) [11,12] are the most popular time–frequency distributions. The time
frequency representation of non stationary signals was represented using the new
type of wavelet transform, in which discrete wavelet transform was used to get de-
tails and approximations information about the signal to identity the fault [13,14].
The discretes wavelets transform was also used for energy evaluation of known
bandwidth with time scale analysis and acoustic emission signals in detection of
faults in gearbox [15,16]. The gear fault identification based on Hilbert-Hung
transform was proposed to overcome the problem of reconstructing a feature ma-
trix of singular value decomposition [7]. Djebala [17] proposed an approach that
benefits the local nature of wavelet analysis and its ability to detect transients.
They first calculated the Hilbert transform of the signal then calculate an envelope
of the signal will be processed by wavelet analysis. It then extracts the detail cove-
ring the rotation frequency of the defective wheel. Parey [18] using Empirical
Mode Decomposition (EMD) used as a processing tool for non-stationary and
nonlinear signals. In their work, they expose its operation from a real signal. They
exploit this method to detect an early stage the presence of a fault using global vi-
bratory indicators that are based on statistical analysis of a time signal such that
Early Detection of Gear Failure by Vibration Analysis 71
the kurtosis and the crest factor. Indicators are sometimes limited and may not
give a reliable verdict on the state of the gear. Random noise and other machine
components pollute the signal and make it difficult or impossible to detect in some
cases. One solution is to measure signals in narrow bands at high frequencies with
a band pass filter [19]. A second approach is to use modern tools noising vibratory
signals developed from the wavelet transform [20,21], spectral subtraction [22] or
adaptive filtering [23]. A comparison of the three methods was applied to the si-
gnals of defective bearings [24]. Indeed, the use of these methods allows to denoi-
sed and filtered signals whose kurtosis is more significant. This will help detect
any defects that the detection was not possible in the raw signals or broadband.
2 Numerical Simulation
The analytical model used in the present paper is available in literature and the
corresponding graphical representation is shown in fig. 1.
The equation of motion is rearranged to the form shown in the Eq. 1. The dy-
namic response is simulated by numerically solving the second order term (accele-
rations) for each step of incremented time. The effect of the mesh stiffness varia-
tion is implemented in the model by updating its value for each time increment.
∆
1 cos (1)
(3)
and
2
2 (6)
2 (7)
2
where,
: teeth stiffness (mean), 2 peak-to-peak teeth stiffness, : contact ratio
and : phase angle
Gear tooth crack is a defect that has severe impact on healthy machine functio-
ning. Detecting this defect as soon as possible has an extreme importance. In tooth
root crack modeling, most of the studies consider some modifications in meshing
stiffness profile [26]. For instance, faulty tooth stiffness decreases and it is repea-
ted once per contact cycle of this tooth. Therefore, the authors proposed to model
a localized crack on a tooth taking into account its angular position on the base
circle. On this position, meshing stiffness decrease is considered and this pheno-
menon is repeated at each time the cracked tooth enters in contact with the teeth of
the opposed gear. In this case, the condition that governs the stiffness expression
in the presence of a localized fault on a tooth is as follows:
• If the contact is between two healthy teeth, the meshing stiffness is given by
.
• If the contact is between the cracked tooth and the tooth of the opposed gear,
the stiffness is given by 1 .
where is a periodic function representing the amount of stiffness degradation
expressed as:
Early Detection of Gear Failure by Vibration Analysis 73
sin (8)
1
· cos Ω
where Ω 2 , 1,2, and represents the rotation frequency of the gear af-
fected by the crack.
is the amplitude of the function , it represents the ratio stiffness degrada-
tion or the crack evolution.
represents the cyclic ratio.
Figure 2 shows the kurtosis variation according to the default severity taken for
different values of in the expression of the mesh stiffness.
The results obtained show that it is difficult to detect the defect at birth by the
kurtosis (for a previously fixed level). An analysis of this indicator shows that the
fault only appears when the degradation ratio reaches a value of 3,18 %. It is the-
refore necessary to decompose the signal into several approximations using the
wavelet transform.
Fig. 4 Kurtosis comparison between original signal and the first two approximations
A comparison between the values of Kurtosis before and after the signal de-
composition (Fig. 4) shows the contribution of wavelet transforms in improving
the sensitivity of these indicators with respect to the conventional case.
3 Experimental Validation
The vibratory signals on which we made the calculation come from a gearbox (test
bench) installed at the CETIM (France). The dimensioning of the gear wheels of
the tempered hardened, and the operating conditions (speed, torque) were set so as
to obtain a peeling over the entire length of a tooth. [27,28]. Every day after the
acquisition of vibration signals, the bench is stopped, to appraise the state of the
teeth of wheels. Testing of reducing spread over 12 days with registration of a vi-
bratory signal daily. The gear consists of a pinion 20 and teeth of a tooth wheel
21. The speed of the drive motor is 1000 or approximately 16.67 , the
rotational speed of the wheel is 15.87 and the meshing frequency is about
Early Detection of Gear Failure by Vibration Analysis 775
Fig. 5 Kurtosis evolution off the of Fig. 6 Evolution of cepstral peaks A1 andd
experimental signal [18] A2, their sum and the indicator vs. the dayys
acquisition. [29]
El-Badaoui [29] found d that the peak of the cepstral peak of the wheel whicch
develops the defect (A1) increases
i at the expense of the other. The sum of the firrst
peak close to 0,5, tells theem that the processed signals are not very noisy. The evoo-
lution of the indicator clearly shows that from the 8th day, the gear develops a dee-
fect. This result is consisstent with the observation made on the expertise of thhe
gearbox, which indicates that the 8th day of occurrence of chipping on the toop
tooth 15/16 (Fig. 6).
the fault on the 8th day as specified above. To demonstrate the performance of this
technique in the diagnosis of gear systems compared with those of the method of
synchronous cepstrum and the expert report was completed.
3.1 Results
A preliminary analysis of the temporal signals allowed us to see the occurrence of
a fault very marked the 12th day, it is not possible to early detect the occurrence of
a fault with the times signal (Fig. 8).
4 Conclusion
The approach just described can trace the origin of some defects after decomposi-
tion of the original signal details and approximations because a preliminary analy-
sis of temporal signals allowed us to visualize the occurrence of a defect very
marked in the 12th day. Therefore it’s not possible to early detect how the defect at
the time signals. Also, we hope that all the results on the monitoring of gears, seals
those obtained with respect to other machines, the engineer will make a better
diagnosis and that in order to improve the transmission quality.
References
1. Ozguven, H.N., Houser, D.R.: Mathematical models used in gear dynamics a review.
Journal of Sound and Vibrations 121, 383–411 (1988)
2. Weber, C.: The deformation of loaded gears and the effect on their load carrying ca-
pacity. Report Number 3, Spons. Res. (Germany). British Dept. of Sci. and Ind. Res.
(1949)
3. Cornell, R.W.: Compliance and stress sensitivity of spur gear teeth. Journal of Me-
chanical Design, Transaction of the ASME 103, 447–459 (1981)
4. Kasuba, R., Evans, J.W.: An extended model for determining dynamic loads in spur
gearing. Journal of Mechanical Design, Transaction of the ASME 103, 398–409
(1981)
5. Yang, D.C.H., Lin, J.Y.: Hertzian damping, tooth friction and bending elasticity in
gear impact dynamics. Journal of Mechanical Design, Transaction of the
ASME 109(2), 189–196 (1987)
6. Chaari, F., Fakhfakh, T., Haddar, M.: Analytical modelling of spur gear tooth crack
and influence on gearmesh stiffness. Europ. Jou. of Mech. A/Solids 28, 461–468
(2009)
7. Jardine, K.S., Lin, D., Banjevic, D.: A review on machinery diagnostics and prognos-
tics implementing condition-based maintenance. Journal of Mechanical Systems and
Signal Processing 28, 1483–1510 (2006)
8. McFadden, P.D., Smith, J.D.: A signal processing technique for detecting local defects
in a gear from the signal average of the vibration. Proceedings of the Institute of Me-
chanical Engineers 28, 287–292 (1985)
78 M. Merzoug et al.
9. Randall, R.B.: A new method of modelling gear faults. Journal of Mechanical De-
sign 104, 259–267 (1982)
10. Wang, W.J., McFadden, P.D.: Early detection of gear failure by vibration analysis I.
Calculation of the time–frequency distribution. J. Mec. Design 7, 193–203 (1993)
11. Meng, Q., Qu, L.: Early detection of gear failure by vibration analysis I. Calculation of
the time–frequency distribution. J. of Mec. Sys. and Sig. Proc. 5, 155–166 (1991)
12. Pan, M.C., Van Brussel, H., Sas, P., Verbeure, B.: Fault diagnosis of joint backlash.
Journal of Vibration and Acoustics. Transactions of the ASME 120, 13–24 (1998)
13. Peng, Z.K., Chu, F.L.: Application of the wavelet transform In machine condition
monitoring and fault diagnostics: a review with bibliography. Journal of Mechanical
Systems and Signal Processing 18, 199–221 (2004)
14. Douglas, H., Pillay, P., Ziarani, A.K.: A new algorithm for transient motor current sig-
nature analysis using wavelets. IEEE Trans. on Ind. App. 40, 1361–1368 (2004)
15. Kia, S.H., Henao, H., Capolino, G.: A Diagnosis of broken-bar fault in induction ma-
chines using discrete wavelet transform without slip estimation. IEEE Trans. on Indus-
trial Applications 45, 1395–1403 (2009)
16. Gu, D., Kim, J., An, Y., Choi, B.: Detection of faults in gearboxes using acoustic emis-
sion signal. Journal of Mechanical Science and Technology 45, 1279–1286 (2009)
17. Djebala, A., Ouleaa, A., Kebbabsa, T., Ramdane, Y.: Détection des défauts non sta-
tionnaires de dentures d’engrenages par application de l’analyse en ondelettes et la
transformée de Hilbert. In: Congrés Algérien de Mécanique, Mascara, Algeria, No-
vember 27-29 (2013)
18. Parey, A., El Badaoui, M., Guillet, F., Tandon, N.: Dynamic modelling of spur gear
pair and application of empirical mode decomposition-based statistical analysis for
early detection of localized tooth defect. Journal of Sound and Vibrations 294,
547–561 (2006)
19. Pachaud, C., Salvetat, R., Fray, C.: Crest factor and kurtosis contributions to identify
defects inducing periodical impulsive forces. Journal of Mechanical Systems and Sig-
nal Processing 11, 903–916 (1997)
20. Donoho, D.L.: De-noising by soft-thresholding. IEEE Transactions on Information
Theory 41, 613–627 (1995)
21. Ling, J., Qu, L.: Feature extraction based on morlet wavelet and its application for me-
chanical fault diagnosis. Journal of Sound and Vibrations 234, 613–627 (2000)
22. Dron, J.P., Bolaers, F., Rasolofondraibe, L.: Improvement of the sensibility of scalars
indicators thanks to de-noising method by spectacle subtraction. Application to the de-
tection ball bearing defects. Journal of Sound and Vibrations 270, 61–73 (2004)
23. Lin, J., Zuo, M.J.: Gearbox fault diagnosis using adaptive wavelet filter. Journal of
Mechanical Systems and Signal Processing 17, 1259–1269 (2003)
24. Dron, J.P., Rasolofondraibe, L., Chiementin, X., Bolaers, F.: A comparative experi-
mental study on the use of three denoising methods for bearing defect detection. Jour-
nal Meccanica 45, 265–277 (2010)
25. Lin, J., Parker, R.G.: Mesh stiffness variation instabilities in two-stage gear system.
Journal of Vibration and Acoustics 124, 68–76 (2002)
26. Abbes, M.S., Trigui, M., Chaari, F., Fakhfakh, T., Haddar, M.: Dynamic behaviour
modelling of a flexible gear system by the elastic foundation theory in presence of de-
fects. Mechanics A/Solids 29, 887–896 (2010)
27. Drouiche, K., Menad, S., Grenier, Y.: Détection de défauts d’engrenages par analyse
vibratoire. Journal du Traitement du Signal 8, 333–343 (1991)
Early Detection of Gear Failure by Vibration Analysis 79
28. Laspi: Documents du Laboratoire d’Analyse des Signaux et des Processus Industriels,
Université Jean Monnet, St Etienne, France, http://www.laspi.fr
29. El-Badaoui, M.: Contribution au diagnostic vibratoire des réducteur complexes à en-
grenages par l’analyse cepstrale. Thèse de doc., Université Jean Monnet St Etienne
(1999)
30. Haloui, N., Chikouche, D., Benidir, M.: Diagnosis of gear systems by spectral analysis
of vibration signals. Int. J. of Comp. Sci. and Network Security 7, 285–293 (2007)
31. Breneur, C.: Eléments de maintenance préventive de machines tournantes dans le cas
de défauts combinés d’engrenages et de roulements. Thèse de doc. INSA de Lyon
(2002)
Effect of Water Hammer on Relatively Long
Inclined Pipelines
1
Laboratoire de Mécanique des Fluides Appliqués, Génie des Procédés et Environnement,
Ecole Nationale d’Ingénieurs de Sfax, BP 1173, 3038, Sfax, Tunisie
manel.dalleli@gmail.com, ezed.hadj@enis.rnu.tn
2
Laboratoire de Biomécanique, Polymères et Structures (LaBPS),
Ecole Nationale d’Ingénieurs de Metz, 57070 Metz, France
{Schmitt,azari}@enim.fr
1 Introduction
The analysis of transient flows in relatively long pipelines is rather complex be-
cause the pipeline friction head losses are large compared to the instantaneous
head rise caused by a sudden variation of the flow velocity that is by the water
hammer phenomenon.
In this paper, transient flows are accurately analysed to examine if, under unfa-
vourable circumstances damage due to water hammer may occur in relatively long
pipelines.
A computing procedure based on the method of characteristics is developed.
Transient flow is provoked by an instantaneous valve closure at the downstream
end of a relatively long pipeline supplied by a reservoir at a constant level. In this
case, the gravity lift may have an important effect on the maximum pressures,
2 Mathematical Model
The equations which describe transient one-dimensional flow in a cylindrical pipe,
of linear elastic behaviour according to the Hooke law, can be adapted from the
analytical model developed by (Wylie et al. 1993). Applying the law of conserva-
tion of mass and momentum yields to the following equations of continuity and
motion:
2
∂H Q ⎛ ∂H ⎞ C ∂Q
+ ⎜ − sin γ ⎟ + =0 (1)
∂t S ⎝ ∂x ⎠ gS ∂x
∂Q Q ∂Q ⎛ ∂H ⎞
+ + gS ⎜ +J⎟=0 (2)
∂t S ∂x ⎝ ∂x ⎠
where Q is the fluid discharge, H is the head, C is the pressure waves celerity,
( )
J = (λ D)Q Q 2gS 2 : is the linear head loss by unit of length of the pipe, λ is
the coefficient of friction, S is the cross section area of the pipe, γ is the pipe
slope, D is the circle pipe diameter inside, t is the time and x is the distance along
the pipe.
gS g ⎛Q ⎞
dQ ± dH ∓ Q sin γ + gSJdt = 0 and dx = ⎜ ± C ⎟dt (3)
C C ⎝S ⎠
the + is for the waves coming from the upstream while the – is for the waves com-
ing from the downstream.
Equations (3) determine the evolution of the head pressure and the discharge
according to the time and the space. They are much appropriated to be solved nu-
merically on a microcomputer. The obtained solution constitutes a solution to the
original system of the equations (1) and (2).
Effect of Water Hammer in Relatively Long Inclined Pipelines 83
Δx
P
t+Δt
C+ C-
Δt
t
i-1 R i T i+1
After the first time step, the extreme points of the system begin to influence the
calculation of the interior points. Thus, in order to get the solution until an ele-
vated time level, it is necessary to introduce the appropriate boundary conditions.
At either end of a single pipe only one of the characteristic equations (4) or (5) is
available in the two unknowns Q P and H P . An auxiliary equation is therefore
necessary in every case to specify Q P or H P .
4 Numerical Results
An example is presented illustrating the accurate solution of the water hammer phe-
nomenon in a relatively long pipe. A computer program was developed using the
FORTRAN language. The transient flow is produced in a long cast iron pipe which
data are taken from the textbook of (Wylie and Streeter, 1978) (see figure 2):
L = 201 km, D = 0.762 m, e = 10.2 mm, λ = 0.018, H0 = 426.72m ,
sinγ = 0.002 , V0 =1.2954m s , Q 0 = 0 . 59075 m 3 s and C = 1005 m s .
H0
L
Pipeline
Valve
Initial conditions
The initial conditions at t=0 s are given by the initial steady-state flow. They are
defined, for i = 1,2, , N + 1 , by: Q i = Q 0 = 0 .59075 m 3 s and
H i = H 0 − λ (i − 1)Δ xQ 02 (2 gDS 2 )
Boundary conditions
The transient flow depends on the conditions imposed at the ends of the pipe. In
this study, the pipe is supplied, at the upstream end, by a reservoir at fixed level,
so HP1 = Hresevoir(t ) = 426,72m , ∀t ≥ 0 . At the downstream end, the pipe is
equipped with a valve closed suddenly, so QPN+1 = Qvalve(t) = 0 , ∀t > 0 .
Effect of Water Hammer in Relatively Long Inclined Pipelines 85
Results
By using the method of characteristics (MOC) for the time step (increment)
( )
Δt = L / N * C (N = 4000), one can determine the changes in pressure at dif-
ferent sections of the pipe and then the maximum stresses can be readily calcu-
lated via thin-walled hollow cylinder assumptions. In figure 3, the pressures are
plotted at some equidistant sections of the pipe. The maximum pressures and
stresses at these sections are presented in table 1.
50
p (Bar)
x =0
40 x = L/8
x = L/4
30 x = 3L/8
x = L/2
x = 5L/8
20
x = 3L/4
x = 7L/8
10 x =L
0
0 2 4 6 8 10 t*C/L 12
5 Security Analysis
The risk of failure of the long pipeline in presence of a notch-type defect, under
the effect of water hammer, can now be enjoyed by the value of the safety factor.
This failure may occur in case of the presence of a critical defect, of size a/e = 0.5,
subjected to critical stress corresponding to the breaking strength of the pipe mate-
rial or tenacity (Pluvinage , 2003) .
The break condition can be written:
f (K r , L r ) = 0 (8)
K app
Kr = (9)
K Ic
L r is the dimensionless applied stress defined as the quotient between the
maximum applied stress σ max and the flow stress Rc :
σ max
Lr = (10)
Rc
Any loading condition in the structure containing a defect, leading to a non-
critical situation, is described by the calculated point in plan ( L r , K r ).
The coordinates of this point are ( Lr * , K r * ). The failure begins or occurs
when the coordinates of this point increases until it reaches the critical values
( Lrc , Krc).
All critical operating points leading to failure are located on a curve given out
by the code SINTAP level 1 (SINTAP, 1999) and defined by the relationship:
− 0 .5
⎛
⎜
L2
f (l rc ) = ⎜ 1 + rc
2
⎞
⎟
⎟
[0.3 + 0.7 exp (− 0.6 L )]
6
rc
0 ≤ Lrc ≤ Lr max (11)
⎝ ⎠
2.5
⎡150 ⎤
where Lr max = 1 + ⎢ ⎥ .
⎣σY ⎦
As illustrated in Figure 4, the Failure Assessment Diagram sets the safety factor
defined by the report OC on OA for an operating point A. If this point is inside the
boundary lines of the diagram which is limited by the interpolation curve, the
structure is safe. If not, the assessment point is situated outside of the interpolation
curve and failure may occur (Rouabeh et al. 2012).
In this study, the pipes are considered as ductile cast iron having the mechani-
cal properties shown in Table 2.
Effect of Water Hammer in Relatively Long Inclined Pipelines 87
Breaking pipes is initiated from a semi elliptical defect, for instance a corrosion
crater. The facture toughness is determined by a local failure criterion, and not by
a conventional criterion of fracture mechanics which assumes that the defect is a
crack (Pluvinage, 2001). This criterion assumes that the fracture process requires a
certain level of development of the failure process, Figure 5. This volume is as-
sumed cylindrical of diameter equal to the effective distance X eff . Inside this vo-
lume weighted mean stress prevails, equal to the effective opening stress.
The elastic–plastic stress and gradient constraints are shown in this diagram.
Three specific areas can be distinguished (Bouaziz et al. 2014):
Zone I: The elastic–plastic stress notch root increases to the maximum stress
σ Max. Note that the maximum stress is not at the bottom notch but at a dis-
tance X Max.
Zone II: The elastic–plastic stress decreases until the effective distance
X eff which corresponds to the minimum of stress gradient.
Zone III: The evolution of the elastic–plastic strain has a linear behavior in the
log–log graph.
This stress distribution, before failure, is described by the equation:
Kρ
σ yy = (12)
(2πr )α
where σ yy is the opening stress, K ρ the notch stress intensity factor, r the dis-
tance and α the exponent of the power stress distribution.
The effective distance is defined as the boundary between zones II and III
(figure 5).
The toughness is then defined as equal to the critical notch stress intensity
c
factor Kρ :
( )
Kρc = 2πX eff α σ eff (13)
For brittle materials such as cast iron and a failure to low notch radius, tough-
C
ness Kρ is equated to toughness KIc.
88 M. Dalleli et al.
Log(σ yy (r))
I II III
X eff Log(r)
Geometrical defect
Fig. 5 Typical stress distribution around Fig. 6 Mesh density around the semi-
notch tip in bi-logarithmic diagram elliptical defect
The effective stress and the effective distance are determined by the volumetric
method. Then using equations (9) and (10), the coordinates of evaluation points
( k r and L r ) are determined, see Table 3
Figure 7 presents the Failure Assessment Diagram using the SINTAP code for
a long pipeline. All the assessment points are given on the diagram for different
positions of the defect along the pipe for defects place on different positions along
the pipe.
Effect of Water Hammer in Relatively Long Inclined Pipelines 89
Examining Figure 7, it appears that the points placed after the point X = L / 4
are in a critical situation (located in the failure zone) while the other points lie in
the Zone of security. In fact, the greatest value of safety factor obtained is equal to
7.49 which is the point of coordinate X = 0 (the lowest pressure).
6 Conclusion
In this study, pressure wave propagation in a long pipeline system was investi-
gated numerically by the method of the characteristics. These waves were caused
by the sudden closing of a valve at the downstream end of an inclined long pipe. A
mathematical model, represented by a system of two partial differential equations,
taking into account friction and gravity forces, was developed. The mathematical
equations were solved in the time domain by the method of characteristics using
linear integration.
The results indicate that the gravity action has an important effect in the predic-
tion of the maximum pressures, which are very important near the valve and may
provoke the failure of the pipe especially in presence of defect. Maximum pres-
sures are calculated to see if the chosen pipe-wall thickness is reasonable to resist
these pressures and thus to avoid the pipeline failure. In this study the safety factor
has been computed at equidistant sections of the pipe. It determines the distance
between the reservoir and the defect from which the failure may happen. A simu-
lation of a grey cast iron pipe with a surface defect was made to analyze the ef-
fects of the change in pressure. An example of application in respect of a long pipe
was proposed. The criticality of a semi-elliptical defect in the cast iron pipes and
the depth half the thickness was studied using the finite element method pro-
grammed on the ABAQUS software. The results drawn in the diagram DIR, based
on the code SINTAP level 1, show that all points within the three quarters of the
downstream conduit are located outside the field of integrity. It can be concluded
that the studied system does not work in the field of security and we must
90 M. Dalleli et al.
References
Abstract. The aim of this paper is to study experimentally by particle image velo-
cimetry system (PIV) the effect of the pumping direction introduced with a 60°
pitched blade turbine. In fact, an Up- and a down-pitched blades turbine were
used. The particle image velocimetry technique is equipped with a double pulsed
Nd:YAG laser, a CCD camera resolution and a mini-synchronizer. Therefore, sev-
eral results were carried out to investigate the hydrodynamic structure and the
energetic parameters in a stirred vessel.
1 Introduction
used laser doppler velocimetry (LDV) data to validate results of the single phase
turbulent flow in a stirred tank generated by a down- and an up-pumping pitched
blade turbine. Gabriele et al. (2009) used angle resolved particle image velocime-
try to compare the turbulent flow characteristics generated with a 45° up- and
down-pumping pitched blade. Driss et al. (2010) developed specific computational
fluid dynamics (CFD) code to analyze numerically the turbulent flow field in a
stirred tank equipped with the pitched blade turbines. Driss et al. (2011) used spe-
cific computational fluid dynamics code to compare the laminar hydrodynamic
flow generated by mono and twin screw profile with simple and modified profiles.
Ben Amira et al. (2011) used particle image velocimetry to analyze the Reynolds
number effect in a cylindrical stirred tank generated with a six flat blade turbine.
Feng et al. (2012) used an explicit algebraic stress model (EASM) to simulate ani-
sotropic turbulent flows in baffled stirred tanks equipped with a standard Rushton
turbine.
This work is an attempt to enhance the studies of the pitched blades turbine to
improve the mixing in a stirred vessel instead the 45° turbine. PBTU 60° and the
PBTD 60° are proposed according to these previous studies and as a continuation
of a previous works.
2 Experimental Setup
The schematic of the experimental setup is shown in figure 1. Obviously, the ex-
periments were carried out in a cylindrical stirred vessel. The diameter of the
stirred tank is equal to T = 0.3 m. Four baffles with width of 0.03 m are equipped
equally-spaced at the wall. The liquid height for all experiments is set at H=T.
The off-bottom clearance (from the middle of the impeller to the vessel bottom) is
T/3 in experiments. Two different types of stirrers are used, namely, 60° pitched
blade turbine down flow (PBTD) and 60° pitched blade turbine up flow (PBTU).
The experimental data were taken from a particle image velocimetry (PIV) system
which equipped with a 532 nm green light Nd:YAG laser, a 1600 x 1200 pixel² of
CCD camera resolution and a mini-synchronizer. In this study, we are interested in
the analysis of the hydrodynamic structure of turbulent flow. Up- and down-
pumping pitched six-blade turbines with a blade inclination equal to 60° were used
(figure 2). In fact, we present results associated to the vertical measurement plan
which located midway between two blades. The diameter of the impeller is equal
to 102 mm and the tangential velocity at the tip of the blade is equal to 0.8 m.s-1.
3 Experimental Results
the up-pumping configuration. Two main recirculation loops are presented in the
up-pumping configuration. Thus, one is localized in the third upper zone of
the vessel and attached to the free surface. Otherwise, the second one is attached
the wall of the vessel, which it is larger than the first recirculation loop. However,
one main recirculation loop is presented in the down-pumping configuration
which localized in the bottom of the vessel and near the vessel wall. In addition,
there are no vortices presented in the volume upper the impeller which characte-
rizes by the lowest velocity. Close to the impeller, the stream discharged from the
blade as the fluid accelerates from a downward to an upward direction in the upper
reaches of the vessel. In fact, the sloping stream turns vertically close the impeller,
then, included velocity direction appears between the main recirculation loops.
Otherwise, the stream discharged from the blade causing a local velocity in the
down-pumping configuration. Whereas, the velocity increases as the fluid accele-
rates from an upward to a downward direction in the inferior reaches of the vessel.
The sloping stream confined by the bottom wall and the side wall.
Velocity (mm.s-
Velocity (mm.s-
(a) PBTU 60° (b) PBTD 60°
3.3 Vortices
Figure 5 shows the distribution of the vortices for both up- and down- pumping
pitched six-blade turbines. The maximum values are observed in two locations
corresponding to the location of highest value and the lowest values for both up-
and down- pumping configuration. Consequently, the highest and the lowest local
vortex values were found at the impeller level and become medium moving both
upward and downward from this level. For the up-pumping configuration, the
trailing vortices cross the plane to the back side at the impeller tip and to the front
side above the blade. However, for the down-pumping configuration, the trailing
vortices cross the plane to the back side above the blade and to the front side at the
impeller tip. Indeed, it has been observed that the high vortex value of the down-
pumping configuration is greater than the high vortex value of the up-pumping
configuration. However, the values of the lowest vortex of the down-pumping
configuration and the up-pumping configuration are almost the same.
Fig. 5 Vortices
96 B. Ben Amira, Z. Driss, and M.S. Abid
0.16 0.35
0.14
0.3
0.12
0.25
0.1
0.2
0.08
0.15
0.06
0.1
0.04
0.02 0.05
0.045 0.11
0.04 0.1
0.09
0.035
0.08
0.03
0.07
0.025
0.06
0.02 0.05
0.015 0.04
0.03
0.01
0.02
0.005
0.01
0
3.6 Profiles
In this section we are interested to the radial revolution velocity profile, the radial
velocity component profile, the radial rms velocity profile, the turbulent kinetic
energy profile, the dissipation rate of the turbulent kinetic energy profile and the
turbulent viscosity profile in several positions in the bulk tank for both the up- and
down- pumping pitched six-blade turbines. Figure 8 shows the radial revolution of
the profiles at the axial position equal to z=1/3 H from the free surface. Therefore,
for all configurations, it has been seen that the hydrodynamic structure of the ves-
sel has the same distribution along the vessel. The up- and the down-pumping con-
figurations have the similar effect at this region. However, it has been seen that the
up-pumping turbine is greeter that the down-pumping turbine at the upper region
of the vessel. Then, the highest value of the up-pumping configuration were pre-
sented at the half of the vessel near the vessel wall which described by the main
recirculation loop. A quick decrease of the velocity near the vessel wall has been
noted. However, the first half region besides the shaft is described by the absence
of the recirculation loop for both configurations. Consequently, at this region, the
up-and the down-pumping configuration were looked confounded.Figure 9 shows
the radial revolution of the profiles at the axial position equal to z=1/2 H. It has
been seen that the up-pumping turbine is greeter that the down-pumping turbine at
the medium of the vessel. Moreover, the down-pumping configuration has sinu-
soidal reparation along the vessel. Then, the highest value of the up-pumping con-
figuration were presented at the medium region of the vessel which described by
the discharge flow region. A progressive decrease of the velocity has been noted.
98 B. Ben Amira, Z. Driss, and M.S. Abid
(d) Turbulent kinetic energy (e) Dissipation rate of the (f) Turbulent viscosity
turbulent kinetic energy
(d) Turbulent kinetic ener- (e) Dissipation rate of the (f) Turbulent viscosity
gy turbulent kinetic energy
Figure 10. shows the radial revolution of the profiles at the axial position equal
to z=1/3 H from the vessel bottom. Therefore, for all configurations, it has been
seen that the hydrodynamic structure of the vessel has the same distribution along
Experimental Study of the 60° PBT6 Pitching Blade Effect with a PIV Application 99
the vessel unless both velocity and vortex change direction near the vessel wall
which described by the second recirculation loop.
(d) Turbulent kinetic energy (e) Dissipation rate of the (f) Turbulent viscosity
turbulent kinetic energy
4 Conclusion
In this work, we have studied the effect of the pumping direction on the hydrody-
namic structure of a stirred vessel. In fact, it has been noted that the vessel which
equipped by the PBTU configuration is more stirred than the PBTD configuration.
In addition, the velocity change direction only for the up-pumping configuration
unless at the upper region of the vessel. Moreover, the trailing vortices and the
highest values were found in the same direction of the flow discharge.
In the future, we propose to develop a numerical simulation to study a new
geometrical parameters.
References
Aubin, J., Fletcher, D.F., Xuereb, C.: Modeling turbulent flow in stirred tanks with CFD:
the influence of the modeling approach, turbulence model and numerical scheme. Expe-
rimental Thermal and Fluid Science 28, 431–445 (2004)
Ben Amira, B.: Experimental characterization of a stirred vessel equipped with eight flat
blades turbine using a PIV system: Effect of the Reynolds number. Récents Progrès en
Génie des Procédés, Numéro 101–2011. Ed. SFGP, Paris, ISSN 1775-335X - ISBN
2-910239-75-6
100 B. Ben Amira, Z. Driss, and M.S. Abid
Driss, Z., Bouzgarrou, G., Kchaou, H., Abid, M.S.: Simulation numérique de l’écoulement
laminaire induit dans des cuves agitées générées par les mobiles de proximité de type
mono et double vis à profils simple et modifié. Mécanique & Industries 12, 109–121
(2011)
Driss, Z., Bouzgarrou, G., Chtourou, W., Kchaou, H., Abid, M.S.: Computational studies of
the pitched blade turbines design effect on the stirred tank flow characteristics. European
Journal of Mechanics B/Fluids 29, 236–245 (2010)
Feng, X., Cheng, J., Li, X., Yang, C., Mao, Z.S.: Numerical simulation of turbulent flow in
a baffled stirred tank with an explicit algebraic stress model. Chemical Engineering
Science 69, 30–44 (2012)
Gabriele, A., Nienow, A.W., Simmons, M.J.H.: Use of angle resolved PIV to estimate local
specific energy dissipation rates for up-and down-pumping pitched blade agitators in a
stirred tank. Chemical Engineering Science 64, 126–143 (2009)
Guillard, F., Fritzon, R., Revstedt, J., TraKgasrdh, C., AldeHn, M., Fuchs, L.: Mixing in a
confined turbulent impinging jet using planar laser induced fluorescence. Exp. Flu-
ids 25, 143–150 (1998)
Guillard, F., TraKgasrdh, C., Fuchs, L.: A study of turbulent mixing in a turbine-agitated
tank using a fluorescence Technique. Experiments in Fluids 28, 225–235 (2000)
La Fontaine, R.F., Shepherd, I.C.: Particle Image Velocimetry Applied to a Stirred Vessel.
Experimental Thermal and Fluid Science 12, 256–264 (1996)
Mohan, P., Emery, A.N., AI-Hassan, T.: Review Heat Transfer to Newtonian Fluids in Me-
chanically Agitated Vessels. Experimental Thermal and Fluid Science 5, 861–883
(1992)
Paglianti, A., Pintus, S.: An impedance probe for the measurements of liquid hold-up and
mixing time in two/three-phase stirred tank reactors. Experiments in Fluids 31, 417–427
(2001)
Incidence Angle Effect on the Aerodynamic
Structure of an Incurved Savonius Wind Rotor
1 Introduction
Savonius wind rotor has many advantages over others in that its constructions are
simpler and cheaper. It is independent of the wind direction and has a good start-
ing torque at lower wind speeds (Altan et al. 2008, Saha et al. 2008, Mohamed et
al. 2010, Damak et al. 2013, Driss et al. 2014). Therefore, available in the litera-
ture are a lot of studies that have been conducted to increase the performance of a
Savonius wind rotor. For example, Grinspan et al. (2001) developed a new blade
shape with a twist for the Savonius rotor. Saha and Rajkumar (2006) performed
work on twist bladed metallic Savonius rotor and compared the performance
with conventional semi-circular blades having no twist. Saha et al. (2008)
conducted wind tunnel tests to assess the aerodynamic performance of single,
two and three-stage Savonius rotor systems. Both semicircular and twisted
blades have been used. In this paper, we are interested on the study of the inci-
dence angle effect on the aerodynamic structure characteristics of an incurved
Savonius wind rotor. Results have been visualized and commented.
*
Corresponding author.
2 Cases Study
3 Numerical Results
For θ=45°, a large recirculation zone is formed in the downstream of the rotor.
However, for θ=90° two recirculation zones have been observed in the down-
stream of the advancing and returning buckets. The direction of the infiltration has
an effect on the formation of the recirculation zones. Indeed, the wake characteris-
tic of the maximum values has been observed above and under the rotor both for
incidence angle θ=45° and θ=90°. Two acceleration zones has been observed in
the downstream of the rotor where the maximum velocity values are recorded. The
largest wake is formed for θ=45° and is located in the exterior endplates of the re-
turning bucket, while the wakes formed in the exterior sides of the two buckets for
θ=90° and θ=135° and that of the advancing bucket for θ=45° are less important.
For θ=0°, an acceleration zone where the maximum velocity values are recorded is
located in the convex surface of the two buckets.
which is weak for θ=0° and relatively weak for θ=135°, while it becomes impor-
tant for θ=45° and θ=90°. In the other hand, the total pressure is found to be great-
er whenever the convex surface of the advancing bucket of the rotor is facing the
air flow. For an incidence angle θ=0°, the total pressure is uniform and globally
low with a zone of minimum pressure value located in the concave surface of the
rotor buckets. While for θ=45° and θ=135°, the total pressure increases and is
greater for an incidence angle θ=45°. This fact can be explained by the bucket po-
sition. In fact, in this case the returning bucket is facing the flow while for θ=135°
it’s the advancing bucket. For θ=90° the configuration of the rotor is allowing it to
play the role of a large obstacle and the total pressure is high in the collector and
in the upstream of the rotor. A rapid decrease in the total pressure in the down-
stream of the rotor causes a depression zone and the total pressure keeps increas-
ing in the way out of the wind tunnel. The formation of an important compression
zone has been noted for θ=90°, θ=45° and is less important for θ=135°. However,
for θ=0° the wake formed is restricted to the zone located in the concave surface
of the rotor buckets.
In this case the diffuser is globally considered as a zone that holds the maxi-
mum value of the turbulent kinetic energy with the exception of the lateral surface
of the diffuser’s exit in which the energy’s value drops. For θ=45° as well as
θ=135°, the turbulent kinetic energy’s maximum value is located in the exterior
endplates of the rotor’s buckets and around the shaft. Although the turbulent kinet-
ic energy is greater in the rotor’s area for θ=135°, the opposite is noted in the zone
of the diffuser.
Indeed, two symmetrical wakes are formed in the downstream of the rotor for
θ=45° and θ=90°. The greater one is found to be for θ=90°. In this case, the test
section in the rotor downstream is globally considered as a zone that holds the
maximum value of the turbulent dissipation rate. Then, it drops through the diffus-
er’s exit. For θ=45° as well as θ=135°, the maximum value of the turbulent dissi-
pation rate is located in the exterior endplates of the rotor buckets and around the
shaft.
For θ=0°, the viscosity increases slightly in the concave surface of the two
buckets, keeps decreasing out of the test section and increases again forming a
wake characteristic of the maximum values of the turbulent viscosity in the lat-
eral walls of the diffuser’s exit. A relatively similar behavior has been noted for
θ=135°. In fact, the viscosity increases in the concave surface of the buckets and
again in the diffuser exit. Moreover, another wake is found to be formed in the
convex surface of the advancing rotor bucket and in the diffuser exit. This wake
is greater than in θ=0° and it covers not only the walls but the whole exit area.
For θ=45° and θ=90°, the wake holds the maximum viscosity values and extends
on the diffuser. For θ=45°, the wake is created in the exterior endplate of the ad-
vancing bucket while for θ=90° the wake covers the whole convex surface of the
advancing bucket.
V (m/s)
15
10
Experimental
Numerical
0 x(mm)
-200 -100 0 100 200
5 Conclusion
Numerical investigation has been developed to study the incidence angle effect
on the aerodynamic structure of an incurved Savonius wind rotor. Numerical re-
sults, such as velocity fields, pressure and turbulent characteristics are presented in
the wind tunnel, considered as a control volume. According to the obtained re-
sults, the the incidence angle of the incurved Savonis wind rotor has a direct
effect on the local characteristics.
In the future, we intend to change the bucket geometrical parameters and pro-
pose new configurations in order to increase the performance of the Savonius
wind rotor.
References
Altan, B.D., Atılgan, M., Özdamar, A.: An experimental study on improvement of a Savo-
nius rotor performance with curtaining. Experimental Thermal and Fluid Science 32,
1673–1678 (2008)
Damak, A., Driss, Z., Abid, M.S.: Experimental investigation of helical Savonius rotor with
a twist of 180°. Renewable Energy 52, 136–142 (2013)
Driss, Z., Mlayeh, O., Driss, D., Maaloul, M., Abid, M.S.: Numerical simulation and expe-
rimental validation of the turbulent flow around a small incurved Savonius wind rotor.
Energy 74, 506–517 (2014)
Mohamed, M.H., Janiga, G., Pap, E., Thévenin, D.: Optimization of Savonius turbines us-
ing an obstacle shielding there turning blade. Renewable Energy 35, 2618–2626 (2010)
Saha, U.K., Thotla, S., Maity, D.: Optimum design configuration of Savonius rotor through
wind tunnel experiments. Journal of Wind Engineering and Industrial Aerodynamics 96,
1359–1375 (2008)
Grinspan, A.S., Kumar, P.S., Saha, U.K., Mahanta, P., Ratnarao, D.V., Veda Bhanu, G.:
Design development & testing of Savonius wind turbine rotor with twisted blades. In:
Proceedings of National Conference on Fluid Mechanics & Fluid Power, India, vol. 28,
pp. 28–31 (2001)
Saha, U.K., Jaya Rajkumar, M.: On the performance analysis of Savonius rotor with twisted
blades. Renewable Energy 31, 1776–1788 (2006)
Saha, U.K., Thotla, S., Maity, D.: Optimum design configuration of Savonius rotor through
wind tunnel experiments. Journal of Wind Engineering and Industrial Aerodynamics 96,
1359–1375 (2008)
Experimental Characterization of a NACA2415
Airfoil Wind Turbine
Zied Driss1,*, Tarek Chelbi1, Ahmed Kaffel2, and Mohamed Salah Abid1
1
Laboratory of Electro-Mechanic Systems (LASEM), National School of Engineers of Sfax
(ENIS), Univrsity of Sfax, B.P. 1173, Road Soukra km 3.5, 3038 Sfax, TUNISIA
2
University of Maryland College Park, MD 20742, USA
Zied.Driss@enis.rnu.tn
1 Introduction
*
Corresponding author.
Hirahara et al. (2005) developed a unique and very small wind turbine with a di-
ameter of 500 mm and a small aspect ratio for wide use in urban space. The basic
performance of μF500 was tested for various free stream and load resistance. The
airflow around the turbine was investigated using a particle image velocimetry
(PIV). Schreck and Robinson (2007) showed that wind turbine blade aerodynamic
phenomena can be broadly categorized according to the operating state of the ma-
chine, and two particular aerodynamic phenomena assume crucial importance. At
zero and low rotor yaw angles, increasing rotation determines blade aerodynamic
response. At moderate to high yaw angles, dynamic stall dominates blade aerody-
namic. Wright and Wood (2004) showed that the acceleration and deceleration of
the rotor at speeds below its controlled maximum speed for a range of wind speeds
were calculated and compared with data. The main goal of the Mirzaei et al.
(2009) investigation was to understand the flow field structure of the separation
bubble formed on NLF-0414 airfoil with glaze-ice accretions using CFD and hot-
wire anemometry and comparing these results with previous researches performed
on NACA 0012 airfoil. Driss et al. (20014) investigated the complex flow field
developing around a NACA2415 airfoil wind turbine and studied the wedging an-
gle effect of the airfoil. This will help improving the aerodynamic efficiency in the
design of packaged installations of the NACA2415 airfoil type wind turbine. In
other paper, Driss et al. (20014) studied the turbulent flow around a small in-
curved Savonius wind rotor. Experimental results are conducted on an open wind
tunnel equipped by a small incurved Savonius wind rotor to validate the numerical
method.
In this paper, we are interested on the study of the experimental characteriza-
tion of the horizontal axis NACA2415 airfoil type wind turbine. Particularly, a de-
tailed description of the used wind tunnel and the various manipulations per-
formed will be presented to study the aerodynamic characteristics of the wind
turbine.
7 Wind turbine
6 Support
5 Ventilation chamber
4 Diffuser
3 Test vein
2 Collector
1 Plenum
Ref. Designation
2.3 Anemometer
To initiate the various processes necessary for the characterization of the rotor
with the wind tunnel, several measuring instruments were used. Among which in-
clude the tachometer, the anemometer, the static torque meter and the dynamome-
ter. The experimental study conducted as part of this work involved the realisation
of a horizontal axis wind turbine, a wind tunnel. Experimental trials, allowing the
determination of the static torque Ms, the dynamic torque Md and the velocities
profiles in different directions pre-selected in the test section. Such a system is
equipped with a specific instrument for the development of various experimental
tests required at the laboratory scale. The anemometer measures wind speed in dif-
ferent positions in the pre-selected test section or outside. This is feasible, since
the range of variation of speed is between 0.2 and 20 m.s-1 with a resolution which
reaches 0.1 m.s-1. The different characteristics of the anemometer used are
summarized in table 1.
3 Experimental Results
meter types TQ 8800 mounted directly on the rotor axis. Particularly, we chose to
study different flow regimes defined by Reynolds numbers equal to Re=165093,
Re=194350, Re=217338, Re=242415, Re=257044 and Re=265403. Figures 5 and
6 present the variation of the static torque Ms according to the wedging angle β.
These results relate well to the study of the influence of pitch angle on the static
torque for different Reynolds numbers. According to these results, it has been
noted that these curves have the same allure. With the increase of the wedging an-
gle β, the value of the static torque increases and reaches a maximum and then it
decreases. Figure 6 shows the variation of static torque as a function of the wedg-
ing angle β for different Reynolds numbers. From these results, it has been noted
that the maximum static torque Ms is obtained for a Reynolds number equal to
Re=265403. In this case, the value of the static torque is equal to Ms=5 N.cm at
the corresponding wedging angle equal to β=50 °. All these curves reach their
maximum for the same wedging angle. Moreover, it has been noted that the mini-
mum value of the static torque Ms is obtained for a Reynolds number equal to
Re=165093. In this case, the static torque value is equal to Ms=0.5 N.cm with a
wedging angle β=5°.
4 Conclusion
References
Driss, Z., Chelbi, T., Kaffel, A., Abid, M.S.: Static Study of the Wedging Angle Effect of A
NACA2415 Airfoil Type Wind Turbine. World Research Journal of Engineering and
Technology 3, 1–8 (2014)
Driss, Z., Mlayeh, O., Driss, D., Maaloul, M., Abid, M.S.: Numerical simulation and expe-
rimental validation of the turbulent flow around a small incurved Savonius wind rotor.
Energy 74, 506–517 (2014)
Hu, D., Hu, O., Du, Z.: A study on stall-delay for horizontal axis wind turbine. Renewable
Energy 31, 821–836 (2006)
Hirahara, H., Hossain, M.Z., Kawahashia, M., Nonomura, Y.: Testing basic performance of
a very small wind turbine designed for multi-purposes. Renewable Energy 30,
1279–1297 (2005)
Mirzaei, M., Ardekani, M.A., Doosttalab, M.: Numerical and experimental study of flow
field characteristics of an iced airfoil. Aerospace Science and Technology 13, 267–276
(2009)
Schreck, S.J., Robinson, M.C.: Horizontal Axis Wind Turbine Blade Aerodynamics in Ex-
periments and Modeling. IEEE Transactions on Energy Conversion 22, 61–70 (2007)
Sicot, C., Devinant, P., Loyer, S., Hureau, J.: Rotational and turbulence effects on a wind
turbine blade: investigation of the stall mechanisms. Journal of Wind Engineering and
Industrial Aerodynamics 96, 1320–1331 (2008)
Wrigh, A.K., Wood, D.H.: The starting and low wind speed behaviour of a small horizontal
axis wind turbine. Journal of Wind Engineering and Industrial Aerodynamics 92,
1265–1279 (2004)
DFMM Approach: Modelling, Formalization
and Choice of Material and Manufacturing
Processes
Abstract. In a very competitive industrial world, the integrated design and par-
ticularly the design for manufacturing make progress the profitability of the prod-
uct. The DFM (design for manufacturing) improves the transition between the
design and the manufacturing phases and reduces the time and the costs of product
manufacturing. Within this framework of research, the DFMM approach (design
for material and manufacturing) takes into account the constraints of manufactur-
ing, as of the phase of design, and makes it possible to choose consequently the
most adequate material and processes for the manufacturing of the mechanical
parts. This approach is composed primarily of the modelling of the product, the
manufacturing knowledge management, the choice of material and the choice of
manufacturing processes.
1 Introduction
Separation between design and manufacturing phases generates the realization of
not manufacturable concept. Then, the product is taken back to the design phase
for making the necessary modifications to be manufacturable. With these iterative
loops, the companies lose much time and money. As part of the integrated design
and simultaneous engineering, design for manufacturing solves this problem by
integrating the manufacturing parameters at the design stage. According to (Ker-
brat and Al 2011), the DFM treats the objectives of design and the constraints of
manufacturing simultaneously such is the case of the modular hybrid concept.
Other DFM applications help the designer in the choice of the manufacturing
processes such as the data-processing demonstrator “DFM_Peaux_Squelettes”,
(Skander and Al 2008). In this context, the DFMM is a more complete approach
which makes it possible to choose the most adequate material and processes for
the manufacturing of the mechanical parts.
The choice of the processes is based on the comparison of the attributes of the enti-
ties skin and skeleton of the part with the attributes of the compatibles processes of
the selected material, this method is composed of the following stages (cf Fig.2):
While following these stages (cf Fig.2), the designer can identify all the proc-
esses able to manufacture the totality of the part. Then, for each entity skin and
each entity skeleton, there is a list of ranked appropriate processes. Similarly, the
primary processes are identified and classified.
3 Example of Use
To implement this work, we used the example of the piston of a Schrader grip.
While following all the stages of the DFMM approach, we will determine the most
adequate material and processes for the manufacture of this part.
DFMM Approach 125
Table 3 Choice of the manufacturing processes for the two entities skin (1, 2) and the two
entities skeleton (1, 7)
The processes compatible with selected material (Super alloys based on nickel)
will be used like database for the choice of the processes. The comparison be-
tween the attributes of these processes and the attributes of the entities skin and
skeleton allows to identify the processes able to carry out these entities. For exam-
ple, the capacities (attributes) of the processes of milling and turning make it
possible to manufacture the entity skin 1. Thereafter, the primary processes are
identified by taking the matter addition processes present in all the matter addition
entities (sens of material =1).
While following these stages, we identified the processes able to manufacture
the part such is the case of the two entities skin (1, 2) and the two entities skeleton
(1, 7). Thereafter, these processes are classified according to their average relative
cost (cf Table 3).
References
Kerbrat, O., Mognol, P., Hascoët, J.Y.: A new DFM approach to combines machining
and additive manufacturing. Computers in Industry 62(2011), 684–692 (2011),
ISSN: 0166-3615
Skander, A., Roucoules, L., Meyer, J.S.K.: Design and manufacturing interface modelling
for manufacturing processes selection and knowledge Synthesis in design. The Interna-
tional Journal of Advanced Manufacturing Technology 37(5-6), 443–454 (2008)
Assembly magazine, DFMA Software Generates Cost-Efficient Product Design (2013),
http://www.assemblymag.com/articles/91090-dfma-software-
generates-cost-efficient-product-design (accessed April 1, 2013)
DFMM Approach 129
Abstract. In this work, we are interested on the study of the aerodynamic struc-
ture around an obstacle with inclined roof for a Reynolds numbers equal to
Re=2666. The software "SolidWorks Flow Simulation" has been used to present
the local characteristics. The numerical model considered is based on the resolu-
tion of the Navier-Stokes equations in conjunction with the standard k-ε turbu-
lence model. These equations were solved by a finite volume discretization
method. The numerical model is validated with experimental results conducted on
an open wind tunnel.
1 Introduction
The charactristics of the flows around building-shaped obstacles are different from
those in laminar boundary immersed in turbulent boundary layers. It is important
for the appropriate estimation of wind loading to investigate these characterestics
of flows observed under oncoming turbulence (Nozawa and Tamura, 2002; Ka-
wai, 1999). Computational simulations have an advantage to clarify the process
and mechanism of these localized phenomena. In recent years, several methods to
generate turbulent inflow data were proposed and some of them were used as in-
flow data (Kondo and Murakami,1997; Kataoka, and Mizuno,1999; Noda and
Nakayama, 1989). In this context, Nasrollahei et al. (2013) used a builder software
to study the thermal and cryogenic performance. The results of the research show
that the temperature of Shevaduns in summer is less than the average maximum
temperature and the average minimum temperature of outdoor, and the thermal
condition of the Shevadun space during the measurement is lower than the thermal
comfort limit which is specified in the standard. Ould Said et al. (2014) dedicated
to the numerical simulation of thermal convection in a two dimensional vertical
conical cylinder partially annular space. The governing equations of mass,
momentum and energy are solved using the CFD FLUENT code. Tominaga and
Stathopoulos (2013) reviewed current modeling techniques in computational fluid
dynamics (CFD) simulation of near-field pollutant dispersion in urban environ-
ments and discussed the findings to give insight into future applications. De Paepe
et al. (2013) simulated five different wind incidence angles using a turntable, in
order to quantify their effect on indoor air velocities. The responses in local air ve-
locities could largely be attributed to the relative position of the end walls of the
scale models orientated towards the wind. This crucial position allows the meas-
ured air velocity trends to be explained. The estimated airflow rates gradually de-
creased for larger wind incidence angles. Lim et al. (2009) presented a numerical
simulation of flow around a surface mounted cube placed in a turbulent boundary
layer which, although representing a typical wind environment, to match a series
of wind tunnel observations. The simulations were carried out at a Reynolds num-
ber, based on the velocity at the cube height, of 20,000. The results presented in-
clude detailed comparison between measurements and large eddy simulation
(LES) computations of both the inflow boundary layer and the flow field around
the cube including mean and fluctuating surface pressures. Melo et al. (2012) de-
veloped two Gaussian atmospheric dispersion models, AERMOD and CALPUFF.
Both incorporating the PRIME algorithm for plume rise and building downwash,
are intercompared and validated using wind tunnel data on odour dispersion
around a complex pig farm facility comprising of two attached buildings. The re-
sults show that concentrations predicted by AERMOD are in general higher than
those predicted by CALPUFF, especially regarding the maximum mean concen-
trations observed in the near field. Comparison of the model results with wind
tunnel data showed that both models adequately predict mean concentrations fur-
ther downwind from the facility. However, closer to the buildings, the models may
over-predict or under-predict concentrations by a factor of two, and in certain cas-
es even larger, depending on the conditions. Meslem et al. (2014) observed
changes in the prediction of local and global mean-flow quantities as a function of
the considered turbulence model and by the lack of consensus in the literature on
their performance to predict jet flows with significant three-dimensionality. The
study reveals that none of the turbulence models is able to predict well all jet cha-
racteristics in the same time.
According to these anteriors studies, it’s clear that the study of the aerodynamic
around the obstacle is very interesting. For thus, we have studied the aerodynamic
structure around an obstacle with inclined roof.
2 Geometrical System
The computational domain is shown in figure 1. It is defined by the interior vo-
lume of the wind tunnel blocked by two planes. The first one is in the tranquilliza-
tion chamber entry and the second one is in the exit of the diffuser. The test vein is
equipped by the considered inclined roof obstacle.
Study of the Aerodynamic Structure around an Obstacle with Inclined Roof 133
3 Numerical Results
3.8 Vorticity
Figure 9 presents the distribution of the vorticity in the longitudinal plane defined by
X=0 mm. According to these results, the vorticity is at its minimum in the collector
region and starts increasing after crossing the obstacle. The greatest vorticity values
are reached after hitting the obstacle blade. In these conditions, the maximum value
of the vorticity is equal to 6.2 s-1 for a Reynolds number equal to Re=2666.
5 Conclusion
References
De Melo, A.M.V., Santos, J.M., Mavroidis, I., Reis Junior, N.C.: Modelling of odour dis-
persion around a pig farm building complex using AERMOD and CALPUFF. Compari-
son with wind tunnel results, Building and Environment 56, 8–20 (2012)
De Paepe, M., Pieters, J.G., Cornelis, W.M., Gabriels, D., Merci, B., Demeyer, P.: Airflow
measurements in and around scale-model cattle barns in a wind tunnel: Effect of wind
incidence angle. Biosystems Engineering 115, 211–219 (2013)
Driss, Z., Bouzgarrou, G., Chtourou, W., Kchaou, H., Abid, M.S.: Computational studies of
the pitched blade turbines design effect on the stirred tank flow characteristics. European
Journal of Mechanics B/Fluids 29, 236–245 (2010)
Lim, H.C., Thomas, T.G., Castro, I.P.: Flow around a cube in a turbulent boundary layer:
LES and experiment. Journal of Wind Engineering and Industrial Aerodynamics 97,
96–109 (2009)
Meslem, A., Bode, F., Croitoru, C., Nastase, I.: Comparison of turbulence models in simu-
lating jet flow from a cross-shaped orifice. European Journal of Mechanics B/Fluids 44,
100–120 (2014)
140 S. Driss, Z. Driss, and I.K. Kammoun
Nozawa, K., Tamura, T.: Large eddy simulation of the flow around a low-rise building im-
mersed in a a rough-wall turbulent boundary layer. J. Wind Eng. Ind. Aerodyn. 90,
1151–1162 (2002)
Kawai, H.: Local peak pressure and conical vortex on building. In: Proceedings of the 10th
International Conference on Wind Engineering, pp. 1807–1812 (1999)
Kataoka, H., Mizuno, M.: Numerical flow computation around 3D square cylinder using in-
flow turbulence. J. Archit. Plan Environ. Eng., AIJ 523, 71–77 (1999)
Kondo, K., Murakami, S., Mochida, A.: Generation of velocity fluctuation for inflow boun-
dary condition of LES. J. Wind Eng. Ind. 67&68, 51–64 (1997)
Nasrollahei, N., Mahdavinejad, M., Hadiyanpour, M.: Studying the Thermal and Cryogenic
Performance of Shevadun in Native (Local) Buildings of Dezful Based on Modeling and
Environmental Measuring. American Journal of Energy Research 1(3), 45–53 (2013)
Noda, H., Nakayama, A.: Reproducibility of flow past two-dimensional rectangular cylind-
ers in a homogeneous turbulent flow by LES. J. Wind Eng. 89, 321–324 (2001)
Ould Said, B., Retiel, N., Bouguerra, E.H.: Numerical Simulation of Natural Convection in
a Vertical Conical Cylinder Partially Annular Space. American Journal of Energy Re-
search 2(2), 24–29 (2014)
Tominaga, Y., Stathopoulos, T.: CFD simulation of near-field pollutant dispersion in the
urban environment: A review of current modeling techniques. Atmospheric Environ-
ment 79, 716–730 (2013)
Contribution to Inverse Kinematic Modeling
of a Planar Continuum Robot Using a Particle
Swarm Optimization
1
Department of Mechanical Engineering, University1, Constantine, Algeria
{ammar_amouri,azaatri}@yahoo.fr
2
Department of Mechanical Engineering,
University Labri Ben M’ Hidi, Oum el Bouaghi, Algeria
mahfoudi.chawki@gmail.com
1 Introduction
An important research theme for study the continuum robots is the calculation of
kinematic models, where the traditional approaches to modeling, in which frames
are associated with each joint, are inappropriate for this case, because of the ab-
sence of discrete and rigid links in their architecture and the complexity of the
problem which reside in the coupling of operational variables. However, the hy-
per-redundant robotic systems, discrete links or continuum form cannot be control
by considering a finite number of degrees of freedom, thus admitting a reduced set
of physical solutions (Robinson and Davies 1999). Among contributions on the
inverse kinematic model of these structures, we can cite the model for one bending
section with a special configuration proposed by (Hannan and Walker 2003), to
validate the model; the authors used a specific software symbolic computation and
have simulated the kinematic behavior of the robot at two sections of Air-Octor
and Oct-Arm. The same authors studied the workspace and configuration space
describing the constraints due to limitations with respect to the lengths of robots.
(Escande et all. 2011), synthesized and validated an inverse kinematic model for
one bending section. The experimental results of this model show an accurate fol-
low-up identification of all existing uncertainties on the bench test and measure-
ment systems for both models and also in (Lakhal et all. 2014), the authors used
an optimization algorithm, based on a Sequential Quadratic Program (SQP). The
case of a multi-section is studied in (Neppalli et all. 2009), where the end-point of
each section is assumed to be known. The overall points can be chosen by solving
a system of inequalities. The authors did not validate experimentally the model but
they simulated the overall behavior using appropriate software. (Iqbal et all.
2009), presenting a complete inverse kinematic model for one bending section
based on the development realized by (Escande et all. 2011).This inverse kinemat-
ic model is formulated as an optimization problem with a cost function and con-
straints using the principle of interval analysis. The validation of this model has
done on a micro-robot and showed that the model was not sufficiently robust
against uncertainties of environment and inaccuracies of materials. The solution of
inverse kinematic model is posed also in terms of optimizing an objective function
under inequality constraints by (Bailly and Amirat 2005). The validation of this
model was made on a micro-robot dedicated to surgery. Finally, a state of the art
report in modeling these classes of robots is presented by (Robert et all. 2010).
The lack of analytical models for the kinematic of continuum robots, guide us to
use a powerful numerical method such as Particle Swarm Optimization (PSO).
The first idea based on swarm intelligence algorithm was introduced in 1995
(Kennedy and Eberhart 1995), who studied social relationships in a group to de-
velop new methods of calculation. This method has showed its capacity to solve
different optimization problems compared to other heuristics (Elbeltagi et all.
2005). The main contribution of this paper is the calculation of operational coor-
dinates of each section, by using an optimization method, under the constraints of
conserving the length of each section. The rest of the paper is organized as fol-
lows: Section 2 describes the bending section modeling. Section 3 shows the op-
timization technique for inverse kinematic modeling of a PCR. Finally, simulation
examples are performed to validate the proposed approach.
y k −1
Pk ( x k , y k )
yk
sk
xk
lth ,k
κk
δk θk
x k −1
o k −1
⎧δ k = a tan 2 ( y k , x k )
⎨ (1)
⎩ θ k = π − 2δ k
⎛θ ⎞
2sin ⎜ k ⎟
κk = ⎝ 2 ⎠ (2)
x k + y k2
2
θk
sk = (3)
κk
3.1 PSO
Particle Swarm Optimization (PSO) is a biologically inspired computational
search and optimization method based on the social behaviors of some biological
organisms, especially the group's ability of some animal species to locate a desira-
ble position in the given area. It was proposed first by (Kennedy and Eberhart
1995) and its fast convergence compared to other Evolutionary Algorithms (Elbel-
tagi et all. 2005). Thus, using the simple rules of displacement (in the space of the
solutions), the particles can progressively converge to a global optimum. During
each iteration t, the velocity changes by applying equation (4) to each particle:
( ) (
v p i (t + 1) = ω v p i (t ) + c1 ρ1 Pi ,best (t ) − Pi (t ) + c 2 ρ 2 Pg ,best (t ) − Pi (t ) ) (4)
144 A. Amouri, C. Mahfoudi, and A. Zaatri
The position update is applied by equation (5) based on the new velocity and
the current position.
Pi (t + 1) = Pi (t ) + v Pi (t + 1) (5)
{ }
T
tor v Pi (t ) = v Px ,i v Py ,i . The best position visited by the particle is denoted
{ }
T
Pi ,best (t ) = Px ,i ,best Py ,i ,best and the best position found by the swarm
{ }
T
is Pg ,best (t ) = Px , g ,best Py , g ,best .
n
(
F lth ,k , l r ,k =) ∑ f k ( lth ,k , l r ,k ) (6)
k =1
With:
( ) 1
( )
2
f lth ,k , l r ,k = lth ,k − l r ,k (7)
2
Such that:
⎧ θk
⎪lth ,k = ( x k − x k −1 ) + ( y k − y k −1 ) sin (θ
2 2
⎨ k ) (8)
⎪ θ = a tan 2 ( ( y − y ) , ( x − x ) )
⎩ k k k −1 k k −1
Where lth ,k , l r ,k : the real length and the theoretical length calculated using eq-
uation (8) of a bending section k respectively and where n is the number of
bending section.
The intermediate operational coordinates x k∗ , y k∗ ( ) of each section, are calcu-
lated by solving the following optimization problem:
⎧ min
⎪ x ∗ , y ∗ ; k =1,…,n (
F x k∗ , y k∗ )
⎪ k k
⎨ under the constraints g (9)
⎪ j
⎪ g j ≤ 0; j = 0,1,… , m
⎩
Where g j the constraints are could be used for example: obstacle avoidance,
taking objects, energy minimization, etc. And m is the number of added con-
straints.
Initially, the population of particles is located with random positions in the
problem space. During each iteration t, each particle changes their velocity ac-
cording to equation (4) and move to a new position using (5). Once the displace-
ment of the particles carried out, the new positions are evaluated using equation
(6) and the vector Pi ,best and Pg ,best are determined. The algorithm stops if
maximum number of iterations is achieved or any other stopping criteria are satis-
fied. The solution of this algorithm represents the intermediate Cartesian coordi-
nate researched with respect to the reference frame R 0 .
Such that:
4 Simulation Results
The effectiveness of proposed approach is verified by some simulations through a
personal computer with i3 3.30 GHz processor. The first example was performed
for a planar robot consists of two bending sections for the following settings of
⎛ 1 ⎞
the trajectory in operational space: ⎜ x 2∗ = t 2 , y 2∗ = 200 − 4 t ⎟ The trajectory
⎝ 2 ⎠
has been accomplished in 10 sec of time duration, corresponding to 51 positions
captures. Fig. 2 represents the configurations (solutions) for a desired point of the
given trajectory, while respecting the imposed assumptions. The execution time of
the optimization process for each point found is shown in Fig. 3. Fig. 4 shows the
convergence of cost function. In this simulation, the tolerance error is fixed to
10−6 while the time execution is equal to 23.91 msec.
For the second case, knowing the general movement of a superior mobile plat-
form generated by the trajectory in operational space ( x 2∗ = 0, y 2∗ = 200 − 10t )
where t is time variable varying from 0 to 10 sec, with a step equal to 2 sec.
Fig.5.a represents the possible configurations for a desired point of the given tra-
jectory. The problem of infinite solutions is solved by adding some other con-
straints. To illustrate this, two cases be imposed. The first case uses the constraint
of g1 = θ1 〈 0 and the second that uses g 2 = θ1 〉 0 , as shown in Fig.5.b and 5.c,
respectively.
(
In the third example, the trajectory of equation: x 2∗ = 0, y 2∗ = 200 − 10 t ) is
used as input in the proposed model and lengths variation is determined. This tra-
jectory has been accomplished in 10 sec of time duration, corresponding to 121
position captures. Fig. 6 shows the length profiles obtained by unconstrained
( )
model g j = 0 . Fig. 7 show the length profiles, where the constraint is given
by g1 = θ1 〈 0 .
Contribution to Inverse Kinematic Modeling of a Planar Continuum Robot 147
200
150
Y ( mm ) 100
50
0
-40 -20 0 20 40 60
X ( mm)
30
Time ( msec )
20
10
0
0 2 4 6 8 10
Found Position
200
Cost Function
150
100
50
0
0 100 200 300 400 500
No. of Iterations
Fig. 4 Convergence of the cost function
148 A. Amouri, C. Mahfoudi, and A. Zaatri
Y ( mm )
Y ( mm )
Y ( mm )
50 50 50
0 0 0
-100 -50 0 50 -100 -50 0 50 -50 0 50 100
X ( mm ) X ( mm ) X ( mm )
L21
100
90
0 1 2 3 4 5 6 7 8 9
Time (sec )
Second bending section
120
len g th ( m m )
L12
100
L22
80
0 1 2 3 4 5 6 7 8 9
Time ( sec )
L11
100
L21
90
0 1 2 3 4 5 6 7 8 9
Time ( sec )
Second bending section
120
len g th ( m m )
L12
100
L22
80
0 1 2 3 4 5 6 7 8 9
Time ( sec )
5 Conclusion
This article presents a new approach for the synthesis of the inverse kinematic
model for a planar continuum robot. The mathematical formulations defining the
kinematic of a bending section are given. This approach allows us to calculate the
Cartesian coordinates of the origin of intermediate platforms using the PSO opti-
mization method, where the search space that is chosen arbitrarily. Future work is
to improve the execution time, one taking the space of research in the workspace
of each section, in order to find a solution close to real time and the generalization
of the approach in the three-dimensional.
References
Robinson, G., Davies, J.B.C.: Continuum Robots: A State of the Art. In: IEEE International
Conference on Robotics and Automation, Detroit, Michigan, pp. 2849–2854 (1999)
Hannan, M., Walker, I.D.: Kinematics and the Implementation of an Elephant’s Trunk Ma-
nipulator and Other Continuum Style Robots. Journal of Robotic Systems 20(2), 45–63
(2003)
Escande, C., Pathak, P.M., Merzouki, R., Coelen, V.: Modeling of Multi-section Bionic
Manipulator: application to Robotino XT. In: 2011 IEEE International Conference on
Robotics and Biomimetics (ROBIO), Phuket, Thailand, pp. 92–97 (2011)
Lakhal, O., Chibani, A., Escande, C., Merzouki, R.: Contribution to Inverse Kinematic
Modeling of a class of Continuum Bionic Handling Arm. In: 2014 IEEE/ASME, Intern-
tional Conference on Advanced Intelligent Mechatronics (AIM), Besançon, France
(2014)
150 A. Amouri, C. Mahfoudi, and A. Zaatri
Neppalli, S., Csencsits, M.A., Jones, B.A., Walker, I.D.: Closed-Form Inverse Kinematics
for Continuum Manipulators. Advanced Robotics 23, 2077–2091 (2009)
Iqbal, S., Mohammed, S., Amirat, Y.: A guaranteed approach for kinematic analysis of con-
tinuum robot based catheter. In: IEEE International Conference on Robotics and Bio-
mimetics (ROBIO), Guilin, China, December 19-23, pp. 1573–1578 (2009)
Bailly, Y., Amirat, Y.: Modeling and control of a hybrid continuum active catheter for aor-
tic aneurysm treatment. In: Proc. of the International Conference on Robotic and Auto-
mation, Barcelona, Spain, pp. 936–941 (2005)
Webster III, R.J., Jones, B.A.: Design and Kinematic Modeling of Constant Cur-vature
Continuum Robots: A Review. The International Journal of Robotics Research 29(13),
1661–1683 (2010)
Kennedy, J., Eberhart, R.C.: Particle Swarm Optimization. In: Proceeding of the IEEE In-
ternational Conference on Neural Networks IV, pp. 1942–1948 (1995)
Elbeltagi, E., Hegazy, T., Grierson, D.: Comparison among five evolutionary-based optimi-
zation algorithms. Advanced Engineering Informatics 19(1), 43–53 (2005)
Experimental and Numerical Study of Beams
under Low-Velocity Impact
Abstract. This paper is an experimental and numerical study of the low velocity
impact on beams. It is shown that the numerical model adopted, for the impact
force history, produces numerical results in good agreement with experimental da-
ta. The analysis establishes that the contact point position is insignificant on the
main impact and rebounds are closer in time when we approach the clamped edge.
The increasing of the incident energy (mass or celerity of the striker) has the effect
of amplifying the impact force without changing its duration in the case of the ve-
locity and with increasing the amplitude of the impact force in the case of the
mass. However elevated hardness of the contact aggravates the impact effects by
increasing the magnitude and decreasing the contact force duration.
1 Introduction
The study of the structures behaviour under impact remains an area of interest to
many researchers in recent years because of the complexity of modeling these
structures and the phenomena occurring during impact. Determining the impact
force is very difficult. Knowledge of the evolution of the impact force is necessary
to study the behaviour of structures in shock. Several researches has been done us-
ing numerical approaches or experimental studies but still unable to properly
define the exact evolution of the impact force.
Several modelling of the impact have been adopted, we can note those consist
in representing the impact globally by an energy balance. The main advantage of
this type this presentation type, is the fact that the input parameters are parameters
characteristic of the impact: incident speed, the mass of the striker , geome-
tric and mechanical properties of the plate and impactor. In its work, Greszczuk
[1] modeled the impact by summing the contact energy than bending Shi-
vakumar [2] improved the earlier model by matching the energy of fexion, the
energy of shear and introducing the energy of membrane in the case of
thin plates. Thus the energy balance becomes:
1 2 (1)
2
Caprino and al. [3] presented the impact by a discrete mass-spring model ha-
ving a single degree of freedom, where the spring stiffness corresponds to the
static stiffness of the plate. A good correlation is observed up to 2.5 / speed,
between the calculated by the model and the measured.
In this model type, the effects of contact and plate inertia are not considered,
hence the usage limit for higher speeds. Sjoblom [4] then suggested a model with
two degrees of freedom that allows the integration of non-linearity in the contact.
With this model, it is possible to uncouple the rigidity of the contact plate. Never-
theless, the effects of shear localized around the point of impact of the large, for
thick plates or for very high speeds, are not taken into account. Therefore Lal [5]
developed a model with three , and representing rigidities punching,
bending and shear springs respectively. This model is valid in the case of small
displacements. Shivakumar and al. [2] then took into account the membranes ef-
fects using mass models-springs of one or two degrees of freedom with four
springs stiffness , , and . The indices , , and respectively
represent the membrane, shear, and bending of the plate contact with the striker.
Fig. 2 Mass-springs models: (a) one degree of freedom, Caprino and al. [3], (b) two degrees
of freedom model, Shivakumar and al. [2]
In this model, the two masses and are connected by a spring represen-
ting the law wireless loading-penetration of the contact. The strength properties-
transverse displacement are represented by the combination of the flexural
springs, and shear membrane. A previous study [6] on the plate free vibrations
with a concentrated mass attached to the plate center showed that the effective
mass, contributing to the plate inertia effect is a quarter of the total plate mass. In
the present document, we present an experimental study of the low-speed impact
Experimental and Numerical Study of Beams under Low-Velocity Impact 153
To treat the impact problem, normally the contact force method and energy ba-
lance method are used. Among various contact force methods, Hertzian method is
the simplest in terms of formulation by using Hertzian’s contact law or its modifi-
cation. An homogeneous beam subjected to bending vibration is governed by the
equation of motion follows:
, ,
, (2)
where and are respectively the area and the moment of inertia of the cross
section, and are the density and Young's modulus of the beam.
The impact is caused by a spherical striker of mass end radius animated
by an incident velocity , as shown in fig. 1.
On contact with the beam, the striker will be modeled by a mass attached to
by non-linear beam spring cf. fig. 3, describing the law of Hertzian contact,
given by:
/ (3)
where is the indentation of the sphere in the beam and the contact stiffness
is given by:
4√
(4)
3
where is the striker radius and the index is related to the striker.
The dynamic balance of the spherical striker and the beam is given respectively
by the equation of dynamic equilibrium of the projectile, driven by an incident
velocity upon contact and initial conditions are written:
154 M. Neder and A. Miloudi
0
(5-a)
0 0 0
Those of the beam, initially at rest, are written:
, ,
. (5-b)
,0 0 ,0 0
is the displacement of the projectile, is the contact force of Hertz eq. (3)
with .
is the deflection of the beam at the impact point.
is the Dirac pulse which introduces a concentrated force equation of
the beam.
These two coupled equations (5-a) and (5-b) are non-linear differential equa-
tions which can be solved by a numeric integration method (Newmark integration
scheme) until 0.
3 Experimental Study
The experimental device consists of two parts: structural and acquisition parts.
1. The structural part consists in turn of two compartments:
• The compartment for positioning of the structure, subject of study, which
allows for several combinations of boundary conditions,
• The fund implementation force that is a pendulum which, by the law of po-
tential energy to kinetic energy conversion, to estimate the impact speed.
2. The acquisition part allows to increase the signal strength and the response of
the structure. It consists of:
• Acceleration sensor "accelerometer" piezoelectric brand Wilcoxon-
Research; model 736,
• A hammer pulse "modal hammer" brand Endvico model: 2302-10,
• A data acquisition card 4-input, National Instrument Model: NI9234,
• Acquisition and data processing Labview software.
2 (6)
4 Results
Fig. 5, shows a comparison between the measured force and that determined by
a numerical approach. It can be seen that the force calculated numerically reached
it's peak faster than the measurement one. It may be caused by the boudaries con-
ditions in the experimental study whitch it is pratically impossible to realise per-
fect clamped edges. Also we note that the main impact is generally followed by a
less important amplitude rebounds.
Fig. 6 Influence of the impact force position: steel clamped-clamped beam with 0.66m
length
Fig. 7 Influence of the impact force position: aluminum cantilever beam with 0.5m length
We have retained three different masses for the same incidental velocity impact.
Fig 8. Show the influence of the masses on impact force evolution.
We have retained four different masses for the same incidental velocity impact.
Fig 8. Show the influence of the masses on impact force evolution.
Fig. 8 and 9 showed that when the incident energy is more important, it has the
effect of increasing the intensity of the impact force. The duration of the impact is
the same if the velocity changes but increase with increasing the mass of the pro-
jectile.
5 Conclusion
References
1. Greszczuk, L.B.: Damage in Composite Materials due to Low Velocity Impact. In:
Zukas, Z.A., et al. (eds.) Impact Dynamics, pp. 55–94. Wiley editions, New York
(1982)
2. Shivakumar, K.N., Elber, W., Illg, W.: Prediction of Impact Force and Duration due to
Low Velocity Impact on Circular Composite Laminates. Journal of Applied Mechan-
ics 52, 674–680 (1985)
3. Caprino, G., Crivelli-Visconti, I., Di Ilio, A.: Elastic Behaviour of Composite Struc-
tures Under Low Velocity Impact. Composites 15(3), 213–234 (1984)
4. Sjoblom, P.: Simple Design Approach Against Low Velocity Impact Damage. In: Pro-
ceedings of the 32nd SAMPE Symposium, Anaheim, CA, pp. 30–52 (1987)
5. Lal, K.M.: Residual Strength Assessment of Low Velocity Impact Damage of Graphite
Epoxy Laminates. Journal Reinf. Plast. Comp. 2, 226–238 (1984)
6. Leissa, A.W.: Vibrations of Plates, NASA SP-160 (1969)
7. Blevins, R.D.: Formulas for Natural Frequency and Mode Shape. R.E. Kriegen Pub-
lishing (1984)
8. Miloudi, A.: M., Comportement Dynamique de Plaques Impactées. In: 8ème Congrès
de Mécanique, El Jadida, Maroc., Avril 17-20, pp. 17–20 (2007)
9. Wali, M., Abdennadher, M., Fakhfakh, T., Haddar M.: Comportement Dynami-que des
Plaques Sandwichs Impactées: Estimation du Coefficient de Restitution. In: 3ème
Congrès de Conception et Modélisation des Systèmes Mécaniques CMSM 2009,
Hammamet, Tunisie, Mars 16-18 (2009)
10. Neder, M., Miloudi, A.: Dynamique des plaques sandwichs sous impact à faible vi-
tesse: Approche numérique. In: 5ème Congrès de Conception et Modélisation des
Systèmes Mécaniques, CMSM 2013, Tunisie (2013)
Modeling of Cutting Forces and Roughness
by the Response Surface Method (RS)
1 Introduction
Metal cutting, machining processes whose origin is very old, has expanded greatly
with the mechanization of increasingly sophisticated manufacturing facilities. De-
spite the emergence of new technologies in recent years, it remains today one of
the methods most widely used (Bouacha 2011).
Machining by cutting means all techniques to obtain a surface by removing ma-
terial with a sharp tool. This ancient technique is often called cutting expensive in
so far as the shaping of the part causes a noble processing waste material (chip).
However, it remains an important manufacturing technique and answered. In shap-
ing metal by cutting, lathing, in our study, alone accounts for 33% of the area of
machining by chip removal (CETIM 2000).
Metal cutting is strongly coupled thermo mechanical process in which plastic
deformation, heat and friction phenomena plays a critical role in terms of wear
(Gautier 2004).
These occurring in a very small region around the tip of the tool, the phenome-
na related to tribology of cutting are difficult to study and model [Gokkaya and
Nalbant ).
2 Experimental Study
The objective of this study is to examine the behavior do matter-torque tool when
shooting Dry Steel FK 20 Mn Cr 5 with a metal coated carbide tool P40 Brand
Sandvik (reference TPMN 16 03 08).
The experiment was done as follows:
• The tests are performed on a lathe roughing and model Maximat Emco V13
thread.
• Measurement of cutting forces during machining consists of an acquisition
channel consists of a dynamometer Kistler 9257 B and a charge amplifier Kistler
5019 B three channels.
To study the criteria roughness Ra, Rt and Rz, we used a roughness (2D) Mitu-
toyo, with printer profile roughness. The experimenter, whatever their field of
study, is still facing the problem difficult for the optimal organization of his es-
says. How to get the right information as soon as possible and for the least cost?
(Remadna 2001).
The use of experimental designs, especially Taguchi orthogonal tables (Taguchi
and konishi 1987) allows the determination of the influence of the simultaneous
effects and interaction of operating parameters on the cutting forces and rough-
ness.
In this work, the standard orthogonal table L27 (313) Taguchi is suitable as an
experimental plane. Parameter levels were chosen as recommended by the cutting
tool manufacturer intervals.
The parameters to be studied and the allocation of the respective levels are
shown in Table 1.
3 Resultats
From this figure, we see that the lowest value of cutting forces obtained with
different combinations of the parameters of cutting regime that is provided with
Test No19 (Vc = 155 m / min, f = 0, 08 mm / rd and ap = 0.1 mm).
In this section we will determine the mathematical models of the cutting forces.
These models highlight the relationship between the elements of the regime cut-
ting (cutting speed, feed and depth of cut) and responses studied (cutting force).
To achieve these objectives, several methods and models are available.
Initial analysis of responses from the response surface method RSM includes all
parameters and their interactions. Models are reduced by eliminating the terms
that have no significant effect on responses.
By this process of elimination back end quadratic models of the equation of
responses in terms of actual factors are presented as follows:
2
Fa=343 – 2 * Vc – 2130 * f - 249 * ap + 9716 *f + 300 * ap2 +
Vc * ap + 1096 * f * ap (1)
Fc = 657 – 3 * Vc – 5136 * f – 806 * ap + 21243 * f2 + 753 * ap2 +
2 * Vc * ap + 3668 * f * ap (2)
Fp = 842 – 4 * Vc – 5033 * f – 743 * ap + 18663 * f2 + 1357 *
ap2 + 3 * Vc * ap + 4445 * f * ap (3)
164 F. Bentaleb and I. Amara
Fa = e15,102 * Vc 0,654 * ap - 3,216 * f ln (f 2,403) * apln (ap 0,301) * Vc ln (ap 0,79) (7)
The comparison between the experimental values of the cutting forces and
those estimated by the models obtained by the response surface methodology
RSM is presented graphically in Figure 2.
Modeling of Cutting Forces and Roughness by the Response Surface Method (RS) 165
(a) (b)
(c)
Fig. 2 Curves of experimental efforts and estimated cut: (a) F, (b) Fc and (c) Fp.
3.2 Roughness
Results of the criteria of roughness obtained during the tests are shown in figure 3.
Fig. 3 Change in roughness (Ra, Rt and Rz) based on the parameters of cutting regime
In this section we will determine the mathematical models of the criteria chosen
roughness. These models highlight the relationship between the elements of the
regime cutting (cutting speed, feed per revolution and depth of cut) and responses
studied (roughness of the machined surfaces). For a number of methods and mod-
els are proposed.
The analysis of responses obtained from RSM includes only linear terms. The
final equation linear response models in terms of actual factors were as follows:
Ra = -2,1479 - 0,0018 * Vc + 34,3606 * f + 0,0204 *ap (19)
Rt = 1,5255 - 0,0011 * Vc + 18,31106 * f + 0,2913 *ap (20)
Rz = 3,0169 - 0,0108 * Vc + 7,2217 * f + 0,5834 *ap (21)
In order to explain the variation in cutting forces based on the factors identified in
the study (Vc, f and ap) in an exponential form, a linear regression (multiple) can
be used.
The comparison between the experimental values of the different roughness pa-
rameters Ra, Rt and Rz and those estimated by the models obtained by the
response surface methodology is presented graphically in Figure 4, although qua-
dratic and exponential models give very good results. It was found that clearly
those following the methodology of genetic algorithms are better.
(a) (b)
(c)
Fig. 4 Curves of different criteria of experimental and estimated roughness: (a) Ra, (b) and
Rt (c) Rz
4 Conclusion
References
Journal article only by DOI
Taguchi, G., Konishi, S.: Taguchi metods, orthogonal arrays and lineair graphs, tools for
quality engineering. American supplier institute (1987)
Gokkaya, H., Nalbant, M.: The effects of cutting tool geometry and processing parameters
on the surface of AISI 1030 steel of roughness. Materials and Design
Book
Documentation CETIM, hard turning and high-speed machining (HSM), a technical center
for mechanical industries (AAJ), Report, p. 86 (2000)
Dissertation
Bouacha, K.: Behavior of tool-material pair when machining hard materials. PhD thesis,
defended in January 2011 in the University of May 8, 1945, Guelma-Algeria
Gautier, L.: Study of damage carbide tools WC-CO by the characterization of the mechan-
isms chip tool interface application to dry machining of aerospace aluminum alloy
AA2024 T351. PhD thesis submitted in December 2004 to the National Higher School
of Arts and Crafts Centre in Bordeaux-France (2004)
Remadna, M.: The system behavior in machining hard turning. Application to the case of a
hardened steel machined with CBN. Doctoral thesis submitted in 2001 at INSA Lyon
France (2001)
Analysis of Inelastic Behavior of Amorphous
Polymers
1 Introduction
The sub-Tg plastic deformations of amorphous polymers have been successfully
described using a theory [Perez J 1984, Perez J et al 1988, Cavaillé JY et al 1989]
based on:
• The concept of quasipoint defects qpds in concentration Cd corresponding to
positive or negative microfluctuations of specific volume ;
• The assumption of hierarchically constrained molecular motions. Once
activated, these qpd may be transformed into shear microdomains smd after an av-
erage time period τ 1 . As long as the stress is applied, the smd grow larger and ul-
timately merge irreversibly into one another after a average time period τ2,
which can expressed as [Cavaillé JY et al 1989]:
dn(t ) n(t ) − n(∞ ) n(0) − n(t ) (1)
=− +
dt τ1 τ2
where n(t) is the number of inactivated defects at time t. τ 2 is the mean time
of the order of the time characterising the elementary molecular motions responsi-
ble to the secondary, or β , relaxations, τ β and the time τ 2 was found to be of
the order of the mean time needed for a translational elementary molecular motion
τ mol , which is related to τ β according to:
1
⎛ τβ ⎞κ
τ mol = ⎜ 1−κ ⎟ (2)
⎜t ⎟
⎝ 0 ⎠
where t0 is a scaling parameter. κ is a measure of the degree of hierarchical
constraints on the correlated molecular motions (0< κ <1): κ =0 corresponds to a
fully constrained situation (closely packed molecular arrangements) and κ =1
corresponds to a constrained free situation, that is τ mol = τ β . After τ 2 the ma-
terial begins to deform plastically. The fact that the correlation parameter κ de-
pends on the polymers microstructure. A simple linear relationship was originally
proposed [Cavaillé et al 1991] to relate κ and the structural parameter Cd:
κ = a 0 .C d (3)
where a0 is a proportionality factor.
⎡U ⎤
τ β = τ 0 β . exp .⎢ β ⎥ is the characteristic time for the elementary molecular mo-
⎣ KT ⎦
tions, which has been related to the β relaxation ; where U β is the molecucular
activation energy and τ 0 β is a time of vibration ( τ 0 β =5.2 10-16 s). The time was
found to lie between τ β and τ 2 . On the basis of these arguments, the time de-
pendent shear compliance was expressed as:
⎧ ⎡ ⎛ t ⎞
κ
⎤ ⎫⎪ ⎛ t ⎞
κ'
1 ⎪ (4)
J (t ) = + Aan ⎨1 − exp ⎢− ⎜⎜ ⎟⎟ ⎥ ⎬ + Avp ⎜⎜ ⎟⎟
Gu ⎪⎩ ⎢⎣ ⎝ τ mol ⎠ ⎥⎦ ⎪⎭ ⎝ τ mol ⎠
where Gu is the unrelaxed modulus and Aan and Avp are proportionality factors
that depend mainly on the concentration Cd of defects and κ ' (0< κ ' <1) is an
exponent that takes into consideration the fact that segmental mobility is affected
by strain above a certain threshold of strain. Calculating the Fourier transform of
equation (4) and adding the relaxed modulus, Gr , for the primary relaxation which
corresponds to the value of the modulus at the rubbery plateau, yields the dynamic
modulus:
Gu − G r
G * (iω ) = Gu + (5)
1 + (iωτ mol ) + C.(iωτ mol )
−κ −κ '
2 Theoretical Considerations
In previous developments, the elementary molecular motions were assumed to be
independent of stress, for the stress levels applied in mechanical spectrometry
were low. If strain is no longer linear with stress ( σ 〉 106 Pa ), a part of the basic
hypothesis must be completed: these elementary molecular motions become ther-
momechanically activated and the amount of purely thermal energy needed has
been shown to decrease, this leads to the expression:
⎧ ⎡ ⎤3 ⎫
2
⎪U ⎢1 − ⎛⎜ σ ⎞⎟⎥ ⎪
⎜ ⎟
⎪⎪ ⎢ ⎝ σ 0 ⎠⎦⎥ ⎪⎪
β
(6)
⎛ ΔH a ⎞
τ β = τ 0β . exp.⎨ ⎣ ⎬ = τ 0β . exp.⎜ ⎟
⎪ KT ⎪ ⎝ KT ⎠
⎪ ⎪
⎪⎩ ⎪⎭
dCd ⎡ dε vp ⎤ ⎡ 3 C − C (T ) ⎤ (7)
= ⎢ B0 (1 − Θ)⎥ − ⎢ Cd d d eq (1 + Θ)⎥
dt ⎣ dt ⎦ ⎢⎣ 2 τ 2 (σ ) ⎥⎦
where B0 and Θ = τ 2 (σ ) / τ 2 (0) is a constant.
172 N. Ouali and A. Ahmed Benyahia
where a1 is a constant.
During a mechanical test with a constant speed of the compressing cross-head Vt,
the experimental considerations lead to the following relationship:
dσ ⎡ dε dε ⎤ (9)
Vt = R + L0 ⎢ an + vp ⎥
dt ⎣ dt dt ⎦
from which dσ can be isolated as follows:
dσ 1 ⎡ ⎛ dε dε ⎞ ⎤
= ⎢Vt − L0 ⎜⎜ an + vp ⎟⎟⎥ (10)
dt R⎣ ⎝ dt dt ⎠⎦
where L0 is the sample length and R is a factor including the rigidity of both test
machine and the sample. Thus, incrementing the time and calculating the incre-
mental values of anelastic deformation, dε an , and viscoplastic deformation, dε vp ,
from equation (4) modified, it becomes possible to numerically obtain σ (t ) .
A computer program was accordingly written to calculate the time evolutions of
σ and ε .
A uniaxial compression test was performed for the study of the large-stress de-
formation plastic response of PMMA. In order to apply a uniaxial stress, cylin-
drical samples were selected. Samples 12.00 mm thick were thus machined with a
turn from a PMMA rod of 8.02 mm diameter. A good parallelism between the two
circular faces was necessary to avoid bending during compression. The two faces
were carefully polished in order to minimise interfacial friction between the sam-
ple and the press. The simples thus prepared were placed inside a thermoregulated
cell (200 – 400 K).
Analysis of Inelastic Behavior of Amorphous Polymers 173
The True stress, σ , and the strain, ε , were directly obtained from the meas-
ured values of compressive force, F, and the displacement, ΔL , according to the
following relationships:
F
σ= [1 − ε ] (11)
S0
ΔL
ε= (12)
L0
where S0 and L0 are respectively the cross-section and the length of the sample be-
fore testing.
Figure 1 shows the stress-strain curves during compression tests at a constant de-
formation rate ε =14.10-5 s-1, measured at several temperatures between 293 and
363 K. These curves resemble those usually found in the literature for glassy po-
lymers [Ward IM & Sweeney J 2004, Richeton J 2006, Richeton J 2007] and can
be divided in four parts:
• An initially linear part corresponding to the elastic response ;
• A part where the curve reaches a maximum in stress, indicating a softening of
the simple with an accompanying drop in stress ;
• A part where the stress is minimum, corresponding to the stationary plastic
flow mode ;
• A final part where the stress increases, corresponding to homogeneous defor-
mation with increasing work hardening prior to breaking.
Figure 1 further presents a comparison between data and the theoretical model.
We remark that all the features of the experimental plastic responses are remarka-
bly well predicted by the molecular model which is therefore able to describe a
full set of large-stress plastic response data.
Figure 2 shows the evolutions of σ , Cd corresponding to T=293 K and
ε =14.10-5 s-1 together with the experimental activation volume, Vexp, determined
according to:
⎡ ⎛ dε vp ⎞⎤
∂ ⎢ln⎜⎜ ⎟⎥
⎟
⎣⎢ ⎝ dt ⎠ ⎦⎥
Vexp = KT (13)
∂σ
174 N. Ouali and A. Ahmed Benyahia
Fig. 1 Effect of the temperature on the plastic behavior of the PMMA in compression:
Results of simulations (solid line), compared with the experimental curves.
Cd is seen to increase right after the appearance of the peak in stress, implying
that, no plastic deformation occurs prior to it. Both Cd and Vexp tend to stabilise
Analysis of Inelastic Behavior of Amorphous Polymers 175
following parallel trends around a time when σ is minimum in the plastic flow
region. The Cd and Vexp values corresponding to such stationary conditions are re-
ferred to as Cd stat and Vexp stat, respectively, and are plotted as a function of temper-
ature in figure 3 together with the stationary experimental activation enthalpy
values, ΔH exp stat , which was calculated from:
⎡ ⎛ dε vp ⎞⎤
∂ ⎢ln⎜⎜ ⎟⎥
⎟
⎣⎢ ⎝ dt ⎠⎦⎥ (14)
ΔH exp stat = KT 2
∂T
Both the time and temperature evolutions of Vexp stat and ΔH exp stat closely re-
semble those measured by LEFEBVRE JM 1982.
Fig. 3 Variation of the stationary values: (a) Cd stat , (b) Vexp stat , (c) ΔH exp stat with the
temperature. Comparison with the experimental values (curves d and e) obtained by
LEFEBVRE [13]
The Cd stat evolutions convincingly show that, while at low temperatures the dy-
namic equilibrium between creation and annihilation of defects stabilises around a
value mainly independent of temperature, that equilibrium becomes more and
more biased towards the annihilation of defects as the temperature increases so
that Cd stat rapidly decreases towards its equilibrium value Cd eq (T) when T ap-
proaches the glass transition temperature, Tg (Tg=380K). The model is thus capa-
ble to explain both regimes using only one set of equations, as is clearly seen in
figure 4 where the stationary value of compression plastic flow stress, , norma-
lised to the shear modulus, G (G=2500 MPa), is plotted as a function of tempera-
ture normalised to Tg.
Figure 4 shows that log (σ stat ) initially decrease linearly with the increasing
G
temperature and decreases more rapidly as T approaches Tg. The good agreement
between the calculated curve and the experimental data points obtained from
figure 1 confirms the validity of the model. The temperature which separates the
two deformation regimes (figure 4) indicates the same separation temperature as
that shown in figure 3 and formally separates the low from the high temperature
behavior.
Fig. 4 Variation of the plastic flow stress with the temperature: comparison of the
calculated results (solid line) with the experimental results of figure 1
Analysis of Inelastic Behavior of Amorphous Polymers 177
5 Conclusion
The model Based on the concepts of quasipoint defects, qpds, nucleation and ex-
pansion of shear microdomains smd, and hierarchically constrained molecular dy-
namics, provides a satisfactory description both of the dynamic mechanical
response, and the sub-Tg mechanical behavior of amorphous polymers when sub-
jected to large stresses under varying conditions of temperature and deformation
rates. Plastic deformation of amorphous polymers could thus be described in terms
of dynamic recovery phenomena.
References
Cady, C., Blumenthal, W., Gray III, G., Idar, D.: Journal de Physique IV 110, 27–32 (2003)
Cavaillé, J., Perez, J., Johari, J.P.: Phys. Rev. B 39(4), 2411 (1989)
Cavaillé, J., Perez, J., Johari, J.P.: J. Non-Crystalline Solids 133, 935 (1991)
Haward, R.N., Young, R.J.: The physics of glassy polymers, 2nd edn. Chapman & Hall,
London (1997)
Lefebvre, J.M.: Ph. D. Thesis, UST Lille, France (1982)
Mangion, M.B.M., Cavaillé, J.Y., Perez, J.: Phil. Mag. A 66, 773 (1992)
Mulliken, A.D.: M.S. Thesis, Dept. of Mech. Eng. Massachusetts Institute of Technology,
Cambridge, MA (2004)
Ouali, N.: Ph. D. Thesis, INSA-Lyon, France (1992)
Perez, J.: Acta. Mat. Sci. 32(12), 2163 (1984)
Perez, J., Cavaillé, J.Y., Etienne, S., Jourdan, C.: Rev. Phys. Appl. 23, 125 (1988)
Perez, J.: Physics and mechanics of amorphous polymers, 324 p. Editions CRC Press, Be-
kema AA, Rotterdam (1998)
Richeton, J., Sahzi, S., Vecchio, K.S., Jiang, F.C., Adharapurapu, R.R.: International Jour-
nal of Solids and Structures 43, 2318 (2006)
Richeton, J., Sahzi, S., Vecchio, K.S., Jiang, F.C., Makradi, A.: International Journal of
Solids and Structures 44, 7938
Ward, I.M., Sweeney, J.: An introduction to the mechanical properties of solid polymers,
2nd edn. Wiley, Chichester (2004)
First Principal Calculations of Optical
Properties of InGaN2 Using in Solar Cells
Applications
Abstract. The aim behind this work is to look into the optical properties of In-
GaN2 with chalcopyrite structure in ternaries compounds. To obtain precise
results, we have used the first-principles calculations, using the full potential-
linearized augmented plane wave method (FP-LAPW) within the density func-
tional theory (DFT). in order to get best and accurate values of the band gap, as
exchange–correlation potential, , we have used the modified Becke – Johnson
(mBJ) of Tran and Blaha, which are based on the optimization of total energy and
corresponding potential. The InGaN2 chalcopyrite demonstrates semiconducting
behavior, with the direct-band gap of 1.68 eV. We have studied the dielectric
function, refractive index, reflectivity, absorption coefficient, and the optical con-
ductivity function. The results we have obtained indicate that InGaN2 is an attrac-
tive and promising material for optoelectronic and photovoltaic applications.
1 Introduction
2 Computational Details
The present calculations are performed using the full-potential linearized aug-
mented plane wave (FP-LAPW) method (G. K. H. Madsen et al, 2001) (K.
Schwarz et al, 2002) integrated in Wien2k package (P. Blaha et al, 2001).
We take into consideration the InGaN2 ternary compounds in body-centered te-
tragonal chalcopyrite structure (space group I ¯ 42d) with 8 atoms in the unit cell,
shown in fig 1, the lattice constant a corresponding to the lattice constant of a cu-
bic zincblende structure, the c/a ratio, and the internal displacement parameter u
revealing the distortion of the anion sublattice due to different surroundings. In the
case of ideal structure, c/a = 2 and u = ¼ (J.P. Perdew et al, 2008).
In order to achieve energy eigenvalues convergence, the wave functions in the
interstitial regions were expanded in plane wave with a cutoff RMT x Kmax = 8,
First Principal Calculations of Optical Properties of InGaN2 181
where RMT is the minimum radius of the muffin-tin spheres and Kmax gives the
magnitude of the largest k vector in the plane wave expansion. Inside the spheres,
the valence wave functions were expanded up to lmax =10. In order to keep the
same degree of convergence for all the lattice constants, we kept the values of the
sphere radii and Kmax constant over all the range of lattice spacing considered. The
RMT chosen values in our calculations for Ga, In and N are respectively 1.9, 2.0
and 1.6 a.u.
The exchange-correlation potential for electronic and linear optical properties
was treated by the modified Becke-Johnson (mBJ-LDA) (F. Tran and P. Blaha,
2009) (D.Koller, F. Tran, and P. Blaha, 2011) (A. D. Becke and E. R. Johnson,
2006) which is seen as a promising tool for accurate determination of the funda-
mental band gaps of wide-band-gap materials (D.Koller et al, 2011) (A. D. Becke
and E. R. Johnson, 2006) (P. E. Blochl et al, 1994)( V. L. Shaposhnikov et al,
2012).
The self-consistent calculations are considered to be converged when the total
energy of the system is stable (P. E. Blochl et al, 1994) (F. Annane et al, 2010)
within 10-4 Ryd. The integrals over the Brillouin zone (BZ) are performed up to
102 k-points, (grid of 10_10_10 meshes, and equivalent to 1000 k-points in the
entire BZ).
In the optical properties calculations we used the equivalent to 20000 k-points
in the entire BZ by the OPTIC code (C. Ambrosch-Draxl and J.O. Sofo, 2006),
implemented in the all-electron Wien2K method (P. Blaha et al, 2001). We have
therefore made a correction ∆ 0.057 eV to the gaps.
DOS
16
14 InGaN2
12 mBJ-LDA
10
8
6
4
Energy (eV)
2
0
-2
-4
-6
-8
-10
-12
-14
-16
Z Γ XP N Γ 20 40 60 80 100
Fig. 2 Electronic band structure of density of state of InGaN2 calculated within mBJ-LDA.
The optical properties are deduced from the dielectric function of the semiconduc-
tor, given by:
Respectively 1(ω) and 2(ω), are, its real and imaginary parts. The dielectric
function can be calculated based on the band-structure results. The imaginary part
of the dielectric function 2(ω) is directly obtained from the electronic structure,
using the joint density of states and the optical matrix overlap. The real part of the
dielectric function 1(ω) is then calculated by the Kramers-Kronig relation (S.
Berrah et al, 2009) (M. Fox, 2001).
For calculation of frequency dependent dielectric function we need energy
eigenvalues and electron wave functions. The energy eigenvalues and electron wave
function are the natural output for the calculation of electronic band structure.
The imaginary dielectric function is an indispensable quantity; it indicates the
varied transitions inter-band in the semiconductor (S. Berrah et al, 2009).
The real and imaginary parts of InGaN2 between 0 and 18 eV are shown in Fig-
ures 3 and 4 respectively, along two main crystallographic directions E//c and
E c. Our calculated ɛ2 (ω) curve of InGaN2 shows a sharp increase at about 1.74
eV at the first onset E0 of the direct optical transitions. The first main peak is si-
tuated at about 2.02 eV, the second one at 5.92 eV, whereas the third, the fourth
and the fifth respectively located at 6.87, 9.82 and 12.21 eV. Our calculated static
value of ɛ1(0) is 5.060.
First Principal Calculations of Optical Properties of InGaN2 183
8
InGaN2
E⊥c
6
E//c
ε1(ω)
2
-2
mBJ-LDA
-4
0 2 4 6 8 10 12 14 16 18
Energy (eV)
Table 1 The calculations peaks (E c) in the imaginary dielectric function ε2(ω), static
optical parameters ε1(ω) and n (ω)
E0 E1 E2 E3 E4 E5 ε1(0) n(0)
InGaN2 1.74 eV 2.02 eV 5.92 eV 6.87 eV 9.82 eV 12.21 eV 5.060 2.253
10
mBJ-LDA E4 InGaN2
E⊥c
8 E//c
E5
E3
6
ε2(ω)
E2
4
E1
E0
2
0
0 2 4 6 8 10 12 14 16 18
Energy (eV)
The refractive index n is a very important physical parameter related to the micro-
scopic atomic interactions. Its values are often required to interpret various types
of spectroscopic data. The refractive index of a semiconductor, n (ω) is computed
through the real dielectric function (S. Berrah et al, 2009):
n = ε1 (2)
We have presented on fig 5 the refractive index of InGaN2. The calculated static
components of refractive index are presented in Table 1. The static values of the
184 H. Bennacer et al.
refractive index for low frequency described by the relation n(0)=√ε1(0), n(ω)
having value 2.253 and n( ) having value 2.248 satisfy the relation n(0)=√ε1(0).
The two polarizations non zero components of refractive index show maximum
value at 5.77 eV at n(ω) and 9.74 eV at n(ω) .
3.0
InGaN2 E⊥c
E//c
refraction index 2.5
2.0
1.5
1.0
0.5
mBJ-LDA
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26
Energy (eV)
3.1.3 Reflectivity
In fact, reflectivity of light given by equation (3) (Sadao Adachi, 2009), is one of
the most important parameters in linear optical calculations, it is sensitive to a
complicated combination of 1(ω) and 2(ω) (Mohamed Issam Ziane et al, 2015).
We plot the reflectivity R(ω) calculated within mBJ-LDA functional as a function
of photon energy in Fig. 6. The zero frequency reflectivity is 0.148. To be used in
photovoltaics, a material should satisfy some requirements: to be a direct-gap sem-
iconductor, and to have a sizable band gap value (more than 1.2 eV), small
reflectance, and large absorption coefficient to absorb the essential part of the visi-
ble light spectrum (V.L.Shaposhnikov et al, 2012).
0,4
Reflectivity
0,3
0,2
0,1
0,0
0 2 4 6 8 10 12 14 16 18 20
Energy (eV)
However, the compound with a band gap higher than 1.5 eV may absorb less of
visible spectrum (V.L.Shaposhnikov et al, 2012). Obviously from Fig. 6, the max-
imum static value of reflectivity (reflectivity at 0 eV) located at about 7.42 eV.
The optical absorption coefficient α(ω) is one of the most important evaluation
criterions for the photoelectric materials (A. Boukortt et al, 2010) (A. Bouhema-
dou et al, 2013) (Omkar Jani et al, 2005) (B. Amin et al, 2012). Our calculated ab-
sorption coefficient spectra α(ω) show in fig 7, that InGaN2 have a good optical
absorption in a wide energy range (2-17 eV). The absorption is determined by the
expression:
4π
α (ω ) = k (ω ) (4)
λ
The maximum value of the absorption coefficient shifts from 12.36 eV (E c) to
12.98 eV (E//c). There is a high absorption between 1.76 eV and 13 eV, corres-
ponding to wave lengths between 704.455 to 95.372 nm.
275
InGaN2 E⊥c
250
E//c
225
absorption in [10^4 / cm)]
200
175
150
125
100
75
50
25
mBJ-LDA
0
0 2 4 6 8 10 12 14 16 18 20 22
Energy (eV)
The optical conductivity given by equation (5) is deduced from the dielectric func-
tion (S. Berrah et al, 2009):
iω
σ (ω ) = − ε (ω ) (5)
4π
The curve of the optical conductivity versus energy is shown in fig 8. We
represent the optical conductivity spectra using the imaginary part of the dielectric
function in range of energies between 0 and 18 eV. The curve presents several
peaks corresponding to the bulk plasmon excitations which are caused by elec-
trons crossing from the valence to the conduction band. The main peak is situated
in 9.80 eV and the total width of the optical conductivity is 13.33 eV.
186 H. Bennacer et al.
14000
InGaN2 E⊥c
12000 mBJ-LDA E//c
8000
6000
4000
2000
0
0 2 4 6 8 10 12 14 16 18
Energy (eV)
4 Conclusion
References
Laref, A., Altujar, A., Luo, S.J.: Eur. Phys. J. B 86, 475 (2013)
De Vos, A.: Endoreversible Thermodynamics of Solar Energy Conversion, p. 90. Oxford
Univ. Press, Oxford (1992)
Jani, O., Ferguson, I., Honsberg, C., Kurtz, S.: Design and characterization of GaN/InGaN
solar cells. Appl. Phys. Lett. 91(13), 132117 (2007)
Jampana, B.R., Melton, A.G., Jamil, M., Faleev, N.N., Opila, R.L., Ferguson, I.T., Hons-
berg, C.B.: IEEE Electron Device Letters 31(1) (January 2010)
Xu, B., Han, H., Sun, J., Yi, L.: Physica B 404, 1326–1331 (2009)
Mao, Y., Jun-Jie, S., Wei-Ning, M., Lu, W., Min, Z.: Chin. Phys. Lett. 29(10), 107301
(2012)
Madsen, G.K.H., Blaha, P., Schwarz, K., Sjostedt, E., Nordström, L.: Phys. Rev. B 64, 64
(2001)
Schwarz, K., Blaha, P., Madsen, G.K.H.: Compute. Phys. Commun. 147, 147 (2002)
Blaha, P., Schwarz, K., Madsen, G.K.H., Kvasnicka, D., Luitz, J.: WIEN2k, An Augmented
Plane Wave Plus Local Orbital’s Program for Calculating Crystal Properties (Vienna
University of Technology, Vienna, Austria (2001) ISBN 3-9501031-1-2
Perdew, J.P., Ruzsinszky, A., Csonka, G.I., Vydrov, O.A., Scuseria, G.E., Constantin, L.A.,
Zhou, X., Burke, K.: Phys. Rev. Lett. 100, 136406 (2008)
Tran, F., Blaha, P.: Phys. Rev. Lett. 102, 226401 (2009)
First Principal Calculations of Optical Properties of InGaN2 187
Koller, D., Tran, F., Blaha, P.: Phys. Rev. B 83, 195134 (2011)
Becke, A.D., Johnson, E.R.: J. Chem. Phys. 124, 221101 (2006)
Blochl, P.E., Jepsen, O., Anderson, O.K.: Phys. Rev. B 49, 16223 (1994)
Shaposhnikov, V.L., Krivosheeva, A.V., Borisenko, V.E.: Phys. Rev. B 85, 205201 (2012)
Annane, F., Meradji, H., Ghemid, S., El Haj Hassan, F.: Computational Materials
Science 50, 274–278 (2010)
Ambrosch-Draxl, C., Sofo, J.O.: Computer Physics Communications 175, 1–14 (2006)
Jansen, H.J.F., Freeman, A.J.: Phys. Rev. B 30, 561 (1984)
Hohenberg, P., Kohn, W.: Phys. Rev. 136, B64 (1964)
Sham, L.J., Kohn, W.: Phys. Rev. 145, 561 (1965)
Perdew, J.P., Wang, Y.: Phys. Rev. B 45, 13244 (1992)
Berrah, S., Boukortt, A., Abid, H.: Physica E 41, 701–704 (2009)
Fox, M.: Optical Properties of Solids, p. 6. Oxford University Press, New York (2001)
Boukortt, A., et al.: Physica B 405, 763–769 (2010)
Bouhemadou, A., et al.: Solid State Sciences 20, 127–134 (2013)
Jani, O., Honsberg, C., Asghar, A., Nicol, D., Ferguson, I., Doolittle, A., Kurtz, S.: The
31st IEEE Photovoltaic Specialists Conference, January 3-7 (2005)
Amin, B., et al.: Opto-electronic response of spinels MgAl2O4 and MgGa2O4 through
modified Becke-Johnson exchange potential. Physica B 407, 2588–2592 (2012)
Adachi, S.: Properties of Semiconductor Alloys: Group-IV, III–V and II–VI Semiconduc-
tors. John Wiley & Sons, Ltd. (2009) ISBN: 978-0-470-74369-0
Ziane, M.I., Bensaad, Z.: Tarik Ouahrani and Hamza Bennacer. Materials Science in Semi-
conductor Processing 30, 181–196 (2015)
Agent-Based Approach for the Optimal Design
of Mechatronic Systems
1
Engineering Laboratory of Mechanical Systems and Materials LISMMA, SUPMECA-
Paris, 3 rue Fernand Hainaut, 93407 Saint-Ouen, France
amir.guizani@live.fr,
{moncef.hammadi,jean-yves.choley,thierry.soriano}@supmeca.fr
2
Mechanics Modeling and Production Research Laboratory LA2MP,
National School of Engineers of Sfax (ENIS), B.P, 1173, 3038 Sfax, Tunisia
{slim.abbes,mohamed.haddar}@enis.rnu.tn
1 Introduction
We suppose that this process begins with DA1, then DA2 and terminates with
DA3. In the first step, the coordinating agent defines the initial values of y13 and y12
for DA1 to perform the first local optimization and generate y31 and y21. Then DA2
uses y21 and an initial value of y23 for its local optimization to generate y12 and y32.
Then, DA3 uses y31 and y32 to generate y13 and y23. In the next iteration, every DA
uses the previously generated values for the new local optimizations. For mono-
objective problems, CA calculates the difference between yij and yji (||yij-yji||2) at
very iteration and the optimization process stops when the error is less than a pre-
defined minimum error. For multi-objective partitions, the Pareto-front is nar-
rowed by determining the intersection of fronts between the previous iteration and
the next one until convergence. For both cases of mono-objective or multi-
objective partitions, if no solution is found the stop criteria could be a number of
iterations not to be exceeded.
4.1 Introduction
We consider the case of a preliminary design of an electric vehicle. The objective
of this study is to optimize the electric motor and a one-ratio gear box to fulfill the
performance requirements related to the maximum velocity, gradeability and acce-
leration test.
The diameter d1 and the number of teeth on the pinion Z1 are given by:
2τ m .KV .K B .Kαβ 1/3
d1 ≥ ( )
Rg
Ω0 .ψ . 2 2
.Z p .ZV
Rg + 1 (2)
⎛ σ ⎞ Rg + 1 1
Z1 < Z1max = ⎜ 0 ⎟ . . 2 2
⎝ Ω 0 ⎠ Rg Z p .ZV
Where τm is the resultant torque, Ω0 is the criterion of superficial resistance of a
gear tooth, σ0 is the criterion of resistance to the root, ψ is the length coefficient of
a pinion, Rg is the gearbox ratio, KV is the dynamic speed factor, KB is the service
factor, Kαβ is the load distribution factor, ZV is the oil film factor and ZP is the
roughness factor.
The diameter d2 and the width b of the gear are determined by:
d 2 = Rg .d1
(3)
b = ψ .d1
In our case, cementation alloy steel 20 MC 5 is chosen. The characteristics of the
material used to make the gear are shown in (Le Borzec 1992).
Fig. 3 Comparison between input NEDC profile and output vehicle velocity
Using a classic optimization approach, by defining one analysis model for the
three test cases (Acceleration test, Maximum velocity, Gradeability), is a complex
task and computationally expensive. Therefore we will use the multi-agent ap-
proach for this purpose.
1
http://www.phoenix-int.com/software/phx-modelcenter.php
196 A. Guizani et al.
agent compares the two optimal Pareto fronts generated by DA1 and DA2 and
keeps the common solution for both design agents.
The process can be repeated until convergence. If the problem has no solution
an empty Pareto front will be returned by the coordinating agent.
2
http://jade.tilab.com/
Agent-Based Approach for the Optimal Design of Mechatronic Systems 197
The designer could choose one optimal configuration in the Pareto-front. For
example, results of simulation for the optimal configuration corresponding to
Km=0.31 and Rg=10.5, for which we have V10@0%=100.1 km/h, Vmax@0%=140.08
km/h and V10@30%=20.02 km/h (figure 5). The maximum pick of the required elec-
tric power is 190.6 kW, the maximum torque is 202 N.m and the maximum volume
of the gearbox is 0.00205 m3.
5 Conclusion
In this paper we have presented a new approach for optimizing mechatronic sys-
tems based on multi-agent technology. We described our approach and we
presented an example on how to deal with the optimization of an electric vehicle.
Instead of integrating all models relative to all test cases, we have decomposed the
design process to two design agents and we have affected to each agent its objec-
tives and constraints. The coordinating agent has also been considered to
coordinate between design agents to find the overall optimal design. Our main
contribution in this study is to propose an agent approach that allows designers to
determine optimal mechatronic design by decomposing the complex problem and
by coordinating the local optimizations performed by design agents.
Acknowledgements. The authors would like to thank the researchers of the Linz Center of
Mechatronics (LCM), who have contributed in this work with their helpful comments and
suggestions.
198 A. Guizani et al.
References
Allison, J., Kokkolaras, M., Papalambros, P.Y.: Optimal partitioning and coordination deci-
sions in decomposition-based design optimization. Journal of Mechanical De-
sign 131(8), 081008 (2009)
Fritzson, P.: Principles of object-oriented modeling and simulation with Modelica 2.1. John
Wiley & Sons (2010)
Hammadi, M., Choley, J.Y., Penas, O., Louati, J., Rivière, A., Haddar, M.: A and Haddar
M Layout optimization of power modules using a sequentially coupled approach. Inter-
national Journal of Simulation Modelling (IJSIMM) 10(3) (2011)
Hammadi, M., Choley, J.Y., Penas, O., Rivière, A.: Mechatronic system optimization based
on surrogate models-application to an electric vehicle. In: SIMULTECH, pp. 11–16
(2012)
Guizani, A., Hammadi, M., Choley, J.Y., Soriano, T., Abbes, M.S., Haddar, M.: Multidis-
ciplinary approach for optimizing mechatronic systems: Application to the optimal de-
sign of an electric vehicle. In: 2014 IEEE/ASME International Conference on Advanced
Intelligent Mechatronics (AIM), pp. 56–61 (2014)
Guizani, A., Hammadi, M., Choley, J.Y., Soriano, T., Abbes, M.S., Haddar, M.: Multidis-
ciplinary optimization of mechatronic systems: Application to an electric vehicle. In:
Mechatronic Systems: Theory and Applications, pp. 1–14. Springer (2014)
Le Borzec, R.: Réducteurs de vitesse à engrenages. Ed. Techniques Ingénieur (1992)
Plume, J., Mitchell, J.: Collaborative design using a shared ifc building modellearning from
experience. Automation in Construction 16(1), 28–36 (2007)
A Digital Pattern Approach to the Design
of an Automotive Power Window
by means of Object-Oriented Modelling
Abstract. The paper deals with a digital pattern (DP) approach to the design of an
automotive power window, using object-oriented modelling. Therefore, the paper
faces the designing of a mechatronic system by using an integrated approach to
product development. Then, Dymola/Modelica environment is used as a tool of a
decision support system that makes possible the DP approach. The paper briefly
sum up the results of simulations related to a power window system characterized
by a double bowden sliding mechanism. Finally, the paper highlights the parame-
ters that could be easily integrated in a graphical user interface, aimed to reduce
both the development time of new power window system and to increase the accu-
racy of design activities.
1 Introduction
The object-oriented modelling (OOM) aimed to the design of multi-domain systems
is, today, widely used. The OOM, in fact, is used to model and simulate possible de-
sign solutions related to different domains (e.g. a mechanical or a controller
domain), not only in a bounded domain. The OOM of multi-domain systems is per-
formed by continuously switching between the domains. Then, simulations allow
easily exploring and comparing several design changes (van Amerongen and
Breedveld, 2003).
The OOM is significant, for example, in case of mechatronic systems defined
as “engineering systems with mechanical, electrical, hydraulic or pneumatic com-
ponents, including interactions of physical effects from various energy domains”
(Borutzky, 2009). Such systems, in fact, due to their multi-disciplinary nature in-
volve several competencies and need for a team of designers with different back-
grounds. Therefore, the modelling of mechatronic systems must support standard-
ized languages and tools for model management and model abstraction (Sinha
et al, 2001). In the last year, several research activities have been focused on such
integration of languages and tools, as well as on the possibility to design a mecha-
tronic system as a whole (Chhabra and Reza Emami, 2011). In particular, the
powerful bond graph representations have been used to deem mechatronic systems
as energy systems and to define three holistic design criteria (namely energy, en-
tropy and agility) to be used in the designing of the mechatronic system when the
power energy, the thermodynamic aspects or the operative aspects of the system,
respectively, are the most significant.
On the other side, industrial companies do prefer to use OOM and multi-
domain tools based on Modelica language for their design activities, without ac-
complishing a deep insight into modelling and simulation (Zupancic and
Sodja, 2011). In this context, also the availability of powerful libraries can help
companies to develop complex models using Dymola-Modelica environment, but
methodological tools are necessary to reduce times and improve target of design-
ing activities. Some large companies, in fact, in automotive and railway fields,
have deepen the tools and methods to accomplish “digital patterns for product de-
velopment” viz. geometrical data and models, to be reconfigured and re-used in
different but similar design activities (Lanzotti, 2013) (Patalano, 2013). These pat-
terns are generally supported by a series of tools, as for example, “quality and
standard checkers”, “predictive engineering wizards” and “cost advisors” that are
part of a decision support system. Then, Dymola-Modelica environment and re-
lated tools could be used to accomplish the decision support system able to guide
the designer during his activities.
A typical mechatronic system to be frequently developed in automotive field is
the power window system. Some research activities were carried out on general
methodological approaches aimed to design the power window controller, as for
example in (Mustafiz et al, 2012). On the opposite, further activities were carried
out for the development of pinch detection algorithm related to the control system
(Ra et al, 2008). No contributes are known as regards the OOM of the whole sys-
tem, according the above mentioned digital pattern approach for product devel-
opment.
The paper is arranged as follows. Section 2 presents the approach and simula-
tion tools used to accomplish the simulations of a mechatronic system. Section 3
summarizes the specifications of the power window system taken into account.
Finally, Sections 4 and 5 draw results of simulations and conclusions.
The Digital Pattern (DP) approach aims to develop models, to be reconfigured and
re-used in similar designing tasks. Therefore, the DP models are geometrical and
numerical models, preconfigured and parametrical, able to be adapted according
the constraints of a new designing task. Then, the designer, by means of DP
models, reduces the development time of new complex products and increases the
A Digital Pattern Approach to the Design of an Automotive Power Window 201
accuracy of his results as the tools of a decision support system drive him. Specif-
ic models, international standards and company rules are used to develop such
tools. Figure 1 depicts the DP approach. In particular, the exchange of contents be-
tween parametrical CAD models and the tools of the decision support system is
cyclic as it deals with the continuous development of each product aimed to fulfill
the system requirements. A graphical user interface (GUI) could be developed to
support designers’ choices, to collect company knowledge and to accomplish veri-
fications.
Mathematical Models
Tools for Decision
Company Rules Support System
Quality and standard checkers
International Predictive Engineering Wizard
Standards
⊆X Cost advisor
Dymola/Modelica simulations
.....
Fig. 1 Contents and operational flow of the DP approach for product/process development
Using this holistic vision is then possible to decompose each subsystem to accom-
plish the modeling of groups and components, respectively, taken into account es-
sential simplification hypotheses.
The window is fixed on two supports respectively constrained along two rails.
The user, moving upwards or downwards the button, generates a signal, which is
processed by controller that drives the DC motor. The motor wraps two bowden
cables, which move two supports and, consequently, the automotive window. The
basic operations of the motor are accomplished by reversing the polarity of its
power and ground input.
The power window system could be modelled by applying a top down ap-
proach to OOM. A series of levels of detail could be taken into account (Figure 3).
First, the system was divided into three subsystems related to Control, Sensor and
Physic subsystem, respectively. Then, each subsystem was further detailed using
functional groups. Finally, the level related to the components was defined.
A Digital Pattern Approach to the Design of an Automotive Power Window 203
Level 0
system
Power Window
System
subsystems
Level 1
Level 2
groups
components
Level 3
Fig. 3 Functional levels related to power window system using a top-down approach
The power window system was modeled using Modelica language within Dy-
mola environment. Modelica Standard Libraries and ad hoc developed objects
were used. The model of the power window system is showed in Figure 4.
4 Results
The results of simulationss are briefly summed up in this section. In particular, thhe
trends of some dependentt parameters are represented when the power of DC moo-
tor and speed ratio changee.
The trends of dependeent parameters were obtained through a simulation of thhe
button tap and release, vizz. when the window goes all the way up and down. Winn-
dow mass has been set on o 3 kg, max supply voltage on 12V, speed ratio on 1.8,
winding pulley diameter on 130mm and window stroke on 400mm. The requireed
time of transition of the window
w is about 3s.
Figure 6 shows, respecctively, the window displacement and speed as well as thhe
forces acting on front andd rear support. It can be seen how the lifting (lower) tim
me
decreases with increasing g powerful, thus also for the speed and force amplitudees.
Then, designers could diirectly use values of simulated DC motor power, to ac-
complish the dimensionin ng of the system. Overall dimensions of DC motor changge
when power increases. Th herefore, also fixture shape and pins change, consequennt-
ly. The characteristics of assembling surfaces and pins are then automatically proo-
vided to designers, by meeans of a datasheet, accomplished in Excel environmennt
(Figure 7).
Fig. 7 Operative flow existiing between DC motor independent parameters and dependeent
parameters plus assembly ch
haracteristics
Figure 8 depicts the trrends of the parameters for three different values of thhe
speed ratios. It highlights the need for a more powerful DC motor when speed ra-
tio decreases. Powers of 96W,
9 50W and 38W are necessary when the speed ratiio
is equal to 1, 2 and 3, resspectively. This behavior is due to the fact that the exteer-
nal loads do not vary; therrefore, it is necessary to balance the external loads to fuul-
fill the requirement of lifting and lower of the window within 3s.
Fig. 8 a) Window displacem ment; b) Window speed; c) Force on front support; d) Force oon
rear support. Curves related to speed ratio values equal to 1 (solid line), 2 (dashed line) orr 3
(dotted line), respectively.
5 Conclusions an
nd Further Works
Acknowledgements. The present work was developed with the economic support of MIUR
(Italian Ministry of University and Research) performing the activities of the project
PON01_01268 “Digital pattern product development: a pattern driven approach for indus-
trial product design”.
References
Borutzky, W.: Bond graph modelling and simulation of multidisciplinary systems – An in-
troduction. Simulation Modelling Practice and Theory 17, 3–21 (2009),
doi:10.1016/j.simpat.2007.08.008
Chhabra, R., Reza Emami, M.: Holistic system modeling in mechatronics. Mechatron-
ics 21, 166–175 (2011), doi:10.1016/j.mechatronics.2010.10.003
Lanzotti, A., Patalano, S., Vitolo, F.: A Graph-based Software Tool for the CAD Modeling
of Mechanical Assemblies. In: Proc. of 8th International Conference on Computer
Graphics, Theory and Applications, GRAPP 2013 IVAPP 2013, Barcelona, Spain, Feb-
ruary 21-24, pp. 60–69. SciTePress (2013), doi:10.5220/0004299000600069
Mustafiz, S., Joachim Denil, J., Lucio, L., Vangheluwe, H.: The FTG+PM framework for
multi-paradigm modelling: An automotive case study. In: Proceedings of 6th Interna-
tional Workshop on Multi-Paradigm Modeling – MPM 2012, Innsbruck, Austria (2012)
Patalano, S., Vitolo, F., Lanzotti, A.: A graph-based approach to CAD modeling: a digital
pattern application to the sizing and modeling of manual transverse gearboxes. In: Proc.
of International Conference on Graphic Engineering, Madrid, Spain, June 19-21 (2013)
Ra, W.S., Lee, H.J., Park, J.B.: Practical pinch detection algorithm for smart automotive
power window control systems. IEEE Transactions on Industrial Electronics 55, 1376–
1384 (2008)
Shina, R., Peredis, C.J.J., Liang, V.C., Khosla, P.K.: Modeling and simulation methods for
design of engineering systems. Journal of Computing and Information Science in Engi-
neering 1, 84–91 (2001), doi:10.1115/1.1344877
Van Amerongen, J., Breedveld, P.: Modelling of physical systems for the design and con-
trol of mechatronic systems. Annual Review Control 27, 87–117 (2003),
doi:10.1016/S1367-5788(03)00010-5
Zupančič, B., Sodja, A.: Advanced multi-domain modelling: advantages and dangers. In:
UKSim 13th International Conference on Modelling and Simulation (2011),
doi:10.1109/UKSIM.2011.56
Numerical Study of the External Excitation
Amplitude Effect on Liquid Sloshing
Phenomenon
Abstract. In this work, we are interested on the study the liquid sloshing pheno-
mena in a tank subjected to a sinusoidal excitation. Particularly, we have investi-
gated the amplitude movement effect. The hydrodynamic parameters describing
the flow like the velocity and the static pressure are presented. The numerical re-
sults are investigated for four different values of sinusoidal excitation amplitude.
The results show that the behavior of the sloshing phenomenon strongly depends
on the amplitude value of the external excitation. A good agreement has been
shown by comparing the numerical and the experimental results.
1 Introduction
The liquid sloshing in rigid containers has been the interest of many studies be-
cause of its frequent application in many engineering sections such as aerospace
vehicles, transportation engineering and maritime applications. The sloshing phe-
nomenon in the tank subjected to external movement was investigated analytical-
ly, numerically and experimentally. Many researchers are interested to studying
the effect of the baffle on the sloshing. For example, Liu et al. (2009) studied the
effect of the vertical and horizontal baffle in the liquid sloshing in a tank. They
showed that sloshing is more reduced using the vertical baffle than using the hori-
zontal baffle. Bouabidi et al. (2013) studied the effect of the vertical baffle height
on liquid sloshing in an accelerating rectangular tank. They showed that the liquid
sloshing decrease with the increase of the vertical baffle height. Panigrahy et al.
(2009) studied experimentally the effect of the three different baffles, horizontal,
vertical and ring baffle. They showed that the ring baffle reduced the sloshing
more than the conventional baffle. The type of the external excitation also affects
the behavior of the liquid motion. Therefore, the sloshing is studied for many
types of excitations. Yan et al. (2009) investigated experimentally the effect a lat-
eral and longitudinal excitations on the sloshing of the liquid in a partially filled
tank. Koh et al. (2013) studied the phenomena of the liquid sloshing in a container
subjected to a sinusoidal movement. Akyıldız et al. (2013) analyzed the sloshing
phenomena in a tank subjected to rotational flow for different values of rolling
frequencies. Ji et al. (2012) investigated the motion of the liquid in rectangular
tank subjected to a lateral oscillation. Jin et al. (2014) carried out a series of expe-
riments to study the phenomenon of sloshing subjected to a sinusoidal excitation.
They studied the effect of the frequency and the amplitude of the external excita-
tion on the liquid sloshing. They showed that the characteristic of the excitation
affects considerably the sloshing in the tank.
In this paper, we are interested to study numerically the effect of the amplitude
excitation on the sloshing phenomena in a rectangular tank subjected to a sinu-
soidal movement. The VoF method is used to predict the free surface elevation
and the Navier- Stockes equations are solved in conjuction with k-ε turbulence
models. The free surface elevation and the magnitude velocity are predicted and
analyzed for the different cases.
2 Geometry System
3 Numerical Method
The commercial CFD code "Fluent" has been used in the numerical simulation of
the liquid sloshing phenomenon in the tank. The VoF method (Volume of Fluid) is
Numerical Study of the External Excitation Amplitude Effect 211
used to define the deformation of the free surface over time. For the motion of the
tank, a User Defined Function (UDF) is developed on language “C++” and inter-
preted in “Fluent”. The Navier-Stokes equations and the standard k-ε turbulence
model equations were solved using a finite volume discretization method.
The problem of liquid sloshing is governed by the continuity and the momen-
tum equations as follows:
∂ρ ∂
+ ( ρ ui ) = 0 (1)
∂t ∂xi
∂ ∂ ∂p ∂ ∂u ∂u j
( ρ ui ) + ( ρ ui u j ) = − + μ( i + ) + Fi (2)
∂t ∂x j ∂xi ∂x j ∂x j ∂xi
Ф ( xi , t ) > 0 , in air
Ф ( xi , t ) = 0 , on surface (6)
Ф ( xi , t ) < 0 , in water
4 Numerical Results
show that the violence of the sloshing increases with the increase of the amplitude
value. The most free surface deformation has been observed for A=A4 and the
minimum for A=A1. For the t=T/4, the liquid moves from right to left. For t=T/2
and t=T, when the external excitation changes the sense, the liquid appears stable
and the free surface evolution has been observed minimal. However for t=3T/4, it
moves from the left to the right.
and the recirculation zones become important for the different amplitudes. The re-
sults show that the value of the velocity depends on the amplitude value. It in-
creases with the increase of the amplitude. Except at t=T/4, the velocity for A1 is
more important than A2 and A3. The location of the maximum value of the veloci-
ty varies over time and also according to the external excitation amplitude.
The evolution of the static pressure over time is presented in figure 10 for the dif-
ferent values of amplitude considered in this study A1, A2, A3 and A4. According
to these results, it has been observed that the static pressure varies periodically
with time. These results are expected since the tank is subject to a periodic sinu-
soidal movement. Indeed, it is clear that the static pressure increases with the
increase of the external excitation amplitude. The maximum value has been ob-
served for A=A4. However, the minimum pressure is obtained for A=A1.
p (Pa)
t (s)
The period of the static pressure variation depends also on the external excitation
amplitude. For the case of A=A1, the numerical results are compared with the ex-
perimental results of Panigrahy at al. (2009). According to these results, a good
agreement was shown which confirms the validity of our numerical model.
6 Conclusion
In this work, we have studied the external excitation amplitude effect on liquid
sloshing phenomenon in a rectangular tank. The tank is subjected to an external
sinusoidal excitation. Our numerical results are compared with experimental re-
sults and a good agreement is shown. The free surface evolution, the distribution
of the magnitude velocity and the static pressure profile are presented and ana-
lyzed. It appeared that the motion of the liquid in the tank subjected to the exter-
nal excitation considerably depends to the value of the amplitude. When the
amplitude value of the external motion increase, the sloshing become more impor-
tant. The static pressure also depends on the amplitude value. It increases with the
increase of the amplitude value. For the velocity, it appears that the recirculation
zones become more important over time and generally located on the free surface
of the liquid. The location of maximum value of the velocity changes over time
and with the amplitude value. The highest value of the velocity has been observed
for the maximum value of amplitude. However, the minimum value has been
noted for the minimum amplitude value.
In the future, we propose to study the liquid sloshing phenomenon in a battery cell
with a mixing element.
References
Akyildiz, H., Ünal, N.E.: Sloshing in a three-dimensional rectangular tank: numerical simu-
lation and experimental validation. Ocean Engineering 33, 2135–2149 (2006)
Bouabidi, A., Driss, Z., Abid, M.S.: Vertical Baffles Height Effect on Liquid Sloshing in an
Accelerating Rectangular Tank. International Journal of Mechanics and Applications 3,
105–111 (2013)
Ji, Y.M., Shi, Y.S., Park, J.S., Hyun, J.M.: Experiments on non-resonant sloshing in a rec-
tangular tank with large amplitude lateral oscillation. Ocean Engineering 50, 10–22
(2012)
Liu, D., Lin, P.: Three dimensional liquid sloshing in a tank with baffles. Ocean Engineer-
ing 36, 202–212 (2009)
Jin, H., Liu, Y., Li, J.H.: Experimental study on sloshing in a tank with an inner horizontal
perforated plate. Ocean Engineering 82, 75–84 (2014)
Koh, C.G., Luo, M., Gao, M., Bai, W.: Modelling of liquid sloshing with constrained float-
ing baffle. Computers and Structures 122, 270–279 (2013)
Panigrahy, P.K., Saha, P.K., Maity, U.K.: Experimental studies on sloshing behavior due to
horizontal movement of liquids in baffled tanks. Ocean Engineering 36, 213–222 (2009)
Yan, G., Rakheja, S., Siddiqui, K.: Experimental study of liquid slosh dynamics in a partial-
ly-filled tank. J. Fluids Eng. 131(7) (2009)
The Study of the Hardening Precipitates and the
Kinetic Precipitation. Its Influence on the
Mechanical Behavior of 2024 and 7075
Aluminum Alloys Used in Aeronautics
Ahmed Ben Mohamed, Amna Znaidi, Moez Baganna, and Rachid Nasri
Abstract. The Research about the aluminum alloys has been in progress. Alumi-
num alloys, in contrast to the pure aluminum have very high mechanical characte-
ristics of substantial resistance. These mechanical performances combined with
lightness of the structure there of has a variety of uses especially in the transport
sector. The aluminum alloys of the 2000 series (Cu is a main part of addition) and
7000 (Zn is the main part of addition) are essentially aeronautics alloys, hence its
strong presence Air in the design of bus where Al 2024 and Al 7075 are the major
part (Figure 1).
Both alloys (2024-7075) are hardening structural alloys and not by strain work-
ing. Their mechanical properties are of particular microstructure obtained from
thermo-mechanical treatments. Among other things, this is a complicated micro-
structure which is responsible for these mechanical performances. It is interesting
to examine the description of this complex phenomenon of the hardening precipi-
tation, including areas of Guinier-Preston (GP Zone) which are very dependent on
the alloy’s composition and thermo-mechanical treatments they undergo. The
hardening phases (A.Deschamps2001) of 2024 and 7075 aluminum alloys are re-
spectively (Al2Cu - Al2CuMg) and (MgZn2). These phases are obtained by preci-
pitation of the sequence which depends on the proportions of elements alloying
elements. An experimental tensile tests companion of these heat treated aluminum
alloys (quenching) will allow us to understand the kinetic precipitation and its in-
fluence on the mechanical properties (Rp0.2, Rm and A%), dice quenching cool
until maturation (T4).
1 Introduction
7075 following materials and different load tests, which will later on visualize the
anisotropy of these materials and to accurately describe the evolution of hardening
precipitates that are responsible for the evolution of the mechanical properties of
the aluminum alloys. These experimental tests are carried out for different direc-
tions of test pieces in the following sequences ( fresh quenching, after 1 hour of
quenching, after 2 hours of quenching, one day after quenching and after 7 days of
quenching). The 2024 and 7075 aluminum alloys are hardening alloys (age har-
dening) which makes interesting description of the evolution of hardening precipi-
tates and their effects on the mechanical properties of alloys to achieve maturation
and stabilization.
Al 2024 Cu Mn Mg Rp0.2 Rm A%
T3 3.8- 4.9 0.3-0.9 1.2-1.8 300MPa 440MPa 9
Al 7075 Cu Mg Zn Rp0.2 Rm A%
T7 1.2- 2.0 2.1-2.9 5.1-6.1 390MPa 475MPa 7
Fig. 3a
Fig. 3b
Fig. 3c
224 A.B. Mohamed et al.
Fig. 4a
Fig. 4b
Fig. 4c
The Study of the Hardening Precipitates and the Kinetic Precipitation 225
Fig. 5a
Fig. 5b
Fig. 5c
226 A.B. Mohamed et al.
Fig. 6a
Fig. 6b
Fig. 6c
The Study of the Hardening Precipitates and the Kinetic Precipitation 227
4 Conclusion
In this paper, we studied the behavior of 2024 and 7075 aluminum alloys sub-
jected to tensile tests in different directions. The change in directions of loading of
the specimens and the precipitation kinetics showed a significant influence on the
mechanical characteristics of these aluminum alloys.
From these interpretations, we can summarize the behavior of aluminum Al
2024 and Al 7075 plastic deformation which is a basic process for making alloys
in the aerospace industry:
Al 2024 is a material forming and deep drawing by excellence and especially to
its equilibrium state T4 (after 7 days of fresh quenching). While the 7075 Al is
228 A.B. Mohamed et al.
better for plastic forming and deep drawing between fresh quenching and 1 hour
to fresh quench.
This work will allow us later to better optimize the behavior of aluminum al-
loys to plastic deformation and forming by deep drawing by adopting new identi-
fication behavior laws and calculating the coefficients of anisotropy strategies and
hardenings. Further visualizations and by transmission electron microscopy mi-
crographs explorations (MET) in these types of alloys will be discussed.
References
Barralis, J., Maeder, G.: Précis métallurgie: élaboration, structure, propriété, normalisation.
Édition Afnor/Nathan, Paris (2002)
Wilm, A.: Recherche sur la métallurgie physique des alliages d’aluminium-magnésium.
Métallurgie 8, 225 (1911)
Deschamps, A., Militzer, M., Poole, W.J.: Precipitation kinetics and hardening in a Fe-Cu
alloy. ISIJ International (Japan) 41(2), 196–205 (2001)
Deschamps, A., Bigot, A., Auger, P., Bréchet, Y., Livet, F.: A comparative study of preci-
pitate composition in Al-Zn-Mg using Tomographic Atom Probe and Small-Angle X-
Rays Scattering. Phil. Mag. A 81(10), 2391–2414 (2001)
Dumont, D., Deschamps, A., Bréchet, Y.: On the relationship between microstructure,
strength and toughness in AA7050 aluminum alloy. Materials Science and Engineering:
A00 1-/11 (2003)
Znaidi, A.: Stratégie d’identification de lois de comportements anisotropes: Applications
aux alliages d’aluminium. Toulouse (2010)
Znaidi, A.: Formulation et identifications des lois de comportements anisotropes d’un al-
liage d’aluminium 2024T3. Marseil (2009)
Baganna, M., Ben Mohamed, A., Znaidi, A., Guellouz, S.: Nasri, Etude des alliages
d’aluminium utilisés dans le domaine aéronautique: comportement mécanique de
l’alliage d’aluminium, T351. Cotume, Tunisie (2012)
Multidisciplinary Design Optimization in Small
Satellite
Abstract. Many industrial tools for simulation developed during these last years
still have an impact limited on the complex systems design such as satellites, air-
crafts, engines or rockets. Whereas the modeling tools of each discipline or sub-
systems are refined gradually, the total optimization of these systems encounters
difficulties of orders at the same time methodological, organizational, data-
processing and numerical. Actually and in the best of the cases, the complex sys-
tems are optimized by subsystems. Savings of time of design and performances
are accessible if one manages to optimize in a collaborative and robust way the
unit of the subsystems. The aerospace structures generally require light designs.
The goal of these later is to be optimized the force by weight or effectiveness of
the design. The satellite structural design developed considerably during the last
four decades. Traditionally, the effectiveness was achieved by a combination of
various designs and structural materials. Our objective is to build methods and
tools for the robust design and multidisciplinary optimization in the small satel-
lites mechanics design, taking into account the aspects related to attitude and orbit
control subsystem, thermal control, static, dynamic and thermo-elastic analysis
and including the design of mechanical housing for RF systems. Our proposed
method is demonstrated through multidisciplinary design optimization application.
In this study, the multidisciplinary design optimization in small satellite mechani-
cal design is formulated as optimization of sub problem involving the appropriate
selection of a configuration. The different results obtained in this paper show that
the multidisciplinary design optimization methods present a large interest in the
space design with an aim of having a reliable space system with lower cost.
1 Introduction
The space vehicle design process is complex and difficult due to the physical and
operating constraints, uncertain environments, and extreme operating conditions].
The Multidisciplinary design optimization (MDO) approaches have significant ef-
fects on aerospace vehicle design methodology]. The complexity and the multidis-
ciplinary design make it difficult to have a complete survey and a deep under-
standing of the whole design process. A recent step forward towards a
multidisciplinary approach in the early phases of space system design has been
achieved by concurrent engineering (Blaine 2004).
One of the most important considerations in designing the space vehicle is
weight. By reducing the weight of a space vehicle, it is possible to increase the
payload. In many industrial applications, reducing the weight of a structure with-
out compromising its strength and stiffness is considered as one of the most im-
portant design criteria (Kalavalapally et al. 2006). Today, the search of the best
performance, best quality, and the best cost for space vehicles became a complex
process. The required optimum in a total way, on the level of the system, implies
choices of compromise between the various elements which make it up in order to
answer increasingly many and sometimes contradictory requirements. This con-
cern the Multidisciplinary Optimization (MDO) hence of many disciplines (me-
chanisms, thermal, control of flexible structure, propulsion, cost, structure, opto-
mechanical design, functional architecture, and architecture system) must be taken
into account. These disciplines interact in a very complex way, often according to
some criteria's, and the objective of the MDO consists in optimizing them simul-
taneously. The optimum must be found globally, at the system level, by trade off
choices among the various systems’ components, in order to achieve homogeneity
and perfect integration. Multidisciplinary design optimization (MDO) has
emerged as an engineering discipline that focuses on the development of new de-
sign and optimization strategies for complex systems. MDO researchers strive to
reduce the time and cost associated with the coupling interaction among several
disciplines. As stated in (AIAA Technical Committee for MDO 1991) decomposition
approaches provide many advantages for the solution of complex MDO problems,
as they enable a partitioning of a large coupled problem into smaller, more mana-
geable sub-problems. The resulting computational benefits, besides the obvious
one associated with the solution of smaller problems, include creating a potential
distributed processing environment.
The primary benefit, however, pertains to the savings in personal hours, be-
cause groups are no longer required to wait around for other groups in the process
to complete their design tasks. Considering the number of disciplines involved and
the level of knowledge acquired in each of them, the "engineer’s trade" in the
sense of man’s capability to master all of the disciplines and their interactions is
no longer enough. It is now necessary to develop design and optimization tech-
niques based on mathematical tools, which allow effective integration of the me-
thods used and developed in the various disciplines. These techniques can then be
Multidisciplinary Design Optimization in Small Satellite 231
used in searching for the best tradeoffs leading to a complete optimized system.
This study develops a system sensitivity analysis method to optimize multidiscip-
linary design of space vehicle. Suitable design variables, technological, and func-
tional constraints are considered. Appropriate combinations of disciplines such as
space vehicle layout, thermal, structure; opto-mechanical design, thermo elastic
analysis and electronics are used.
Structural
Structural Attitude, orbit determination Software
mechanics and control
mechanics Reliability
Reliability
Optimization
Optimization Operations and
Geometry
Geometry andand Data handling ground systems
satellite
satellite layout
layout Instruments
Fluid
Fluid Electrical power Telemetry
mechanics
mechanics Wet mass Tracking and
Dry mass command
multidisciplinary
multidisciplinary design
design
optimization
optimization Thermal control
Materials
Materials
Mission analysis Launch mass
Mechanisms
Mechanisms andand
Structure and
control
control systems
systems mechanisms
Propulsion Adapter
Opto-mechanics
Opto-mechanics Heat
Heat transfer
transfer
Propellant mass
Δ Design = (∑ Δ
i Discipline i
)+ Δ MDO
(1)
The proposed algorithm is given on figure 3. This makes our process a powerful
MDO technique because of:
- The quality of the solution obtained compared to the selected criteria of
optimization,
- The robustness of the solution with respect to uncertainties of the models
and the variations of parameters of design,
- The CPU time in the process of optimization.
The general form of a traditional MDO problem can be summarized in the
following way:
Multidisciplinary Design Optimization in Small Satellite 233
Where:
– f : minimization criterion,
– z : shared variables between different disciplines,
– x : local variables of each discipline,
– y : coupling variables between different disciplines,
– g : problem equality constraints,
− h : problem inequality constraints.
Several algorithms already exist are used to solve MDO problems like gradient
based algorithms, game theory, genetic and evolutionary algorithms, multi-agent
systems and many other numerical methods which existed before even the appear-
ance of the MDO.
We can gather the various methods in two categories: single level and multi-
level. For the Single level methods, these methods use only one optimizer and
consider the disciplines as analysis tools.
Among these methods (Clément 2009): MDF (Multi-Discipline Feasible)], IDF
(Individual Discipline Feasible), AAO (All At Once). The multilevel methods
use a global optimizer on the system level and a local optimizer for each disci-
pline. We finds in this category: CSSO (Concurrent SubSpace Optimization),
BLISS (Bi-Level Integrated Systems Synthesis), Collaborative Optimization.
systemlevel
system levelDesign
Design
System level analysis Coordination
System level analysis Coordination
Design Vector
Experts
Experts
⎢ ⎥
⎢x 2 ⎥
Thermal
Knowledge Acquisition
⎢ . ⎥
Knowledge Acquisition
Component
Propulsion Component
⎢ . ⎥
Input
Input
⎢ ⎥
AODCS
Interactive Component
Interactive Component
Explanation
⎢ . ⎥
Explanation
Power Component
Component
Users
⎢ . ⎥
Users
Knowledge
⎢ ⎥
Structures & Knowledge
mechanisms base
base
Inference
⎣x n ⎦ Inference
Component
Component
Opto-Mechanical (Instruments)
Output
Output
Objectif
Objectif vector
vector
Plastic cup
Mast
Helixes
Brass tube
Fig. 5 QFH antenna design problem Fig. 6 QFH antenna CO optimizer architecture
The global optimizer developed in order to minimize the total mass of the an-
tenna (mass of helices and mass of the support supposed as a hollow cylinder –
monocoque structure). The formulas used for the mechanical optimizer are de-
scribed in (Maxwell 2001) by using the Matlab built-in function “fmincon” used
for such constrained nonlinear problems. The radiation pattern optimizer is based
on a FORTRAN genetic algorithm combined with a nonofficial Numerical Elec-
tromagnetic Code which is largely used for antenna calculation (Burke, 1981).
Our optimization tool use two databases, one for launchers (Falcon1, Delta
II/IV, Soyuz, Dnepr, Start1 and LM2C) and a second for materials (1- Al 7075, 2-
Al Alloy 2024 T, 3- Austenitic Stainless Steel 304, 4- Ti Alloy Ti-6Al-4V and 5-
Graphite Epoxy).
The figure 7 presents the GUI designed to enter the design variables (antenna
operating frequency, orbit height and launcher) and shows values used in this test.
With the values shown in figure 7, we obtained the configuration shown in figure
8 and a radiation pattern illustrated in figure 9.
Figures 10 to 12 present parameters evolution in 9 iterations. We can note that
all constraints are respected except the VSWR value which is equal to 1.2. In spite
of that, the resultant antenna presents a small footprint (3.6cm of diameter and
18.75cm of height) and has a light weight (294.7gr). We also remarque that the
thickness of the support is equal to its diameter, it means that this support is not a
monocoque structure but it’s a cylinder. With this configuration (figure 8), we got
natural frequencies equal to 84.66 Hz/lateral and 665.43Hz/axial which are greater
than launcher natural frequencies. The material of the support is graphite epoxy.
We should remarque those three parameters still constant during all the 9 itera-
tion. Those parameters are the antenna helices diameter, the antenna pitch distance
and the antenna number of turns. According to (Larson and Wertz 2005), the two
parameters are dependant and the number of turns must be greater than 1.25 to
have a conical shaped pattern. This configuration is sufficient to resist to mechani-
cal choc due to launcher’s rocket motors load and can be used practically
independently to available space both in the satellite and the launcher’s fairing be-
cause the small dimensions, it results a reduced cost compared to other types of
antennas.
6 Conclusion
Recognizing the need for multidisciplinary optimisation method, the research pre-
sented in this paper introduced a new approach for multidisciplinary optimization
in space vehicle and in the particular in small satellite mechanical design for re-
mote sensing missions. The multidisciplinary design optimisation methods present
a large interest in space vehicle design and in particular the related aspects with
the small satellite mechanical design. This is an important characteristic of the
conceptual design for high-cost systems. Collaborative Optimization approach
was used to solve a QFH antenna design problem. To have a reliable configuration
of the antenna, space environment constraints were taken into account. The an-
tenna design respects both radiation pattern and mechanical constraints. The opti-
miser developed during this study presents an environmental aided tool for the
multidisciplinary design optimization of small satellite by integrating all the ex-
perience gained from the previous space projects in order to obtain a powerful
product and minimize the design risk.
238 A. Boudjemai et al.
References
Aglietti, G.S.: QFH qualification vibration test report, SSTL internal report (2002)
AIAA Technical Committee for MDO. Current State of the Art of Multidisciplinary Design
optimization, AIAA White Paper, approved by AIAA Technical Activities Committee,
Washington, D.C. (September 1991)
Boudjemai, A., Bouanane, M.H., Merad, L., Si Mohammed, M.: A Small Satellite Structur-
al Optimisation Using Genetic Algorithm Approach. In: 3rd International Conference on
Recent Advances in Space Technologies (2007), doi:10.1109/RAST.2007.4284021, 1-
4244-1057-6
Blaine, B.T.: Multifunction structures in satellite design: thermoelectric elements embedded
into structural honeycomb panels. Dissertation, Santa Clara University (2004)
Burke, G. (1981), Numerical Electromagnetic Code developed at Lawrence Livermore Lab
& Livermore CA (January 1981), http://www.si-list.net
Clément, J.: Optimisation multidisciplinaire: étude théorique et application à la conception
des avions en phase d’avant projet. Dissertation, Toulouse University (2009)
Kalavalapally, R., Penmetsa, R., Grandhi, R.: Multidisciplinary optimization of a
lightweight torpedo structure subjected to an underwater explosion. Finite Elements in
Analysis and Design (2006), doi:10.1016/j.finel.2006.07.005.
Kilgus, C.C.: Multielement Fractional Turn Helices. IEEE Transactions on Antennas and
Propagation (1968), doi:10.1109/TAP.1968.1139231
Kilgus, C.C.: Resonant Quadrafilar Helix. IEEE Transactions on Antennas and Propagation
(1969), doi:10.1109/TAP.1969.1139459
Larson, W.J., Wertz, J.R.: Space Mission Analysis and Design, the space technology li-
brary. Published jointly by Microcosm Press and Kluwer Academic Publishers (2005)
Mechanical Vibration Cancellation
Using Impact Absorber
Abstract. The free and forced vibrations of a mechanical system equipped with a
ball absorber were studied. The modeling of the impact damping and the problem
formulation and resolution were conducted. The effect of the absorber parameters
on the vibration attenuation of the primary system was then examined. The absor-
ber parameters considered are: the mass, the stiffness, the damping, the clearance,
and the impact velocity. The excitation parameters were also evoked. Physical and
mathematical modeling, then numerical simulation led to determine the free and
forced responses of the considered system equipped with the impact absorber, for
different situations. This helped to identify the absorber favorable parameters,
leading to reduce considerably the primary system vibration, and to advance de-
sign recommendations.
1 Introduction
The ball absorber is considered as a small mass linked to a primary system. The
vibration attenuation is due to the impacts between the two masses leading to vi-
bratory energy dissipation (Marhadi and Kinra). The impacts are modeled by a
spring and a viscous damper (Cheng and Xu).
Ball absorbers are used to damp excessive vibrations, as for turbine blades and
machine tools: analytical, numerical, and experimental studies were conducted
(Cheng and Wang, Ema and Marui). Also, ball absorbers were used to decrease
industrial machine noise (Xu et al.). A well designed ball absorber is more effi-
cient, less complicated to realize, and more robust to parameters changes then a
classic dynamic absorber (Sankar and Bapat). Other experimental studies on ball
absorbers, also called impact absorbers, were also elaborated (Masri, Desen).
In this work, we are interested in studying the free and forced responses of a
mechanical system equipped with an impact absorber. Firstly, we present the vi-
bration equations and their resolution. Secondly, we exhibit the free and forced
responses numerically obtained using Matlab software. Thirdly, we examine the
influence of the impact absorber parameters variation on the vibration reduction to
identify the favorable design. Finally, we advance some recommendations.
The results are displacement-based responses, in contrast with known works
(energy-based), which are more useful for mechanical applications.
2 Free Vibration
2.1 Formulation
We model the mechanical system equipped with the ball absorber in free vibration
as a 2ddl vibratory system, as shows Fig.1:
The impacts are modeled by two springs, with a stiffness k1, and two viscous
dampers, with a damping coefficient c1. x1 and x2 are respectively the displace-
ments of the primary mass (M) and the impact mass (m), and d represents the
clearance (the width of the free zone).
Three cases are possible:
• Case 1: x2(t) – x1(t) < 0
The impact mass hits the primary mass left side, the vibration equations are as fol-
lows:
⎡M 0 ⎤ ⎧ x1 ⎫ ⎡c + c1 − c1 ⎤ ⎧ x1 ⎫ ⎡k + k1 − k1 ⎤ ⎧ x1 ⎫
⎨ ⎬+ ⎨ ⎬+ ⎨ ⎬=0
⎢0
⎣ m⎥⎦ ⎩x2 ⎭ ⎢⎣ − c1 c1 ⎥⎦ ⎩x2 ⎭ ⎢⎣ − k1 k1 ⎥⎦ ⎩x2 ⎭
The velocities of masses M and m just before and immediately after the collision,
verify these equations:
(1 − μα ) − μ (1 + α ) − + (1 + α ) − (μ − α ) −
x + 1 = x 1 + x 2 ; x 2 = x 1 + x 2
(1 + μ ) (1 + μ ) (1 + μ ) (1 + μ )
Where the mass ratio and the restitution coefficient are respectively:
m ( x + 2 − x +1 )
μ= ;α = − −
M ( x 2 − x −1 )
Mechanical Vibration Cancellation Using Impact Absorber 241
The indices – and + correspond respectively to just before and immediately after
the collision.
Therefore, the vibration equations become:
x1 (t ) + 2ξω x1 (t ) + ω ² x1 (t ) = μ [2ξ 1ω1 y (t ) + ω1 ² y (t ) ]
y(t ) + 2ξ 1ω1 y (t ) + ω1 ² y (t ) = − x1 (t )
Where:
c c
ω = k M ; ω1 = k1 m ; ξ = ; ξ 1 = 1 ; y (t ) = x 2 (t ) − x1 (t )
2Mω 2 mω 1
The system initial conditions, at t=0, are:
x1 (0) = x10 ; x2 (0) = x20 ; y(0) = y0 = x20 − x10
x1 (0) = x10 ; x2 (0) = x20 ; y (0) = y 0 = x20 − x10
To simplify the study, we consider the following hypotheses:
a. The two natural frequencies can be approximated by their average; b. The im-
pact mass is small relatively to the primary mass; c. The primary mass damping
factor (ξ) is small relatively to the impact mass damping factor (ξ1).
Three cases show up:
ξ1 < μ ; The approximate response is:
⎛1 ⎞ ( − ξ1ω a t )
1
x1 (t ) = A cos ⎜ μ − ξ12 ω a t − θ ⎟e 2 sin(ω a t )
⎝2 ⎠
1/ 2
⎡ ⎛ x ⎞
2
⎛ μ x ⎞
2
⎤
A = ⎢ ⎜⎜ 10 ⎟⎟ + ⎜⎜ μ y 0 + ⎛⎜ ξ 1 + ⎞⎟ 10 ⎟⎟ (μ − ξ 1 ² )− 1 ⎥
⎢⎣ ⎝ ω a ⎠ ⎝ ⎝ 2 ⎠ ωa ⎠ ⎥⎦
⎡⎛ μω a y 0 μ⎞ ⎤
θ = tan −1 ⎢⎜⎜ + ξ1 + ⎟(μ − ξ1 ² )−1 / 2 ⎥
⎢⎣⎝ x10 2 ⎟⎠ ⎥⎦
ξ1 = μ ; The approximate response is:
⎡ x μ x10 + μωa y0 ⎤ (−
1
μω a t )
x1 (t ) = ⎢ 10 + t ⎥e 2
sin(ωa t )
⎢⎣ ωa 2 ⎥⎦
ξ1 ; μ ; The approximate response is:
⎧⎡ ξ + ξ ²−μ ⎫
(
⎪ ⎢ − μy0 + − ξ1 + ξ1 ² − μ 10 ⎥ e
ωa ⎦
)
x ⎤ − 1 21 ωat
+⎪
⎪⎣ ⎪ sin(ω a t )
x1 (t ) = ⎨ ⎬
⎪ 2 ξ1 ² − μ
−ξ + ξ ² − μ
⎪⎡
( )
x10 ⎤ 1 2 1 ωat
⎪ ⎢ μy0 + − ξ1 + ξ1 ² − μ ω ⎥ e ⎪
⎩⎣ a ⎦ ⎭
2.2 Optimization
To examine the effect of the absorber parameters on the primary system vibration
amplitudes, we conduct a numerical study. The basic characteristics of the primary
system equipped with the absorber are:
M=1Kg; m=0.4Kg; K=1N/mm; k1=0.4N/mm; d=6mm; α=0.2
The initial conditions are:
x10=0; x20 =0; x 10 =15.10-3m/s; x 20 =0
• Undamped System
We consider a negligible damping either for the primary system or the absor-
ber: C=c1=0.01Ns/m.
Figure 2, shows the effect of the impact absorber on the primary system re-
sponse. The primary mass vibration amplitudes are neatly reduced.
-4
x 10
5
1
X1(m)
-1
-2
-3
avec absorbeur
-4 sans absorbeur
-5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
t(s)
Fig. 2 Primary free system response without (Black) and with (Red) absorber
The numerical results show that the favorable stiffness ratio is: k1/K=0.8 (Fig.
3). Other results, not presented, show that the favorable mass ratio is: m/M=0.8
(maximal), whereas the clearance (d) and the restitution coefficient (α) have no
clear effect.
Mechanical Vibration Cancellation Using Impact Absorber 243
Fig. 3 Primary free mass displacement for k1/K= a)0.2; b)0.4; c)0.8
• Damped System
We consider a damped primary system: ξ=2.5%.
The system mass maximal displacement is the lowest when the damping ratio
is: c1/C=3 (Fig. 4); and when the clearance is: d=12mm (Fig. 5).
Other results, not presented, show that the effect of d depends on ξ, whereas the
effect of α is notable and depends also on ξ.
Fig. 4 Primary free mass displacement for c1/C= a)0.3; b)1; c)3
3 Forced Vibration
3.1 Formulation
The mechanical system equipped with the ball absorber in forced vibration is
modeled as a 2ddl vibratory system, as shows Fig.6:
Where:
⎡ ⎛ω ⎞ ⎤
⎢ 2ξ ⎜⎜ ⎟⎟ ⎥
F0 / K ⎢ ⎝ ω0 ⎠ ⎥
X1 = and ϕ = − arctg ⎢ 2 ⎥
⎢1 − ⎛⎜ ω ⎞⎟ ⎥
2
⎡ ⎛ ω ⎞2 ⎤ ⎡ ⎛ ω ⎞⎤
2
⎢1 − ⎜⎜ ⎟⎟ ⎥ + ⎢2 jξ ⎜⎜ ⎟⎟⎥ ⎢ ⎜ω ⎟ ⎥
ω ⎝ ω 0 ⎠⎦⎥ ⎣ ⎝ 0⎠ ⎦
⎣⎢ ⎝ 0 ⎠ ⎥⎦ ⎣⎢
• Situation 3: x2(t) – x1(t) > d
The impact mass hits the primary mass right side, the responses are similar to
those obtained in Case 1, when the impact mass hits the primary mass left side.
3.2 Optimization
We examine the effect of the absorber parameters on the primary system vibration
amplitudes through a numerical study. The basic characteristics of the primary
system equipped with the absorber are:
M=1Kg; m=0.4Kg; K=1N/mm; k1=0.4N/mm; d=6mm; F0=1N; f=ω/2π=10Hz
The initial conditions, at t=0, are null.
• Undamped System
Negligible damping is considered either for the primary system or the absorber:
C=c1=0.01Ns/m.
Figure 7, shows the effect of the impact absorber on the primary system re-
sponse. The primary mass vibration amplitudes are clearly reduced.
0.02
0.015
0.01
0.005
X1(m)
-0.005
-0.01
-0.015
avec absorbeur
sans absorbeur
-0.02
0 5 10 15
t(s)
Fig. 7 Primary forced system response without (Black) and with (Red) absorber
The numerical results show that the favorable stiffness ratio is: k1/K=0.8 (Fig.
8). Other results, not presented, show that the favorable mass ratio is: m/M=0.8
(maximal), the clearance (d) has no clear effect, and the favorable excitation fre-
quency and modulus are: f=50Hz (maximal) and F0=0.2N (minimal).
246 K. Chehaibi, C. Mrad, and R. Nasri
Fig. 8 Primary forced mass displacement for k1/K= a)0.2; b)0.4; c)0.8
• Damped System
We consider a damped primary system: ξ=5%.
The system mass maximal displacement is the lowest when the damping ratio
is: c1/C=1 (Fig. 9); the clearance is: d=12mm (Fig. 10); and when the excitation
frequency is: f=50Hz (Fig. 11).
Other results, not presented, show that the favorable excitation modulus is:
F0=0.2N (minimal).
Fig. 9 Primary forced mass displacement for c1/C= a)0.3; b)1; c)3
Mechanical Vibration Cancellation Using Impact Absorber 247
4 Conclusion
The ball absorber is found efficient in reducing free and forced vibrations of me-
chanical systems especially when the absorber parameters are optimized.
When the system is free, large mass and stiffness ratios lead to small primary
mass displacements. Besides, a large relative damping improves the primary sys-
tem response. The incidence of the clearance and the impact quality is not clear
and needs to be examined closely.
When the system is forced, also large mass and stiffness ratios lead to small
primary mass displacements. A medium relative damping is favorable but masks
the impact quality, and needs further examination. Besides, the clearance effect is
clearer when the primary system is damped, a large clearance is favorable. In ad-
dition, excitation with large frequency and small modulus is favorable.
Acknowledgements. This work was conducted within the Systems Mechanics Team di-
rected by Charfeddine MRAD. The authors gratefully acknowledge the helpful comments
and suggestions of the reviewers, which have improved the presentation.
References
Marhadi, K.S., Kinra, V.K.: Particle impact damping: effect of mass ratio, material, and
shape. Texas Agricultural and Mechanical University, USA (2004)
Cheng, J., Xu, H.: Inner mass impact damper for attenuating structure vibration. Interna-
tional Journal of Solids and Structures 43, 5355–5369 (2006)
248 K. Chehaibi, C. Mrad, and R. Nasri
Cheng, C.C., Wang, J.Y.: Free vibration analysis of a resilient impact damper. International
Journal of Mechanical Sciences 45, 589–604 (2003)
Xu, Z., Wang, M.Y., Chen, T.: A particle damper for vibration and noise reduction. Journal
of Sound and Vibration 270, 1033–1040 (2004)
Ema, S., Marui, E.: Damping characteristics of an impact damper and its application. Inter-
national Journal of Machine Tools and Manufacturing 36, 293–306 (1996)
Sankar, S., Bapat, C.N.: Single unit impact damper in free and forced vibration. Journal of
Sound and Vibration 99, 85–94 (1985)
Masri, S.F.: General motion of impact dampers. University of Southern California, USA
(1972)
Masri, S.F.: Analytical and experimental studies of impact dampers. Dissertation, Califor-
nia Institute of Technology, USA (1965)
Desen, I.C.: Experimental study on an impact vibration absorber. Dissertation, Texas Tech-
nical University, USA (2000)
RFLP Approach in the Designing of Power-
Trains for Road Electric Vehicles
1 Introduction
The System Engineering (SE) approach is today widely used in current designing
activities as well as in research activities. During last years, the approach has been
extended also to product development lifecycle to gain the contents of Axiomatic
Design and cover early factors that affect the entire product lifecycle (Gumus et
al., 2008). Furthermore, SE was joined to Model Based (MB) approach as a
unique and integrated methodology to be used especially when designers have to
accomplish modelling and simulation integration, during the whole design process
(Graignic et al., 2013). Typical complex systems to be developed using the MBSE
approach are the mechatronic systems because they act on multi-domains and ex-
hibit interactions of physical effects due to different energy domains. For these
The RFLP approach represents the baseline for Systems Engineering in model-based
designing processes, enabling a strong interaction and collaboration between differ-
ent engineering disciplines. The RFLP can be considered as a top-down approach,
which starts analyzing the whole complex system. Then the system is divided into
different sub-systems and the approach proceeds with the analysis of the single sub-
systems (Lanzotti et al., 2013). The acronym RFLP stands for Requirements engi-
neering, Functional design, Logical design and Physical design, i.e. the 3D CAD
modelling (Kleiner and Kramer, 2013). The first step consists in the definition of the
system requirements and is characterized by the analysis of customer needs and
technical requirements. During this phase the main features and services offered by
the analyzed system are defined in a specific documentation. The second step of the
RFLP approach is represented by the functional analysis. This phase is focused on
the analysis of the specific functions that each component needs to perform in order
to satisfy the requirements described in the above requirement phase. In the third
step of the RFLP approach, represented by the logical analysis, the logical connec-
tions between the subsystems are analyzed in order to carry out a logical model of
the system. This logical step represents the first quantitative analysis of the complex
system. In particular the behavior of the system components is modeled, starting
from the knowledge of the expected inputs and outputs of each subsystem as already
analyzed in the functional diagram. The last step of the RFLP approach consists in
the physical model, in fact the design process of a system is not considered complete
without its geometric and dimensional model. During this phase a tridimensional
models of sub-systems are created and their parametrical relationships, which allow
the physical design, are defined.
RFLP Approach in the Designing of Power-Trains for Road Electric Vehicles 251
In this work, the electric power-train of an electric urban bus is considered as case
study. In order to simplify the designing procedure, the propulsion system under
study is considered mechanically coupled to a laboratory test bench for the simula-
tion of the base operative conditions of the vehicle. The main characteristic of the
vehicle are reported in Table 1.
Number of Passengers 18
The definition of Requirements starts with the need for the propulsion systems
to power a specific electric vehicle. This requirement can be divided in the follow-
ing sub-requirements: the test-bench has to represent the mechanical load of the
vehicle on the road; the energy storage system has to power the electric drive; the
power-train on the test-bench has to follow the driving cycle in terms of speed and
resistant torque references. These requirements can be further divided in more de-
tailed sub-requirements. In particular the simulation of the real mechanical load
has to represent the rolling, advancing, aerodynamic and inertia resistant forces,
which depend on the road slope and resistant coefficient, air density, weight, fron-
tal area, speed and acceleration of the vehicle. Moreover, the energy storage sys-
tem has to guarantee both the required driving range and dynamic performance for
the specific mission of the vehicle. In addition, the test-bench simulating requires
the proper control of resistant torque, motor torque and speed.
The functional map, realized using CATIA V6, as result of this functional anal-
ysis described in Section 2, is reported in Figure 1 for the considered urban bus.
In order to be concise, only the highest level of the functional diagram is
represented in the functional map of Figure 1. According to this diagram the main
functionalities the propulsion system has to perform are: storing and supplying
electric energy; supplying mechanical energy; simulating the driver; simulating
the resistant forces; monitoring and controlling the operative parameters. The
functionalities represented in the above diagram are connected each others through
lines and arrows, which represent different type of incoming and outgoing flows.
In particular the energy flows are represented by red lines, and flows of informa-
tion are represented by green lines.
252 C. Capasso et al.
Simulate resistant
forces
Actual resistant torque Reference
torque
Reference speed
Recharge
energy Voltage and
Electric current measure
energy
Torque and Monitor, control and
speed supervise working
Actual speed and torque
The logical analysis aims to characterize the behavior of the whole propulsion
system of the urban bus and of its components on the test-bench. The logical mod-
el is firstly implemented by means of Dymola software, which is based on the
Modelica simulation environment and is specifically devoted to modeling and si-
mulating integrated and complex systems. Then, the integration into the logical
environment of CATIA v6 is obtained by means of DBM Workbench of CATIA
V6 (Smolek et al, 2013).
In order to model and simulate the behavior of the whole power-train, the first
component taken into consideration is the energy storage system. In this regard,
the energy storage modeling procedure is based on simplifying hypothesis, using
the specific ‘Electric Energy Storage’ Modelica library (Einhorn et al. 2011).
The input parameters of this block are set on the base of the main operative
conditions related to the Li(NiCoMn)O2 battery pack considered to power the pro-
pulsion system under study. The main characteristics and recommended operative
conditions of a single Li(NiCoMn)O2 cell are reported in Table 2 (Capasso and
Veneri, 2014).
In this paper, the electric brake model is implemented representing only the
mechanical behavior, without taking into account thermal and electrical aspects.
Moreover the maximum brake mechanical torque is considered constant with the
angular speed. According to these hypotheses the electric brake model is built us-
ing the Dymola mechanical libraries (Hammadi et al., 2012) taking into account
that the vehicle inertia is simulated considering a flywheel of equivalent inertia in-
cluded inside the "brake" block. The related inertia value is evaluated at the elec-
tric motor shaft using a Modelica script.
Battery
Electric Drive
Ground
Driving Cycle
Before starting the physical step, a verification phase of the whole design pro-
cedure is required to optimize performance and size of the power-train. The flow-
chart of this verification procedure is reported in Figure 3.
As described in the flow-chart of Figure 3, after the design of the logical model,
the input parameters of the specific case study are required to be properly set.
Then it is possible to run the simulations and the results are evaluated. The first
evaluation is related to the relative difference between actual speed reached by the
vehicle and reference speed imposed by the required driving cycle. If this differ-
ence is higher than 15% the propulsion system is not considered suitable to satisfy
the dynamic performance requirements and a better solution for the electric drive
254 C. Capasso et al.
needs to be evaluated. The next step consists in evaluating the vehicle autonomy,
in terms of hours, on a specific driving cycle, which is required to be greater than
4 hours when the vehicle runs on a plane road and greater than 2 hours on a road
slope of 10%. This last evaluation is directly related to the design of the vehicle
battery pack in terms of capacity. The simulation is stopped when the battery pack
SoC reaches the value of 20%, in order to avoid damages to the storage system. At
the end of this iterative verification procedure, the power-train is considered well
designed and its dimensions can be used in the parametrical 3D CAD model to
complete the physical step of the RFLP process.
Table 4 Parameters of First Configuration of propulsion system for the Trapezoidal Driving
Cycle 0% Slope Simulation
Battery Cells in Series Cells in Parallel Cell Capacity Cell Weight Battery Pack Weight
Pack 140 1 50 Ah 1.25 kg 175 kg
Electric Power Maximum Torque Rated Current Rotor Inertia Weight
Drive 54.6 kW 470 Nm 101 A 0.123 kgm2 130 kg
The first analysis, reported in Figure 4, is related to the actual vehicle speed
profile compared to the reference speed profile for a trapezoidal driving cycle on a
plane road, derived by a standard ECE driving cycle.
RFLP Approach in the Designing of Power-Trains for Road Electric Vehicles 255
40
Reference Speed
Actual Speed
30
Speed (km/h)
20
10
0
0 20 40 60 80
Time (s)
Fig. 4 Comparison among the actual and reference vehicle speed on a trapezoidal driving
cycle for the first configuration
In this case it is clear that the urban bus, in this configuration is not able to sa-
tisfy the dynamic performance required by the driving cycle, because of a percen-
tage error of about 20%. This result depends on the motor torque reported to the
wheels, which results too low if compared with the vehicle inertia combined with
the resistant forces.
For the above reasons, following the designing verification procedure, a second
test is performed choosing an electric drive with a nominal power of 63.4 kW and
a battery cell capacity of 150 Ah. In this case, the electric bus follows, with a per-
centage error of 12%, the dynamic requirements of the driving cycle of Figure 4,
although the parameters chosen for this second solution affect the vehicle weight.
For this last configuration, Figure 5A shows the vehicle mechanical power in
comparison with the battery pack electric power during the driving cycle. For this
set of parameters, as reported in Fig. 5B, the vehicle autonomy on the driving
cycle, with a road slope of 0%, is evaluated equal to 5.2 hours. This means that the
dynamic performance and autonomy requirements are satisfied.
100 100
A B
60
40
-50
20
-100 0
0 20 40 60 80 0 1 2 3 4 5 6
Time (s) Time (h)
Fig. 5 Vehicle mechanical and electric power (A) reported to the motor shaft and battery
SoC (B) vs time during a trapezoidal driving cycle
256 C. Capasso et al.
200 100
A B
Battery Electric Power
Motor Shaft Mechanical Power
150 80
Battery SoC (%)
Power (kW)
100 60
50 40
0 20
-50 0
0 20 40 60 80 0 0.5 1 1.5 2 2.5
Time (s) Time (h)
Fig. 6 Vehicle mechanical and electric power (A) reported to the motor shaft and battery
SoC (B) vs time during a trapezoidal driving cycle with a road slope of 10%
At the end of the design verification process, the geometries of 3D CAD mod-
els related to the electric motor, joint, flywheel and brake are automatically rege-
nerated. These models are used to design the supports for the motor and flywheel.
Figure 7 shows the 3D CAD models of the whole test bench of the power-train
powered by the battery pack and connected to the eddy current brake, running in
CATIA V6 environment.
The designing approach presented in this paper, enriched by means of parame-
tric relations building the final test bench CAD model implemented in CATIA V6,
represents a design procedure which is expected to achieve a drastic reduction of
designing times and costs for the whole development process of propulsion sys-
tems for road electric vehicles.
5 Conclusions
The present paper briefly summarizes the preliminary method and main steps ac-
complished in the designing process of a propulsion system for road electric ve-
hicles, by means of the RFLP approach. The results obtained through simulations,
compared to those related to reference driving cycles, evidence the possibility to
accomplish adequate simulation models. Also the verification and improvement
procedures, in the sense of RFLP approach, as well as the possibility to directly
link the changes of physical models to the ones of logical model were well
accomplished. At the present, further developments need for the management of
system requirements, as regards the availability of tools related to the educational
environment CATIA V6 used. Further activities will deal with the validation of
actual simulation models by means of a comparison with experimental data. In
this way it will be possible to extend the same RFLP approach and related models
to the designing of different propulsion systems for road electric vehicles.
Acknowledgements. The present work was developed with the contribution of the French
Ministry for Higher Education performing the activities of the PLACIS (Plateforme Colla-
borative d’Ingénierie Système, Collaborative Platform for Systems Engineering) project.
The authors gratefully acknowledge Mr. Antonio Rossi, technician of Istituto Motori, for
his cooperation in the drawings of this paper, and engineers Antonio Amodio and Daniela
De Vivo for their assistance in carrying out preliminary aspects of the simulation phase.
References
Capasso, C., Veneri, O.: Experimental analysis on the performance of lithium based batte-
ries for road full electric and hybrid vehicles. J. Applied Energy (2014),
doi:10.1016/j.apenergy.2014.04.013
Einhorn, M., Conte, F.V., Kral, C., Nikla, S.C.: A Modelica Library for Simulation of Elec-
tric Energy Storages. In: Proc. of the 8th International Modelica Conference, Technical
Univeristy, Dresden, Germany, March 20-22 (2011), doi:10.3384/ecp11063436
Graignic, P., Vosgien, T., Jankovic, M., Tuloup, V., Berquet, J., Troussier, N.: Complex
System Simulation: Proposition of a MBSE Framework for Design-Analysis Integration.
Procedia Computer Science 16, 59–68 (2013)
258 C. Capasso et al.
Gumus, B., Ertas, A., Tate, D., Cicek, I.: Trans disciplinary Product Development Life-
cycle. Journal of Engineering Design 19(03), 185–200 (2008),
doi:10.1080/09544820701232436
Hammadi, M., Choley, J., Penas, O., Riviere, A.: Multidisciplinary approach for modelling
and optimization of Road Electric Vehicles in conceptual design level. In: Proc. of Elec-
trical Systems for Aircraft, Railway and Ship Propulsion (2012),
doi:10.1109/ESARS.2012.6387488
Hehenberger, P., Poltschak, F., Zeman, K., Amrhein, W.: Hierarchical design models in the
mechatronic product development process of synchronous machines. Mechatronics 20,
864–875 (2010)
Kleiner, S., Kramer, C.: Model Based Design with Systems Engineering Based on RFLP
Using V6. In: Proc. of the 23rd CIRP Design Conference (2013), doi:10.1007/978-3-
642-30817
Lanzotti A, Patalano S, Vitolo F (2013) A Graph-based Software Tool for the CAD Model-
ing of Mechanical Assemblies. Proc. of 8th International Conference on Computer
Graphics, Theory and Applications, GRAPP 2013 IVAPP 2013, pp. 60-69, Barcelona,
Spain, 21th-24th February,SciTePress Digital Library, doi 10.5220/0004299000600069
Mhenni, F., Choley, J.-Y., Penas, O., Plateaux, R., Hammadi, M.: A SysML-based metho-
dology for mechatronic systems architectural design. Advanced Engineering Informat-
ics 28, 218–231 (2014)
Smolek, P., Heinzl, B., Ecker, B., Breitenecker, B.: Exploring the Possibilities of Co-
Simulation with CATIA V6 Dynamic Behavior Modeling. In: Proc. of Electrotechnical
and Computer Science Conference ERK (2013)
Veneri, O., Migliardini, F., Capasso, C., Corbo, P.: Experimental performance assessment
of Pb, Li[NiCoMn]O2 and LiFePO4 batteries for road vehicles. In: Proc. of Power Elec-
tronics, Electrical Drives, Automation and Motion, SPEEDAM (2012),
doi:10.1109/SPEEDAM.2012.6264626
Mechanical Vibration Elimination Using
Friction Absorber
1 Introduction
Various studies were done to reduce mechanical systems vibration using friction
absorbers, however the absorber parameters effect on the mechanical system vi-
bration amplitude is barely considered.
The friction presents a damping behavior (Balachandran and Magrab, Lopeza et
al.), the vibration reduction by friction needs understanding the excited system (Hi-
nrichs et al., Stein et al.), this behavior is nonlinear relatively to velocity (Pratt,
Liang), whereas the Coulomb model is basic to dry friction (Hundal): the friction
force is proportional to the velocity, it is confined within two limits to form an im-
pair function. To study 2ddl vibratory systems containing friction, the velocity re-
sponse is periodic due to physical considerations, verified experimentally (Hundal).
The work consists on conducting a parametric study on the vibratory response
of a mechanical system equipped with a friction absorber, with and without damp-
ing. The effect of the absorber parameters: mass, stiffness, friction coefficient, and
2 Undamped Absorber
2.1 Formulation
We consider a vibratory mechanical system with a base excitation. The system is
modeled as a mass (m1) linked to the base through a spring (K1). To the primary
mass, a secondary mass is added (m2), it is linked through a spring (K2). The sec-
ondary mass slides with dry friction (Fig. 2).
m 0 x K K K x f cos ωt F sign x x
0 m x K K x F sign x x
f1 et ω are respectively the excitation modulus and pulsation.
Mechanical Vibration Elimination Using Friction Absorber 261
2.2 Optimization
A parametric study was conducted to determine the absorber parameters effect on
the primary system response.
Figure 3, shows the increase of mass m2 conducts to a decrease of the primary
mass vibration amplitude. Figure 4, shows the primary system vibration amplitude
increases while increasing the stiffness K2. Figure 5, shows the rise of friction
conducts to a drop of the primary mass vibration amplitude.
Fig. 3 X1 function of m2
Fig. 4 X1 function of K2
2.3 Equivalence
• Existance
The linearization consists on replacing all forms of energy dissipation by a
viscous damper, dissipating the same quantity of energy.
262 A. Nasr, C. Mrad, and R. Nasri
Fig. 5 X1 function of
The equivalence existence is investigated using the force ratios R1 and R2, link-
ing friction and equivalent forces. Figures 6, 7, 8, show the development of R1 and
R2 with m2 and K2, for m1=1Kg, K1=1N/mm, f1=1N, ω=20πrad/sec, and =0.1.
R1 and R2 are always under 1, which indicates equivalence when the secondary
mass slides without stopping.
• Comparaison
Figure 9, shows the primary mass vibration amplitude using the linear and non-
linear methods; when m2=0.8Kg, K2=0.4N/mm, and m2 slides without stopping.
The two responses are close, which confirms that the linear method can replace
the nonlinear method when the secondary mass slides without stopping.
Mechanical Vibration Elimination Using Friction Absorber 263
3 Damped Absorber
3.1 Formulation
The mechanical system vibrates due to a base excitation. To the primary mass, a
secondary mass is added (m2), it is linked through a spring (K2) and a viscous
damper (C). The secondary mass slides with dry friction (Fig. 11).
m 0 x C C x K K K x
0 m x C C x K K x
f cos ωt F sign x x
F sign x x
f1 et ω are respectively the excitation modulus and pulsation.
3.2 Optimization
A parametric study was also conducted to determine the absorber parameters ef-
fect on the primary system response.
Figure 12, shows the rise of mass m2 conducts to a drop of the primary mass
vibration amplitude. Figure 13, shows the primary system vibration amplitude de-
creases while decreasing the stiffness K2. Figure 14, shows the augmentation of
friction conducts to a diminishment of the primary mass vibration amplitude.
Mechanical Vibration Elimination Using Friction Absorber 265
Fig. 11 X1 function of m2
Fig. 12 X1 function of K2
Fig. 13 X1 function of
266 A. Nasr, C. Mrad, and R. Nasri
3.3 Linearization
• Existence
The equivalence existence is examined using the force ratios R1 and R2, linking
friction and equivalent forces. Figures 15, 16, show the change of R1 and R2 with
m2 and K2, for m1=1Kg, K1=1N/mm, f1=1N, ω=20πrad/sec, =0.1, and with an
absorber relative damping of 15%.
R1 and R2 are continually under 1, which indicates equivalence when the sec-
ondary mass slides without stopping.
• Comparaison
Figure 17, shows the primary mass vibration amplitude utilizing the linear and
nonlinear methods; when m2=0.8Kg, K2=0.4N/mm, and m2 slides without stop-
ping.
Mechanical Vibration Elimination Using Friction Absorber 267
The two responses are close, which prove that the linear method can replace the
nonlinear method when the secondary mass slides without stopping.
4 Conclusion
Acknowledgements. This work was conducted within the Systems Mechanics Team di-
rected by Charfeddine MRAD. The authors gratefully acknowledge the helpful comments
and suggestions of the reviewers, which have improved the presentation.
References
Balachandran, B., Magrab, E.B.: Vibrations. Toronto, Canada (2009)
Lopeza, I., Busturiab, J.M., Nijmeijera, H.: Energy dissipation of a friction damper. Journal
of Sound and Vibration 278, 539–561 (2004)
268 A. Nasr, C. Mrad, and R. Nasri
Hinrichs, N., Oestreich, M., Popp, K.: On the modeling of friction oscillators. Journal of
Sound and Vibration 216, 435–459 (1998)
Stein, G.J., Zahoransky, R., Mucka, P.: On dry friction modelling and simulation in kine-
matically excited oscillatory systems. Journal of Sound and Vibration 311, 74–96 (2008)
Pratt, T.K., Williams, R.: Non-linear analysis of stick/slip motion. Journal of Sound and
Vibration 74, 531–542 (1981)
Liang, J.: Damping estimation via energy-dissipation method. Journal of Sound and Vibra-
tion 307, 349–364 (2007)
Hundal, M.S.: Response of a base excited system with Coulomb and viscous friction. Jour-
nal of Sound and Vibration 64, 371–378 (1979)
AC Flashover: An Analysis with Influence
of the Pollution, Potential and Electric Field
Distribution on High Voltage Insulator
1
L’I3CUB university of Biskra, Algeria
hanibenguesmia@yahoo.fr
2
Laboratoire de Recherche en Electrotechnique (L.R.E).
Ecole Nationale Polytechnique, Algiers, Algeria
Abstract. This paper presents a work devoted to the impact study of the pollution
severity on the energy level dissipated on high voltage insulator. Pollution of high
voltage insulators is a factor of prime importance in the quality and reliability of
power transmission. It causes the circumvention of high voltage insulators. Howev-
er, when it is wet, it causes the dissolution of salts and the formation of an electro-
lyte layer on the surface of insulators, thus reducing dielectric strength. Pollution
generally poses few problems. However, when it is humid, it causes the dissolution
of salts and the formation of an electrolytic layer on the insulator surfaces; as a
result, it reduces their dielectric rigidity.
The aim of this paper is the experimental study of the flashover process and the
distributions of the potential and the electric field pollution carried out under 50
Hz applied voltage on a circular model simulating the 1512 L outdoor insulator
largely used by the Algerian company of electricity and gas (SONELGAZ). This
circular model is under different surface conductivity and different level of uni-
form discontinues pollution.
The computer simulations are carried out by using the COMSOL Multiphysics
3.5 software. This program uses the finite element method to solve the partial dif-
ferential equations that describe the field. Original experimental results made in
the laboratory are presented and compared with a real insulator, and simulations
presented in this paper are originals.
1 Introduction
High voltage insulators form an essential part of high voltage electric power
transmission systems. Any failure in the performance of high voltage insulators
will cause considerable loss of capital. (B.Moula et al).
2 Experimental Techniques
The laboratory’s model used in this study is a glass disc (400 “mm” diameter and
5 “mm” thick) having the feature that resists to the heat due to the electric dis-
charge. This model is equipped with two cylindrical electrodes made of alumi-
num. The first (26 “mm” diameter) placed at the high voltage and the second (172
“mm” diameter) placed at the ground (fig.1(a)). The disposition of the clean and
polluted zones is illustrated on (fig.1(b)). (p: indicates polluted zone, c: indicates
clean zone). The longitudinal dimensions of the clean “zone (c)” and “polluted
ones (p)” (fig.2) are indicated in the table1.
The tension of test is measured using a capacitive tension divider connected to
the secondary of a transformer of test 140kV (fig. 3).
AC Flashover: An Analysis with Influence of the Pollution, Potential and Electric Field 271
(a) (b)
Fig. 1 Experimental laboratory model (a), Determination of clean and polluted areas (b)
HV Transformer
Measurement
capability
Test object
(model insulator)
N1 51 19 17 18 9 38 40 8
N2 53 23 21 22 5 34 36 6
N3 83 0 25 26 0 30 32 4
N4 85 0 29 30 0 26 28 2
N5 87 0 33 34 0 22 24 0
N6 89 0 37 34 0 18 22 0
N7 91 0 41 34 0 14 20 0
N8 93 0 45 34 0 10 18 0
The degree of pollution used in the laboratory’s model surface is a saline solu-
tion (water and NaC1). The different conductivities studied are: 1.823 “mS/cm”,
3.33 “ms/cm”, 8.02 “ms/cm”, and 12.61 “ms/cm”, 16.32 “mS/cm”, 30.05
“ms/cm”, 50.04 “ms/cm”, and 93.70 “ms/cm”.
AC Flashover: An Analysis with Influence of the Pollution, Potential and Electric Field 273
(1)
0 2
0 3
The profile of the insulator is introduced in the software. The mesh density is
higher in the critical regions of the insulators where higher accuracy is required. The
number of meshes for insulator string is 8644 elements. Fig(4) (B.M’hamdi et al).
Fig. 4 Discretization in finite elements and determination of mush elements of the insulator
1512L
(a)
(b) (c)
Fig. 5 initiation of arcs (a), evollution of arcs (b) and flashover total (c)
3 Experimental Results
In this section we focused on the variation of the flashover according to the con-
ductivity and the pollution width (polluting layer). This circumvention is characte-
rized by the non-generation of partial arcs, is due to the pollution used and to the
fact that the total width of the clean equivalent tape (representing the amount of
different clean partial tape) exceeds the critical value from no stable discharge is
propagated (R. J Wilkins et al), (H. Streubel),( A. Mekhaldi et al).
AC Flashover: An Analysis with Influence of the Pollution, Potential and Electric Field 275
Fig. 6 Presentation of the real model on the software (clean model), (a) FEMM, (b) Comsol
4 Result of Simulation
creepage distance. In addition, the lines of the field are divergent (outgoing of the
electrode of high voltage).
It comes out from these characteristics the following remarks: (fig.9)
• The electric field is intense at the end of the electrode of high voltage.
• The field will not vanish in the glass but gives a very low value.
• The electric field is practically null inside the two electrodes, because the two
electrodes are conductive.
• The electric field is important in the zone meadows of the active electrode.
• The electric field is null in the clean zones.
The lines of the electric field being perpendicular to equipotential surfaces
leave the electrode under tension and arrive at the level of the electrode ground.
The vectors represented on fig.9 (a) are tangential with the lines of electric field,
we notice that the electric field is more important with dimensions interior of the
electrodes.
64.00 80.00
1.823 1.823
3.33
8.02 3.33
8.02
62.00 12.61
16.32 12.61
30.5 16.32
70.00 30.5
50.4
60.00 93.7 50.4
mS/cm 93.7
mS/cm
58.00
60.00
56.00
Flashover voltage (kV)
54.00 50.00
Flashover voltage (kV)
52.00
40.00
50.00
48.00
30.00
46.00
44.00 20.00
0.00
1.00 3.00 5.00 7.00
Width2.00
1.00
of layer
3.00
pollution (mm)6.00
5.00 7.00
Width 2.00 4.00
of layer pollution 6.00
(mm) 8.00 0.00 4.00 8.00
(a) (b)
Fig. 7 Flashover voltage-width of the layer pollution for different surface conductivity (a)
model insulator), (b) real insulator
(a)
(b)
Electric field (kV/cm)
Fig. 8 Electric field, (a) vector of the electric field,(b) distribution (echograms) of the
electric field, (c)Electric field along the leakage distance,(d) equipotential lines
Electric potential (V)
(a)
Fig. 9 Distribution (echograms) of the potential (a), electric potential along the
cleakage distance (b)
278 H. Benguesmia, N. M’ziou, and A. Boubakeur
The figure10 respectively illustrate the variation of the potential along the cree-
page distance and the distribution of the potential of the insulators. The potential is
very important on the level of the electrode of high voltage then decreases as one
moves away from the electrode credit.
• In a dry state, the potential takes the maximum value 30 Kv on the active
level of the electrode then decrease in a linear way as one moves away from this
electrode until the attack of the electrode of mass where the potential is cancelled.
5 Conclusion
In this paper, we have presented the results of different tests on the laboratory’s
model high voltage insulator as well as the real one and simulation of electric field
and potential distribution under normal operating conditions in 220kV AC trans-
mission lines has been carried out using a two dimensional finite element model.
According to the results:
• The flashover reduces with the conductivity of the polluted environment. Oth-
erwise the insulator is more rigid when the conductivity of the surface
increases as well as its width.
• For the dry state, the potential takes the maximum value 30 Kv on the active
level of the electrode then decrease in an almost linear way as one moves
away from this electrode until the attack of the electrode of mass where the
potential is cancelled.
• We also saw that the potential decreases as one moves away from the active
electrode.
• The maximum value of the electric field is obtained on the level of the clean
layers of lower width.
• Field calculations help in improving the design and identification of vulnera-
ble areas.
• Usually in electric field distribution, the electric field magnitude is high at the
energized end and decreases exponentially along the creepage distance.
References
Moula, B., Mekhaldi, A., Teguar, M.: Wavelet Packet Analysis of the Applied Voltage
Waveform as a Monitoring Technique of the Outdoor Insulation Surface. In: IEEE In-
ternational Conference on Solid Dielectrics, Bologna, Italy, pp. 174–177 (2013)
Moula, B., Mekhaldi, A., Teguar, M., Haddad, A.: Characterization of Discharges on non-
Uniformly Polluted Glass Surfaces Using a Wavelet Transform Approach. IEEE Trans-
actions on Dielectrics and Electrical Insulation 20(4), 1457–1466 (2013)
Jaafar, S., Ahmad, A.S., Ghosh, P.S., Aljunid, S.A.K.: A new approach in modeling AC
flashover voltage for polluted insulator. In: 2002 IEEE Conference on Electrical Insula-
tion and Dielectric Phenomena, Universiti Tenaga Nasional, Kajang, Selangor, Malay-
sia, pp. 558–561 (2002)
AC Flashover: An Analysis with Influence of the Pollution, Potential and Electric Field 279
Zaffanella, L.E., Scheinder, H.M., Dunlap, J.H.: Perfermances des isolateurs polluées pour
lignes ccht, CIGRE, rapport 33-05 (1986)
Dhahbi-Meghriche, N., Beroual, A.: Flashover dynamic model of polluted insulators under
ac voltage. IEEE Transactions on Dielectrics and Electrical Insulation 7(2), 283–289
(2000)
M’hamdi, B., Teguar, M., Mekhaldi, A.: Potential and Electric Field Distributions on HV
Insulators String Used in The 400 kV Novel Transmission Line in Algeria. In: 2013
IEEE International Conference on Solid Dielectrics, Bologna, Italy, June 30-July 4,
pp. 190–193 (2013)
Huang, D., Ruan, J., Cai, W., Li, T., Wei, Y., Liu, J.: Flashover Prevention on High-altitude
HVAC Transmission Line Insulator Strings. IEEE Transactions on Dielectrics and Elec-
trical Insulation 16(1), 88–98 (2009)
Jowkar, M., Gholami, A.: Corona ring designation for 400 kv ac transmission line with
com-posite insulators by FEM. In: 26th International Power System Conference, PSC
2011, Tehran, Iran, October 31-November 02 (November 2011)
Wilkins, R.J., Al-Baghdadi, A.A.J.: Arc Propagation along an electrolyte surface. Proc.
IEEE 118(12), 1886–1894 (1971)
Streubel, H.: The influence of non-uniform contamination on the flashover voltage of insu-
lators. In: 14th International Symposium on High Voltage Engineering, Athens, rapport
46-10 (September 1983)
Mekhaldi, A., Namane, D., Bouazabia, S., Beroual, A.: Flashover of discontinuous pollu-
tion laye ron hv insulators. IEEE Trans. on Dielec. and Elect. Insul. 6(6), 900–906
(1999)
Evaluation by Wide-Angle X-ray Scattering
of the Particle Evolution in Polypropylene
Processed by Equal Channel Angular Extrusion
1
LMA, FGMGP/USTHB, BP.32 El-Alia, Bab-Ezzouar, 16111 Alger, Algérie
{ramdane.boulahia,t-boukha}@gmail.com
2
UMET, Université de Lille 1, Rue Longevain, Lille, France
jean-michel.gloaguen@univ-lille1.fr
Abstract. The equal channel angular extrusion (ECAE) is a process used to im-
prove the mechanical properties of materials. Key parameters to achieve this
objective are mainly the geometrical dimensions of the extrusion die and the ex-
trusion parameters such as extrusion rate, the back pressure, the applied load,
etc…. The characterization by the use of X-ray showed that the sample extruded at
45 mm/min, and which has light and dark bands due to the stick-slip phenomenon
polymer, has sheared in the direction of approximately 45° to the extrusion direc-
tion. The using of the Fit2d software can locate the families of crystal planes (hkl)
of the PP. Observations by X-ray in the direction of flow have shown the exis-
tence of three planes (040), (130) and (111). On the other hand, by the observation
in the transverse direction revealed the existence of another plane (131).
1 Introduction
For the last 20 years, a lot of research efforts in materials science-related fields of
metallurgy, mechanical engineering, and chemical engineering have been focused
on wider development, implementation, commercialization and improvement of
new material forming methods known as severe plastic deformation (SPD)
schemes (Alexander V. Perig et al, 2014). The classical SPD processing method is
Segal’s equal channel angular extrusion (ECAE) material forming technique,
which have found application to processing of metals, alloys, composites, powder
and polymer semi manufactured articles, etc. The ECAE method, developed in the
beginning of the 1980s (Segal et al 1981), is based on one or several extrusion
passes of a lubricated metal, polymer, powder, or composite material through a die
with two intersecting channels of equal cross-section (Fig 1). The works on ECAE
of polymers were initiated relatively recently (Sue H-J, Li CK-Y 1998). In this
Fig. 1 Schematic illustration of a 90° ECAE die (flow direction: FD, load direction: LD,
transverse direction: TD)
1 ⎡ ⎛ϕ ψ ⎞ ⎛ ϕ ψ ⎞⎤
εp = ⎢2 cot⎜ 2 + 2 ⎟ + ψ cos ec⎜ 2 + 2 ⎟⎥
(1)
3⎣ ⎝ ⎠ ⎝ ⎠⎦
the friction between the sample and the die, as a result of an unstable balance
between the yield stress and the friction force. This interpretation borrows from
previous studies regarding solid-state extrusion (Krjutchkov et al. 1986).
The effect of backpressure on the plastic flow instabilities is shown in (Fig. 4).
The application of back-pressure significantly contributes to reduce the gap
between the sample and the outer corner of the die. The wave formation is
concomitantly reduced and the alternating stripes tend to disappear with increasing
back-pressure.
Fig. 3 Macrographs of PP samples after ECAE at a ram speed of 45 mm/min without back-
pressure
Fig. 4 Macrographs of PP samples after ECAE at a back-pressure of 900 N and ram speeds
of 0.45 mm/min
(a)
(b)
Fig. 5 (a) Lamellae of PP for WAXS measurements, (b) Samples after ECAE
Fig. 6 2D-WAXS patterns, viewing along TD axe of an as-received PP (a) top, (b) middle,
(c) bottom
top to bottom of the slice, as sketched in (Fig. 5.b). However, much different is the
situation for the edge-view with the X-ray beam normal to the FD-LD plane. In
the case of the higher ram velocity 45 mm/min and no back-pressure, (Fig. 8.)
displays three 2D-patterns recorded from consecutive dark-light-dark stripes
according to the sketch of (Fig. 5b.). The middle light stripe looks roughly
isotropic whereas the dark ones exhibit a strong diagonal reinforcement of the 3
first reflections. This is relevant to a preferred orientation of the (110), (040) and
(130) planes containing the chain axis parallel to the shear direction.
Fig. 7 2D-WAXS patterns, viewing along FD axe, of PP after ECAE at a ram speed of 45
mm/min and without back-pressure: (a) top, (b) middle, (c) bottom.
Fig. 8 2D-WAXS patterns, viewing along TD axe, of PP after ECAE at a ram speed of 45
mm/min and without back-pressure: (a) dark stripe, (b) light stripe, (c) dark stripe (the shear
direction is parallel to the beam stop arm).
(a) (b)
Table 1 Tab. Directions 1 and corresponding plans for each diffraction peaks
7 Conclusion
The characterization by means of X-ray showed that the polymer extruded with a
speed of the 45mm/min, which has light and dark bands due to the phenomenon
stick-slip has sheared in the direction of approximately 45° to the direction of
extrusion. The treatment with Fit2d software allowed us to locate the families of
crystal planes (hkl) of the PP. The observations in the flow direction showed the
existence of three planes (040), (130) and (111). Against this, the observation in
the transverse direction revealed the existence of another plane (131)."
References
Perig, A.V., Golodenko, N.N.: CFD 2D simulation of viscous flow during ECAE through a
rectangular die with parallel slants. Int. J. Adv. Manuf. Technol. (2014),
doi:10.1007/s00170-014-5827-2
Sue, H.-J., Li, C.K.-Y.: Control of orientation of lamellar structure in linear low-density po-
lyethylene via a novel equal channel angular extrusion process. J. Mater. Sci. Lett. 17,
853–856 (1998)
Segal, V.M., Reznikov, V.I., Drobyshevskiy, A.E., Kopylov, V.I.: Russ. Metall. 1, 99–105
(1981)
Qiu, J., Murataa, T., Wu, X., Kitagawa, M., Kudo, M.: Plastic deformation mechanism of
crystalline polymer materials in the equal channel angular extrusion process. Journal of
Materials Processing Technology 212(2012), 1528–1536 (2012),
doi:10.1016/j.jmatprotec.2012.02.015
Zuo, F., Keum, J.K., Chen, X., Hsiao, B.S., Chen, H., Lai, S.Y., Wevers, R., Li, J.: The role
of interlamellar chain entanglement in deformation-induced structure changes during un-
iaxial stretching of isotactic polypropylene. Polymer 48, 6867–6880 (2007)
Phillips, A., Zhu, P.-W., Edward, G.: Simple shear deformation of polypropylene via the
equal channel angular extrusion process. Macromolecules 39, 5796–5803 (2006)
Bartczak, Z., Argon, A.S., Cohen, R.E.: Texture evolution in large strain simple shear de-
formation of high-density polyethylene. Polymer 35, 3427–3441 (1994)
Li, C.K.Y., Xia, Z.-Y., Sue, H.J.: Simple shear plastic deformation behavior of polycarbo-
nate plate II. Mechanical property characterization. Polymer 41, 6285–6293 (2000)
Kitagawa, M., Kozaki, T.: ECAP process in polypropylene. J. Soc. Mater. Sci. 55, 37–41
(2006)
Iwahashi, Y., Wang, J., Horita, Z., Nemoto, M., Langdon, T.: Principle of ECA pressing for
the processing of ultra-fine grained materials. Scripta Mater. 35, 143–146 (1996)
Krjutchkov, A.N., Dorfman, I.Y., Prut, E.V., Enikolopyan, N.: Solid state extrusion of se-
micrystalline polymers. Polym. Compos. 7, 413–420 (1986)
A Multi-site Supply Chain Planning Using
Multi-stage Stochastic Programming
1 Introduction
are divided into two sets. The decision variables of the first stage called “here and
now” decisions have to be made before the realization of uncertainty. Subsequent-
ly, the second stage decision variables are chosen after the presence of uncertain
parameters in order to correct the infeasibilities caused by uncertainty realization
(“wait and see” decisions). Therefore, the value of the objective function is the sum
of first stage decision variables and the second stage expected recourse variables.
Many works in the literature have been interested in two-stage stochastic pro-
gramming model for supply chain planning problem. Gupta and Maranas (2003)
studied a multi-product, multi-period multisite supply chain tactical planning
problem under demand uncertainty based on a two-stage stochastic programming
approach. Leung et al. (2006) proposed a two-stage stochastic programming mod-
el in order to optimize a multi-site aggregate medium-term production planning
problem under an uncertain environment. Mirzapour Al-e-hashem et al. (2013)
proposed a multi-site multi-period multi-product aggregate production medium-
term planning problem in a green supply chain under demand uncertainty using a
two-stage stochastic programming approach.
However, adopting a two-stage stochastic approach to address the uncertain
multi-period production planning problem cannot effectively reflect the dynamic
variations of the system condition. The multi-stage stochastic programming for-
mulation is an extension of the two-stage stochastic formulation (Birge and Lou-
veaux, 1997). The multi-stage stochastic model allows the decision maker to
revise and to update the production plan through time period evolution and when
more information about the customer demand scenarios becomes available. Many
papers have addressed such situation in the literature. Ahmed et al. (2003) pro-
posed a multi-period capacity expansion problem under demand and investment
costs uncertainties. A multi-stage stochastic programming model was developed in
order to incorporate the uncertain parameters. Karabuk (2008) developed a multi-
stage stochastic programming formulation to deal with a yarn production planning
problem under demand uncertainty in textile industry. Kazemi Zanjani et al.
(2010) proposed a multi-product, multi-period production sawmill planning prob-
lem under quality of raw materials and demand uncertainty using a multi-stage
stochastic programming model. To the best of our knowledge, the multi-stage sto-
chastic programming approach has been rarely applied in multi-site supply chain
planning problem.
The objective of this paper is to develop a new mathematical model in order to
maximize the net profit of a multi-product, multi-period, multi-site supply chain
planning problem under demand uncertainty using a multi-stage stochastic pro-
gramming approach. A numerical example is illustrated to demonstrate the effec-
tiveness of the proposed model compared to the traditional deterministic model.
The remainder of the paper is organized as follows. In the next section, we
present the problem statement. Section 3 details the stochastic mathematical for-
mulation of the multi-site manufacturing network planning problem. Section 4
provides an illustrative example and corresponding analyses followed by the
conclusion and future work in section 5.
A Multi-site Supply Chain Planning Using Multi-stage Stochastic Programming 291
2 Problem Statement
In this paper, we consider a multi-site supply chain wherein the finished product is
processed by means of different production stages. Each production stage may in-
clude more than one plant forming a multi-site supply network manufacturing
environment as illustrated in Fig. 1. Within such multi-plant supply network, pro-
duction and supply decisions should be coordinated in order to achieve high per-
formance. Since these plants are situated in different areas, the transportation of
semi-finished and finished products between the entities of the supply chain is
considered. It should be noted, that a delivery lead time is taken into account in
the transportation of semi-finished product.
The fluctuation of customers’ demand is one of the most important sources of
uncertainty that should be considered in supply chain planning problems. Indeed,
the exact customer demand cannot be determined before the end of the planning
horizon. The under-estimation of overall demand leads either to loss sales and un-
satisfied customer. However, the over-estimation of the products demand results
in high production and inventory costs.
The objective of the supply network planning problem is to maximize the ex-
pected profits, computed by subtracting the expected total costs from the total rev-
enues. Decisions to be decided include the production amount, the inventory and
backorder sizes and the flows of materials between the different plants and cus-
tomers taking into account the demand uncertainty.
every stage may include more than one period. The evolution of the uncertainties
is represented using a scenario tree. The scenario tree is formed of many nodes
and arcs at each stage as illustrated in Fig. 2. The different nodes in each stage are
used to represent the stochastic parameters and the arcs denote the scenarios for
the following stage. Every arc of the scenario tree is associated with a probability,
which is the probability of scenario that correspond to this arc.
Indices
Li Set of direct successor plant of plant i.
ST j Set of stages (j= 1,2, ..., N).
i, i’ Production plant index (i,i’ = 1, 2 , ...,I) where i belongs to stage n
and i’ belongs to stage n+1.
A Multi-site Supply Chain Planning Using Multi-stage Stochastic Programming 293
Decision variables
Pikt (n) Production amounts of product k at plant i in period t at node n.
Amounts of end of period inventory of product k for scenario s at
S ikts ( n ) plant i in period t at node n.
s Amounts of end of period inventory of semi-finished product k
JSikt (n) for scenario s at plant i in period t at node n.
s Backorder amounts of finished product k for scenario s in period t
BDkt (n) at node n.
TRi −>i ',kt (n) Amounts of product k transported from plant i to i’ in period t at
node n.
s Amounts of product k transported from the last plant i to customer
TRi −>CUS ,kt (n) for scenario s in period t at node n.
Amounts of product k received by plant i for scenario s in period t
Qi ,k ( n)
at node n.
Parameters
s
D kt Demand of finished product k for scenario s in period t.
The occurrence probability of scenario s for the random demand
S
π s
where ∑π s
=1
s =1
−csik (S (n) + JS (n)) − cti −>CUS ,kTRis−>CUS ,kt (n) − cti −>i ',k TRi −>i ',kt (n)
s
ikt
s
ikt (1)
Nn T K I
−∑π n ∑∑∑ cpik Pikt (n)
n =1 t =1 k =1 i =1
294 H. Felfel, O. Ayadi, and F. Masmoudi
Siks ,t (n) = Siks ,t −1 (m) + Pikt (n) − ∑ TRi −>i ',kt (n), ∀i ∈ STj < N , ∀k , t, s, n (2)
i '∈Li
I I
∑b P
k =1
k ikt (n) ≤ cappit , ∀i, t , n (7)
∑S
k =1
s
ikt (n) + JSikts (n) ≤ capsit , ∀i, t , s, n (8)
∑ TR
k =1
i −> i ', kt (n) ≤ captrit , ∀i, t , s, n (9)
Pikt ( n), S ( n), JSikts ( n), TRi −>i ',kt ( n), TRis−>CUS ,kt ( n), Qi ,k ( n),
s
ikt
(10)
BDkts ( n) ≥ 0, ∀i , k , t , s , n
The objective function (1) aims to maximize the expected profit for different
scenarios of product demand by subtracting the total expected cost from the ex-
pected revenue. The equation (2) provides the balance for the inventory of prod-
ucts in every production stage except for the last stage. Constraint (3) represents
the balance for the inventory for the last production stage, considering customer
demands. Constraint (4) is the inventory balance equation for the semi-finished
products. Constraint (5) provides the balance equation for shortage in end product
demand. Equation (6) represents the balance equations for the transportation be-
tween the different plants. The set of constraints (7), (8) and (9) makes sure that
the production capacity, the storage capacity and the transportation capacity
are respected. Constraint (10) is the non-negativity restriction on the decision
variables.
A Multi-site Supply Chain Planning Using Multi-stage Stochastic Programming 295
5 Illustrative Example
T6 T8 T10
Scenario
P1 P2 PROB P1 P2 PROB P1 P2 PROB
S1 3350 3230 0.25 3300 3310 0.25 3430 3210 0.2
S2 2650 2520 0.35 2730 2770 0.4 2720 2530 0.3
S3 1860 1550 0.2 1930 2040 0.15 1950 1860 0.25
S4 1240 1020 0.2 1180 1060 0.2 1010 930 0.25
According to Fig. 3 and Table 2, the expected net profit of the supply chain
production plan of (4SM), (3SM) and (2SM) models are significantly higher than
those of the deterministic model. Moreover, the expected inventory and backorder
costs of the deterministic model are considerably greater than those of the (4SM),
(3SM) and (2SM) models. It is worth remarking that when the number of stages
increases in the demand scenario tree, the expected net profit increases too. In-
deed, the expected profit of (4SM) model is higher than those of (3SM) and (2SM)
models. Furthermore, Fig. 3 shows that the expected profit of (4SM) and (3SM)
models are close to each other. So, there is no need to consider more stages in the
scenario tree.
Where WS represents the objective value of the “wait and see” model and 2SP
is the objective value of two-stage stochastic programming model.
The second measure used in order to evaluate the performance of multi-stage
programming model is the relative value of the multi-stage stochastic solution
(RVMSS) (Nickel et al. 2012). This measure shows how much saving can be rea-
lized using the stochastic programming model in comparison with the determinis-
tic model. It is formulated as follows:
A Multi-site Supply Chain Planning Using Multi-stage Stochastic Programming 297
SP − EEV (12)
RVMSS = 100
EEV
Where EEV represents the optimal solution of deterministic model.
The obtained results are reported in Table 2 and Table 3. As shown in Table 2,
the EVPI/WS ratio shows a big influence of demand uncertainty on the obtained
solution (19.26%). So, it is worth to have better forecast about the demand scena-
rios. We can see also from Table 3 that the relative value of the multi-stage sto-
chastic solution RVMSS of the 4SM, 3SM and 2SM models are 20.33%, 19.19%
and 16.53%, respectively. It is clear that the multi-stage programming solution of
4SM and 3SM outperforms the two-stage programming solution for the supply
chain planning problem.
6 Conclusion
In this paper, we have developed a multi-site supply chain production planning
model under demand uncertainty. A multi-stage stochastic programming approach
is applied to incorporate the effects of uncertainty in the production planning prob-
lem. The computational results show that the quality of the four-stage and three
stage stochastic model solutions is higher than those of the two-stage stochastic
and deterministic models. Managing risk by incorporating a risk metric into the
stochastic programming model is one of the directions for future research.
References
Ahmed, S., King, A., Parija, G.: A multi-stage stochastic integer programming approach for
capacity expansion under uncertainty. Journal of Global Optimization 26, 3–24 (2003)
Birge, J.R., Louveaux, F.: Introduction to stochastic programming. Springer, New York
(1997)
298 H. Felfel, O. Ayadi, and F. Masmoudi
Felfel, H., Ayadi, O., Masmoudi, F.: Multi-objective Optimization of a Multi-site Manufac-
turing Network. In: Mechatronic Systems: Theory and Applications. Springer (2014)
Gupta, A., Maranas, C.D.: Managing demand uncertainty in supply chain planning. Com-
puters and Chemical Engineering 27(8-9), 1219–1227 (2003)
Kall, P., Wallace, S.W.: Stochastic Programming. John Wiley and Sons, New York (1994)
Kazemi Zanjani, M., Nourelfath, M., Ait-Kadi, D.: A multi-stage stochastic programming
approach for production planning with uncertainty in the quality of raw materials and
demand. International Journal of Production Research 48(16), 4701–4723 (2010)
Leung, S.C.H., Tsang, O.S., Ng, W.L., Wu, Y.: A robust optimization model for multi-site
production planning problem in an uncertain environment. European Journal of Opera-
tional Research 181, 224–238 (2007)
Leung, S.C.H., Wu, Y., Lai, K.K.: A stochastic programming approach for multi-site ag-
gregate production planning. Journal of the Operational Research Society 57, 123–132
(2006)
Lin, J.T., Chen, Y.Y.: A multi-site supply network planning problem considering variable
time buckets– A TFT-LCD industry case. The International Journal of Advanced Manu-
facturing Technology 33(9-10), 1031–1044 (2006)
Lin, J.T., Chen, T.L., Chu, H.C.: A stochastic dynamic programming approach for multi-
site capacity planning in TFT-LCD manufacturing under demand uncertainty. Interna-
tional Journal of Production Economics 148, 21–36 (2014)
Mirzapour Al-e-hashem, S.M.J., Baboli, A., Sazvar, Z.: A stochastic aggregate production
planning model in a green supply chain: considering flexible lead times, nonlinear pur-
chase and shortage cost functions. European Journal of Operational Research (ISI) 230,
26–41 (2013)
Nagar, L., Jain, K.: Supply chain planning using multi-stage stochastic programming.
Supply Chain Management: An International Journal 13(3), 251–256 (2008)
Nickel, S., Saldanha-da-Gamac, F., Zieglera, H.P.: A multi-stage stochastic supply network
design problem with financial decisions and risk management. Omega 40(5), 511–524
(2012)
Shah, N.K., Ierapetritou, M.G.: Integrated production planning and scheduling optimization
of multisite, multiproduct process industry. Computers and Chemical Engineering 37,
214–226 (2012)
Wang, R.C., Fang, H.H.: Aggregate production planning in a fuzzy environment. Interna-
tional Journal of Industrial Engineering-Theory, Applications, and Practice 7(1), 5–14
(2000)
A System Engineering Conception of Multi-
objective Optimization for Multi-physics System
Mian Chen1 and Omar Hammami2
1
IRT SystemX, France
mian.chen@iry-systemx.fr
2
ENSTA ParisTech, France
omar.hammami@ensta-paristech.fr
Abstract. At present, a system involves more and more domains and becomes
more and more complex. In this paper, a system engineering conception of multi-
objective optimization for a multi-physics system is formed. An example of multi-
physics system design illustrates the presented approach.
Keywords: System Engineering, Multi-Physics Simulation, Multi-Objective
Optimization.
1 Introduction
A system recently deals with more and more domains like mechanical, electronic
and thermal etc. In addition, the interaction and information exchanges among
them increase rapidly. Therefore, the complexity of system is increasing gradual-
ly. As the complexity of a system augments, the work of system optimization be-
comes not only heavier but also more difficult. Besides, the cost and time for the
whole system engineering design augment [9].
System engineering flow plays a crucial role because it should have a global
view of the whole system and optimize the whole system in consideration of mul-
tiple objectives in different domains. Also, it ought to insure that the system satis-
fies some requirements [8].
For the moment, in order to develop a multi-physics system, the experts in dif-
ferent domains have to hold a lot of meetings to reach agreement. However, a
method of automatic system engineering working flow with multi-objective opti-
mization simultaneously is expected.
In order to realize the expected system engineering flow and take into account
the feasibility, several software are chosen as Fig.2. Rhapsody is chosen as a tool
to deal with system engineering model which presents the system’s requirements
and describes all the structural, parametric information by language SysML [6].
At the bottom level, several different software could be applied. Catia is chosen
to simulate the geometrical model, Dymola is used to simulate the mechatronic
model and Comsol is chosen to do the mechanical and thermal simulation.
Certainly, other software of physical simulation would be added in this multi-
objective system engineering working flow to realize different simulations.
Besides, the software ModelCenter is taken use to do the multi-objective opti-
mization, which would harmonize multiple simulations in different software and
on the other side, it could collaborate with Rhapsody [7].
respectively two jacks to open or close the whole liftgate automatically which
could replace traditionally manual operation.
Fig. 5 Comsol model - Stress distribution Fig. 6 Comsol model - Displacement distri-
bution
After modeling of motor and liftgate, those two parts could be combined. In the
Dymola model, the motor control part is simplified, a constant tension is supposed
to supply those two motors which drive the whole liftgate. Finally, a 3D animation
simulation has been done. The whole process of opening or closing is between 7
seconds and 8 seconds (Fig.9 and Fig.10).
304 M. Chen and O. Hammami
4 Multi-objective Optimization
Before running the whole system multi-objective optimization, one important con-
ception should be emphasized: an optimization problem with more than one objec-
tive, there is no single optimum solution. It exists a number of solutions which are
all optimal. Without any further information, no solution from the set of optimal
solutions can be said to be better than any other.
4.1 NSGA-II
Compared with other multi-objective algorithms like Multi-Objective Simulated
Annealing (MOSA) [5], Multi-Objective Optimization using Evolutionary Algo-
rithms (MOEA) [4] and Multi-Objective Genetic Algorithm II (MOGA-II) [3], an
algorithm called Non-dominated Sorting Genetic Algorithm II (NSGA-II) [1] [2]
is selected to achieve the multi-objective optimization in this paper.
The algorithm NSGA-II involves with three main conceptions: non-dominated
sort, crowding distance and genetic operators. Its main loop of optimization
process is presented in Fig.11.
Fig. 13 Results
5 Conclusion
Even though the main conception is realized in this paper, from the industrial
viewpoint, this study case is too simple. Therefore, in the future work, some more
connections problems between those software and models synchronization should
be solved.
Acknowledgment. For the members in the laboratory U2IS and SUPMECA, here we ex-
press all our esteem and sincere acknowledgement to them for their encouragement, friend-
ship moral support and their scientific assistants for all the time.
References
[1] Deb, K.: Multi-Objective Optimization Using Evolutionary Algorithms, Chichester,
U.K. (2001)
[2] Deb, K., Pratap, A., Agarwal, S., Meyarivan, T.: A Fast and Elitist Multiobjective
Genetic Algorithm: NSGA-II. IEEE Trans. Evol., Comput. 6(2), 182–197 (2002)
[3] Poles, S.: MOGA-II An improved Multi-Objective Genetic Algorithm. Esteco, Italy
(2003)
[4] Seshadri, A.: Multi-Objective Optimization using Evolutionary Algorithms (MOEA)
[5] Rigoni, E.: MOSA Multi-Objective Simulated Annealing, Esteco, Italy (2003)
[6] Systems Engineering Tutorial for Rational Rhapsody, IBM (2009)
[7] MBSE PAK for Rhapsody USER GUIDE, Phoenix Integration, Blacksburg, USA
(2014)
[8] Hammami, O.: SYNSYS-ME: Seamless System Engineering to mechanical flow
through multiobjective optimization and requirements analysis. In: 8th Annual IEEE
Systems Conference (SysCon), Ottawa, Canada (2014)
[9] Hammadi, M., Choley, J., Penas, O., Riviere, A., Louati, J., Haddar, M.: A new mul-
ti-criteria indicator for mechatronic system performance evaluation in preliminary de-
sign level. In: 9th France-Japan Mechatronics, Paris, France (2012)
[10] Raka, S.: Methodes et outils ensemblistes pour le pre-dimensionnement de systemes
mecatroniques. Ensea, Cergy-Pontoise, France (2011)
[11] Peak, R.S., Burkhart, R., Friedenthal, S., Wilson, M., Bajaj, M., Kim, I.: Simulation-
Based Design Using SysML Part 1: A Parametrics Primer. In: INCOSE International
Symposium, San Diego, U.S.A. (2007)
[12] Peak, R., Burkhart, R.M., Friedenthal, S.A., Wilson, M.W., Bajaj, M., Kim, I.: Simu-
lation-Based Design Using SysML Part 2: Celebrating Diversity by Example. In:
INCOSE Intl. Symposium, San Diego, U.S.A. (2007)
[13] Hammadi, M.: Contribution a l’integration de la modelisation et la simulation multi-
physique pour la conception des systemes mecatroniques. In: LISMMA and
SUPMECA, Paris, France (2012)
Falling Film in a Heated Micro-channel
1
Research unit of heat and thermodynamics of industrial processes
National Engineering School of Monastir
Rue Ibn Al Jazzar 5000 Monastir, Tunisia
sabrynener@yahoo.fr
2
Research Center of Albi in Process Engineering, Divided Solids, Energy and Environment
Ecole des Mines d'Albi Campus Jarlard - Route de Teillet
81013 Albi CT Cedex 09, France
fudym@mines-albi.fr
Abstract. The main objective of this work is to study experimentally and numeri-
cally a falling film in a micro-channel. The experimental section involves in creat-
ing a temperature gradient within the liquid, while monitoring the temperature
using an infrared camera. A numerical model is established and solved by a semi-
analytical method called the thermal quadrupole method. Finally, we conclude
with a comparison between the experiments and the numerical study.
1 Introduction
Miniaturization is currently a promising field of both theoretical and experimental
research (Aiello et al. 2009; Foerster et al. 2013). It is to reduce the dimensions of
some industrial equipment in order to reduce the technological constraints in favor of
environmental chemistry and processes safety (Brendner et al. 2013). It is targeting,
in general, refrigeration equipments, air conditioning and heat pumps which have
their effects on the environment, the layer ozone and global warming (Luke et al.
2010). Thus, comes the idea of increasing processes. This means: developing and
creating methods and devices offering improved production quality, reducing the size
/ capacity, lowering effluent quantities and / or catalysts, and reducing costs and
energy losses with a greater respect for environmental constraints (Mhiri.N 2006).
Micro-reactors operate on the principle of a continuous process and are highly
different from those several traditional characteristics such that the increase of the
exchange surface through the ratio S / V (Perry et al. 2010), the improvement of heat
and mass transfer and a high temperature gradient and pressure (Baldassari et al.
2013; Conti et al. 2012). They help maintain better control of reaction conditions,
improved security conditions, the gain in energy and reducing congestion (Hung et al.
2013). The falling films are generally used as an intermediary for the transfer of heat
Our experiment consists in: creating a temperature gradient, in the back of the plate,
and then managing the liquid flow (water and ethanol). An infrared camera is used in
order to visualize the effect of the temperature gradient on the flow.
2.1 Preparation
First of all, the plate was cleaned well and then sprayed with a black spray to create a
black thin layer in order to maximize radiation, since an infrared camera is used for
the visualization of the heat transfer on the surface of the falling film. Then, an
electrical resistance is implanted in the back of the plate to generate heat.
40
30
20
10 Ethanol surface
0 Temperature in a
micro-channel
10.0625
10.9375
11.8125
12.6875
13.5625
14.4375
2.1875
3.0625
3.9375
4.8125
5.6875
6.5625
7.4375
8.3125
9.1875
Lenght (cm)
Fig. 3 Curve of the surface temperature of the ethanol falling film in a single channel
For the water (Fig. 4 and Fig. 5), the temperature reached 33°C, and then went
down to 18°C. In both experiments, we calculated the temperature gradient and found
it 2.8°C per centimeter which is close to 3°C/cm, the value we wanted to reach.
Falling Film in a Heated Micro-channel 311
5.544
6.468
7.392
8.316
0
10.164
11.088
12.012
12.936
4.62
9.24
Lenght (cm)
Fig. 5 Curve of the surface temperature of the falling film of ethanol for one channel
3 Mathematical Model
y
u = um (1 − (1 − )²) (1)
δ
∂T ∂ ²T ∂ ²T
u = α( + ) (2)
∂x ∂x ² ∂y ²
312 S. Mejri et al.
∂ ²T ∂ ²T
+ =0 (3)
∂x² ∂y ²
Equation (2) is the result of the energy equation in the falling film and Equation
(3) is the result of the energy equation in the micro-channel.
Transforming all the equations previously obtained along with the boundary
conditions gave us these results:
In the plate:
⎛ 1 ⎞
⎛θ 0 ⎞ ⎜ cos pl sin pl ⎟⎛ θ l ⎞
⎜⎜ ⎟⎟ = ⎜ λ p ⎟⎜⎜ ⎟⎟ (4)
⎝ φ 0 ⎠ ⎜ − λ p sin p l φ
cos pl ⎟⎠⎝ l ⎠
⎝
In order to use the thermal quadrupoles method on the liquid, we proceed by
dividing the film into N layers having each one the same “Eq. 1”. So the matrix in
each layer is:
Falling Film in a Heated Micro-channel 313
(5)
⎛ 1 ⎞ ⎛ δ 1 δ⎞
⎛θw ⎞ ⎜ cos pl sin pl⎟ N ⎜ coshKi N λK sinhKi N⎟⎛θδ ⎞
⎜⎜ ⎟⎟ =⎜ pl ⎟*∏⎜ ⎟⎜ ⎟ (6)
δ ⎟⎜⎝φδ ⎟⎠
i
With the inverse Fourier transform, we find the expression of the surface tem-
perature of the falling liquid film which is simulated then to make the profile:
The Figure (Fig. 7) shows the curve of the temperature profile at the surface of the
flow. We can see the increase in the temperature in the heating zone.
314 S. Mejri et al.
Note that the temperature rises from 22.8 ° C, the input channel temperature and
the beginning of the heating zone, at x = 0.018 m, reached its peak at x = 0.052 m
with a temperature of 27.4 ° C, and begins to fall back to its initial temperature. This
drop is due to the periodicity of the proper function (temperature), since the resolution
was made by the integral Fourier transform. Same thing for the ethanol graph
(Fig. 8), rising from a temperature close to 22°C, the liquid exceeded 28°C in the
heating zone and then come down to the cooling temperature.
We studied also the changes in temperature within the liquid along the y axis. We
note that there is not a great change throughout the y axis which means we can avoid
the division of our multilayer medium, that is to say we can keep a single mesh.
There is, also, a resemblance between the numerically determined temperature profile
and that determined experimentally. From the figure that the temperature profile in
the flow, we can conclude that the independence with respect to y means that the
temperature can be considered as a single unit with one average speed.
We managed to achieve a temperature gradient close to 3 ° C / cm; this number
was calculated by the laboratory of Nancy LRGP for diabatic distillation.
Falling Film in a Heated Micro-channel 315
4 Conclusion
References
Zhang, F., Wub, Y.T., Gengb, J., Zhangb, Z.B.: An investigation of falling liquid films on a
vertical heated/cooled plate. International Journal of Multiphase Flow 34, 13–28 (2008),
doi:10.1016/j.ijmultiphaseflow.2007.08.003
Mhiri, N.: Etude d’un micro-absorbeur Gaz/Liquide à film tombant. Récents progrès en
génie des procédés, Dissertation, Institut national polytechnique de Lorraine (2006)
Fudym, O., Ladevie, B., Batsale, J.C.: A seminumerical approach for heat diffusion in hete-
rogeneous media: one extension of analytical quadrupole method. Numerical Heat
Transfer, Part B 42, 325–348 (2002), doi:10.1080/10407790190053978
Bu, X., Ma, W., Huang, Y.: Numerical study of heat and mass transfer of ammonia-water in
falling film evaporator. Heat Mass Transfer 48, 725–734 (2011), doi:10.1007/s00231-
011-0923-4
Perry, I., Jannot, Y., Maillet, D., Fiers, B.: Effect of velocity distribution on external wall
temperature field for a flat microchannel. Experimental Heat Transfer 23, 27–43 (2010),
doi:10.1080/08916150903402724
Baldassari, C., Marengo, M.: Flow boiling in micro-channels and micro-gravity. Progress in
Energy and Combustion Science 39, 1–36 (2013), doi:10.1016/j.pecs.2012.10.001
Hung, T.C., Huang, Y.X., Yan, W.M.: Thermal performance analysis of porous-
microchannel heat sink with different configuration designs. International Journal of Heat
and Mass Transfer 66, 235–243 (2013), doi:10.1016/j.ijheatmasstransfer.2013.07.019
Wiesegger, L.E., Knauss, R.P., Guntsching, G.E., Brandner, J.J.: Vapor-liquid phase sepa-
ration in micro-/ministructured devices. Chemical Engineering Science 93, 32–46
(2013), doi:10.1016/j.ces.2013.01.030
Conti, A., Lorenzini, G., Jeluria, Y.: Transient conjugate heat transfer in straight micro-
channels. International Journal of Heat and Mass Transfer 55, 7532–7543 (2012),
doi:10.1016/j.ijheatmasstransfer.2012.07.046
316 S. Mejri et al.
Verma, N., Mewes, D., Luke, A.: Lattice Boltzman study of velocity, temperature, and con-
centration in micro-reactors. International Journal of Heat and Mass Transfer 53,
3175–3185 (2010), doi:10.1016/j.ijheatmasstransfer.2010.03.009
Foerster, M., Lam, K.F., Sorensen, E., Gavriilidis, A.: In situ monitoring of microfluidic distil-
lation. Chemical Engineering Journal 227, 13–21 (2013), doi:10.1016/j.cej.2012.11.125
Marenzana, G., Perry, I., Maillet, D.: Mini and micro-channels: influence of axial conduc-
tion in the walls. International Journal of Heat and Mass Transfer 47, 3993–4004 (2004),
doi:10.1016/j.ijheatmasstransfer.2004.04.016
Buonomo, B., Manca, O.: Transient natural convection in a vertical micro-channel heated at
uniform heat flux. International Journal of Thermal Sciences 56, 35–47 (2012),
doi:10.1016/j.ijthermalsci.2012.01.013
Aiello, G., Ciafolo, M.: Natural convection cooling of a hot vertical wall wet by a falling
liquid film. International Journal of Heat and Mass Transfer 52, 5954–5961 (2009),
doi:10.1016/j.ijheatmasstransfer.2009.08.007
Interfacial Stresses in FRP-Plated RC Beams:
Effect of Adherend Shear Deformations
1
Département de Génie Civil, Université Ibn Khaldoun, BP 78 Zaaroura, 14000 Tiaret, Algérie
master1513@hotmail.fr,
bem13@live.fr,
daouadjitah@yahoo.fr
2
Département de Génie Civil, Université Hassiba Benbouali, Chlef, Algérie
chersel826@gmail.com
1 Introduction
The structural plating of reinforced concrete (RC) beams with fibre reinforced po-
lymer (FRP) laminates represents a new technology in the civil engineering field.
In fact, until today, the most common method in structural strengthening has been
the hermitian technique using the application of steel plates.
The new method has many advantages:
• High strength to weight ratio;
• Lightness;
• Corrosion resistance.
2 Method of Solution
the beam is loaded, vertical separation occurs between RC beam and FRP platte.
This separation creates ann interfacial normal stress in the adhesive layer. We noote
that this assumption is ussed in several works, e.g., Tounsi et al., Smith and Tenng
and Stratford and Cadei.
7. Aparabolic shear sttress distribution through the depth of both the concreete
beam and the bonded platte is assumed.
8. The section propertiies of the RC beam were based on the uncracked section,
excluding the conventionaal steel reinforcement. It is known that for uncracked secc-
tion, the concrete can susstain tension. However, for cracked section, the concreete
cannot sustain tension andd this is why the effect of steel reinforcement in concreete
is not neglected.
du2 (x) − y2 N 5t dτ a
ε 2 (x) = = M2 (x) + 2 − 2 (2)
dx E2 I 2 E2 A2 12G2 dx
Where N(x) is the axial force in each adherend and A is the cross- sectional area.
The shear stress in the adhesive can be expressed as follows:
τ a = τ ( x ) = K s [u 2 ( x ) − u1 ( x )] (3)
Where MT(x) is the total applied moment. The bending moment in each adherend,
expressed as a function of the total applied moment and the interfacial shear stress,
is given as
R ⎡ ⎤
x
M1 (x) = ⎢MT (x) − b2 ∫τ (x)( y1 + y2 + ta )dx⎥ (5)
R +1 ⎣ 0 ⎦
1 ⎡ ⎤
x
(6)
M ( x) =
2 ⎢
R +1 ⎣
T 2 ∫
M ( x ) − b τ ( x )( y + y + t )dx
0
1 2 a ⎥
⎦
d 2τ (x) ⎛ ( y + y )( y + y + t ) 1 1 ⎞ ⎛ y +y ⎞ (10)
− K1b2 ⎜⎜ 1 2 1 2 a + + ⎟⎟τ ( x) + K1⎜⎜ 1 2 ⎟⎟VT ( x) = 0
dx2 ⎝ E I
1 1 + E I
2 2 E A
1 1 E2 2⎠
A ⎝ E1I1 + E1I1 ⎠
Where
1 (11)
K1 =
⎛ ta t 5t ⎞
⎜⎜ + 1 + 2 ⎟⎟
⎝ a
G 4G 1 12G2 ⎠
⎛ ( y1 + y2 )(y1 + y2 + ta ) 1 1 ⎞
λ2 = K1b2 ⎜⎜ + + ⎟ (13)
⎝ E1I1 + E2 I 2 E1 A1 E2 A2 ⎟⎠
K1 ⎛ y1 + y2 ⎞
m1 = ⎜ ⎟ (14)
λ2 ⎜⎝ E1 I1 + E2 I 2 ⎟⎠
The expression for the interfacial shear stress at any point
− λx
⎡ m2 a ⎤ qe ⎛L ⎞ (15)
τ ( x) = ⎢ ( L − a) − m1 ⎥ + m1q⎜ − a − x ⎟ 0 ≤ x ≤ LP
⎣ 2 ⎦ λ ⎝2 ⎠
Where q is the uniformly distributed load and x; a; L and Lp are defined in Fig. 1.
K1 y1
m2 = (16)
E1I1
Where Kn is the normal stiffness of the adhesive per unit length and can be written
as
σ n ( x) σ n ( x) ⎛ 1 ⎞ E
Kn = = ⎜ ⎟= a (18)
Δ w( x ) Δw( x) / t a ⎜⎝ t a ⎟⎠ t a
W1(x) and W2(x) are the normal displacements of adherends 1 and 2, respectively.
Differentiating Eq. (17) twice results in
d 2σ n ( x ) ⎡ d 2 w 2 ( x ) d 2 w1 ( x ) ⎤ (19)
= n⎢ −
dx 2 ⎥⎦
K
dx 2 ⎣ dx
2
Adherend 1:
d 4 w1 ( x) 1 y dτ ( x) q
4
= b2σ ( x) + 1 b2 +
dx E1I1 E1 I1 dx E1 I1 (20)
Adherend 2:
d 4 w2 ( x) 1 y dτ ( x)
4
=− b2σ ( x) + 2 b2 (21)
dx E2 I 2 E2 I2 dx
322 A. Rabahi et al.
Substitution of Eqs. (20) and (21) into the fourth derivation of the interfacial
normal stress obtainable from Eq. (19) gives the following governing differential
equation for the interfacial normal stress:
d 4σ n ( x) Eab2 ⎛ 1 1 ⎞ Eb ⎛ y y ⎞ dτ ( x) qEa
4
+ ⎜⎜ + ⎟⎟σ n ( x) + a 2 ⎜⎜ 1 − 2 ⎟⎟ + =0 (22)
dx ta ⎝ E1I1 E2 I 2 ⎠ ta ⎝ E1I1 E2 I 2 ⎠ dx ta E1I1
E a b2 ⎛ 1 1 ⎞
β =4 ⎜⎜ + ⎟ (25)
4t a ⎝ E1 I1 E 2 I 2 ⎟⎠
⎛ y E I − y2 E1I1 ⎞
n1 = ⎜⎜ 1 2 2 ⎟⎟ (26)
⎝ E1I1 + E2 I 2 ⎠
E2 I 2
n2 = (27)
b2 ( E1 I1 + E 2 I 2 )
As is described by Tounsi 2006, the constants C1 and C2 in Eq.(24) are deter-
mined using the appropriate boundary conditions and they are written as follows:
Ea
C1 = [VT (0) + βMT (0)] − n3 3 τ (0) +
2β ta E1I1
3
2β (28)
n1 ⎛ d 4τ (0) d 3τ (0) ⎞
⎜
3 ⎜
+β ⎟
2β ⎝ dx 4
dx3 ⎟⎠
Ea n1 d 3τ (0)
C2 = M ( 0 ) − (29)
2β 2 t a E1 I1 2 β 2 dx 3
T
Where:
Eat2 ⎛ y1 y2 ⎞
n3 = ⎜⎜ − ⎟⎟ (30)
ta ⎝ E1 I 1 E 2 I 2 ⎠
Interfacial Stresses in FRP-Plated RC Beams: Effect of Adherend Shear Deformations 323
The effect of the adherend shear deformations using the as sumption of parabolic
shear stress distribution across the thickness of the adherends on the maximum
shear and normal stress is examined by comparing the results obtained with the
present theory and those determined by two other theories [Smith S.T 2001 a,b,c].
This difference is due to the assumption used in the present theory which is in
agreement with the beam theory (parabolic distribution of shear stresses through
the depth of the beam). Hence, it is apparent that the adherend shear deformation
reduces the interfacial stresses concentration and thus renders the adhesive shear
distribution more uniform. The interfacial normal stress is seen to change sign at a
short distance away from the plate end.
4 Parametric Study
5 Conclusion
New theoretical interfacial stress analysis has been presented for simply supported
RC beams bonded with a thin composite or steel plate and subjected to a uniform-
ly distributed bending load. The salient features of the new analysis include the
consideration of the adherend shear deformations by assuming a parabolic shear
stress through the thickness of both the concrete beam and bonded plate. The solu-
tion methodology is general in nature and may be applicable to the analysis of
other types of composite structures. By comparing with experimental results, this
new solution provides satisfactory predictions to the interfacial stress in the plated
beams. The classical solutions which neglect the adherend shear deformations
over-estimate the non-uniformity of the adhesive stresses’ distributions and max-
imum interfacial stresses. In the final part of this paper, extensive parametric
studies were undertaken by using the new solution for strengthened beams with
various ratios of design parameters. Observations were made based on the numeri-
cal results concerning their possible implications to practical designs.
326 A. Rabahi et al.
References
[Smith S.T 2001a] Smith, S.T., Teng, J.G.: FRP – strengthened RC beams. I: Review of
debonding strength models. Engineering Structures 24(4), 385–395 (2002)
[Smith S.T 2001b] Smith, S.T., Teng, J.G.: FRP – strengthened RC beams. II: Assessment
of debonding strength models. Engineering Structures 24(4), 397–417 (2002)
[Smith S.T 2001c] Smith, S.T., Teng, J.G.: Interfacial stresses in plated RC beams. Engi-
neering Structures 23(7), 857–871 (2001)
[Yu 1989] Yu, T.Q., Miao, X.S., Xiong, J.M., Jiang, H., Lee, H.: An orthotropic dam-age
model for concrete at different temperatures. Engng. Fract. Mech. 32, 775–786 (1989)
[Tounsi 2006] Tounsi, A.: Improved theoretical solution for interfacial stresses in con-crete
beams strengthened with FRP plate. Int. J. Solids Struct. 43, 4154–4174 (2006)
A Sensitivity Analysis of Multi-objective
Cooperative Planning Optimization Using
NSGA-II
1 Introduction
The genetic algorithms (GA) have proven their robustness and powerful search
mechanism to solve a variety of single as well as multi-objective problems. It
features by the large search space and the multiple alternative trade-offs generated in
a single run. NSGA-II is one of the most popular multi-objectives GAs that shows
high performance and was successfully applied in various disciplines ((Murugan et
al. 2009), (Kanagarajan et al. 2007), (Kannan et al. 2009), (Agrawal et al. 2007),
(Bekele et al. 2007), (Atiquzzaman et al. 2006), (Hnaien et al. 2010), (Huang et al.
*
Corresponding author.
2010)). Several GA parameters, like the population size and the mutation and
crossover operators, are involved in the optimization process and interact in a
difficult way. Many studies tried to give some guidelines for the choice of the GA
parameters, like the population size or the probabilities operators ((Harik et al.
1999), (Goldberg et al. 1993), (Deb & Agrawal, 1999)...). However, most of these
works made general observations about the parameters setting. According to Hart &
Belew (1991) the parameters interaction mostly depends on the function being
optimized. The problem arises especially in the choice of the adequate population
size. Many researchers scruple about using small population sizes that may risk the
possibility of finding good solutions, or using large population sizes which require
more processing time, increase computational costs and may slow the performance.
That is why in most of works, repeated simulations are required to select the suitable
parameters values. Deb & Agrawal, (1999) investigated the effect of GA operators
for solving simple and complex problems. The considered problems are
Himmeblau's function, Rastrigin's function and deceptive function. They observed
that mutation operator plays a significant role as well the crossover operator, in the
case of simple problems. But, each operator has its working zone. In the case of
complex problem, the most significant operator is the crossover operator, but this
assertion necessitates the choice of the adequate population size. Pongcharoen et al.
(2002) developed a factorial experiment to identify the most efficient genetic
parameters (the probabilities of crossover and mutation, the population size and the
generations’ number) to solve a scheduling problem that aims to minimize a total
cost function. According to their study, the only significant parameters are the
population size, the generation number and the mutation probability. Tran (2005)
developed parameterless NSGA-II, to make NSGA-II easier. The developed
algorithm performs quite well than the original NSGA-II, while solving a
benchmark problem. Nevertheless, it is slower and has worse distribution of the
solutions. According to Zeng et al. (2010), the performance of multi-objective
evolutionary algorithms is limited by the suitability of their corresponding parameter
settings. Thus, they develop a self-adaptive mechanism adjustment of the
distribution index of the simulated binary crossover (SBX) operator to improve
NSGA-II performance.
Most of previous researches, consider benchmark functions in order to study the
GA parameters. This paper focuses on cooperative planning problem, and considers
the seven genetic parameters involved in the optimization process of NSGA-II.
In Ben Yahia et al. (2013), the NSGA-II is adapted to solve the proposed multi-
objective planning model. The model is a bi-objective one, which aims to minimize
the total production cost and the average of inventory level of a cooperative supply
chain (SC). While the proposed model is applied to experimental examples in order
to generate corresponding optimal Pareto solutions, optimization process failed to
generate Pareto of more than one solution and the sensitivity of the results to
NSGA-II parameters variations is not addressed. Our interest here is to investigate
and to analyze this sensitivity during the optimization process. The idea is to
generate optimal production plans while varying the NSGA-II parameters and to
explore their effects.
A Sensitivity Analysis of Multi-objective Cooperative Planning Optimization 329
3 Mathematical Formulation
optimal production plan for a whole manufacturing supply chain, within a fixed
time horizon with a finite capacity of personnel and machines.
The multi-objective optimization problem must satisfy the following assump-
tions:
Several resources, with limited availabilities, can process several items. Raw
materials are always available.
The sequence of operations required to produce an item is fixed, and any
alternative routing is prohibited.
Inventories at the initial period are void.
Items can be only produced if all their predecessor components are available.
Periodic external demand of each item is known.
Overtime is allowed to extend fixed capacity availabilities. Backlogging is not
allowed.
Inventory is calculated at the end of a time period.
External demand has to be fully met in time and quantity.
M in ( C , Im oy ) (1)
T T
St C = ∑ ∑ [( cv . x
t =1 j∈ J
j j, t ) + ( cf j. y j , t )] + ∑ ∑ co r .o r , t
t =1 r ∈ R
(2)
1 T
Im oy = ∑ ∑ ij , t
T t =1 j ∈ J
(3)
i j , t −1 + x j , t = D j , t + r j , k x k , t + i j , t ∀j ∈ J , ∀t (4)
∑ j a r , j. x j , t ≤ C r , t + o r , t ∀j ∈ J , ∀t, ∀r ∈ R (5)
x j , t ≤ L j , t. y j , t ∀j ∈ J , ∀t (6)
x j , t ≥ 0, ij , t ≥ 0 ∀j ∈ J , ∀t (7)
or , t ≥ 0 ∀t, ∀r ∈ R (8)
y j , t ∈ {0,1} ∀j ∈ J , ∀t (9)
The first objective function (2) consists of the minimization of operations, setup
and overtime costs. The second objective (3) aims to minimize the average level of
inventory with respect to the number of planning periods. The constraint (4)
represents the flow balance between output, inventory and consumption by external
demand or successor operations. The constraints (5) ensure the capacity restrictions
in using the resources to produce the different items. Lot-sizing relationships and the
choice of the items to be produced at each time-period and at each plant location are
expressed in (6). The constraints (7), (8) and (9) specify the domains of the different
variables.
4 Tests Description
The purpose of this study is to investigate the influence of the NSGA-II parameters
on the previously proposed model. Two cases will be studied, where only the
demand typology changes. In each test, customer demands are supposed to be
332 W. Ben Yahia, O. Ayadi, and F. Masmoudi
deterministic and e to be fulfilled during the planning horizon while facing finite
resources capacities. The first case is labelled 2PU.2P.ID and is represented by a
supply chain of two production units (2PU) planning its production activities over
two periods (2P) to satisfy the external increasing demand (ID). As shown in table 1,
the demand of the second period exceeds the main available capacity. The second
case (2PU.2P.DD) has the same configuration, but the external demand has a
decreasing trend (DD). This time, the demand of the first period exceeds the main
available capacity. These two tests have the same number of decision variables and
constraints.
The search strategies of the both studied cases are completely different.
First, the 2PU.2P.ID case has a larger solutions region and the conflict between
considered objectives occurs especially in this case. In fact, for the increasing
demand, in order minimize the total production cost the use of the overtime must
be avoided. But like that, the production must be made in advance, which will
increase the average of inventory level. However, in the 2PU.2P.DD case, the
demand has a decreasing pattern, which does not exceed the main available
capacity at the end of the planning period. So there is no need to produce or to
storage. Therefore, during the optimization the algorithm has to find a balance
between the use of the overtime and the inventory level. Ultimately, the choice
of the most suitable production plan is made by the partners according to their
preferences.
2PU.2P.ID 2PU.2P.DD
The parameters are examined separately, i.e. changing the value for the given
parameter while keeping the values of the rest of parameters at a constant level.
During the investigation of the effect of each parameter is examined separately.
In other words, when the considered parameter is varied, the other parameters
keep constant values. The adopted constant values of each parameter are shown in
table 2.
A Sensitivity Analysis of Multi-objective Cooperative Planning Optimization 333
Generation number
generation number
1000 800 600 400 200 20
Inventory level
Inventory
30
40
50
60 000
40000
Total cost
Cost total
20 000
takes a regular pattern in other word the found solution does not improves while
increasing the generation number. During the convergence to the optimal solution,
the total production cost is minimized by 91% in the 2PU.2P.DD case and 96% in
the 2PU.2P.ID case; while, the average of inventory level is minimized by 52% in
the 2PU.2P.DD case and 15% in the 2PU.2P.ID case. Therefore, in- creasing the
generation number improves the solution quality.
Inventory
1 00
Inventory
00
Inventory
Inventory 50
50
25 000
25000
100
0.2 20 000 20 000
20
40
Inventory
Inventory 60
100
80
100
Inventory
50 20 000
15 000
Cost total
Total cost 50 10 000
0.6 10 000 Cost total
crossover distribution index
0.8
ηc Total cost
5000
crossover probability
Pm 1.0
0
Inventory
80
60
100
40 Inventory
Inventory
20
50 Inventory
25000
0.06
20 000
0.04
15 000
15 000 mutation probability
totalCost
Total cost
Pm 0 10000
10 000 0.02
Cost total 200
Total cost distribution index mutation 5000
5000 ηm 400
6 Conclusion
References
Agrawal, N., Rangaiah, G., Ray, A., Gupta, S.: Design stage optimization of an industrial
low-density polyethylene tubular reactor for multiple objectives using NSGA-II and its
jumping gene adaptations. Chemical Engineering Science 62(9), 2346–2365 (2007)
Atiquzzaman, M., Liong, S., Yu, X.: Alternative decision making in water distribution
network with NSGA-II. Journal of Water Resources Planning and Management 132(2),
122–126 (2006)
Bekele, E.G., Nicklow, J.W.: Multi-objective automatic calibration of SWAT using NSGA-
II. Journal of Hydrology 341(3-4), 165–176 (2007)
Ben Yahia, W., Cheikhrouhou, N., Ayadi, O., Masmoudi, F.: A Multi-objective
Optimization for Multi-period Planning in Multi-item Cooperative Manufacturing
Supply Chain. In: Haddar, M., Romdhane, L., Louati, J., Ben Amara, A. (eds.) Design
and Modelling of Mechanical System, pp. 635–643. Springer, Heidelberg (2013)
Deb, K., Agrawal, S.: Understanding interactions among genetic algorithm parameters. In:
Foundations of Genetic Algorithms V, pp. 265–286. Morgan Kaufmann, San Mateo
(1999)
Deb, K., Agrawal, S.: A fast elitist non-dominated sorting genetic algorithm for multi-
objective optimization: NSGA-II. IEEE Transactions on Evolutionary Computation 6(2),
182–197 (2002)
Harik, G., Cantú-Paz, E.: The gambler’s ruin problem, genetic algorithms, and the sizing of
populations. Evolutionary Computation 7, 231–253 (1999)
A Sensitivity Analysis of Multi-objective Cooperative Planning Optimization 337
Hart, W.E., Belew, R.K.: Optimizing an Arbitrary Function is Hard for the Genetic
Algorithm. In: Proceedings of the Fourth International Conference on Genetic
Algorithms, pp. 190–195 (1991)
Hnaien, F., Delorme, X., Dolgui, A.: Multi-objective optimization for inventory control in
two-level assembly systems under uncertainty of lead times. Computers & Operations
Research 37(11), 1835–1843 (2010)
Huang, B., Buckley, B., Kechadi, T.-M.: Multi-objective feature selection by using NSGA-
II for customer churn prediction in telecommunications. Expert Systems with
Applications 37(5), 3638–3646 (2010)
Kanagarajan, D., Karthikeyan, R., Palanikumar, K., Davim, J.P.: Optimization of electrical
discharge machining characteristics of WC/Co composites using non-dominated sorting
genetic algorithm (NSGA-II). The International Journal of Advanced Manufacturing
Technology 36(11-12), 1124–1132 (2007)
Kannan, S., Baskar, S., McCalley, J.D., Murugan, P.: Application of NSGA-II Algorithm to
Generation Expansion Planning. IEEE Transactions on Power Systems 24(1), 454–461
(2009)
Murugan, P., Kannan, S., Baskar, S.: NSGA-II algorithm for multi- objective generation
expansion planning problem. Electric Power Systems Research 79(4), 622–628 (2009)
Pongcharoen, P., Hicks, C., Braiden, P.M., Stewardson, D.J.: Determining optimum
Genetic Algorithm parameters for scheduling the manufacturing and assembly of
complex products. International Journal of Production Economics 78(3), 311–322
(2002)
Tran, K.D.: Elitist Non-Dominated Sorting GA-II (NSGA-II) as a Parameter-Less Multi-
Objective Genetic Algorithm. In: Proceedings of the IEEE SoutheastCon 2005,
pp. 359–367 (2005), doi:10.1109/SECON.2005.1423273
Zeng, F., Low, M., Decraene, J.: Self-adaptive mechanism for multi- objective evolutionary
algorithms. In: Proceedings of the International MultiConference of Engineers and
Computer Scientists, IMECS 2010, vol. I, pp. 7–12 (2010)
Transient Flow Analysis through Centrifugal
Pumps
Abstract. A theoretical study has been carried out on the transient flow through a
centrifugal pump during the starting period. The problem is governed by two
hyperbolic partial differential equations which the continuity and the motion equa-
tions. The mathematical model is solved numerically by using the method of cha-
racteristics with specified time intervals. The comparison between the obtained
numerical results and those obtained by experiment has shown a good concor-
dance. In this study, the effect of the starting time, the impeller diameter, the
number and the height of blades on the pressure increase were analyzed. The nu-
merical results have shown that the pressure increase is inversely proportional to
the starting time. However, during the starting period, these results have shown
that the number of blades has no significant effect on this pressure increase.
1 Introduction
2 Mathematical Formulations
dρ JJG
+ ρ divW = 0 (1)
dt
G
dW JJJJJG JJJG G G G G G G G G
ρ
dt
= grad p − p + divτ + ρ F − ρ ⎡ 2Ω ∧ W + Ω
⎣⎢
(
∧ oM + Ω ∧ Ω ∧ oM ⎤
⎦⎥
) (2)
G
Where ρ the density of the fluid is, τ is the stress tensor of viscosity, F is the
G G
external body force, W is the relative velocity and Ω is the rotational speed.
Figure 1 shows the control volume defined by two successive blades limited by
suction and discharge sides of the impeller. The principle of the study of turbo
machinery is based on the velocity diagram for determining the absolute velocity
at any point of the blade. In particular, point 2 which represent the outlet of the
G
impeller, the absolute velocity V2 is given by the following equation:
G G G
V2 = W2 + U 2 (3)
G
where U 2 the entrainment velocity at that point.
Transient Flow Analysis through Centrifugal Pumps 341
This relationship allows the passage from the rotating frame to the absolute
frame. The moving frame is used as reference for the flow study through the im-
peller. Taking into account the assumption of radial flow through the different
sections of thicknesses Δr of the impeller and the shear forces act in the radial
direction, the equations of motion (1) and (2) are simplified [Analogy to unidirec-
tional flow (Wylie et al, 1993; Thanapandi and Prasad, 1994)]:
∂H C 2 ∂Q
+ =0 (5)
∂t gA ∂r
1 ∂Q ∂H QQ
+g +λ − r Ω2 = 0 (6)
A ∂t ∂r 2 Dh A2
⎧ gA
⎪ dH + dQ + JAdt − r Ω Adt = 0
2
C+ ⎨ C (7)
⎪ dx = Cdt
⎩
⎧ gA
⎪−
− dH + dQ + JAdt − r Ω 2 Adt = 0
C ⎨ C (8)
⎪dx = −Cdt
⎩
where J = λQ Q 2 DA2 is the head loss by unit of the impeller passage length.
342 I. Chalghoum et al.
gAi −1
QPi − Qi −1 + ( H Pi − Hi −1 ) + Ai −1 Ji −1Δt − ri −1Ω2 Ai −1Δt = 0 (9)
C
gAi +1
QPi − Qi +1 − ( H Pi − H i +1 ) + Ai +1 J i +1Δt − ri +1Ω 2 Ai +1Δt = 0 (10)
C
1
H Pi = ⎡C1H i −1 + C 2 H i +1 + ( Qi −1 − Qi +1 ) − G ⎤⎦ (11)
C1 + C 2 ⎣
By substituting equation (11) in equation (10), one can obtain the discharge at
section i as follows:
These equations permit to determine QPi and H Pi at any interior point of the
pump. In order to get the solution until an elevated time level, it is necessary to in-
troduce the appropriate boundary conditions. These conditions will be introduced
in the next section.
Transient Flow Analysis through Centrifugal Pumps 343
To study the transient flow through centrifugal pumps, we consider the hydraulic
facility (TE-83), available at ENIS, represented by the simplified schematic (Fig-
ure 3). This installation has two identical centrifugal pumps permitting the study
of a single pump, two pumps in parallel and two pumps in series. The test facility
represented by figure 3 contains two digital pressure gages for measuring the pres-
sure at the suction side and the discharge side of the pump. The rotational speed of
the motors is controlled by potentiometers. A mechanical device is used to deter-
mine the engine torque.
Figure 4 illustrates the centrifugal pump impeller. It is composed of 8 blades
with a base suction diameter of 42.8 mm and a discharge diameter of 128 mm.
When the valve is placed between the upstream end and the downstream end of
the discharge pipe, the resolution by the method of characteristics requires that the
main discharge pipe must be treated as if it is formed by two conduits (figure.5).
344 I. Chalghoum et al.
Compatibility equations that solve this problem are: equation (9) for the pipe 1
and equation (10) for the pipe 2. These equations are written for the pipes 1 and 2
respectively as follows:
QP 2,1 − Q2,2 −
gA
C
( H P 2,1 − H 2,2 ) + AJ 2,2 Δt = 0 (14)
gA gA
where K 2 = Q1, N 1 + H1, N 1 − AJ1, N 1Δt , K 3 = Q2,2 − H 2,2 − AJ 2,2 Δt and
C C
K1 = gA C
r/R 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Cc 0.611 0.612 0.616 0.622 0.633 0.644 0.662 0.687 0.722 0.781
Transient Flow Analysis through Centrifugal Pumps 345
By substituting equations (15) and (16) in the sum of equations (11) and (12),
we obtain the following quadratic equations:
K1
QP21, N1+1 − 2QP1, N1+1 + ( K1 + K2 ) = 0 (19)
( Cd AG )
2
K1
QP21, N1+1 + 2QP1, N1+1 − ( K1 + K2 ) = 0 (20)
( Cd AG )
2
Equation (19) is used to determine the positive flow rate while equation (20) is
used to determine the negative flow rate.
r
G
W2 G
β 2 V2
Ω
α2
G
U2
5 2,5
pressure (m)
2 1,0
startup time 0.15 s - N = 1450 rpm
startup time 0.15 s/ N =1450 rpm
1 0,5
0 0,0
0,0 0,1 0,2 0,3 0,4 0,5 0,6 t (s) 0,7 0,0 0,1 0,2 0,3 0,4 0,5 0,6 t (s) 0,7
Fig. 8 the pressure difference versus time Fig. 9 Variation of flow rate versus time for
for different valve opening positions different valve opening positions
Figure 9 shows the evolution of the pressure difference between inlet and outlet
sides of the pump as a function of time and for the different valve opening posi-
tions. These curves are considered for a rapid starting period (≈ 0.15 s ) and a
rotational speed of 1450 rpm. This figure shows that the pressure is inversely pro-
portional to the valve opening percentage. Indeed, when the valve is fully open,
the pressure drops to 0.2 m. This pressure drop is due to the reduction of the flow
resistance caused by the friction at the suction and discharge circuits and mainly at
the valve. This figure also shows an oscillating effect at the pressure curve when
the valve is in the closed position. This phenomenon is due to the wave propaga-
tion along the pipes and its reflection at both ends of the hydraulic circuit. It can
be seen also in figure 9 the presence of peak of pressure from the mid-open posi-
tion up to complete valve opening. This peak is due to the pump start-up time and
to the mass of water which increases when the resistance in the circuit decreases
with the valve opening.
that when is increased the rotational speed, the pressure range and the flow rate
range is increased too. This result is logic since the centrifugal force which per-
mits the flow through the impeller is proportional to the square root of the rota-
tional speed. Figure 12 shows the effect of the diameter of the impeller on charac-
teristic curve of the pump. It can be noted that the head range of the characteristic
curve is proportional to the diameter of the impeller. This is due to the effect of
centrifugal force which increases with the radius. For a constant impeller diameter
and by changing the blades height, the flow rate range increases while the pressure
remains unchanged as shown in Figure 13.
6 20
pressure (m)
pressure (m)
startup time 0.15 s - N = 1450 rpm startup time 0.15 s
5 16 N = 2900 rpm
4
12
3 Pump TE-83
Pump TE-83
8
2 theoretical model
theoretical m odel
4
1
0 0
Q(l/s) Q(l/s)
0,00 0,25 0,50 0,75 1,00 1,25 1,50 1,75 0,00 0,50 1,00 1,50 2,00 2,50 3,00 3,50
Fig. 10 Comparison of theoretical and experimental curves for N = 1450 rpm and N = 2900
rpm
10 6
pressure (m)
pressure(m)
s tartup time 0.15 s - N =1450 rpm s tartup time 0.15 s - N =1450 rpm
8
R = 84 mm
R = 74 mm 4
6 h = 5 mm
R = 64 mm h = 7 mm
4
2
2
0 0
0 0,4 0,8 1,2 1,6 Q (l/s ) 2 0 1 2 3 4 Q (l/s) 5
Fig. 11 Effect of the impeller diameter on Fig. 12 Influence of the blades height on the
the pump characteristics curve pump characteristics curve
3 4
pressure (m)
pressure (m)
L2 = 1.5 m - N =1450 rpm startup time 0.15 s - N =1450 rpm
0 0
0,0 0,1 0,2 0,3 0,4 0,5 0,6 t (s ) 0,7 0,0 0,1 0,2 0,3 0,4 0,5 0,6 t (s) 0,7
Fig. 13 Effect of starting time on the Fig. 14 Effect of the impeller diameter on
pressure evolutions the pressure curve
4 Conclusion
In this paper, the transient flow inside a centrifugal pump has been conducted. The
transient is created as a result of a linear variation of pump speed from rest to the
permanent regime. This type of flow is governed by two partial differential equa-
tions: the momentum and the continuity equations. These equations are solved
numerically by the use of the method of characteristics. The pump impeller was
modelled as conduits connected in parallel. The obtained results have shown a
good concordance with those obtained by experiment. In this study the effect of,
the starting time, the geometric proprieties of the impeller and the valve openings,
on the pressure evolution has been also analysed. These results have shown that
the pressure increase during the starting period was influenced by the starting time
and the valve openings. However, these parameters have no influence on the
steady state pump characteristic curve. These results have shown that this curve is
well influenced by the radius and the blade height of the impeller.
References
Tsukamoto, H., Ohashi, H.: Transient characteristics of a centrifugal pump during starting
period. ASME Journal of Fluids Engineering 104, 6–13 (1982)
Dazin, A., Caignaert, G., Bois, G.: Transient behavior of turbomachineries: application to
radial flow pumps startups. ASME Journal of Fluids Engineering 129, 1436–1444
(2007), doi:10.1115/1.2776963
Saito, S.: The transient characteristics of a pump during startup. Bulletin of JSME 25,
372–379 (1986)
Thanapandi, P., Prasad, R.: Centrifugal pump transient characteristics and analysis using
the method of characteristics. Int. J. Mech. Sci. 37, 77–89 (1995)
Streeter, V.L., Wylie, E.B.: Hydraulic Transients. McGraw-Hill Book Company, New York
(1967)
Vibro-Acoustic Analysis of Laminated
Double-Wall: Finite Element Formulation
and Reduced-Order Model
Abstract. This paper presents a finite element model for sound transmission anal-
ysis through a double sandwich panels with viscoelastic core inserted in an infinite
baffle. The proposed model is derived from a multi-field variational principle in-
volving structural displacement of the panels and acoustic pressure inside the fluid
cavity. To solve the vibro-acoustic problem, the plate displacements are expanded as
a modal summation of the plate’s real eigenfunctions in vacuo. Similarly, the cavity
pressure is expanded as a summation over the modes of the cavity with rigid bound-
aries. Then, an appropriate reduced-order model with mode acceleration method
by adding quasi-static corrections is introduced. The structure is excited by a plane
wave. The radiated sound power is calculated by means of a discrete solution of
the Rayleigh Integral. Fluid loading is neglected. Various results are presented in
order to validate and illustrate the efficiency of the proposed reduced finite element
formulation.
1 Introduction
Double-wall structures are widely used in noise control due to their superiority over
single-leaf structures in providing better acoustic insulation. Typical examples in-
clude double glazed windows, fuselage of airplanes and vehicles, etc. Various the-
oretical, experimental and numerical approaches have been investigated to predict
the sound transmission through double walls [1, 2, 3].
By introducing a thin viscoelastic interlayer within the panels, a better acous-
tic insulation is obtained. In fact, sandwich structures with viscoelastic layer are
commonly used in many systems for vibration damping and noise control. In such
structures, the main energy loss mechanism is due to the transverse shear of the
viscoelastic core [4, 5].
ΩF
Γt
Fd Ωs2
Ωs1
u d
Γu
Sandwich panel with
ud viscoelastic core
inviscid fluid occupying the domain ΩF . The cavity walls are rigid except those in
contact with the flexible wall structures noted Σ.
The harmonic local equations of this structural-acoustic coupled problem can be
written in terms of structure displacement u and fluid pressure field p [6]
divσ(u) + ρS ω 2 u = 0 in ΩS (1)
σ(u)nS = Fd on Γt (2)
σ(u)nS = pn on Σ (3)
u=u d
on Γu (4)
2
ω
Δp + p=0 in ΩF (5)
c2F
∇p · n = ρF ω 2 u · n on Σ (6)
where ω is the angular frequency, nS and n are the external unit normal to ΩS and
ΩF ; ρS and ρF are the structure and fluid mass densities; cF is the speed of sound
in the fluid; and σ is the structure stress tensor.
σ = C∗ (ω)ε (7)
where ε denotes the strain tensor and C∗ (ω) is the complex elasticity tensor. It is
generally complex and frequency dependent (∗ denotes complex quantities). It can
be written as:
C∗ (ω) = C (ω) + iC (ω) (8)
√
where i = −1.
Furthermore, for simplicity, a linear, homogeneous, viscoelastic constitutive
equation will be used in this work. In the isotropic case, the viscoelastic material
is defined by a complex and frequency dependence shear modulus in the form:
G∗ (ω) = G (ω) + iG (ω) (9)
where G (ω) is know as shear storage modulus, as it is related to storing energy and
G (ω) is the shear loss modulus, which represents the energy dissipation effects.
352 W. Larbi, J.-F. Deü, and R. Ohayon
With these assumptions, the stress tensor of the sandwich structure is complex and
frequency dependent.
where U and P are the vectors of nodal values of u and p respectively; Mu and
K∗u (ω) are the mass and stiffness matrices of the structure; Mp and Kp are the mass
and stiffness matrices of the fluid; Cup is the fluid-structure coupled matrix; F is
the applied mechanical force vector. Note that since the elasticity modulus of the
viscoelastic core of the sandwich panels is complex and frequency dependent, the
stiffness matrix K∗u (ω) is also complex and frequency dependent.
Due to the frequency dependent of the stiffness matrix, this eigenvalue problem is
complex and nonlinear. It is assumed that vibrations of the damped structure can be
represented in terms of the real modes of the associated undamped system if appro-
priate damping terms are inserted into the uncoupled modal equations of motion.
Thus, the complex stiffness matrix is decomposed in the sum of two matrices:
where Ku0 = Re[K∗u (0)] is the real and frequency-independent stiffness matrix
calculated with a constant Young module’s of the viscoelastic core and δK∗u (ω) is
the residual stiffness matrix.
The ith real eigenmode is obtained from the following equation
2
Ku0 − ωsi Mu Φsi = 0 for i ∈ {1, · · · , Ns } (13)
where (ωsi , Φsi ) are the natural frequency and eigenvector for the ith structural mode.
Vibro-Acoustic Analysis of Laminated Double-Wall 353
where (ωf i , Φf i ) are the natural frequency and eigenvector for the ith acoustic
mode.
It is important
to note that the physical acoustic modes in a rigid fixed cavity
are such that ΩF p dv = 0 which excludes the ωf i = 0, p = constant solution
which is not physical but had to be introduced in the formulation of the coupled
problem. Thus, the variational formulation of this kind of system, in order to be
regularized for zero frequency situation, i.e. valid for a static problem,
has to be
modified by adding the following constraint ρF c2F Σ u.n ds + ΩF p dv = 0 (see
[6] for details). When doing this, on one hand the static pressure is defined precisely
by
ρF c2F
p =−
s
u.n ds (15)
|ΩF | Σ
and on the other hand, the reduced order formulation will be carried only by projec-
tion on the physical acoustic modes.
Substituting these relations into Eq. (10) and pre-multiplying the first row by ΦTs
and the second one by ΦTf , we obtain
• Ns mechanical equations
Ns
Nf
2 ∗ 2
−ω qsi + γik (ω)qsi + ωsi qsi − βij qf j = Fi (17)
k=1 j=1
354 W. Larbi, J.-F. Deü, and R. Ohayon
• Nf acoustic equations
Ns
−ω 2 qf i + ωf2 i qf i − ω 2 βij qsj = 0 (18)
j=1
where Fi (t) = ΦTsi F is the mechanical excitation of the ith mode; βij = ΦTsi Cup Φf j
∗
is the fluid structure coupling coefficient and γik (ω) = ΦTsi δK∗u (ω)Φsk the reduced
residual stiffness complex coefficient.
At each frequency step, the reduced system (Eqs. (17) and (18)) is solved by
∗
updating γik (ω). After determining the complex amplitude vectors qsi and qf i ,
the displacement and pressure fields are reconstructed using the modal expansion
(Eqs. (16)).
L
F= αi (t)F0i (19)
i=1
where F0i is the invariant spatial portion and αj (t) is the time varying portion. For
each invariant spatial load, the static modal eigenvector Ψsi is given by:
Ψsi = K−1
u0 F0i (20)
The truncated basis containing the real and undamped structure modes is enriched
by the static modal eigenvectors such that
The truncated fluid basis is enriched with the static pressure Ps computed from
Eq. (15):
Φ̄f = [Φf Ps ] (22)
Thus, the displacement and pressure are sought as
where the vectors q0s and qf0 are the quasi-static modal amplitudes of the structure
displacement and the fluid pressure respectively. Similar coupled differential equa-
tions than Eqs. (17) and (18) can of course be obtained after modal projection on
the enriched bases Φ̄s and Φ̄f .
Vibro-Acoustic Analysis of Laminated Double-Wall 355
4 Numerical Examples
4.1 Sound Transmission through an Elastic Double-Panel System
In this section, the validation of the proposed coupled finite elements formulation for
sound radiation is presented. The problem under consideration is shown in Fig. 2. A
normal incidence plane wave excites a double-plate system filled with air (density
ρF = 1.21 kg/m3 and speed of sound cF = 340 m/s). The plane wave has a pressure
amplitude of 1 N/m2 and is applied to plate 1 as the only external force to the system.
The plates are identical and simply supported with thicknesses of 1 mm. The density
of the plates is 2814 kg/m3 , the Youngs modulus is 71 GPa, the Loss factor is 0.01
and Poisson ratio 0.33. The surrounding fluid is the air. This example was originally
proposed by Panneton in [2].
Rigid baffle
Air Cav
ity
b=0.22 m
Plate 2
Normal plane
wave Plate 1 m
4
76
a=0.35 m 0.0
c=
When the excitation is applied to the first plate, the second one vibrates and
radiates sound caused by the coupling of air and plate 1. The normal incidence sound
transmission loss is then computed using the Rayleigh’s integral method [1] which
needs the finite element solution of surface velocities of plate 2. For this purpose,
the resolution of the coupled system is done with a modal reduction approach using
the first 10 in vacuo structural modes and the first 10 acoustic modes of the fluid in
rigid cavity. The truncated bases are enriched by the static modal eigenvectors.
Fig. 3 shows the normal incidence transmission loss through a simply supported
plate (dashed line). Due to the modal behavior of the plate, dips in the transmission
loss curve are observed at its resonance frequencies (modes (1, 1), (3, 1) and (1,
3)). When a second plate is used to form an airtight cavity (continuous line), an
increase in the transmission loss is achieved except in the region of the so-called
plate-cavity-plate resonance (mode (1, 1)*). At this frequency, the two plates move
356 W. Larbi, J.-F. Deü, and R. Ohayon
80
Air−filled double wall system
70 Simple panel
60
50
nSTL (dB)
40
30
20
10
mode (1, 3)
mode (1, 1) mode (3, 1)
0
mode (1, 1)*
1 2 3
10 10 10
Frequency (Hz)
Fig. 3 Comparison of the normal incidence sound transmission (nSTL) through an air-filled
double panel and a simple panel
out of phase with each other and the effect of the cavity on the plates is mostly
one of added stiffness. This frequency is similar to the mass-air-mass resonance of
unbounded double panels.
In addition, the variation of the nSTL of an air-filled panels and a simple panel is
in very good agreement with the published date from [2].
100
−12
Radiated Sound Power (dB) − ref 10
80
60
40
20
Fig. 4 Comparison of radiated sound power from a simple glass pane and a laminated glass
with the same mass
A comparison between a simple glass and a laminated glass with PVB inter-
layer with an equivalent surface mass is shown in Fig. 4. Calculation was limited
to 2000 Hz maximum. This comparison shows that laminated glass has a much
lower acoustic radiation compared to conventional glass at resonance frequencies
du to the effect of the viscoelastic layer. The reduction of sound radiation power
is around 10 dB in lower frequencies and around 20 dB in higher frequencies. In
fact, at low frequencies, the viscoelastic material is soft and the damping is small.
At higher frequencies, the stiffness decreases rapidly and the damping is highest.
Moreover, flexural vibrations causes shear strain in the viscoelastic core which dis-
sipates energy and reduces vibration and noise radiation. Note that the thickness of
the viscoelastic layer has a significant influence in terms of attenuation.
120
Modal reduction
Direct method
100
80
nSTL (dB)
60
40
20
0
1 2 3
10 10 10
Frequency (Hz)
Fig. 5 nSTL through an air-filled double panel: comparison between the modal reduction
approach and the direct nodal method
358 W. Larbi, J.-F. Deü, and R. Ohayon
Fig. 5 shows a comparison between the nSTL of the coupled problem, obtained
with the proposed accelerated modal reduction approach with a truncation on the
first twenty structural modes (Ns = 20) and first twenty acoustic modes (Nf = 20)
and the direct nodal method (Eq. (10)) where the displacement and pressure vectors
are calculated for each frequency step. The structural modes are calculated from
Eq. (13) using the constant shear storage modulus G∞ . As can be seen, a very good
agreement between the two methods is proved. In this respect, it should be noted
that the resulting reduction of the model size and the computational effort using the
reduced order method are very significant compared to those of the direct approach.
5 Conclusions
In this paper, a finite element formulation for sound transmission through double
wall sandwich panels with viscoelastic core is presented. A reduced-order model,
based on a normal mode expansion, is then developed. The proposed methodology
requires the computation of the eigenmodes of the undamped structure, and the
rigid acoustic cavity. Quasi-static corrections are introduced in order to accelerate
the convergence. Despite its reduced size, this model is proved to be very efficient
for simulations of steady-state analyses of structural-acoustic coupled systems with
viscoelastic interlayers when appropriate damping terms are inserted into the modal
equations of motion.
References
1. Fahy, F.: Sound and structural vibration, 1st edn. Academic Press, New York (1985)
2. Panneton, R.: Modélisation numérique tridimensionnelle par éléments finis des milieux
poroélastiques: application au problème couplélasto-poro-acoustique. Ph.D., Université
de Sherbrooke (1996)
3. Quirt, J.D.: Sound transmission through windows I. Single and double glazing. Journal of
the Acoustical Society of America 72(3), 834–844 (1982)
4. Akrout, A., Hammami, L., Ben Tahar, M., Haddar, M.: Vibro-acoustic behaviour of lam-
inated double glazing enclosing a viscothermal fluid cavity. Applied Acoustics 70(1),
82–96 (2009)
5. Basten, T.G.H.: Noise reduction by viscothermal acousto-elastic interaction in double wall
panels. PhD-thesis, University of Twente, Enschede, The Netherlands (2001)
6. Morand, H.J.-P., Ohayon, R.: Fluid-structure interaction. John Wiley & Sons, New York
(1995)
7. Larbi, W., Deü, J.-F., Ohayon, R.: Finite element formulation of smart piezoelectric com-
posite plates coupled with acoustic fluid. Composite Structures 94(2), 501–509 (2012)
8. Havriliak, S., Negami, S.: A complex plane analysis of?-dispersions in some polymer
systems. Journal of Polymer Science Part C: Polymer Symposia 14(1), 99–117 (1966)
Time Step Size Effect on the Hydrodynamic
Structure around a Water Darrieus Rotor
Ibrahim Mabrouki, Ahmed Samet, Zied Driss, and Mohamed Salah Abid
Abstract. Turbulent flows are encountered in many hydraulic and water resources
engineering problems. Their understanding is thus a critical prerequisite for de-
signing stream and river restoration projects and a broad range of hydrodynamic
structures. For this purpose, an unsteady Reynolds averaged Navier-Stokes
(URANS) equations with a two-equations turbulence closure model is employed.
The present paper aims to numerically explore the three-dimensional unsteady
flow over a conventional Darrieus type rotor and to study the time step size effect
on the hydrodynamic structure.
1 Introduction
Darrieus turbine was developed in France by Georges Darrieus in the 1920's to
generate power from wind. It consists of a set of long, rectangular airfoils con-
nected to a central rotating shaft. The airfoils may be curved to directly connect to
the shaft or be straight and held parallel to the shaft by struts, arms or discs. These
blades are there by oriented transversely to the fluid flow and parallel to the axis
of rotation. Significant testing in the 1980's and 1990's demonstrated the utility of
this turbine design. Several different models were successfully tested in the la-
boratory and various waterways. This kind of turbine considered for diffrent ap-
plications, such as tidal power, run-of-the-river hydroelectricity, and wave energy
conversion (Shiono et al. (2000), Prabhu et al. (2013) and Maitre et al. (2013)).
Hwang et al. (2009) studied the effect of number Blade on the performance of
Darrieus Water Turbine. It’s one of the important parameter of Darrieus water
turbine. These authors studied the influence of chord length of Darrieus water tur-
bine. They show the evolution of maximum power coefficient and the TSR for a
fixed number of blades. Attempts to improve the performance of H-Blade Dar-
rieus rotor by using a duct placed outside the rotor have been documented. Shimo-
kawa et al. (2010) measured the influence of the stream section with and without
nozzle in the distribution of velocity vectors. In addition, Matsushita et al. (2008)
studied the effect of draft tube on the efficiency in the case with narrow intake Sin
/D=0.8. Malipeddi et al. (2012) studied the effect of changing position of the tur-
bine in the duct by performing simulation with the turbine at various locations in
the duct, for a tip-speed ratio of 2. Rao et al. (2010) studied the performance of
Darrieus rotor equipped with a symmetric two-foiled channeling device (Diffuser
Annular Ring). They changed the channeling profile and, demonstrate their effect
in the performance of the turbine. A ducted Darrieus turbine system as appropriate
one for extra-low head hydropower has been proposed and developed by different
authors. The design guidelines of Darrieus turbine and the characteristics have
been clarified to obtain higher performance As the results of still continued inves-
tigation to simplify the structure for making cost effectiveness higher, it was rec-
ognized for small sized runner that there was no need to install a draft tube and the
sidewalls of runner casing section in the case of runner with inlet nozzle. A nu-
merical investigation was developed in the unsteady state using computational flu-
id dynamic (CFD) in order to characterize the hydrodynamic comportment of the
flow around the rotor. We studied also the effect of time step size on the hydrody-
namics characteristics. The used CFD code "Fluent" is based on the Unsteady
Navier-Stokes equations by a finite volume discretization model. The numerical
approach used is the sliding mesh (SM) model. In these condition, the test section
equipped with 2 parts stationary domain and moving domain in which we have
impose a rotational speed equal to ω = 25.61 rad.s-1. The turbulent flow is defined
by the Reynolds number Re=3.8 106.
2 Geometrical System
The test section under investigation in this work is composed of a duct equipped
with water Darrieus rotor. The duct have a prismatic shape with a height equal to
b=0.2 m, a width equal to s=0.526 m, and a length equal to L=2.886 m .The Dar-
rieus turbine, is composed by three straight blades and is placed inside the duct.
The blades have profiles belonging to the symmetrical NACA 0018 four digit with
a maximal thickness equal to 18% of the chord, a height equal to b=0.2 m and a
diameter equal to D=0.39 m.
The flow over Darrieus rotor is periodic and a proper time step selection is im-
portant in order to ensure a good results. The time step corresponds to one degree
rotation of the rotor and is expressed as:
Ts = / (180 ω) (1)
Where:
ω: Vitesse angulaire (rad/s)
Time Step Size Effect on the Hydrodynamic Structure 361
In this study the used time steps correspond to 1.25° and 2.5° rotation and they
are shown in Table 1.
3 Numerical Model
In the present work, the commercial CFD code Fluent is used to perform 3D simu-
lations, solving the incompressible Unsteady Reynolds-Averaged Navier-Stokes
(U-RANS) equations discretized by means of a finite volume approach. The im-
plicit segregated version of the solver is employed. The pressure-velocity coupling
is achieved by means of the SIMPLE algorithm. The convective terms are discre-
tized using a second order accurate upwind scheme, and pressure and viscous
terms are discretized by a second-order-accurate centered scheme. A first order
implicit time formulation is also used. Turbulence closure is provided by shear
stress transport k-ω turbulence model. The model has been validated, by compar-
ing numerical results with available experimental data (Matsushita et al. 2008).
4 Numerical Results
In this study, we are interested to visualize the velocity field, the static pressure,
the dynamic pressure, and the turbulent kinetic energy. Indeed, the variation of
each result as function of time in different points located in the downstream of the
rotor (figure 4).
points x (m) y (m) z (m)
Point 1 0.22 0 0
rotor and the blades walls. In the downstream of the rotor, the velocity value keeps
decreasing until the out of the test section. Then, a sharp increase has been noted
through the divergent angle where the velocity values are recorded in the lateral
walls of the duct. Globally, similar results are obtained with the different time step
sizes. However, the difference is localized in the duct extremity and in the blades.
Figure 7 shows the evolution of the static pressure in three different points (1),
(2) and (3) localized in the downstream of the rotor for a period of 2 seconds.
Overall, it has been noted a significant fluctuations of the static pressure curves.
Particularly, it has been observed that the evolution of the static pressure is less
important in point (1) than in the points (2) and (3). This fact due to the position of
this point located in the middle of the duct. For the points (2) and (3) located in the
lateral faces of duct, it has been noted a little difference between the results due to
the effect of time steps sizes.
Figure 9 shows the evolution of the dynamic pressure indifferent in three points
(1), (2) and (3) located in the downstream of the Darrieus rotor for a period of 2 s.
Overall, it has been noted a significant fluctuations of the dynamic pressure
curves. Indeed, it’s clear that the fluctuating of the dynamic pressure is less impor-
tant in point (1) than the point (2) and (3). This fact is due to the position of this
point located in the middle of the duct. Indeed, it has been noted a difference be-
tween the results due to the time steps effect.
Figure 11 shows the evolution of the turbulent kinetic energy in different three
points (1), (2) and (3) located in the downstream of the turbine for a period of 2 s.
Overall, it has been noted a significant fluctuations of the turbulent kinetic energy
curves. According to these results, it has been noted that the turbulent kinetic
energy fluctuation is less important in point (1) than the point (2) and (3). This fact
is due to the position of this point located in the middle of the duct. Indeed, it has
been noted a difference between the two results, due to the effect of the time steps.
6 Conclusion
In this paper, we have studied the numerical parameters effect in the unsteady
state. Particularly, we have studied the time step sizes effect and we concluded
that this parameter has an influence in the rotation degree of the water Darrius ro-
tor. Then, we have concluded that the sliding mesh (SM) method can be used to
model the water turbine characteristics. This study allows us to find the adequate
numerical parameters enabling more detailed analyses of the flow around the Dar-
rieus rotor.
References
Shiono, M.I., Katsuyuki, S.U., Seiji, K.I.: An Experimental Study of the Characteristics of a
Darrieus Turbine for Tidal Power Generation. Electrical Engineering in Japan 132(3)
(2000)
Prabhu, S.V., Vimal, P.A., Himanshu, C.H.: Performance Prediction of H-Type Darrieus
Turbine by Single Stream Tube Model for Hydro Dynamic Application. International
Journal of Engineering Research & Technology 2(3) (2013)
Maitre, T., Amet, E., Pellone, C.: Modeling of flow in Darrieus water turbine: wall grid re-
finement analysis and comparison with experiments. Renewable Energy 51, 497–512
(2013)
Huang, S.W., Tsai, Y.D., Liang, S.Y., Hsieh, C.H., Chen, S.J.: Wind-tunnel study on aero-
dynamic performance of small vertical-axis wind turbines (2009)
368 I. Mabrouki et al.
Shimokawa, K.A., Akinori, F.U., Kusuo, O.K., Daisuke, M.A.: Satoshi WA Side-wall ef-
fect of runner casing on performance of Darrieus-type hydro turbine with inlet nozzle
for extra-low head utilization. Technological Sciences 53, 93–99 (2010)
Matsushita, D.A., Okuma, K.U., Watanabe, S.A., Furukawa, A.K.: Simplified structure of
ducted Darrieus-type hydro turbine with narrow intake for extra-low head hydropower
utilization. Journal of Fluid Science and Technology 3, 387–397 (2008)
Malipeddi, A.R., Chatterjee, D.: Influence of duct geometry on the performance of Darrieus
hydro turbine. Renewable Energy 43, 292–300 (2012)
Roa, A.M., Aumelas, V., Maître, T., Pellone, C.: Numerical and experimental analysis of a
Darrieus-type cross flow water turbine in bare and shrouded configurations. In: 25th
IAHR Symposium on Hydraulic Machinery and System (2010)
Simulation of the Dynamic Behavior
of Multi-stage Geared Systems with Tooth
Shape Deviations and External Excitations
1
INSA de Lyon, LaMCoS, France
2
Hispano-Suiza, France
Philippe.Velex@insa-lyon.fr
Abstract. In this paper, a torsional dynamic model of multi-stage idler spur and
helical gears is presented which combines time-varying internal and external exci-
tations such as time-varying external torques. Each contact line in the various base
planes is discretized in elemental cells which are all attributed a time-varying
mesh stiffness element and an initial separation to account for tooth shape devia-
tions from ideal involute flanks. The corresponding non-linear differential system
is solved by combining a Newmark’s numerical scheme and a normal contact al-
gorithm. A number of simulation results are presented on the influence of the
combined effect of errors and shape deviations along with external excitation
sources on dynamic tooth loads.
Keywords: Multi-stage idler gears, Spur and helical gears, Time-varying torques,
Tooth shape deviations.
1 Introduction
In certain conditions, power circulation can influence vibrations as is the case for
the dynamic behavior of multi-stage geared systems with the various possible
combinations of resisting and driving torque positions. In this context, Kuria and
Kihiu (2012) developed a numerical model to study the effect of three design vari-
able on dynamic stress of multi-stage spur gear. Al-Shyyab and Karahman (2005)
used a Harmonic Balance Method in conjunction with continuation technique to
describe the periodical responses of multi mesh gear system. Velex and Saada
(1991), Velex and Raclot (1999), Kubur and al. (2004) proposed FE models of
double stage drives and investigated the influence of key design parameters on the
system dynamics. Carbonelli (2014) used a Finite Difference approach accounting
for both internal and external sources in the time domain such as clearance nonli-
*
Corresponding author.
The normal deflection at any point of contact M mij (m is the stage number), is
equivalent to the normal approach δ mij with respect to rigid-body positions minus
the initial separation δ emij possibly induced by tooth shape modifications:
In what follows, k mij represents the discrete stiffness element associated with
each discrete cell on the contact lines which depends on load and shape deviations
(Velex and Maatar 1996) and which is expressed as:
⎧ H = 0 if Δ
⎪ mij < 0
k mij = kˆmij H ( Δ mij ) with ⎨ (2)
⎪ H =1 if Δ mij > 0
⎩
where H ( Δ mij ) is the Heaviside function used to simulate total or partial contact
losses.
For a pinion-gear pair with one torsional degree of freedom each, the mesh stiff-
ness matrix can be expressed as:
JJJG JJJJGT
⎡ K (t )⎤ = ∑ k (3)
⎣ m ⎦ mij Vm Vm
mij
JJJG
where Vm = ζ Mm Rbm cos βb ζ Mm Rbm +1 cos βb is a structural vector which accounts for
gear geometry.
Simulation of the Dynamic Behavior of a Multi-stage Geared Systems 371
Extending the methodology to a system with one pinion, one idler gear and one
output member, the global stiffness matrix reads:
⎡ ⎤
⎢ k1 ( t ) Rb1 k1 ( t ) Rb1 Rb 2
2 0 0 0 ⎥
⎢ ⎥
⎢k ( t ) R R
⎢1 b1 b 2 ( )
k1 ( t ) + k2 ( t ) Rb22 k2 ( t ) Rb 2 Rb 3 0 0 ⎥
⎥
⎣ ⎦ ( )
⎡ K ( t )⎤ = cos 2 β ⎢
b ⎢ 0 k2 ( t ) Rb 2 Rb 3 ( )
k2 ( t ) + k3 ( t ) Rb 3 "
2 0
⎥
⎥
(4)
⎢ ⎥
⎢ 0 0 # ⎥
⎢ ⎥
⎢⎣ 0 0 k N −1 ( t ) RbN + k h ⎥⎦
2
[ M ] {q} + [C ] {q} + ⎡⎣K ( t ,q )⎤⎦ {q} = {F0( t )} + {F1( t ,q,δ em )} + {F2 ( t )} (5)
where [M] and [K(t,q)] are the global mass and stiffness matrix, the damping ma-
trix is expressed using a Rayleigh model such that [C]=a.[M]+b.[Kav] (Velex
1988) , {F0 ( t )} = Ct1 ( t ) Ct2 ( t ) ... Ct N ( t ) is the external torque vector,
{F1 (t , q ,δ em )} embodies the contributions of shape deviations and errors and its com-
ponents are:
⎧ k1 ( t )δ e1 ( M )ζ M 1Rb1 cos βb ⎫
⎪ ⎪
⎪
⎪ ( )
k1 ( t )δ e1 ( M )ζ M 1 + k2 ( t )δ e2 ( M )ζ M 2 Rb2 cos βb ⎪
⎪
⎪ ⎪
( )
{F1(t ,q,δ em )} ⎪ k2 ( t )δ e2 ( M )ζ M 2 + k3 ( t )δ e3 ( M )ζ M 3 Rb3 cos βb ⎪
= ⎨ ⎬ (6)
⎪ # ⎪
⎪ ⎪
( )
⎪ k N − 2 ( t )δ eN − 2 ( M )ζ M N − 2 + k N −1 ( t )δ eN −1 ( M )ζ M N −1 RbN −1 cos βb ⎪
⎪ ⎪
⎪ k N −1 ( t )δ eN −1 ( M )ζ M N −1RbN cos βb ⎪
⎩ ⎭
372 H. Fakhfakh et al.
Ω1(t)
O1
e1(t): excitation due to geometrical errors
K1: excitation due to
mesh stiffness
c1 K2
O3 ∆l2
c2
O2 e2(t)
kh
∆l1: relative phase shift ON
3 Dynamic Results
The numerical applications were conducted on the 4-stage idler spur gear shown
in Figure 2 where the system is represented with the pinion centres all aligned
such that γ1 = γ2 = γ3 = 180°. The corresponding gear data are listed in Table 1.
The critical speeds of the four stages appear simultaneously on the four curves
which proves the significant inter-stages coupling so that each pinion is excited by
all the mesh stiffness functions. The response curves for a constant load (figure
3.a) exhibit peaks at 775 rpm ≌ (W2/50)/3, 1075 rpm ≌ (W2/50)/2 and 1950 rpm
corresponding to a natural frequency being excited by the mesh frequency and its
harmonics. The introduction of a time-varying torque of frequency 20Ω1 gene-
rates mostly an additional response peak at 5650 ≌ W2/20 rpm. The system also
exhibits contact losses between the teeth at some critical speeds on some of the re-
duction stages. The introduction of short profile corrections (linear relief of depth
Pf=60μm at tooth tips and extents corresponding to 20% of the active profile) on
all the pinions, leads to a substantial improvement of the dynamic behaviour be-
low the main critical speed (≌775 rpm) as shown in figure 3.c.
374 H. Fakhfakh et al.
(a)
(b)
(c)
Fig. 3 Influence of (a) constant torque and (b) time-varying torque (c) profile corrections on
dynamic mesh forces
Simulation of the Dynamic Behavior of a Multi-stage Geared Systems 375
(a)
(b)
Fig. 4 Influence of (a) constant torque and (b) time-varying torque (c) profile corrections on
dynamic mesh forces
376 H. Fakhfakh et al.
(c)
Fig. 4 (continued)
The results in figure 4 show that although the critical rotational speeds are the
same as in the previous case, the different power circulation modifies the dynamic
amplifications. The introduction of the resisting torque on pinion 3 unloads the
meshes downstream (between pinions 3, 4 and 5) over the whole speed range. When
the periodic component is superimposed (Ct3=-500+200.sin (20Ω1t)), an additional
tooth critical speed emerges for all the meshes amplifying contact losses especially
on stage 3. The response spectra (not shown in this paper) exhibit significant mod-
ulation side-bands related to the torque frequency 20Ω1. The response curves for
short profile corrections shown in figure 4.c illustrates tooth forces amplitude and
contact losses decrease below the main critical speed (≌775 rpm).
4 Conclusion
In this paper, a simplified torsional dynamic model of N stage idler spur and heli-
cal gears has been presented. The time-varying mesh stiffness functions attached
to every individual reduction stage are estimated from the classic formulae of We-
ber & Banaschek and their relative phases are determined analytically. The model
can simulate the influence of periodic torques and power circulations (several out-
puts, time-varying resisting loads, etc.) on tooth load amplification and critical
speeds. It has been shown that short profile corrections can improve the dynamic
behavior for a range of rotational speeds. Research is under way to incorporate
three-dimensional degrees of freedom, shafts, bearings and casings along with a
variety of loading conditions.
Acknowledgements. The authors would like to thank the support of Hispano-Suiza from
the Safran Group.
Simulation of the Dynamic Behavior of a Multi-stage Geared Systems 377
Nomenclature
Ωi ,Ω = angular velocity, acceleration of pinion i
i
Rbi = base radius of pinion i
Ji = polar moments of inertia of pinion i
βb = base helix angle
ζ Mm = sign of the driving torque of stage m
2
kh = torsional stiffness added to node N ( k h = 10 Nm / rad )
c0 , cs = average, oscillation amplitude of the periodic torque
a ,b = constants that can be adjusted to get the desired damping factor ξ .
−5
In this case a = 1, 12 ; b = 1, 2.10
⎡K ⎤ = average stiffness matrix
⎣ av ⎦
References
Kuria, J., Kihiu, J.: Effect of Gear Design Variables on the Dynamic Stress of Multi stage
Gears. Innovative Systems Design and Engineering (2012) ISSN 2222-1727
Al-Shyyab, A., Karahman, A.: Non-linear dynamic analysis of a multi-mesh gear train us-
ing multi-term harmonic balance method: sub-harmonic motions. J. of Sound and Vib.
(2005), doi:10.1016/j.jsv.2003.11.029
Velex, P., Saada, A.: A model for the dynamic behavior of multi-stage geared system. In:
Proc. of the 8th World Congress on Theory of Machines and Mechanisms, Prague,
vol. 2, pp. 621–624 (1991)
Raclot, J.-P., Velex, P.: Simulation of the dynamic behavior of single and multi-stage
geared systems with shape deviations and mounting errors by using a spectral method. J.
of Sound and Vib. 220(5), 861–903 (1999)
Kubur, M., Kahraman, A., Zini, D., Kienzle, K.: Dynamic analysis of a multi-shaft helical
gear transmission by finite elements – Model and experiments. J. Vib. Acous. 126(3),
398–406 (2004)
Carbonelli, A.: Hammering noise modelling-Nonlinera dynamic of a multi-stage gear train.
In: International Gear Conference, Lyon, pp. 447–456 (2014)
Yang, J.: Vibration analysis on multi-mesh gear-trains under combined deterministic and
random excitations. Mechanism and Machine Theory 59, 20–33 (2013),
doi:10.1016/j.mechmachtheory.2012.08.005
Velex, P., Maatar, M.: A Mathematical Model for Analysing the Influence of Shape Devia-
tions and Mounting Errors on Gear Dynamic Behaviour. J. of Sound and Vibra-
tion 191(5), 629–660 (1996), doi:10.1006/jsvi.1996.0148
Kubo, A.: Stress condition, vibrational exciting force, and contact pattern of helical gears
with manufacturing and alignment error. J. Mech. Des. 100(1), 77–84 (1978),
doi:10.1115/1.3453898
378 H. Fakhfakh et al.
Gregory, R.W., Harris, S.L., Munro, R.G.: Dynamic Behavior of Spur Gears. Proc. of the In-
stitution of Mechanical Engineers 178(1), 207–218 (1963),
doi:10.1177/002034836317800130
Velex, P.: Contribution à l’analyse du comportement dynamique de réducteurs à engrenag-
es à axes parallèles. Thèse de Doctorat, INSA de Lyon.N°88 ISAL 0032 (1988)
Fuzzy Modeling and Control of Centrifugal
Compressor Used in Gas Pipelines Systems
1
Applied Automation and Industrial Diagnostic Laboratory,
University of Djelfa 17000 DZ, Algeria
hafaifa.ahmed.dz@ieee.org
2
Faculty of Science and Technology, University of Medea 26000 DZ, Algeria
Guemana_M@univ-medea.dz
3
Faculty of Science and Technology, University of Sedik Ben yahia of Jijel, Algeria
r.belhadef@univ-biskra.dz
1 Introduction
Anti Surge
Valve User 1
Zone de fonctionnement
Normal
normale
operating area Zone de Pompage
Instability area
0.5
Mass flow coefficient
Coefficient de débit
N 0
-0.5
-1
Compressor Turbine
Gas Input
-1.5
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Pa Ta Load factor
Coefficient de Charge
Where m is the mass flow of the compressor inlet, mt the output mass flow
of the valve, Pp the output pressure, V p the output gas volume, P2 the pressure
at the compressor outlet, a01 the sound speed at the compressor inlet, Lc the
length of the compressor driving, A1 the goose surface of the impeller, J the
compressor moment of inertia, τ c the compressor torque and τ t is the turbine
torque.
This model is similar to FINK AL model [7], while the two first equations of
(1) are equivalent to Greitzer model [9, 10], where the speed is not included. The
Pp equation is derived from the mass balance assuming the trial isentropic trans-
fer. The third equation is derived from the second law of dynamics:
∑ M = J W ⇒ = 1 (τ -τ )
W
J
t C (2)
With M is the moments applied to the compression system and W is the an-
gular acceleration.
The compressors characteristics can be defined by different methods, in this
work we adopting the definitions suggested by Greitzer in [9, 10], namely:
w V
• The specific speed: η S = ,
Δ H 3/4
t/2
Vm
• The mass flow coefficient: Φ = , D 2 is the outlet impeller diameter,
U2
Δ H t/2
• The load factor: Ψ = .
U 22
The trial is assumed isentropic, so we can write from a mass balance:
382 A. Hafaifa, G. Mouloud, and B. Rachid
dPP dVP dρ
=-k =k P (3)
PP VP ρP
1 dρ dV
We have ρ = which implies =- , where:
V ρ V
dP
P P = k P ρ P = k R TP ρ P (4)
PP
The mass balance is given by the following relationship:
1 T
ρP = ( m - m t ) ⇒ P P = k R P ( m - m t ) = ( m - m t ) (5)
VP VP
Pt η t m tur C p , t Δ Ttur
τt = = (7)
w 2π N
Where: mtur is the steam flow at the inlet of the turbine and η is the mechan-
ical efficiency of the turbine.
Finally, we have the complete dynamical model given by the equations 8.
⎧ 2
a 01 ⎛⎜ m - k
⎪ p
P = Pp - P01 ⎞⎟
⎪ V p
⎝ t
⎠
⎪
⎪ ⎡ k −1
⎤
⎪ A ⎢ ⎛ Δ h ideal ⎞ k
⎥
⎨ m = 1 ⎢ Pp ⎜ 1 + η i ( m , N ) ⎟ - Pp ⎥ (8)
⎪ Lc ⎢ ⎝ ⎜ C p T01 ⎟
⎠ ⎥
⎪ ⎣ ⎦
⎪
⎪
⎪ N
⎩
= 1 τ -τ
2Jπ
(
t c )
Fuzzy Modeling and Control of Centrifugal Compressor Used in Gas Pipelines Systems 383
⎧ k P01 ⎛
⎪ P p = ⎜ m - kt Pp - P01 ⎞⎟
⎪ ρ 01 Vp ⎝ ⎠
⎪
⎪ ⎡ 4 ( k -1 )
⎤
⎪ A1 ⎢ ⎛ Δ h ideal ⎞ k
⎥
⎪ ⎜ ⎟
⎨
m= ⎢ P01 ⎜ 1 + η i (m, N) ⎟
- Pp ⎥ (9)
⎪ Lc ⎢ ⎝ C p T01 ⎠ ⎥
⎪ ⎣ ⎦
⎪
⎪ ⎛ η m C Δ Ttur ⎞
⎪ = 1 ⎜ t tur p , t
N - 2 r22 σ π N m ⎟
⎪⎩ 2 J π ⎜⎝ 2π N ⎟
⎠
The obtained model (9) is complex and difficult to exploit in control. Admitted-
ly, it better describes the main physical characteristics of the system, but it is often
necessary to simplify this complex model by linearization near a nominal operat-
ing point or by neglecting the less dominant dynamic process.
⎡ Pˆ ⎤ ⎡ B m −B ⎤ ⎡
P ⎤ ⎡ 0 ⎤
x = ⎢ pC ⎥ = ⎢ 1 1 ⎥ ⎢ pC ⎥ + ⎢
− V / B ⎥⎦
û b (10)
⎢⎣ mˆ c ⎥⎦ ⎢ B ⎥ m
⎣ c ⎦ ⎣
⎣⎢ Bmte ⎦⎥
~
With t = twH and wH is the Helmholtz frequency defined by the following
AC B
equation: wH = a with a = γRTa and Bm = . The parameters B and
V P LC G
Ut L A
G are defined as following: B = and G = t C , B is the Greitzer
2wH LC LC At
384 A. Hafaifa, G. Mouloud, and B. Rachid
stability parameter [9, 10]. It follows that the linear model assuming (9) achieving
status as follows:
⎧δx (t ) = A δx(t) + B δu(t)
⎪
⎨ (11)
⎪δy (t ) = C δx(t) + D δu(t)
⎩
⎡ ⎤ ⎡ ⎤
⎢ ⎥ ⎢ ⎥
⎢ 0.2076 0.1323 0⎥ ⎢ 0.2076 1.2365 0 ⎥
A = ⎢⎢ ⎥, B = ⎢⎢ ⎥
3.7181 - 0.4892 0⎥ 0 - 3.6891 - 2.312⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢⎣- 9.6870 - 1.539 0⎥⎦ ⎢⎣ 0 - 3.8230 0 ⎥⎦
⎡ ⎤ ⎡ ⎤
⎢ ⎥ ⎢ ⎥
⎢1 0 0⎥ ⎢ 0 0 0⎥
C = ⎢⎢ ⎥, ⎢
D=⎢ ⎥.
0 1 0⎥ 0 0 0⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢⎣0 0 1⎥⎦ ⎢⎣12.10740 0.3582 1⎥⎦
The first goal in centrifugal compressors modeling is to estimate losses, the first
model called "Single-Zone Modeling" includes only global characteristics of the
compressor. The proposed fuzzy model for the examined compression system is
symbolically expressed by a set of IF-THEN rules of the form [5, 26]:
IF Z 1 (t ) est Fi1 et ... .. et Z n est Fin
THEN
⎧
[ ]
r
The obtained results using fuzzy controller, shown in figure 3 are better than
those obtained with conventional control algorithms. The proposed controller is a
system based on special knowledge, using a limited depth of reasoning in a for-
ward chaining procedure rules (rule firing by the premises).
Compression Compression
system output system input
Knowledge Base
Decision mechanism
Fuzzification Defuzzification
Compression
system
Several stages in the processing rules are distinguishing in the proposed fuzzy
controller structure. The variables are the output quantities of gas compression
system and capture decisive action to the dynamic evolution in the examined sys-
tem. The knowledge base is composed of a database and a database of rules.
For the different cases studied, the symbolic representation of the dynamics of
the compression system, we have adopted consists of nine rules type If _Then.
386 A. Hafaifa, G. Mouloud, and B. Rachid
0.3
-0.05
Fuzzy model
Linearized model
-0.1 0.25
Complex model
Mass flow coefficient
-0.2
0.15
Fuzzy model
Linearized model
-0.25 Complex model
0.1
-0.3
0.05
-0.35
0 1.6 3.2 4.8 6.4 8
0
time(s) 0 1.6 3.2 4.8 6.4 8
time(s)
Fig. 4 Fuzzy model of the mass flow coeffi- Fig. 5 Fuzzy model of the mass flow coef-
cient for B = 1.25 ficient for B = 1
From the obtund results it is clear that the proposed fuzzy model is more accu-
rate than the linearized model. In the various cases considered, the characteristics
of the compression system described by the model complex are perfectly repro-
duced by the proposed fuzzy model.
To validate the proposed model, we used a fuzzy controller that allows distri-
buted seven fuzzy sets on the universe of discourse of each input variable and out-
put (NG, NM, NP, EZ, PP, PM, PG). Increasing the number of fuzzy sets to seven
(NG, NM, NP, EZ, PP, PM, PG) requires the processing of 49 rules, so that the
distribution of these sets on the universe of discourse of each variable becomes a
choice. As regards the matrix inference, the sets of the output variable (increment)
are always derived from the Mac-Vicar matrix. By analyzing the responses of
fuzzy controllers we see the smooth continuation of fuzzy controllers to variations
of the operating. The figure 6 shows the evolution of load factor, we see that it is
properly oriented and stabilizes reference and the figure 7 shows the response of
proposed fuzzy controller applied to the load factor.
Fuzzy Modeling and Control of Centrifugal Compressor Used in Gas Pipelines Systems 387
-0.1 2
40
-0.2
1.5
35
-0.3
1
30 -0.4
0 100 200 300 0 100 200 300
0.5
B =d’aspiration
Pression 0.25 / f B = 0.50
2 Résidu sur la pression d’aspiration
45 -0.1 0
-0.2
40 -0.5
-0.3
35 -1
-0.4
30 -0.5 -1.5
0 100 200 300 0 100 200 300 0 200 400 600 800 1000 1200 1400
temps
Time
Fig. 6 Evolution of load factor in the Fig. 7 Response of proposed fuzzy controller
examined system
The increase of the fuzzy sets increasing linguistic rules, therefore the computa-
tion time, the fuzzy controller with five sets in this investigated case represents a
good choice that satisfies the tradeoff between the computation time and the re-
quirement of performance specifications.
4 Conclusion
The obtund results confirm that the proposed fuzzy model is more accurate than
the linearized model. In the various cases considered, the characteristics of the
compression system described by the model complex are perfectly reproduced by
the proposed fuzzy model. Also, the fuzzy controller used appears to have mul-
tiple benefits. It does not require a mathematical programming is difficult and
easy. It should be noted that it requires a relatively slow computation time, while
traditional algorithms often result in very short programs. It may seem at first
glance that the fuzzy controller requires very little knowledge of the system to set-
tle. The controller is very sensitive to variations of system parameters, so its
rugged good looks. In reality should not conceal a certain number of disadvantag-
es of this type of controller.
References
Adamy, J., Schwung, A.: Qualitative modeling of dynamical systems employing conti-
nuous-time recurrent fuzzy systems. Fuzzy Sets and Systems 161(23), 3026–3043
(2010)
388 A. Hafaifa, G. Mouloud, and B. Rachid
Wang, L.-X.: Modeling and control of hierarchical systems with fuzzy systems. Automati-
ca 33(6), 1041–1053 (1997)
Mei, F., Man, Z., Nguyen, T.: Fuzzy modelling and tracking control of nonlinear systems.
Mathematical and Computer Modelling 33(6-7), 759–770 (2001)
Bernal, M., Guerra, T.M., Kruszewski, A.: A membership-function-dependent approach for
stability analysis and controller synthesis of Takagi–Sugeno models. Fuzzy Sets and
Systems 160(19), 2776–2795 (2009)
Guemana, M., Aissani, S., Hafaifa, A.: Use a new calibration method for gas pipelines: An
advanced method improves calibrating orifice flowmeters while reducing maintenance
costs. Hydrocarbon Processing Journal 90(8), 63–68 (2011)
Guemana, M., Aissani, S., Hafaifa, A.: Flow measurement and control in gas pipeline sys-
tem using intelligent sonic nozzle sensor. Studies in Informatics and Control
(SIC) 20(2), 85–96 (2011)
Sugeno, M., Kang, G.T.: Fuzzy modelling and control of multilayer incinerator. Fuzzy Sets
and Systems 18(3), 329–345 (1986)
Burst Test and J-Integral Crack Growth
Criterion in High Density Poly-Ethylene Pipe
Subjected to Internal Pressure
Abstract. In the present work we are interested on the analysis of the severity of
crack defects created by a disc cutter and to study the behavior of a high density
polyethylene pipe (HDPE pipe) when subjected to an internal pressure, either in
the absence or presence of a pre-crack. In order to do this, experimental tests was
performed to measure the toughness and to determine the mechanical behavior of
HDPE. These features were used to perform numerical simulations using
ABAQUS on pipe solicited by an increase in internal pressure. This allows to
compare with burst tests of cracked pipes and to determine the fracture energy that
will be compared to the toughness.
The results show that the crack is initiated in the radial direction. And the size
of the crack has a great influence on the energy of rupture and consequently on the
ultimate pressure. There is a good agreement between experimental and numerical
results.
1 Introduction
High density polyethylene (HDPE) pipe provides the lowest life cycle cost when
compared to other systems due to significantly reduced or no leakage, increased
billable dollars, water conservation, fewer new water-treatment plants, reduced
maintenance crews, reduced seasonal water-main breaks, and its viscoelastic
behavior makes it a good resistant to water hammer (CSIRO 2008).
In the present study, two part investigation will be made for HDPE pipes, with
superficial defect, using the well-known approach “J-integral” of the fracture
mechanics (Rice 1968).
In the first part, experimental tests will be made to determine mechanical
behavior (stress-strain curve) and failure analysis (JIc) of high density
polyethylene.
In the next part, FEM cracked tube model was performed on ABAQUS for
calculating J-integral around the crack type. Placed into the J integral curve JIc
allow to know the internal pressure which leads to the crack initiation.
To validate FEM Model, experimental burst tests are performed on cracked
HDPE pipes. The measurement results will be presented as a curve of pressure
versus volume of the added water.
30
25 V=0.001(1/s)
V=0.005(1/s)
20 V=0.01(1/s)
Stress (MPa)
15
10
0
0 100 200 300 400 500 600 700
Strain (%)
±0.2
W = 41.7 width
±0.2
L = 52.5 overall length (1,25 w ± 0,01 w)
±0.2
d1 = 50 transverse width (1,2 w ± 0,01 w)
R = 5+0.1 Radius (0,125 w ± 0,005 w)
b = 8+0.1 Thickness (0,4 w < b < 0,6 w or b < w/2)
±0.1
a = 18 crack length (0,45 w < a < 0,55 w)
±0.1
d2 = 11.5 distance between the center of two holes, (0,55 w ± 0,005 w)
balanced/Plans crack ± 0.005w
Fig. 2 CT specimen
100
90
80
70
60
data points
J (KJ/m 2 )
50 regression line
40
30
20
10
J0.2=7.69
0
0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8 2 2,2 2,4 2,6 2,8 3 3,2 3,4 3,6
Cra ck extention Δa (mm)
3 Burst Tests
All tests were performed on a sample size of 3 specimens per design type of the
same age and service history (new pipes). And these tests were made according to
the following instructions
- Setting up the specimen and fill it with water until completely filled and
bled from air.
- All specimens have been loaded with a continuous pressure rate of 10 bar
per minute until failure (figures 5a and 5b).
60
50
Pressure (bar)
40
30
20
10
0
0 1 2 3 4 5 6 7
Time (min)
(a) (b)
Fig. 5 Burt test
Unlike metallic structure, the use of strain gauge on HDPE pipes to obtain the
strain-internal pressure curve needs a complicated and delicate pre-treatment and
results will not accurate enough. So a more reliable solution should be adopted
The measurement results will be presented as a curve of pressure versus volume of
water added.
Comparison between experimental data and numerical simulation results will
be made to validate the FEM model.
60
Without crack
a=2mm
a=4mm
50
a=6mm
a=8mm
40
Pressure (bar)
30
20
10
0
0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4
Added volume(l)
58
56
55,44
54 53,8
52,46
Burst pressure (bar)
52
50
48
46,8
46
44
43,1
42
40
0 1 2 3 4 5 6 7 8 9
pre-crack depth (mm)
Burst pressures for different cracked pipes are deduced from figure 7 and
represented on figure 8 as a function of the pre-crack’s depth. These results show
a sharp drop in the burst pressure when the depth of crack is more than 4mm.
The figure 9 shows the evolution of the fracture energy in different contours in
the normal plane to the crack front. They converge to the same value.
The curve of the fracture energy versus pressure and the value of the critical
2
failure energy (JIC=7.69KJ/m ) allow to determine the internal pressure (Pa=46bar)
which lead to the crack initiation (figure 9). It is less than the ultimate pressure
(Pu=52.5 bar) (figure 10).
14
contour2
contour3
12
contour4
contour5
fracture resistance J (KJ/m2)
10
contour6
contour7
8
JIC=7.69
6
-2
0 10 20 30 40 Pa=46 50 60
Pressure (bar)
60
Pu=52.2
50
Pa=46
Numerical test
40
Experimental test
Pressure (bar)
30
20
10
0
0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4
Added Volume(l)
In the case of a cracked pipe with a 4 mm crack’s depth, the ultimate pressure
and maximum admissible pressure are calculated from the burst test using fracture
toughness JIC experimentally obtained. The result is quite logic since Pa < Pu. This
proves the accuracy of the value of JIC.
Numerical calculation of admissible pressure for the other cracked pipe will be
made soon.
In this study, the mechanical propriety of high density polyethylene (PE100) pipe
such as behavior law and fracture toughness JIC were experimentally obtained.
These parameters were used to study the effect of the crack’s depth on the ultimate
pressure of a PE100 pipe.
Burst Test and J-Integral Crack Growth Criterion 399
Results show that the crack is initiated in the radial direction. And the size of
the crack has a great influence on the energy of rupture and consequently on the
ultimate pressure. There is a good agreement between experimental and numerical
results.
In the upcoming study some improvement will be made like:
- Use a concept of visco elastic-plastic fracture mechanics to analyze the
fraction behavior
- Introduce a damage criterion in the numerical simulation
- Perform a dynamic testing to study the effect of transient flow in cracked
pipes
References
CSIRO, Life Cycle Analysis of Water Networks. presented at Plastics Pipe XIV, Budapest
(2008)
Rice, J.R.: A Path Independent Integral and the Approximate Analysis of Strain
Concentration by Notches and Cracks. Journal of Applied Mechanics 35, 379–386
(1968)
ASTM D 6068 Standard Test Method for Determining J-R Curves of Plastic Materials
WARP3D. Release 15.0 manual, civil engineering, reports no UILU-ENG-95-2012.
Urbana, University of Illinois (2004)
Benhamena, A., Bouiadjra, B., Amrouche, A., Mesmacque, G., Benseddiq, N., Benguediab,
M.: Three finite element analysis of semi-elliptical crack in high density poly-ethylene
pipe subjected to internal pressure. Materials and Design 31, 3038–3043 (2010)
Benhamena, A., Aminallah, L., Bouiadjra, B., Benguediab, M., Amrouche, A., Benseddiq,
N.: J integral solution for semi-elliptical surface crack in high density poly-ethylene pipe
under bending. Materials and Design 32, 2561–2569 (2011)
Favier, V., Giroud, T., Strijko, E., Hiver, J.M., G’Sell, C., Hellinckx, S., et al.: Slow crack
propagation in polyethylene under fatigue at controlled stress intensity. Polymer 43,
1375–1382 (2002)
Brostow, W., Fleissner, M., Müller, W.: Slow crack propagation in polyethylene:
determination and prediction. Polymer 3, 419–425 (1999)
Solving the Three-Dimensional Time-Harmonic
Maxwell Equations by Discontinuous Galerkin
Methods Coupled to an Integral Representation
1 Introduction
Electromagnetic phenomena are generally described by the electric and magnetic
fields E and H which are related by the following Maxwell equations:
-ε∂ t E + curl H = 0 (1)
-μ∂ t H + curl E = 0 (2)
Where ε and μ are the complex-valued relative dielectric permittivity and the
relative magnetic permeability, respectively. We are interested in time-harmonic
solutions which satisfy the following equations:
curl E - iωμ H =0 (3)
curl H - iωε E =0 (4)
These two equations are posed in an unlimited domain. Or approach consists of
limiting the obstacle D by a fictitious boundary Γ a where we impose the
absorbent of Silver-Muller condition. On the boundary Γ m of D, we take the
perfect conductor condition.
Where:
• ω : The angular frequency of the problem, n is a normal vector to the
boundary Γ a .
μ
• Z= .
ε
• ℜ( E ) , ℜ( H ) are the expression of the electric and magnetic fields E and
H , respectively, on Γ a . They are given by the following integral
representation:
1
ℜ( E ) = curl x ∫ n( y ) ∧ E ( y )G ( x, y )∂σ y − curlx curlx ∫ n( y ) ∧ H ( y )G ( x, y )∂σ y
Γa iωμ Γa
1
ℜ( H ) = curlx ∫ n( y ) ∧ H ( y )G ( x, y )∂σ y − curlx curlx ∫ n( y ) ∧ E ( y )G ( x, y )∂σ y
Γa iωε Γa
fined as follows:
Solving the Three-Dimensional Time-Harmonic Maxwell Equations 403
⎛ 03×3 N el ⎞
Gl = ⎜ t , for l ∈ {1, 2,3} ,
⎜ N e 03×3 ⎟⎟
⎝ l ⎠
⎛ v1 ⎞
⎜ ⎟
where for a given vector v = ⎜ v2 ⎟ , the matrix N is:
⎜v ⎟
⎝ 3⎠
⎛ 0 v3 −v2 ⎞
⎜ ⎟
N v = ⎜ −v3 0 v1 ⎟ .
⎜v ⎟
⎝ 2 −v1 0 ⎠
⎛E ⎞
It is easy to verify that Gn = G1n1 + G2 n2 + G3 n3 . Let W = ⎜ ⎟ , therefore the
⎝H ⎠
system of equations (5), (6), (7) and (8) can be rewritten in the following
conservative form:
3 Discretization
N
Ωh =
K∈
∪τ h
K
∫
Ωh
(iωQWh )t .Vdx + ∫ (∇.( F (W ))h )t .Vdx = 0 ,
Ωh
∑τ ( ∫ ( ( F (W ) ) ) .∇Vdx + ∫ ( F (W ) .n ) .V ∂σ ) = 0
t
∫Ωh
(iωQWh )t .Vdx −
K∈ h
K h ∂K h
(12)
Using the same technics adopted by Ern and Guermond [1,2], and by adding the
terms of the integral representation , we obtain the following formulation:
∑τ ∫ ( ( F (W )) ) .∇Vdx
t
∫ Ωh
(iωQWh )t .Vdx −
K∈ h
K h
F F h nF h
F ∈Γ0
⎡ ⎛1 ⎞
t
⎛1 ⎞
t
⎤
( )
+ ∑ ⎢ ∫ ⎜ M F , K − I FK GnF Wh ⎟ .V ∂σ − ∫ ⎜ M F , K − I FK GnF ℜ (W ) ⎟ .V ∂σ ⎥ ( )
⎢ ⎝2
F ∈Γa ⎣
F
⎠ F
⎝2 ⎠ ⎦⎥
t t
⎛1 ⎞ ⎛1 ⎞
m
F 2
( ⎠ m ⎝
F 2
)
+ ∑ ∫ ⎜ M F , K − I FK GnF Wh ⎟ .V ∂σ = ∑ ∫ ⎜ M F , K − I FK GnF Whinc ⎟ .V ∂σ
⎝ ⎠
( )
F ∈Γ F ∈Γ
Where:
ℜ (W )( x ) = ∫ m K ( x, y ) W ( y )dy such that K : \ 6 × \ 6 → M 6 ( ^ ) is a Green
Γ
aV b = I FK .V\
1
2 K
K
+ I F Ki .V\ Ki and {V } =
V\ + V\ Ki such as I FK represents the ( )
incidence matrix between oriented faces and elements and they are given by:
We mention here that for the terms on the boundary, we used two classical
numerical flux.
Solving the Three-Dimensional Time-Harmonic Maxwell Equations 405
= ⎨ ⎜⎝ − N nF 03×3 ⎟⎠
t
M F ,K
⎪
⎪⎩ GnF if F ∈ Γ
a
In this case:
⎧ ⎛η F N nF N nt F I FK N nF ⎞
⎛ α N nF N ⎞ ⎪⎪ I FK ⎜⎜ I N t ⎟ if F ∈ Γ m
E t
03×3 ⎟
SF = ⎜ ⎟ , M F ,K
F nF
=⎨ ⎝ FK nF 03×3 ⎠
⎜ α F N nF N nF ⎟⎠
H t
⎝ 03×3 ⎪
⎪⎩ GnF if F ∈ Γ
a
1
For homogeneous media, η F = α FE = α FH =
2
4 Linear System
• A ( j , i ) = B × δ ij
Γ0
ij
Γ0 Γm a 0
Where D , D , D 1
i i
2
i , D i and DiΓ are diagonal blocks, BijΓ defines the
extra-diagonal block.
406 N. Gmati, S. Lanteri, and A. Mohamed
Cij =
1
2
( I3 ⊗ Ψ ia ) .Ki ij .( I 3 ⊗ Ψ im )
5 Numerical Results
Test No. 1
Parameters Value
Outer radius 1.18
Frequency(Ghz) 0.3
H max 0.30
Number of cells 3119
Test No. 2
Parameters Value
Outer radius 1.1
Frequency(Ghz) 0.3
H max 0.16
Number of cells 10512
According to figures we can see that this method leads to a solution which is
very similar to the exact one.
Acknowledgements. The authors also gratefully acknowledge the helpful comments and
suggestions of the reviewers, which have improved the presentation.
References
[1] Ern, A., Guermond, J.-L.: Discontinuous Galerkin methods for Friedrichs systems I.
General theory. SIAM J. Numer. Anal. 44(2) (2006)
[2] Ern, A., Guermond, J.-L.: Discontinuous Galerkin methods for Friedrichs systems II.
Second-order elliptic PDEâs. SIAM J. Numer. Anal. 44(6) (2006)
[3] El Bouajaji, M., Gmati, N., Lanteri, S., Salhi, J.: Coupling of an exact transparent
boundary condition with a DG method for the solution of the time-harmonic Maxwell
equations, vol. 95, pp. 237–247. Springer (2014)
[4] Ben Belgacem, F., Fournie, M., Gmati, N., Jelassi, F.: On the Schwartz algorithms for
the elliptic exterior boundary value problems. ESAIM. Math. Model. Numer. Anal. 39,
639–714 (2005)
[5] Colton, D., Kress, R.: Inverse acoustic and electromagnetic scattering theory. Springer
(June 1997)
Iterative Methods for Steady State Looped
Network Analysis
Abstract. The aim of this paper is to numerically study the steady state of water
flows in looped networks. This study will be performed by the use of Hardy Cross
Method and Newton-Raphson algorithm. The comparison of the numerically
obtained results by these two methods to those obtained by the use the commercial
software Pipe Flow Expert has shown a good concordance between them. The
numerical analysis has shown that the convergence of Newton-Raphson method is
more rapid than that of the classic Hardy Cross Method.
1 Introduction
Fluid transport from production sites to the exploitation sites is a major concern
for manufacturers. Many studies have shown that the most effective method is to
transport via apparent or underground pipelines. These pipelines, connected
together, form what called a "hydraulic network" that can be either branched or
looped or even adopting a hybrid topology. To analyze the steady state, hydraulic
networks flows are to be balanced. For a branched network, by knowing the flow
demand at each terminal, it’s easy to determine the flow rate at each conduit of the
network since the flow direction is evident.
The problem arises when we start facing a purely looped or hybrid network
(containing loops and branches). Several methods and techniques are used to
balance flows in these types of network. Between these methods, one can cite: the
Hardy Cross Method (HCM), Node-Loop Method, Modified Node Method and
Andrijašev Method. A comparative study between the previous techniques has
been carried out (Dejan 2011). In this study, the number of iterations needed to
achieve the network equilibrium is taken as criterion to put in evidence the
efficiency of these methods. Dejan proved that, for complex networks, Node Loop
Method is more efficient as it does not require complex numerical scheme for
algebraic addition of corrections in each of iterations.
These techniques, and as are iterative methods, require an initial guess of flow
rates through pipelines. This guess has to satisfy the first Kirchhoff’s law: nodes
law. In many studies the initial guess is introduced manually.
In this paper, Hardy Cross and Newton-Raphson methods are used to solve a
water looped network. A comparative study between the two methods in terms of
efficiency, elapsed time, and precision will be conducted. In this study, the initial
guess imposed by these iterative methods will be generated automatically.
2 Mathematical Formulations
(1)
(2)
where is the pipe length, is its diameter, is the gravitational acceleration and
is the friction factor.
The friction factor is given by Colebrook-White equation (Giles 1962; Fox
1992; Shames 1989; White 1994):
.
2 (3)
√ . √
where e is the internal roughness of the pipe and is the Reynolds number given
by the following expression:
(4)
The Hardy Cross method is used to balance flows through looped networks. In
this method, an initial guess of flow per conduits has to be assigned, and in that
way chosen values are to be used for first iteration. Then, an iterative process is
conducted with correction of the flow-rate for each loop. The following
assumptions must be satisfied for Hardy Cross calculation:
- Algebraic sum of flows per each node must be exactly equal to zero (first
Kirchhoff”s law: continuity of flow).
- Algebraic sum of head losses per each loop must be approximately zero at the
end of iterative procedure (second Kirchhoff’s law: continuity of potential).
The loop flow correction is given by the following equation:
∑
∑ ⁄
(5)
where ∑ is the algebraic head loss sum of all pipes in the considered loop, is
the flow-rate through each pipe of the loop and is depending on the used
formula for head loss expression.
In general, the iterative correction process is continued until the integrated head
loss around all loops in the network is reduced to a specified limiting value .
The Newton Raphson method is based on the resolution of non linear equation
0 by successive iterations using the following expression:
; 0,1, … (6)
; 0,1, … (7)
Pipe number 1 2 3 4 5 6
Upstream node 1 2 3 3 2 2
Downstream node 2 3 4 5 6 7
(8)
1
1 (9)
0
The dimensions of the matrix are . By respecting equation (9) and table
1, the obtained matrix is:
Iterative Methods for Steady State Looped Network Analysis 413
1 0 0 0 0 0
1 1 0 0 1 1
0 1 1 1 0 0
0 0 1 0 0 0 (10)
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
Vector C is written as follows:
300 ; 0 ; 0 ; 70 ; 150 ; 0 ; 80 (11)
In this vector, the minus sign preceding the first flow-rate indicates that the first
node is a supply node (the node that feeds the network). In a hydraulic network,
we can encounter one or more supply nodes. Respecting the mass continuity
principle, the sum of vector elements has to be equal to zero.
To solve the system of equations (8), matrix must be non singular. However,
its first row can be written
∑ (12)
This equality has a physical signification since the first row concerns the first node
(the supply node) and the other rows concern consumption nodes. To ensure that
matrix is non singular, its first row is removed and subsequently the first
element of the vector . By taking into account the previous considerations, the
resolution of system (8) permits to determine the vector :
⁄ 300 ; 220 ; 70 ; 150 ; 0 ; 80 (13)
All the described steps to solve the branched network equations, in this
section, will be integrated to balance the looped network flows.
Table 2 Tubular pipes material proper- Table 4 Studied looped network characteristics
ties: steel Sch 80 pipe (ANSI standards)
Designation Value Unit Pipe Upstream Downstream Length
number node node (m)
Internal roughness 0.0459 1 2 900
1
Nominal Pipe Size 500 2 2 5 1200
3 6 5 900
Internal diameter 455.625
4 1 6 1200
5 2 3 900
Table 3 Fluid properties: water 6 3 4 1200
7 5 4 900
Designation Value Unit
8 5 8 1200
Temperature 25° ° 9
9 8 900
Density 997 / 10 6 9 1200
Kinematic 11 4 7 1200
0.89 . .
viscosity 12 8 7 900
The initial guess required by the HCM will be generated automatically. As a first
step, a branched network is derived from the initial looped network. As mentioned
in section 5.1 the number of pipes in a branched network is equal to 1. As
consequence, the number of pipes to be removed from a looped network
containing pipes and nodes is : 1 one from each loop. In our case
study, the number of pipes to be removed is equal to 4. Figure 3 shows the
obtained branched network.
Then, an incidence matrix 1 of the eliminated pipes and 2 of the branched
network are built respecting the principles of equation (9) and a vector
solution for the eliminated pipes is generated. This vector can be introduced
randomly or, simply, chosen as null vector.
For the flow-rate vector of the branched network, we have:
Iterative Methods for Steady State Looped Network Analysis 415
1 2 (14)
1 2 1 (15)
100
50
Flow -rate (l/s)
-50
-100
-150
-200
2 4 6 8 10 12 14 16 18 20 22
Iteration number
Fig. 4 Variation of flow-rate through pipe 2 Fig. 5 Pipe Flow Expert simulation result
416 Z. Hafsi et al.
Figure 4 illustrates the correction of flow-rate through pipe 2 from the initial
supposed value to the real one. This figure shows sign respect while applying the
correction procedure, so that the flow-rate value of a common pipe between two
loops is considered positive through one loop negative through the other.
The obtained vector contains no negative values. So, the real flow direction is
the same initially supposed (from the upstream to the downstream node).
considered. In our case, and as mentioned above for both HCM and Newton
Raphson Method, the solution is reached when the difference between two
successive obtained results ( and ) is less than the chosen limiting value
for HCM and for Newton Raphson Method.
To evaluate the efficiency of the used methods of analysis, a simulation of the
studied network was performed by the use of the commercial software Pipe Flow
Expert. The obtained results by that software are shown in figure 5. Table 5 shows
the flow distribution in each conduit of the network obtained by the three used
methods under the same limiting value ( =10 ).
The comparison of Hardy Cross and Newton-Raphson methods results show a
good concordance. Added to that, our numerical results compare very favorably
with those obtained by “Pipe Flow Expert” software.
400 220
Hardy Cross Method Pipe 4 Newton Raphson Method
200
350 Pipe 6
Pipe 9
180 Pipe 4
300 Pipe 12
Pipe 6
160 Pipe 9
Flow-rate (l/s)
250
Flow-rate (l/s)
Pipe 12
140
200 120
150 100
80
100
60
50
40
0
10 15 20 25 30 35 40 45 50 55 60 20
6 8 10 12 14 16 18 20 22
Time (s)
Time (s)
For 4 pipes of the network, this figure shows the variations of flow-rate during
the correction process from the initial guess to the final result. The correction
phase has required more than 53 seconds to achieve the actual solution for HCM
and just 14 seconds for Newton-Raphson Method. Added to that the total elapsed
time for the HCM is about 60 seconds while it’s about 21 seconds for the Newton-
Raphson Method. This difference in elapsed time is explained by the quick
correction phase for the second method. Indeed the flow rate through each pipe is
evaluated once in each iteration, while in the HCM method, we have to correct the
flow-rate value through each loop that means pipes belonging to two loops have to
be calculated twice (in clockwise and anticlockwise directions).
In term of iterations number required to achieve final result, figure 7 shows that
Newton-Raphson Method converges after about 9 iterations and reaches the
desired solution while HCM requires about 21 iterations to find that solution.
Both used methods are efficient to analyze a looped network but in term of
rapidity convergence and simplicity to implement the algorithm, Newton-Raphson
Method is more adequate mainly that the elapsed time will be so important to
reduce when analyzing more complex networks.
418 Z. Hafsi et al.
140
120
Flow-rate (l/s)
100
80 Pipe 2
Final flow- rate: 97.35 l/s
60
40
20
0
2 4 6 8 10 12 14 16 18 20 22
Iteration number
6 Conclusion
References
Cross, H.: Analysis of flow in networks of conduits or conductors (1936)
Brkic, D.: Iterative Methods for Looped Network Pipeline Calculation. Springer
Science+Business Media B.V. (2011)
Giles, R.V.: Theory and Problems of Fluid Mechanics and Hydraulics, p. 274. Schaum
Publishing Co., New York (1962)
Fox, R.W., McDonald, A.T.: Introduction to Fluid Mechanics. Wiley, New York (1992)
Shames, I.H.: Mechanics of Fluids, p. 692. McGraw-Hill Book Co., Singapore (1989)
Walski, T.M., Chase, D.V., Savic, D.A.: Water Distribution Modeling. Haestad Press,
Waterbury (2001)
White, F.M.: Fluid Mechanics, p. 736. McGraw-Hill Book Co., Singapore (1994)
Investigation and Modeling of Surface
Roughness of Hard Turned AISI 52100 Steel:
Tool Vibration Consideration
Mechanics and Structures Research Laboratory (LMS), May 8th 1945 University,
P.O. Box 401, Guelma 24000, Algeria
meddour26@yahoo.fr
Abstract. Since the great interest of the achieved parts quality, this experimental
study focuses on the modeling and the investigation of the surface roughness of
hard turned AISI 52100 steel. The tool vibration was taking into consideration
beside cutting speed, depth of cut, feed rate and tool nose radius. The response
surface methodology (RSM) was employed for modeling process. Models
reliability was established by conducting confirmation tests. Slight divergence
between the experimental and their corresponding predicted values were observed.
The significance of the different factors on surface roughness was established by
applying analysis of variance (ANOVA). The results revealed that the best surface
roughness is obtained by using small feed rate and large nose radius. Furthermore,
a correlation between the surface roughness and tool vibrations was established.
1 Introduction
A numerous researches were carried out to study the effect of cutting conditions
on surface roughness. (Yallese et al. 2005) examined the effect of cutting
parameters, i.e. cutting speed, depth of cut, feed rate on surface roughness in hard
machining of X200Cr12 steel. The results showed that large feed rate deteriorates
surface roughness while depth of cut did not show a remarkable effect. So, they
advised to machine with a large depth of cut in order to increase the material
removal rate. The mixed ceramic tool was used by (Aslan et al. 2007) in their
experimental investigation regarding the hard turning of AISI 4140 steel. They
indicated that the surface roughness was mainly affected by cutting speed-feed
rate and feed rate-depth of cut interactions. They modeled the surface roughness
by multiple linear regressions. (Deruja et al. 2009) observed the best surface
roughness at a small feed rate and high cutting speed in hard turning of AISI H11
steel with coated mixed ceramic inserts. The effect of cutting parameters on the
surface roughness was also investigated by (Lalwani et al. 2008) in finish hard
turning of MDN250 steel using coated ceramic tool. The employment of ANOVA
demonstrated that the feed rate is the prevailing parameter whereas cutting speed
had no significant effect on surface roughness. In hard turning of AISI 52100 steel
with CBN tool, (Bartarya and Choudhuryb 2012) used the multiple linear regression
to develop a surface roughness model prediction based on cutting parameters.
Other experimental studies, as the work of (Horng et al. 2008), considered the
tool nose radius as variable along with the cutting parameters. The ANOVA
revealed that the cutting speed and the tool nose radius influenced significantly the
surface roughness of Hadfield steel hard turned with mixed ceramic insert. Also
they linked the inspected increase of surface roughness when using a large tool
nose radius to the excessive thermal crack and flank wear. In their point of view it
is due to the enlargement of the contact area. The same number of variables was
also adopted by (Saini S et al. 2012) in the case of hard turning of AISI H11 steel
with ceramic tool, the ANOVA results showed that the feed rate influences the
surface roughness most significantly followed by cutting speed and nose radius.
Thus, according to the author the best surface quality was attainable by decreasing
feed rate and increasing cutting speed. In hard turning AISI H11 with CBN tool,
(Aouici et al. 2012) observed that feed rate had a statistical significance on the
surface roughness, because its increase generates helicoids furrows the result of
tool shape and helicoids movement tool-workpiece. They developed mathematical
models for surface roughness by using response surface methodology (RSM).
Other experimental researches were interested by the influence of the vibration
generated during the process on the machined products. This vibration could result
in the instability of the system, known as chattering, if an inadequate selection of the
cutting parameters is done. (Tlusty 1999), consider that this phenomenon occurs
when the uncut chip thickness is great regarding the dynamic stiffness of the system.
(Thomas et al. 1996) related the difference between the measured and the theoretical
surface roughness to the vibrations. Based on this concept, (Jang et al. 1996)
developed a real-time monitoring algorithm to calculate the surface roughness. The
algorithm superpose the theoretical profile and the oscillatory profile
determined by the relative vibration between the cutting edge and the
workpiece. The radial acceleration of the tool was taking into account beside the
cutting parameters by (K.A Risbood et al. 2003), when they developed a neural net
work models to predict surface roughness and dimensional deviation. (Kirby et al.
2004) developed a prediction surface roughness model based on feed rate and the
measured vibration in three directions. (Abouelatta and Madl 2001) disclosed that
the mathematical models of surface roughness including both machining conditions
and tool vibrations were more efficient than those depending on machining
conditions only. (Hessainia et al 2013) correlated between the surface roughness and
tool nose displacement. They developed roughness models containing all cutting
parameters, tool-nose displacements, spindle and machine tool frame.
Investigation and Modeling of Surface Roughness of Hard Turned AISI 52100 Steel 421
The aim of this work is the investigation of the effects of cutting conditions
(cutting speed, depth of cut, feed rate and tool nose radius) and tool vibration on the
surface roughness of hard turned AISI 52100 steel. This, by applying the analysis of
variance (ANOVA). Also, the surface roughness is modeled by employing the
response surface methodology (RSM), first by using only cutting conditions, then,
the tool vibration is included in the model and both models are compared.
2.1 Equipments
Figure 1 shows the experimental procedure and the different equipments used in
this study. The experimentation was performed under dry conditions using AISI
52100 steel as workpiece material. The hardness was raised by quenching and
tempering treatment followed by checking measurement with a digital Micron
Hardness Tester DM2-D390. The average of measured values was 59 HRC. The
workpiece is mounted on a universal lathe, model SN40C, spindle power 6.6 KW.
The ceramic insert tool is a mix of 70% Al2O3 and 30% TiC. This type of inserts is
commonly called CC650, and its ISO designation is SNGN 1204(r) T01020 where
(r) is the tool nose radius ranged from 0.8 to 1.6 mm in this work, indicating a 20°
chamfer angle over 0.1mm width. Tool holders are codified as PSBNR2525M12
with a common active part tool geometry described by: cutting edge angle χr
=+75°, clearance angle α =+6°, rake angle γ =-6° and cutting edge inclination
angle λ =-6°. New cutting edge is used for each test in order to reduce the wear
influence on measured responses. The vibrometer (SmartBalancer2) is used for the
measurement of the acceleration amplitude of the tool vibration in both tangential
and radial directions. The sampling frequency is 12800 Hz. Regarding the
measurement of the surface roughness, a Mitutoyo Surftest–201 (cut-off = 0.8) is
applied after each test in three different positions by going around the workpiece
each 120°. Afterward, the average of the three measurements is estimated. In
addition, the three-dimensional topographic maps of the machined surfaces are
produced using the optical platform of metrology modular Altisurf 500.
∑ ∑ ∑
k k k
Y = a0 + bX + b X Xj + b X2 (1)
i =1 i i i , j ij i i =1 ii i
level Cutting speed (m/min) Feed rate (mm/rev) Depth of cut (mm) Tool nose radius (mm)
-1 100 0.08 0.05 0.8
0 150 0.11 0.15 1.2
+1 200 0.14 0.25 1.6
Investigation and Modeling of Surface Roughness of Hard Turned AISI 52100 Steel 423
The adopted model for surface roughness including cutting parameters is the
linear with interaction. The results given by ANOVA presented in Table 3 shows
that the terms r and f are the most significant with the respective contribution
(41.98, 35.51) %. The term ap has a small contribution with 3.23 %, whereas
cutting speed does not affect the surface roughness. A similar result was obtained
by (Aslan et al. 2007) when turning AISI 4140 steel (63 HRC) with mixed
ceramic tool. The interactions which show significant effect on surface roughness
are ap×r and f×r with the respective contributions (9.67 and 7.34) %. The
determination coefficient model’s R² is 0.94. The final equation in terms of real
factors is given as follow:
Ra = −0.049 + 2.242 × 103Vc − 3.618ap + 12.479 f + 0.051r − 0.015Vc × f − 6.875 × 10 −4 Vc × r + 7.917 ap × f (2)
+ 1.938ap × r − 5.625 f × r
The ANOVA results for surface roughness model including cutting parameters
and tool vibrations in both radial and tangential directions are presented in Table
4. It shows that in addition to the terms cited above, the surface roughness is
influenced by the tool vibration in the radial direction AMPy with 3.71 % of
contribution. The determination coefficient model’s R² is improved to 97.10 %.
The final equation in terms of real factors is given as follow:
Ra = −0.895964 − 2.32358 ap + 16.7492 f + 0.528119 r + 1.19221AMPz + 0.937027 AMPy + 0.0128869
Vc × ap − 0.0190204Vc × f-7.89172ap × f + 0.841952ap × r1.64463ap × AMPz − 1.41237ap × AMPy (3)
− 4.99520 f × r + 0.589721 f × AMPz-4.94334 f × AMPy-0.538542r × AMPz-0.486322r × AMPy-0.988144
AMPz × AMPy
39.11) % and f with the lower contribution of 6.27 %. The term ap² is also
significant (14.97 %). The lack of fit P-value of 0.561 was greater than 0.05.
Therefore, the regression model does not fail to adequately explain the functional
relationship between the experimental factors and the response. Lack of fit may
occur if important terms from the model such as interactions or quadratic terms
are not included. The correlation coefficient R² of about 84.47 is considered as
good. The final equation in terms of real factors acceleration amplitude in radial
direction (AMPy) is given as follow:
AMPy = 1.5015 - 0.0224Vc + 7.6269 ap - 20.6830 f + 2.0932r - 0.0069Vc × ap + 0.0578Vc × f - 0.0024Vc × r (4)
- 13.9167ap × f + 0.7063 ap × r + 3.0625 f × r + 7.7846×10-5Vc2 - 27.0386 ap2 + 60.6823 f 2 - 0.9399r 2
It is well seen in Figure 2(a) that the use of small feed rate and large nose radius
results in the lowest surface roughness. Whereas, the use of large feed rate and
small tool nose radius results in highest surface roughness. Because at large feed
rate the distance between peaks and valleys of the feed marks is much more
important. The ANOVA revealed that cutting speed does not influence the surface
roughness. The depth of cut is found to be statistically significant. The effect of
tool nose radius and depth of cut on surface roughness is depicted in Figure 3(b).
We can observe that using a small depth of cut gives worst surface roughness
especially with tool nose radius of 0.8 mm. We could attribute this to the material
side flow phenomenon, due to temperature elevation and specific cutting pressure
in the cutting zone when using small undeformed chip thickness. It consists of
ploughing and squeezing of the machined material by the tool causing its
deterioration (Kishawy and Elbestawi 1999). Figures 3(a) and 3(b) show the
topography of machined surfaces using the lower feed rate and with both depth of
cut (0.05 and 0.25) mm respectively. We can observe that the valleys of surface
(a) are distant above those of surface (b) due to the material accumulation. This
428 I. Meddour, M.A. Yallese, and H. Aouici
additional volume of squeezed material presses on the tool flank face which can
explain the increase of acceleration amplitude in the radial direction (AMPy) at
small depth of cut (Figure 5(a)). The same trend is observed for the acceleration
amplitude in tangential direction (AMPz) (Figure 5(b)). This figure shows that the
tangential vibration (AMPz) increases with the increase in tool nose radius. Then,
it decreases by using tool nose radius of 1.6 mm. We can explain this by the
stabilization of cutting process when using large nose radius leading to the
improvement of surface roughness.
a) b)
Fig. 3 Topography of AISI 52100 steel hard turned with mixed ceramic insert
Investigation and Modeling of Surface Roughness of Hard Turned AISI 52100 Steel 429
a) b)
4 Models Confirmation
Vc ap f r
run exp Ra pred Ra error (%)
5 Conclusion
Through the results obtained above, the following conclusions could be drawn:
1. The best surface roughness is obtained by using small feed rate and large
nose radius. The use of depth of cut of 0.05 mm with tool nose radius of 0.8
mm gives the worst surface roughness due to the material side flow
phenomenon.
2. The ANOVA results show that depth of cut and cutting speed influence
considerably the tool vibrations on both tangential and radial direction,
whereas the feed rate has a significant effect only on acceleration amplitude
in radial direction (AMPy). It is also revealed that the tool nose radius has
430 I. Meddour, M.A. Yallese, and H. Aouici
References
Ozel, T., Karpat, Y., Figueira, L.: Predictive modeling of surface roughness and tool wear
in hard turning using regression and neural networks. Int. J. Mach. Tools Manuf. 45,
467–479 (2005)
Yallese, M.A., Rigal, J.-F., Chaoui, K., Boulanouar, L.: The effects of cutting conditions on
mixed ceramic and cubic boron nitride tool wear and on surface roughness during
machining of X200Cr12 steel (60 HRC). Proc. IMechE, Part B: J. Eng. Manuf. 219
(2005)
Aslan, E., Camuscu, N., Birgoren, B.: Design optimization of cutting parameters when
turning hardened AISI 4140 steel (63 HRC) with Al2O3 + TiCN mixed ceramic tool.
Mater. Design 28, 1618–1622 (2007)
Dureja, J.S., Gupta, V.K., Sharma, V.S., Dogra, M.: Design optimization of cutting
conditions and analysis of their effect on tool wear and surface roughness during hard
turning of AISI-H11 steel with a coated–mixed ceramic tool. J. Eng. Manuf. 223(B),
1441–1453 (2009)
Lalwani, D.I., Mehta, N.K., Jain, P.K.: Experimental investigations of cutting parameters
influence on cutting forces and surface roughness in finish hard turning of MDN250
steel. J. Mater. Process. Technol. 206, 167–179 (2008)
Bartarya, G., Choudhuryb, S.K.: Effect of cutting parameters on cutting force and surface
roughness during finish hard turning AISI52100 grade steel. Procedia CIRP 1, 651–656
(2012)
Horng, J.-T., Liu, N.-M., Chiang, K.-T.: Investigating the machinability evaluation of
Hadfield steel in the hard turning with Al2O3/TiC mixed ceramic tool based on the
response surface methodology. J. Mater. Process. Technol. (2008)
Saini, S., Ahuja, I.S., Sharma, V.S.: Influence of Cutting Parameters on Tool Wear and
Surface Roughness in Hard Turning of AISI H11 Tool Steel using Ceramic Tools. Int. J.
Precision Eng. Manuf. 13(8), 1295–1302 (2012)
Aouici, H., Yallese, M.A., Chaoui, K., Mabrouki, T., Rigal, J.F.: Analysis of surface
roughness and cutting force components in hard turning with CBN tool: prediction
model and cutting conditions optimization. Measurement 45, 344–353 (2012)
Tlusty, J.: Manuf. Process. Equipment. Prentice Hall (1999)
Thomas, M., Beauchamp, Y., Youssef, A.Y., Masounave, J.: Effect of tool vibration on
surface roughness during lathe dry turning process. Comput. Ind. Eng. 31(3-4), 637–644
(1996)
Jang, D.Y., Choi, Y.G., Kim, H.G., Hsiao, A.: Study of the correlation between surface
roughness and cutting vibrations to develop an on-line roughness measuring technique
in hard turning. Int. J. Mach. Tools Manuf. 36, 453–464 (1996)
Investigation and Modeling of Surface Roughness of Hard Turned AISI 52100 Steel 431
Risbood, K.A., Dixit, U.S., Sahasrabudhe, A.D.: Prediction of surface roughness and
dimensional deviation by measuring citting forces and vibrations in turning process. J.
Mater. Process. Technol. 132, 203–214 (2003)
Kirby, E.D., Zhang, Z., Chen, J.C.: Development of An Accelerometer based surface
roughness prediction system in turning operation using multiple regression techniques.
J. Ind. Technol. 4(20), 1–8 (2004)
Abouelatta, O.B., Madl, J.: Surface roughness prediction based on cutting parameters and
tool vibrations in turning operations. J. Mater. Process. Technol. 118, 269–277 (2001)
Hessainia, Z., Kribes, N., Yalles, M.A., Mabrouki, T., Ouelaa, N., Rigal, J.-F.: Turning
roughness model based on tool-nose displacements. Mechanika 19(1), 112–119 (2013)
ISSN 1392-1207
Kishawy, H.A., Elbestawi, M.A.: Effects of process parameters on material side flow
during hard turning. Int. J. Mach. Tools Manuf. 39, 1017–1030 (1999)
Finite Element Simulation of Fatigue Damage
Accumulation for Repaired Component by Cold
Expansion Method
1
Laboratoire LPQ3M, B.P. 305, Universite de Mascara, Algeria
1aid_abdelkrim@yahoo.com, 2bendoubamos@yahoo.fr
2
Département de Génie Mécanique, Université de Djilali Liabès, Sidi Bel Abèss, Algeria
benguediab_m@yahoo.fr
3
Laboratoire de Génie Civil et Géo-Environnement LGCgE, EA 4515, Faculté des Sciences
Appliquées FSA Béthune, Université d’Artois, France
abdelwaheb.amrouche@univ-artois.fr
Abstract. This manuscript investigates the effectiveness of applying the cold ex-
pansion process to extend the fatigue life of mechanical structures. During the
cold expansion process compressive residual stresses around the expanded hole
are generated. The enhancement of fatigue life and the crack initiation and growth
behavior of a holed specimen were investigated by using the 6082 Aluminum al-
loy. The present study suggests a simple technical method for enhancement of
fatigue life by a cold expansion hole of pre-cracked specimen. This technique pro-
duces beneficial high compressive residual stresses which have been predicted by
means of finite element models, both 3D for proper assessment of thickness ef-
fects. Finite element models have been developed to increase their complexity, Fa-
tigue damage accumulation of cold expanded hole in aluminum alloy which is
widely used in transportation and in aeronautics was analyzed. Experimental tests
were carried out using pre-cracked SENT specimens. Tests were performed in two
and four block loading under constant amplitude. These tests were performed by
using two and four blocks under uniaxial constant amplitude loading. The experi-
mental results were compared to the damage calculated by the Miner’s rule and a
new simple fatigue damage indicator based on an energy criterion. This compari-
son shows that the ‘energy criterion model’, which takes into account the loading
history, yields a good estimation according to the experimental results.
1 Introduction
Over the last 25 years, the cold expansion process has been commonly used to
improve the fatigue life of components containing fastener holes. Cold expansion
In this study, cold expansion is achieved by inserting an oversized rigid ball from
one side (entry face) of the holed plate and removing it from the other side (exit
face). The degree of cold expansion DCE is defined by the relation:
Finite Element Simulation of Fatigue Damage Accumulation 435
(D − d )
DCE % = × 100 (1)
d
A cold expansion of the hole introduces compressive residual stresses that are
beneficial in terms of lifetime until a critical degree of expansion where it
becomes not beneficial for structures. Both aspects are shown in Figure 1.
Fig. 1 The curve of crack initiation according to the numbers of the cycles
Where d is the diameter of the hole drilling and D is the diameter of the rigid
ball. Different technique can be used to repair a cracked component. The cold
working expansion process was realized by forcing a hard steel ball of 6 mm in-
side a predrilled hole (the initial diameters of the hole are: 5.9, 5.8, 5.75, 5.6 and
5.5 mm for aluminium alloy and 5.8, 5.75 for steel), this process is illustrated in
(Aid 2006) .
The material used for this study was aluminum alloy Al 6082 T6, some of the me-
chanical properties are given in (Aid 2006). A difference gap of 10% between the
characteristics of the uncracked specimens and the batch of specimens with hole
was noticed.
The specimens used for this investigation were conforms to ASTM standards
(ASTM 1997). The geometry of the fatigue test specimen cut in the longitudinal
direction (Aid 2006). For getting specimens with an expanded hole of 6 mm in di-
ameter we drilled a hole of Ø5.75 mm at the pre-crack tip and then a cold-working
expansion process was conducted by forcing a steel ball of 6 mm diameter.
The fatigue tests were carried out using a 100 kN capacity Instron hydraulic
machine. The loading frequency was 30 Hz and a stress ratio R of 0.57. During
436 A. Aid et al.
fatigue testing, a video camera with scale of 0.1 mm was used to determine the
crack initiation in the entry and exit faces of specimen.
Fig. 2 3D FEM mesh, with rigid ball for cold-expansion and fatigue critical section
The contact between the surfaces of the steel ball and the hole is simulated by
using the contact elements. Residual stress is obtained at the three surfaces (entry
face, middle and exit face) to analyze the influence of cold expansion direction on
the fatigue crack initiation life. During the simulation, the rigid ball was moved
through the hole incrementally.
The boundary conditions imposed are as following:
The displacements of the nodes located on the X–Y plane are constrained in the
z direction.
The displacements of the nodes located on the X–Z plane are constrained in the
y direction within a rectangular surface on the exit face.
The figure 3 illustrates the hoop stresses at the entry, middle and exit expansion
step. Portrays the finite element prediction of the tangential residual stress varia-
tion at different through thickness positions from the entrance face, this figure re-
veals the large variation in hoop residual stress in the region of influence.
Finite Element Simulation of Fatigue Damage Accumulation 437
Low hoop residual stresses on entrance and exit faces are caused by the shear
stress from the axial movement of the rigid ball. The compressive residual stresses
into the annular region around the hole reduce the origin of cracks. During fatigue
testing, this residual stresses act to change the effective stress intensity factor at
the crack tip, i.e. the crack growth rates are lower than those in absence of residual
stress. The residual stress profiles are in agreement with other studies presented in
previous literature (Poussard et al.1985; Pavier et al. 1999 ; De Matos et al. 2005).
Figure 4 shows the FE prediction of hoop residual stresses through the thickness
of the plate from the entrance face to the exit face. A significant difference in the
magnitude of tangential stress is recorded between the entry side and exit side.
The magnitude of hoop residual stresses on the entrance and the exit faces are
lower than the stresses on the mid-thickness. These results are in accordance with
recent research (Yan et al. 2012 ; Chakherlou 2012).
1
W = σε (2)
2
For uniaxial fatigue loading, authors (Bendouba et al. 2011; Djebli et al. 2013)
introduce a new damage parameter, Di, defined as the ratio of the increment of en-
ergy due to stress damage over the difference between the energy due to
ultimate stress and the applied stress. The damage indicator is defined by:
438 A. Aid et al.
200
150
Hoop residual stress (MPa)
100
50
0
-50 0 5 10 15 20
-100 Entry…
-150 Exit face
-200
-250
-300
Distance from hole edge (mm)
Fig. 4 Finite element predictions of residual hoop stress at entry face and exit face
W edi −W i
D= (3)
W u −W i
where:
Wedi : energy is due to stress damage.
Wi : energy is due to the applied stress.
Wu : energy is due to the ultimate stress of the material.
6 Fatigue Testes
The experimental conditions are given in (Aid 2006). Eight tests were carried out
for increasing loading conditions and as much for decreasing loading. To evaluate
the effect between these loading conditions, the Miner’s Rule was considered for
damage accumulation.
Endurance curves are shown in (Aid 2006). These are based on constant ampli-
tude test and the failure was considered at the crack initiation (Aid 2006).
Two and four cyclic stress levels were considered and two different sequences
were applied (Aid 2006). The aim of this set of tests is to determine the influence
of increasing or decreasing loading conditions on lifetime and to prove that the
proposed model takes into account the history of blocks loading and the nonlinear-
ity of the accumulated damage unlike the Miner’s rule.
Figure 5 demonstrates the comparison between the prediction models (Miner’
Rule, DSM model and the energy approach) with experimental results for different
loading conditions (two and four blocks with increasing and decreasing loads) for
specimens pre cracked and repaired by the technical process of cold expansion.
This figure confirms that the estimates results by the DSM model and the energy
approach used in this investigation are in good agreement with the experimental
results (Aid 2006).
Finite Element Simulation of Fatigue Damage Accumulation 439
Fig. 5 Comparison between theoretical results and experimental results for different loading
conditions
7 Conclusion
It seems that in the case of drilling with a cold expansion, the combination of
the geometrical and mechanical effect attributed to the stress concentration factor
associated with the compressive residual stresses predominate in the life time. In
the other hand the load sequence has a minor effect. In this investigation, the com-
pressive residual stresses at the edge of the hole are around of the yielding stress.
The effective local applied stress is lower than the residual stress; this observation
can explain the raison why there is no significant influence of the sequence load-
ing. Currently, we achieve tests with more important loading in order to evaluate
the sequence loading effect.
References
Lacarac, V.D., Smith, D.J., Pavier, M.J.: The effect of cold expansion on fatigue crack
growth from open holes at room and high temperature. Int. J. Fatigue 23, 161–170
(2001), doi:10.1016/S0142-1123(01)00125-6
Pasta, S.: Fatigue crack propagation from a cold-worked hole. Engng. Frac. Mech. 74,
1525–1538 (2007), doi:10.1016/j.engfracmech.2006.08.006
Chandawanich, N., Sharpe Jr., W.N.: An experimental study of fatigue crack initiation and
growth from cold worked holes. Engng. Frac. Mech. 11, 609–620 (1979),
doi:10.1016/0013-7944(79)90122-X
Su, X., Gu, M., Yan, M.: A simplified residual stress model for predict-ing fatigue crack
growth behavior at cold worked fastener holes. Fat. Frac. Engn. Mat. & Str. 9(1), 57–64
(1986), doi:10.1016/0013-7944(79)90122-X
Pavier, M.J., Poussard, C.G.C., Smith, D.J.: A finite element simulation of the cold work-
ing process for fastener holes. J. Strain Analysis 32, 287–300 (1997)
Poussard, C., Pavier, M.J., Smith, D.J.: Analytical and finite element pre-dictions of resi-
dual stresses in cold-worked fastener holes. J. Strain Analysis 30(4), 291–304 (1995)
Lacarac, V.D., Garcia-Granada, A.A., Smith, D.J., Pavier, M.J.: Prediction of the growth
rate for fatigue cracks emanating from cold expanded holes. Int. J. Fatigue 26(6), 585–
595 (2004), doi:10.1016/j.ijfatigue.2003.10.015
Bernard, M., Bui-Quoc, T., Burlat, M.: Effect of re-cold working on fatigue life enhance-
ment of a fastener hole. Fat. Fra. Eng Mat. Structures 18(17-18), 765–775 (1995),
doi:10.1111/j.1460-2695.1995.tb00902.x
Ghfiri, R., Shi, H.J., Guo, R., Mesmacque, G.: Effect of Expanded and non expanded Hole
on the Delay of Arresting Crack Propagation for Aluminum Alloys. Mat. Sci. Engineer-
ing A 286, 244–249 (2000), doi:10.1016/S0921-5093(00)00805-4
Semari, Z., Aid, A., Benhamena, A., Amrouche, A., Benguediab, M., Sadok, A., Bensed-
diq, N.: Effect of residual stresses induced by cold expansion on the crack growth in
6082 aluminum alloy. Eng. Fra. Mechanics 99, 159–168 (2013),
doi:10.1016/j.engfracmech.2012.12.003
Aid, A., Semari, Z., Benguediab, M.: Cold expansion effect on the fatigue crack growth of
Al 6082: numerical investigation. Structural Engineering and Mechanics 49(2), 225–235
(2014), doi:http://dx.doi.org/10.12989/sem.2014.49.2.225
Miner, M.A.: Cumulative damage in fatigue. J. App. Mech. 67, A159–A164 (1945)
Aid, A., Semari, Z., Amrouche, A., Mesmacque, G., Benguediab, M.: Non linear damage
cumulative criterion on fatigue under random loading. Alg. J. Adv. Materials (2006)
Finite Element Simulation of Fatigue Damage Accumulation 441
Aid, A., Amrouche, A., Bachir Bouiadjra, B., Benguediab, M., Mesmacque, G.: Fatigue life
prediction under variable loading based on a new damage model. Mat. Design 32,
183–191 (2011), doi:10.1016/j.matdes.2010.06.010
Amrouche, A., Su, M., Aid, A., Mesmacque, G.: Numerical study of the op-timum degree
of cold expansion: Application for the pre-cracked specimen with the expanded
hole at the crack tip. J. Mat. Pro. Technology 197, 250–254 (2008),
doi:10.1016/j.jmatprotec.2007.06.030
Aid, A.: Cumul d’endommagement en fatigue multiaxiale sous sollicitations variables.
[Thesis], p. 195. Universitè de Sidi-Bel Abbes, Algérie (2006)
Bendouba, M., Aid, A., Zengah, S., Benguediab, M.: Modèle énergétique de cumul du
dommage et prevision de la dureé de vie en fatigue sous chargements aléatoires. In: Pro-
ceedings of the Cong. Alg. Mécanique, Guelma, Algeria (2011)
Djebli, A., Aid, A., Bendouba, M., Amrouche, A., Benguediab, M., Benseddiq, N.:
A non-linear energy model of fatigue damage accumulation and its verification for
Al-2024 aluminum alloy. Int. J. Non-Linear Mechanics (2013), doi:10.1016
/j.ijnonlinmec.2013.01.007
ASTM E1049-85, Standard practices for cycle counting in fatigue analysis. Annual Book of
ASTM Standards, Philadelphia, vol. 03(01), 614–620 (1997, 1999)
Poussard, G.C., Pavier, M.J., Smith, D.J.: Analytical and finite element predictions of resi-
dual stresses in cold-worked fastener holes. J. Strain Anal. 32(4), 287–300 (1985)
Pavier, M.J., Poussard, G.C., Smith, D.J.: Effect of residual stress around cold worked
holes on fracture under superimposed mechanical load. J. Strain Anal. 63, 751–773
(1999)
de Matos, P.F.P., Moreira, P.M.G.P., Camanho, P.P., de Castro, P.M.S.T.: Reconstitution
of fatigue crack growth in Al-alloy 2024-T3 open-hole specimens using microfracto-
graphic techniques. Engng. Frac. Mech. 72, 2232–2246 (2005),
doi:10.1016/j.engfracmech.2005.02.005
Yan, W.Z., Wang, X.S., Gao, H.S., Yue, Z.F.: Effect of split sleeve cold ex-pansion on
cracking behaviors of titanium alloy TC4 holes. Eng. Fra. Mechanics 88, 79–89 (2012),
doi:10.1016/j.engfracmech.2012.04.008
Chakherlou, T.N., Shakouri, M., Aghdam, A.B., Akbari, A.: Effect of cold expansion and
bolt clamping on fretting fatigue behavior of Al 2024-T3 in double shear lap joints.
Mat., Design 33, 185–196 (2012), doi:10.1016/j.engfailanal.2012.04.008
Smith, K.N., Waston, P., Topper, T.H.: A stress–strain functions for the fatigue on mate-
rials. J. Mater. 5, 767–776 (1970)
Hoffman, H., Seeger, T.: Stress-strain analysis and life predictions of a notched shaft under
multiaxial loading. In: Leese, G.E., Socie, D. (eds.) Multi-axial Fatigue: Analysis and
Experiments AE-14, pp. 81–99. Society of Automative Engineers, Inc., Warrendale
(1989)
Bergman, J., Seeger, T.: On the influence of cyclic stress–strain curves, damage parameters
and various evaluation concepts on the life prediction by the local approach. In: Pro-
ceedings of the 2nd European Conference on Fracture, Darmstadt, Germany. VDI-
Report of Progress, vol. 18(6) (1979)
Geometrically Non-linear Free Vibration of
Fully Clamped Symmetrically Laminated
Composite Skew Plates
1
Université Mohammed V Agdal-Ecole Mohammadia D’ingénieurs, Avenue Ibn Sina,
B.P.765, Agdal, Rabat, Morocco
h.moulay1983@gmail.com, rbenamar@emi.ac.ma
2
FST B.P.2202- Route D’Imouzzer Fes, Morocco
harrasbilal@yahoo.fr
Abstract. The present work concerns the geometrically non-linear free vibration
of fully clamped symmetrically laminated composite skew plates. The theoretical
model based on Hamilton’s principle and spectral analysis previously applied to
obtain the non-linear mode shapes and resonance frequencies of thin straight
structures, such as beams, plates and shells is used. A convergence study has been
performed and has shown that 18 plate functions should be taken into account. Re-
sults are given for the linear and non-linear fundamental mode shape of fully
clamped symmetrically laminated composite skew plates, considering different
parameters such as the skew angle, the number of layers, the fiber orientation, the
vibration amplitudes and plate aspect ratio. It was found that the non-linear fre-
quencies increase with increasing the amplitude of vibration and increasing the
skew angle, which corresponds to the hardening type effect, expected in similar
cases, due to the membrane forces induced by the large vibration amplitudes.
1 Introduction
Composite materials are used in many industrial fields. Resulting composite ma-
terial present the best qualities of their constituents and some properties can be
improved by forming a composite material including reduction in structural
weight, reduction in fatigue and corrosion problems, that’s why it’s present inter-
ested use in aircraft structures and space vehicles. For the potential advantages
listed above, the fundamental carbon fiber reinforced plastic (CFRP) materials
have been used as structural members, taking the place of traditional metals. The
use of composite materials requires complex analytical methods in order to predict
accurately their response to external loading, especially in severe environments,
y η
θ b
x, ξ
a
Fig. 1 Skew plate in x-y and ξ-η co-ordinate system
For the classical plate laminated theory, the strain-displacement relationship for
large deflections are given by:
λ (1)
U W W
k λ
V W
ε ; k k ; λ λ W
(2)
Y
U V k W λ W W
σ Q ε (3)
N A B ε λ
(4)
M B D k
H
A ,B ,D H Q 1, z, z dz (5)
Here the are the reduced stiffness coefficients of the kth layer in the plate
co-ordinates. The transverse displacement function W may be written as in refer-
ence (Ribeiro and Petyt 1999) in the form of a double series:
T
W A W sinkωt (6)
W a w x, y sinωt (7)
Geometrically Non-linear Free Vibration 447
Here the usual summation convention for the repeated indexes i is used. i is
summed over the range 1 to n, with n being the number of basic functions consi-
dered. The expression for the bending strain energy Vb, axial strain energy Va and
kinetic energy T are given in reference (Harras 2001) in the rectangular co-
ordinate (x,y). The skew co-ordinates are related to the rectangular co-ordinate
(ξ,η) by: ξ=x-y tanθ ; η=y/cosθ. So, the strain energy due to bending Vb, axial
strain energy Va and kinetic energy T are given in the ξ-η co-ordinate system. In
the above expressions, the assumption of neglecting the in plane displacements U
and V in the energy expressions has been made as for the fully clamped rectangu-
lar isotropic plates analysis considered in reference (Benamar, Bennouna and
White 1993). Discretization of the strain and kinetic energy expressions can be
carried out leading to:
In which mij, kij and bijkl are the mass tensor, the rigidity tensor and the geome-
trical non-linearity tensor respectively. Non-dimensional formulation of the non-
linear vibration problem has been carried out as follows.
ξ η
w ξ, η Hw , Hw ξ , η (9)
ξ η
Where ξ and η are non-dimensional co-ordinates ξ and η one then
obtains:
H E H E
k k ;b b ;m ρH abm (10)
Where the non-dimensional tensors m*ij, k*ij and b*ijkl are given in terms of in-
tegrals of the non-dimensional basic function wi*, non-dimensionnal extensional
and bending stiffness coefficient A*ij and D*ij , skew angle θ and aspect ratio α.
Upon neglecting energy dissipation, the equation of motion derived from Ham-
ilton’s principle is:
δ V T 0 (11)
Where V=Va+Vb. Insertion of Equations 8 into Equation 11, and derivation
with respect to the unknown constants ai, leads to the following set of non-linear
algebraic equations:
2a k 3a a a b 2ω a m 0, r=1…n. (12)
/
ω (13)
D
ω ρ
ω (14)
To obtain the first non-linear mode shape of the skew plate considered, the con-
tribution of the first basic function is first fixed and the other basic functions con-
tributions are calculated via the numerical solutions of the remaining (n-1)
non-linear algebraic equations.
The aim of this section is to apply the theoretical model presented above to ana-
lyze the geometrical non-linear free dynamic response of skew fully clamped
symmetrically laminated plates in order to investigate the effect of non-linearity
on the non-linear resonance frequencies and non-linear fundamental mode shape
at large vibration amplitudes. Convergence studies are carried out, and the results
are compared with those available from the literature through a few examples of
laminated composite skew thin clamped plates with different fiber orientation and
aspect ratio. The material properties, used in the present analysis are: (1) Plate 1,
5-layers symmetrical angle-ply (45°,-45°, 45°,-45°, 45°); EL/ET = 40.0; GLT/ET =
0.6; νLT = 0.25. (2) Plate 2, 8-layers symmetrically laminated plate (90°,-
45°,45°,0°)sym ; EL=120.5 Gpa; ET=9.63 Gpa; GLT=3.58 Gpa; νLT = 0.32. (3) Plate
3, 16-layers symmetrically laminated plate (45°,-45°,0°,-45°,45°,-45°,0°,45°)sym ;
EL=173Gpa; ET=7.2 Gpa; GLT=3.76 Gpa; νLT = 0.29. Where E, G and ν are
Young’s modulus, shear modulus and Poisson’s ratio. Subscripts L and T
represent the longitudinal and transverse directions respectively with respect to the
fibers. All the layers are of equal thickness. Calculation was made by using 18
functions corresponding to three symmetric beam functions in the ξ direction and
three symmetric beam functions in the η direction, and three anti-symmetric beam
functions in the ξ direction and three anti-symmetric beam functions in the η di-
rection. Table 1 shows the linear results for a fully clamped laminated skew com-
posite plate 2 for skew angle θ=30° and aspect ratio α=0.6 obtained using only 18
well-chosen plate functions. It can be seen a good convergence with results pre-
sented in reference (Harras 2001). The variation of non-dimensional nonlinear
Geometrically Non-linear Free Vibration 449
Table 1 First linear mode shape of a fully clamped laminated skew composite plate 2
(Skew angle =30°, α=0.6): typical numerical results obtained with 18 basic functions
ωl* 9.4343
1 a11 1.05 10 a42 -2.591686E-02
2 a13 -1.549667E-02 11 a44 -2.517736E-03
3 a15 -1.665776E-03 12 a46 -3.854766E-04
4 a22 -1.689070E-01 13 a51 -1.109130E-02
5 a24 -4.314628E-03 14 a53 1.227799E-02
6 a26 -3.644569E-04 15 a55 1.692331E-03
7 a31 -2.185097E-02 16 a62 1.826523E-03
8 a33 2.017160E-02 17 a64 -3.078395E-03
9 a35 1.995894E-03 18 a66 -1.112399E-03
Table 2 Comparison of non-linear frequency ratio ωnl*/ ωl* of plate 2 for θ= 0°, α=2/3,
between results obtained in references (Han and Petyt 1997) and (Harras 2001) associated
with the fundamental non-linear mode shape and present results
Table 3 Convergence and accuracy of the nonlinear frequency ratios (ωnl/ωl) of clamped
5-layered angle-ply [45°/-45°/45°/−45°/45°] skew plate 1 (b/a = 1)
Skew Wmax/h
Angle θ°
0.2 0.4 0.6
0° Present 1.0074 1.0286 1.0641
(Singha, Daripa 2007) 1.0068 1.0271 1.0599
Error% 0.06% 0.14% 0.39%
15° Present 1.0077 1.0299 1.0675
(Singha, Daripa 2007) 1.0076 1.0300 1.0663
Error% 0.01% 0.01% 0.11%
30° Present 1.0086 1.0342 1.0741
(Singha, Daripa 2007) 1.0086 1.0340 1.0749
Error% 0.0016% 0.01% 0.08%
45° Present 1.0093 1.0379 1.0814
(Singha, Daripa 2007) 1.0098 1.0387 1.0849
Error% 0.0519% 0.08% 0.32%
450 H.M. Abdelali, B. Harras, and R. Benamar
Table 4 (continued)
Skew Wmax/h
Angle θ°
0.8 1 1.2
0° Present 1.1076 1.1609 1.2189
(Singha, Daripa 2007) 1.1041 1.1582 1.2214
Error% 0.32% 0.24% 0.20%
15° Present 1.1139 1.1710 1.2372
(Singha, Daripa 2007) 1.1149 1.1744 1.2431
Error% 0.10% 0.29% 0.48%
30° Present 1.1275 1.1944 1.2607
(Singha, Daripa 2007) 1.1297 1.1963 1.2729
Error% 0.19% 0.15% 0.96%
45° Present 1.1401 1.2096 1.2935
(Singha, Daripa 2007) 1.1458 1.2215 1.3076
Error% 0.50% 0.98% 1.08%
frequency ratio ωnl/ωl with respect to non dimensional maximum amplitude wmax/h
is evaluated for clamped square plate and is shown in Tables 2, along with pub-
lished results. It is observed that present results are in close agreement with those
of assumed hierarchical finite element method (Han and Petyt 1997) and Energy
balance method (Harras 2001). Next, the nonlinear free vibration behaviors of ful-
ly clamped angle-ply [45°/-45°/45°/-45°/45°] thin square and skew plates are ex-
amined in Table 3. It is observed that the results obtained here closely matches
with those from (Singha, Daripa 2007) for the case of symmetrically angle ply la-
minated composite skew plates. It is observed that the nonlinear frequency ratio
increases with the increase in amplitude of vibration, indicating hardening type of
nonlinear behavior. Furthermore, for the chosen amplitude, it can be noted that,
with the increase in skew angle, the degree of nonlinearity is high.
Fig. 2 Comparison of backbone curve of plate 1, 2, 3 and isotropic skew plate for θ=30°
Geometrically Non-linear Free Vibration 451
4 Conclusion
The theoretical model established and applied to beams, plates and shells ((Bena-
mar, Bennouna and White 1993) has been successfully applied to calculate the
first non-linear mode shape of fully clamped skew symmetrically laminated plates
for various aspect ratios. The present formulation has been verified with the avail-
able analytical and finite element results. Limited numerical studies are conducted
to examine the effect of skew angle, fiber orientation, and aspect ratio on the
large-amplitude frequency of composite skew plates. As has been shown in this
work, the fundamental non-linear mode shape of the fully clamped skew symmetr-
ically laminated plate can be obtained with sufficient accuracy by using 18 basic
plate functions. This provides a more accurate description of the non-linear mode
shape, compared with previous results based on the use of nine symmetric basic
plate functions, since it allows the non-symmetry induced by the fiber orientation
to be taken into account. Present study reveals hardening type of nonlinearity and
the nonlinear frequency ratio in general increases with the increase in skew angle.
Further work is needed to investigate the behaviour of higher modes, the effects of
fiber orientation, and forced response of fully clamped symmetrically laminated
skew plates.
References
Leissa, A.W.: Vibration of Plates. US Government Printing Office, NASA SP-160, Wash-
ington, DC (1969)
Benamar, R.: Non-linear dynamic behaviour of fully clamped beams and rectangular iso-
tropic and laminated plates. Ph.D. Thesis, Institute of Sound and Vibration Research
(1990)
Chia, C.Y.: Geometrically nonlinear behavior of composite plates: a review. Applied Me-
chanical Reviews 41, 439–451 (1988)
Han, W., Petyt, M.: Linear vibration analysis of laminated rectangular plates using the hie-
rarchical finite element method-I: free vibration analysis. Computers and Structures 61,
705–712 (1996)
Ribeiro, P., Petyt, M.: Multi-modal geometrical non-linear free vibration of fully clamped
composite laminated plates. Journal of Sound and Vibration 225, 127–152 (1999)
Harras, B.: Theoretical and Experimental Investigation of the non-Linear Behaviour of var-
ious Fully Clamped Rectangular Composite Panels at large Vibration Amplitudes. Ph.D.
thesis, University Mohammed V-Agdal (2001)
452 H.M. Abdelali, B. Harras, and R. Benamar
Harras, B., Benamar, R., White, R.G.: Investigation of non-linear free vibrations of fully
clamped symmetrically laminated carbon-fibre-reinforced PEEK (AS4/APC2) rectangu-
lar composite panels. Composites Sciences and Technology 62, 719–727 (2002)
Han, W., Dickinson, S.M.: Free vibration of symmetrically laminated skew plates. Journal
of Sound and Vibration 208, 367–390 (1997)
Malekzadeh, P.: A differential quadrature nonlinear free vibration analysis of laminated
composite skew thin plates. Journal of Thin-Walled Structures 45, 237–250 (2007)
Singha, M.K., Daripa, R.: Nonlinear vibration of symmetrically laminated composite skew
plates by finite element method. Int. J. Non-linear Mech. 42, 1144–1152 (2007)
Ray, A.K., Banerjee, B., Bhattacharjee, B.: Large amplitude free vibrations of skew plates
including transverse shear deformation and rotary inertia-a new approach. J. Sound
Vib. 180, 669–681 (1995)
Upadhyay, A.K., Shukla, K.K.: Large deformation flexural behaviour of laminated com-
posite skew plates: An analytical approach. Composite Structures 94, 3722–3735 (2012)
Benamar, R., Bennouna, M.M.K., White, R.G.: The effects of Large Vibration Amplitudes
on the Fundamental Mode Shape of a Fully clamped, Symmetrically laminated rectan-
gular plate. In: Fourth International Conference of Recent Advances in Structural Dy-
namics, pp. 749–760. Elsevier Applied Science (July 1990)
Mallick, P.K.: Fibre-reinforced composites: Materials, Manufacturing and Design, vol. 152.
Marcel Dekker, Inc., New York (1993); 2nd edn. revised and expanded
Benamar, R., Bennouna, M.M.K., White, R.G.: The effects of large vibration amplitudes on
the fundamental mode shape of thin elastic structures. Part II fully clamped rectangular
isotropic plates. Journal of Sound and Vibration 164, 295–316 (1993)
Han, W., Petyt, M.: Geometrically Nonlinear vibration analysis of thin, rectangular plates
using the hierarchical finite element method-II: 1st mode of laminated plates and higher
modes of isotropic and laminated plates. Computers and Structures 63, 309–318 (1997)
Timoshenko, S., Weinsowsky-Krieger, S., Jones, R.M.: Mechanics of Composite Materials.
International Student Edition, vol. 51. McGraw-Hill Kogakusha, Ltd., Tokyo (1975)
Scratch Adhesion Characteristics of PVD
Cr/CrN Deposited on High Speed Steel and
Silicon Substrates
1 Introduction
Today PVD thin films are widely used in order to increase the machining
performance because PVD coated tools have significantly contributed to higher
production rates and higher tool life's. Several parameters can be controlled to
improve mechanical behavior of thin films. The substrate material is one of the
significant parameters of technology controlling the mechanical properties of PVD
coatings. In order to develop new PVD coatings, it is important to characterize the
mechanical properties of the coating/substrate system and the deformation
mechanisms in a tribological contact. Several methods are available to examine
the efficiency of this system [1]. But the study of adhesion is still essential to
examine coated systems interface. In this context scratch test is probably the most
common and used test to understand coating failure modes of a hard coating
substrate system [2]. During the conventional scratch test a diamond indenter
Rockwell C flattened the surface of the coated material under an increasing
normal load until failure occurs or the damage of the film is noted. The
identification the coating failure mechanisms can be made by the following terms:
LC1, the lower critical load, which is defined as the load where the first cracks
occurred (cohesive failure), LC2, the upper critical load at which the first
delaminating [3] and LC3 the critical load when the damage of the film exceeds
50%. Critical normal loads can be controlled by the coating thickness, coating
hardness and substrate hardness .The present study focuses on the adhesion of Cr
single layer and CrN multilayer deposited on 100C6 Steel substrate and Silicon
substrate by magnetron sputtering system.
2 Experimental Details
Substrates were used in this study is 100C6 steel (AISI 52100) and Si. Cr/Si single
layer, CrN/Cr/Si multilayer and CrN/Cr/100C6 coatings were deposited with a
magnetron sputter unit with a pure chromium target. Thicknesses of thin films
studied were illustrated in table 1. The morphological and microstructural
characterizations were evaluated using an atomic force microscopy (AFM) and a
Scanning electron microscopy (SEM). The scratching response of the coating–
substrate composites were evaluated using the CSM Instruments with Rockwell C
diamond styli with radius 100μm. A normal load range of 3–28N, a scratch length
of 3 mm and a scratching velocity of 6mm/min were used in the experiments.
Hardness of 100C6 steel and Silicon substrate are respectively 11GPa and 8GPa.
Young's modules of both substrates are respectively 200 GPa and 130 GPa.
Fig. 1 Three-dimensional AFM images of Cr/Si single layer, CrN/Cr/Si and CrN/Cr/100C6
multilayer
Table 2 Critical loads and friction coefficients of CrN/Cr/100Cr6, CrN/Cr/Si and Cr/Si
coatings
Coatings CrN/Cr/100C6 CrN/Cr/Si Cr/Si
LC1 5,68 3,23 4,27
LC2 12,79 7,42 9,16
LC3 13,54 12,34 13,16
Figures 3-5 shows the tracks were corresponding to the Cr/Si, CrN/Cr/Si and
CrN/Cr/100C6 coating after scratch test.
Under the same loading conditions, the coatings behavior is different. In the
single layer structure, with the increase of the scratch load, cracks have been
developed throughout the cross section of the film and have spread rapidly. In this
type of coating, adhesive failure occurred for low load comparing with the other
coatings and we observed a total chipping of the film and exposure of substrate at
higher load.
The critical load is the normal load at the first coating failure as detected. This
load was determined with combining observations of scratches by optical
microscopy with measurements of acoustic emission, of normal and tangential
forces and of the residual depth during scratch testing.
Several authors [5,6] have linked the appearance of the first crack (cohesive
failure) at a first critical load noted by LC1 and the upper critical load at which the
first delaminating at the edge of the scratch track occurred (adhesion failure) by LC2.
The critical loads LC1, LC2 and LC3 values were presented in figure 3-5. Higher
critical loads of adhesion were obtained for CrN/Cr/100Cr6. In multilayer
structure coatings, cracks develop mostly in the vicinity of the upper surface, and
the layers interfaces can change the initial crack orientation.
4 Conclusion
In this work, PVD thin films were deposited on different substrate materials,
100C6 (AISI 52100) steel and Si by PVD magnetron sputtering system. The mor-
phological and microstructural proprieties were evaluated using an atomic force
microscopy (AFM) and a Scanning electron microscopy (SEM). Adhesion quality
was studied with scratch tests. Damages mechanisms that were occurred during
scratching were further discussed. The AFM examination revealed the presence of
domes and craters which are uniformly distributed over the surface. CrN/Cr/Si has
a less rough surface because of a small grains size. SEM cross section of all
coatings showed a columnar structure
458 K. Khlifi, H. Dhifelaoui, and A. Ben Cheikh Larbi
The results obtained in the present study show that the coating/substrate
damage mechanisms during scratching and the critical normal loads obtained are
strongly dependent on the mechanical properties of substrate material. The results
obtained in the present study show that the coating/substrate damage mechanisms
during scratching and the critical normal loads obtained are strongly dependent on
the mechanical properties of substrate material.
References
[1] Sveen, S., Andersson, J.M., M’Saoubi, R., Olsson, M.: Scratch adhesion characteristics
of PVD TiAlN deposited on high speed steel, cemented carbideand PCBN substrates.
Wear 308, 133–141 (2013)
[2] Holmberg, K., Laukkanen, A., Ronkainen, H., Wallin, K., Varjus, S.: A model for
stresses, crack generation and fracture toughness calculation in scratched TiN- coated
steel surfaces. Wear 254, 278–291 (2003)
[3] Khlifi, K., Ben Cheikh Larbi, A.: Investigation of the adhesion of PVD coatings using
various approaches. Surface Engineering 29, 555–560 (2013)
[4] Thornton, J.A.: J. Vac. Sci. Technol. 11, 666–670 (1974)
[5] Yıldız, F., Alsaran, A., Celik, A., Efeoglu, I.: Surf. Eng. 26(6), 578–583 (2010)
[6] Kataria, S., Kumar, N., Dash, S., Ramaseshan, R., Tyagi, A.K.: Surf. Coat.
Technol. 205, 922–927 (2010)
Load Sharing Behavior in Planetary Gear Set
1
Laboratory of Mechanics, Modeling and Production (LA2MP),
National School of Engineers of sfax, BP1173 – 3038 – Sfax – Tunisia
ahmed_hammami@voila.fr, fakher.chaari@gmail.com,
mohamed.haddar@enis.rnu.tn
2
Department of Structural and Mechanical Engineering,
Faculty of Industrial and Telecommunications Engineering, University of Cantabria,
Avda de los Castros s/n 39005 – Santander - Spain
miguel.iglesias@unican.es, alfonso.fernandez@unican.es,
fernando.viadero@unican.es
Abstract. The objective of this paper is to study the effects of meshing phase be-
tween planets, the effects of the gravity of carrier and the planet position error on
the load sharing behavior in planetary gear set. These effects will be studied nu-
merically under the stationary condition and will be validated experimentally by a
back-to-back planetary gear test bench. In this test bench, strain gauges are in-
stalled on each planet’s pin hole in order to compare strains in the pinhole of each
planet.
Keywords: Planetary gear, load sharing, meshing phase, gravity, planet position
error.
1 Introduction
Planetary gear can transmit higher power because they use multiple power paths
formed by each planet branches. This allows the input torque to be divided be-
tween the n planet paths, reducing the force transmitted by each gear mesh.
Under ideal conditions, each planet path carries an equal load. Nevertheless,
planetary gears have inevitable manufacturing and assembly errors. So, the load is
not equally shared amongst the different sun-planet and planet-ring paths, which
can be a problem in terms of both dynamic behavior and durability.
Many researchers have done significant works on the subject of planetary gear
load sharing. Their works were based on transmission modeling and assessed by
experimental tests. Kahraman (Kahraman, 1994) used a discrete model to study
the influence of carrier pin hole and planet run-out errors on planet load-sharing
characteristics of a four-planet system under dynamic conditions. He employed
later (Kahraman, 1999) a planet load-sharing model to determine the static planet
load sharing of four-planet systems and presented experimental data for validation
of the model predictions. Iglesias (Iglesias et al, 2013) studied the effect of planet
position error on the load sharing and transmission error. Singh (Singh, 2005)
found that the tangential pin position error has a greater effect on the load sharing
than the radial pin position error. Ligata (Ligata et al, 2008) proved experimentally
that for the same amount of error, the degree of inequality in the planet load-
sharing behavior increases with the number of planets in the system. Guo and Kel-
ler (Guo and Keller 2012) presented a three-dimensional dynamic model which
take into account to the addressing gravity, bending moments, fluctuating mesh
stiffness, nonlinear tooth contact, and bearing clearance. They validated this mod-
el against the experimental data.
In this paper, the effects of meshing phase, the gravity and the error position of
the pin hole of planets on the load sharing characteristics are studied numerically
and validated experimentally with back-to-back planetary gear.
The test bench is composed of two identical planetary gear sets with the same gear
ratio (Fig. 1). The first planetary gear is a “test gear set” and the second planetary
gear is a “reaction gear set” having a free ring. An arm is fixed on this ring and al-
lowing the introduction of external load ((Hammami et al 2015) and (Hammami
et al 2014a)).
The two planetary gears are connected back-to-back in order to provide a me-
chanical power circulation: the sun gears of both planetary gear sets are connected
through a common shaft and the carriers of both planetary gear sets are connected
to each other through a rigid hollow shaft (Hammami et al 2014b).
In order to compare the load sharing between the tests planets, three strain
gauges are installed in the pin holes of each planet in the tangential direction of
the test carrier (Fig. 2).
Fig. 2 Strain gauges mounted in each pin hole of the test carrier planets
The wires from the strain gauges are connected to the acquisition system
through a hollow slip ring (HBM SK5/95) which is installed with the hollow shaft
that connects the carriers (Fig.3).
Strains gauges are used in quarter bridge configuration.
3 Numerical Results
The effect of meshing phase, the gravity of planets carrier and the position error of
planets on the load sharing behavior is studied in this part.
First of all, we define the planet load sharing ratio (LSR) as the ratio of the
meshing torque due to sun-planet(i) and ring-planet(i) meshes of planet (i) by the
meshing torque of all planets.
Tmesh ( Pi )
L Pi = n
(1)
∑ Tmesh ( Pi )
i =1
Z j is the number of tooth of the gear (j). ψ i is the angle position of the planet
(i). n and m are integer.
The gear meshes stiffness of ring-planets gear pairs (Fig. 4) and sun-planets
gear pairs are modeled using the finite element model (Fernandez Del Rincon et al
2013).
The dynamic response is computed according to the procedure given in (Ka-
hraman, 1994) and the load sharing ratio is computed according to equation (1).
Fig.5 shows the planet load sharing ratio for the nominal position of planets (fault-
less system) for 100 N.m of input torque and for speed of motor 165 rpm.
Load Sharing Behavior in Planetary Gear Set 463
8
x 10
7.5
ring-planet1
ring-planet2
7 ring-planet1
5.5
4.5
4
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time (s)
0.355
Planet1
0.35 Planet2
Planet3
0.345
0.34
0.335
LSR
0.33
0.325
0.32
0.315
0.31
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Time (s)
Fig. 5 Planet load sharing ratio for the nominal position of planets
In this case, the LSR for all planets fluctuates slightly around the 1/N value
(N=3: number of planets) because the number of tooth in contact changes. The
fluctuation of LSR of each planet is the same with a phase shift of 2π/3. This
phase is induced by the fact that gear mesh sun-planets and ring-planets are
sequentially phased.
- The first subperiod : where the planet 1 is in the upper position. The values of
gear mesh stiffness are low.
- The second subperiod where the planet 1 is in an intermediate position in (the
left). The values of gear mesh stiffness are medium.
- The third subperiod where the planet 1 is in a lower position. The values of
gear mesh stiffness are high.
- The fourth subperiod where the planet 1 is again in an intermediate position
(the right). The values of gear mesh stiffness are medium.
8
x 10
7
ring-planet1
6.5
Gear mesh stiffness (N/m)
5.5
4.5
4
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Time (s)
0.5
Planet1
Planet2
0.45 3 2 1 3 Planet3
0.4
LSR
0.35
0.3
0.25
The gravity of carrier will lead to a periodic fluctuation of the maximum ampli-
tude of gear mesh stiffness with a period corresponding to that of the rotation of
the carrier Tc. The higher values of the gear mesh stiffness of ring-planets are suc-
cessively on the planet 3 then planet 2 then planet 1.
Load Sharing Behavior in Planetary Gear Set 465
Fig.7 shows the evolution of LSR during one period of rotation of carrier. This
period is divided into three sub-periods of Tc/3. For each sub-period, the LSR of
each planet does not oscillate with the same amplitude; there is always a planet
that will include the maximum (planet1then planet2 then planet3).
Keff is the cumulative stiffness due to meshing stiffness of the contact at the
sun–planet Kps and planet–ring Kpr, and the planet bearing stiffness Kb.
Keff is defined as (Ligata et al, 2009):
1 1 1
= + (5)
K eff K b K ps + K pr
In our case, the planet 1 has an error “e1=62μm” and the planet 2 has an error
“e2=-60μm”. The LSR in this case is represented in Fig.8.
0.7
0.6
Planet 1
0.5 Planet 2
Planet 3
0.4
LSR
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Time (s)
The position errors of planets 1 and 2 have an important effect in the LSR. Pla-
net 1 which has a positive error anticipates the contact, being preloaded before
planets 2 and 3 begin to transmit load; whereas planet 3 which has a negative error
is preloaded after planet 2 and 1.This defects are in tangential direction and they
have an effect more important in the LSR than the effect of gravity of carrier
which is a variable radial error (Bodas and Kahraman 2004).
466 A. Hammami et al.
170.00 1.00
F Time Planet1
F Time Planet2
F Time Planet3
Amplitude
Amplitude
uE
10.00 0.00
0.00 s 1.84
Fig. 9 Variation of the measured planets strains for the speed motor 165 rpm
1
Planet1(Experiments)
0.9 Planet2(Experiments)
Planet3(Experiments)
0.8 Planet1(Numerical)
Planet2(Numerical)
0.7 Planet3(Numerical)
0.6
LSR
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Time (s)
The computational results of the load sharing ratio are compared to the meas-
ured LSR (Fig. 10). In general, the calculated load sharing agrees with the meas-
ured data. In fact, the planet position errors have an important effect. In addition, a
slight deviation of the load sharing evolution due to the effect of gravity of carrier
is observed on the numerical and experimental results whereas the effect of mesh-
ing phase is observed only in the numerical results because the recorded signals
presents noise due to the contact between brushes and slip ring. So, a signal
processing was necessary.
Load Sharing Behavior in Planetary Gear Set 467
5 Conclusion
In this paper, the planets load sharing behaviour is studied and correlated with
those obtained experimentally using strain gages on planets of a back-to-back
planetary gear test bench.
The effect that planets are sequentially phased is included to the model and as
results; the load sharing ratio for all planets fluctuates slightly around the 1/N val-
ue. The time evolution of LSR of each planet bends a little bit every Tc/3 by the
effect of gravity of carrier. However, the planets position errors have an important
effect on the LSR of each planet. The final numerical results with the three effects
agree with the measured strains on planets.
Future investigation will be focused mainly on the planets load sharing under
non-stationary condition (run-up and run-down regimes).
Acknowledgements. This paper was financially supported by the Tunisian-Spanish Joint
Project N° A1/037038/11.
The authors would like also to acknowledge project “Development of methodologies for
the simulation and improvement of the dynamic behavior of planetary transmissions
DPI2013-44860” funded by the Spanish Ministry of Science and Technology.
Acknowledgment to the University of Cantabria cooperation project for doctoral train-
ing of University of Sfax’s students.
References
Bodas, A., Kahraman, A.: Influence of carrier and manufacturing errors on the static lad
sharing behavior of planetary gear sets. Bulletin of the Japan Society of Mechanical En-
gineers 47(3), 908–915 (2004)
Fernandez del Rincon, A., Viadero, F., Iglesias, M., García, P., de-Juan, A., Sancibrian, R.:
A model for the study of meshing stiffness in spur gear transmissions. Mechanism and
Machine Theory 61(2013), 30–58 (2013)
Guo, Y., Keller, J.: Combined effects of gravity, bending moment, bearing clearance, and
input torque on wind turbine planetary gear load sharing. American Gear Manufacturers
Association technical paper, 12FTM05 (2012)
Hammami, A., Fernández, A., Viadero, F., Chaari, F., Haddar, M.: Modal analysis of back-
to-back planetary gear: experiments and correlation against parameter model. Journal of
Theoretical and Accepted Applied Mechanics 53(1) (2015)
Hammami, A., Fernández, A., Viadero, F., Chaari, F., Haddar, M.: Dynamic behaviour of
back-to-back planetary gear in run up and run down transient regimes. Journal of Me-
chanics (2014a), doi:10.1017/jmech.2014.95
Hammami, A., Fernández, A., Chaari, F., Viadero, F., Haddar, M.: Dynamic behaviour of
two stages planetary gearbox in non-stationary operations. Mechatronic System: Theory
and Application, 23–35 (2014b)
Iglesias, M., Fernandez, A., De-Juan, A., Sancibrian, R., Garcia, P.: Planet position errors
in planetary transmission: Effect on load sharing and transmission error. Frontiers Me-
chanical Engineering 8(1), 80–87 (2013)
Kahraman, A.: Load Sharing Characteristics of Planetary Transmissions. Mechanism and
Machine Theory 29, 1151–1165 (1994)
468 A. Hammami et al.
Kahraman, A.: Static Load Sharing Characteristics of Transmission Planetary Gear Sets:
Model and Experiment. SAE Paper No. 1999–01–1050 (1999)
Ligata, H., Kahraman, A., Singh, A.: An experimental study of the influence of manufactur-
ing errors on the planetary gear stresses and planet load sharing. Journal of Mechanical
Design 130 (2008)
Ligata, H., Kahraman, A., Singh, A.: A closed-form planet load sharing formulation for
planetary gear sets using a translational analogy. Journal of Mechanical Design 131,
021007 (2009)
Singh, A.: Application of a system level model to study the planetary load sharing behavior.
Journal of Mechanical Design 127, 469–476 (2005)
Singh, A.: Load sharing behavior in epicyclic gears: Physical explanation and generalized
formulation. Mechanism and Machine Theory 45, 511–530 (2010)
Low Velocity Impact Behavior of Glass Fibre-
Reinforced Polyamide
1
Mechanical Modelisation and Manufacturing Laboratory (LA2MP),
National Engineering School of Sfax, B.P W3038, Sfax, University of Sfax, Tunisia
{Jamelmars,mondherwali}@yahoo.fr, Fakhreddine.Dammak@enis.rnu.tn
2
Laboratory of Industrial and Human Automation control (LAMIH / C2S),
University of Valenciennes
remi.delille@univ-valenciennes.fr
1 Introduction
2 Experimental
2.1 Materials
Two commercial polyamide 66 with 30 wt.% of short glass fibre (HERAMID A
NAT with 30 wt.% glass fibre, provided by RADICI GROUP) and natural polya-
mide 66 (A HERAMID NAT) were used to produce molded glass fibre-reinforced
composites with 00, 10, 20 and 30 wt.% .They were designed as PA66_00,
PA66_10, PA66_20 and PA66_30, respectively. The specimens were made by
injection molding (Figure 1).
Low Velocity Impact Behavior of Glass Fibre-Reinforced Polyamide 4771
Fig. 3 Shape and dimensionss of the specimen, according to ISO 527-2 standard
472 J. Mars et aal.
50
40
30
Stress (MPa)
Test 1
20 Test 2
10 Test 3
0
0 10 20 30 40
Strain %
120
PA66_00
100 PA66_10
PA66_20
80
PA66_30
Stress (MPa)
60
40
20
0
0 10 20 30 40
Strain %
Fig. 7 Uniaxial stress-strain curves of PA66_00, PA66_10 , PA66_20 and PA66_30
474 J. Mars et al.
where σ e is the yield stress, Q, β and K are the mechanical parameters of the elas-
to-plastic model and ε p is the plastic deformation.
Finally, the comparison of the numerical approach with the experimental data
shows the accuracy of the mechanical elasto-plastic model. Figure 8 shows that
the numerical approach is in good argument with the experimental result for
PA66_20 material. This approached law is used in next section when studying the
impact behavior of glass fibre-reinforced polyamide.
Table 2 shows the values of the mechanical parameters for the glass fibre-
reinforced polyamide during the identification phase.
80
60
Stress (MPa)
40 Experience
20 Num Approach
0
0 0.02 Strain0.04 0.06
Proj
Rp
h V0
The presence of the rotational symmetry axis, can simplify the three-
dimensional impact problem in an axisymmetric problem. The problem can thus
be analysed using 2D axisymmetric model. The plate was meshed using the struc-
tured meshing technique and 6000 4-nodes per elements with reduced integration
(CAX4R in ABAQUS). The impactor mesh consisted of 433 4-nodes per elements
(CAX4R in ABAQUS).
476 J. Mars et al.
Given that the impact model needs reliable data for simulation, we need to va-
lidate the mechanical properties and the contact law introduced into numerical
model. For that, a comparative study between the numerical predictions and expe-
rimental response found in literature is established. The plate material is assumed
to be an isotropic and homogeneous, and the elastic, perfectly plastic material and
the Von Mises yield criterion, J2, is employed same as in the work of (Chen et al.
2007). In this section, the deformable impactor is assumed to be elastic. For the
validation, the mechanical properties of both impactor and plate are illustrated in
Table 2.
The impactor velocity predicted by the present numerical model and the result
obtained by experimental data tested through LDA technique (Chen et al. 2007)
are plotted in Figure 10. One can note from this figure that the impactor velocity is
in good agreement with the experimental test.
40
30
Velocity (m/s)
20
10
Fig. 10 Validation: impactor velocity time responses (mp = 54.4 g, Rp=6.35mm, V0 = 29.9 m/s)
h = 4mm whose density, Young’s modulus, Poisson ratio, yield stress and σe are
already mentioned in the previous paragraph. The impactor has the same characte-
ristics described above except that the mass and speed become respectively 1.369
Kg and V0 = 3.418 m / s.
The evolution of plate center displacement according to the time is presented in
Figure 11 for incidental impact energy of 8J for PA66_00, PA66_10, PA66_20
and PA66_30.
Time (ms)
0 2 4 6 8 10 12 14
0.000
-0.002
Displacement (m)
-0.004
-0.006
PA66_00
-0.008
PA66_10
-0.010 PA66_20
-0.012 PA66_30
3
PA66_00
PA66_10
Force MAX (KN)
2 PA66_20
PA66_30
0
0 2 4 6 8 10 12
Time (ms)
Fig. 12 Impact force versus time curves for an impact energy of 8J
4 Conclusion
References
Akay, M., O’Regan, D.F., Bailey, R.S.: Fracture Toughness and impact behaviour of glass-
fibre-reinforced polyamide 6,6 injection mouldings. Composites Science and Technolo-
gy 55, 109–118 (1995)
Benaceur, I., Othman, R., Guegan, P., Dhieb, A., Damek, F.: Sensitivity of the flow stress
of Nylon 6 and Nylon 66 to strain-rate. International Journal of Modern Physics B 22,
1249–1254 (2008)
Cantwell, W.J., Morton, J.: The impact resistance of composite materials-a review. Compo-
sites 22, 347–362 (1991)
Dhakal, H.N., Zhang, Z.Y., Bennett, N., Reis, P.N.B.: The low-velocity impact response of
non-woven hemp fibre reinforced unsaturated polyester composites: Influence of impac-
tor geometry and impact velocity. Composite Structures 94, 2756–2763 (2012)
Chen, L.B., Xi, F., Yang, J.L.: Elastic-plastic contact force history and response characte-
ristics of circular plate subjected to impact by a projectile. Acta Mech. Sin. 23, 415–425
(2007)
Ghorbel, A., Saintier, N., Dhiab, A., Dammak, F.: Etude du comportement mécanique d’un
polyamide 66 chargé de fibres de verre courtes. Mécanique & Industries 12, 333–342
(2011)
Minak, G., Kostic, S.C., Šoškic, Z., Radovic, N.: Equivalent stiffness as measure of low ve-
locity impact damage of complex composite structures. Facta Universitatis: Mechanical
Engineering 10, 91–104 (2012)
Mouti, Z., Westwood, K., Kayvantash, K., Njuguna, J.: Low Velocity Impact Behavior of
Glass Filled Fiber-Reinforced Thermoplastic Engine Components. Materials 3,
2463–2473 (2010)
Low Velocity Impact Behavior of Glass Fibre-Reinforced Polyamide 479
Mouti, Z., Westwood, K., Long, D., Njuguna, J.: An experimental investigation into loca-
lised low-velocity impact loading on glass fibre-reinforced polyamide automotive prod-
uct. Composite Structures 104, 43–53 (2013)
Simeoli, G., Acierno, D., Meola, C., Sorrentino, L., Iannace, S., Russo, P.: The role of in-
terface strength on the low velocity impact behaviour of PP/glass fibre laminates. Com-
posites 62, 88–96 (2014)
Wali, M., Abdennadher, M., Fakhfakh, T., Haddar, M.: Dynamic analysis of an elasto-
plastic sandwich subjected to low velocity impact. Multidiscipline Modeling in Mate-
rials and Structures 7, 184–206 (2011)
Wali, M., Abdennadher, M., Fakhfakh, T., Haddar, M.: Vibration response of sandwich
plate under Low-velocity impact loading. WSEAS Transactions on Applied and Theo-
retical Mechanics 6, 27–36 (2011)
Xu, J., Li, Y., Ge, D., Liu, B., Zhu, M.: Experimental investigation on constitutive behavior
of PVB under impact loading. International Journal of Impact Engineering 38, 106–114
(2011)
Yang, L., Yan, Y., Kuang, N.: Experimental and numerical investigation of aramid fibre
reinforced laminates subjected to low velocity impact. Polymer Testing 32, 1163–1173
(2013)
Burst Pressure Estimation of Corroded Pipeline
Using Damage Mechanics
Djebbara Benzerga
Department of Mines and Metallurgy, LSCMI, Faculty of Mechanical, University of Science and
Technology of Oran Mohamed Boudiaf, B.P. 1505 El M’naouer 31000 Oran, Algeria
djeb_ben@yahoo.fr
1 Introduction
In most applications, the damage is very localized in such a way that the damaged
material occupies a volume small in comparison to the macroscale of the structural
component and even to the mesoscale of the representative volume element RVE.
This is due to the high sensitivity of damage to stress concentrations at the ma-
croscale and to defects at the microscale. This allows us to consider that the effect
of the damage on the state of stress and strain occurs only in very small damaged
regions. In other words, the coupling between damage and strains may be neg-
lected everywhere in the structure except in the RVE(s) where the damage devel-
ops. This is the principle of the locally coupled analysis [Lemaitre and Doghri]
where the procedure may be split into the following two steps as shown in Fig. 1:
upon Damage Mechanics, which currently has reached a stage of maturity enabling
it to model any type of degradation. Using the programmable ANSYS, a subroutine
was developed and implemented in the main code for the determination of the
critical point (M*) in the vicinity of corrosion defect where potential crack may
occurs. In a second step, a post-processor based on the iterative Newton method was
applied to this critical point (M*) for the determination of the maximum internal
pressure that a corroded pipe could support in the case of a longitudinal corrosion
defect. The maximum pressure is the pressure value corresponding to the critical
value of the DC damage (crack initiation). This method was validated by comparing
the results of numerical simulations with experimental and ASME/ B31G results
available in the literature.
2 Modeling
To model the pipeline we used the ANSYS code with two scenarios, a model
without a cavity (no corrosion, Figure 2) and a model with cavity (corroded
pipeline, Figure 3). Taking into account the symmetry of the pipeline, we modeled
only one-fourth of cylinder. The dimensions of the structure are length, inner
radius and thickness. To do this, we have created a rectangular area, which has the
same length as the pipeline, and a width equivalent to the thickness of the pipeline.
By a 90° rotation of this surface around the axis X, a volume was obtained in the
form of a quarter cylinder. For meshing the structure, we used volume elements
with 20 nodes Solid95. These elements have compatible displacement shapes and
are well suited to model curved boundaries. The mesh of the volume is made to
obtain a concentration of elements around the side where the cavity is represented
(Figure 2). For boundary conditions, we have respected the conditions of
symmetry and the experience.
On the corroded pipe (model with cavity), we kept the same data used in the
model without cavity, and created a cavity that represents the corrosion defects on
the pipeline. To do this we created a parabolic surface to the thickness of one
quarter of the cylinder with a length equivalent to the defect length and width
equal to the depth of the defect. The volume of the cavity is obtained by a rotation
of 90 ° around the upper line of the pipeline wall. This volume is subtracted from
the volume of the non-corroded pipeline (Figure 2). The same meshing is used
with a mesh filtering around the cavity. We varied the values of the depth and
extension of the cavity (corrosion defects, see Figure 2). The burst pressure is
determined according to the geometrical characteristics of each cavity. The critical
pressure or the maximum pressure that will support a corroded pipe is that
corresponding to the critical value of the damage Dc (this value depends on the
material and temperature).
484 D. Benzerga
Next, the constitutive law of the critical point (M, *) where the equivalent stress
σ * is maximum, obtained using the ANSYS code, is implanted in the post-
processor based on the iterative Newton method [Benallal et al, 1988.]:
σ* (M ) = Sup⎛⎜ σ* ⎞⎟
⎝ ⎠
1
with σ* = σ eq Rv 2 (1)
2
2 ⎛ σH ⎞
RV = (1 + ν ) + 3(1 − 2ν )⎜⎜ ⎟
⎟
3 ⎝ σ eq ⎠
In most cases, this criterion is satisfied in areas of high stress concentration
with a high coefficient of triaxiality σH / σeq. The final step is to determine the
evolution of damage in solving the following constitutive laws:
e +Ep
E ij = E ij ij
e = 1 + ν σ ij υ σ kk
E ij − δ ij
E 1− D E 1− D
~D (2)
3 σ ij
p
E ij = P Si f = f = 0
2 σ eq
2
σ eq
D= Rv P Si p ≥ p0
2ES
Using a stepwise method simultaneously with the method (scheme) implicit
Newton, which ensures good convergence [Benallal et al, 1988].
486 D. Benzerga
The method developed above gives versus of the internal pressure of the pipe,
the value of the damage, the accumulated plastic strain and stress components at
each instant until a macroscopic crack initiation at the defect corrosion. This
allows determining the new maximum pressure that will support a pipe with a
corrosion defect.
60 60
Burst pressure versus
58
depth for L=21 mm Burst pressure versus
56 55
54 depth for L=41 mm
52
50
50
Burst pressure MPa
48
46 45
44 Simulation Simulation
42 ASME/B31G 40 ASME/B31G
40 Expérimental Experimental
38
35
36
34
30
32
30
28 25
Fig. 4 Burst pressure depending on the depth of corrosion defects for two extensions
L = 21mm and L = 41 mm.
Burst Pressure Estimation of Corroded Pipeline Using Damage Mechanics 487
5 Conclusion
In this paper, we presented a new method based upon damage mechanics, which is
now in its stage of maturity. Our main motivation in this work was to show that
this method is a contribution to the rehabilitation of corroded pipes. In spite of
some differences between proposed method results and the experimental results,
our method have had a good performance in predicting the failure pressure of a
pipeline containing corrosion defects. The ASME/B31G results were more con-
servative than proposed method; however, our method provided more precise re-
sults. Our models proved to be capable of simulating the corroded pipe burst tests
adequately.
488 D. Benzerga
References
Benallal, A., Billardon, R., Doghri, I.: An integration algorithm and the coresponding con-
sistent tangent operator for fully coupled elastoplastic and damage equations. Comm.
Appl. Numer. Methods 4, 731–740 (1988)
Lemaitre, J., Doghri, I.: Damage 90: a post processor for crack initiation. Comput. Methods
Appl. Mech. Eng. 115, 197–232 (1994)
Netto, T.A., Ferraz, U.S., Estefen, S.F.: The effect of corrosion defects on the burst pressure
of pipeline. J. Constr. Steel Res. 61, 1185–1204 (2005)
Numerical Simulation of Incremental Sheet
Metal Forming Process
1 Introduction
The experiences realized with the ISF process are concentrated in the SPIF
process using a CNC machines with a forming tool with specific shape. This tool
replaces the milling tool and it uses the same mounting elements in CN machine.
Many trajectories are used in incremental forming with numerical increments by
the use of CN program in 3 axis or 5 axis machines. The incremental forming tool
trajectories are specific to this process and materials of sheets manufactured in
ISF.
The literature on the subject of ISF: the integrity of surfaces during and after
the incremental forming, the geometry of the specimen manufactured, forming ef-
forts and optimization of forming time are investigated by the majority of re-
searches. The majority of manufactured parts, mentioned in the articles, are not
industrial parts but they are test pieces with cone or pyramidal shape where a he-
mispherical punch is used in the sheet forming process, with a manner close to the
operation of a pocket in a CN milling process. So the study of an industrial piece
will be significant with a good experiment results.
Several studies have been made to understand ISF process taking into account
different material behavior and various methods have been proposed for assessing
the formability in ISF. (Robert et al. 2012) made comparison between incremental
deformation theory and flow rule to simulate sheet metal forming processes. They
used the material user subroutine (VUMAT) to implement two material behavior
laws (incremental deformation theory and classical flow rule plasticity) in their
simulation. At the aim to analyze the stress state and thickness distribution of
sheet metal studied. But in this study they did not take into account kinematic har-
dening in the material behavior for a best of prediction of spring back in sheet
metal incremental forming.
(Ben Ayed et al. 2014) presented a simplified numerical approach to simulate
ISF with efficient CPU time. A shell element DKT12 was implemented and
coupled with an elasto-plastic model based on a classical flow rule, isotropic har-
dening has been considered in their simulations.
(Fang et al. 2014) build an investigation on deformation mechanism and frac-
ture behavior in single point incremental forming. In their analytic model they
make on consideration two parameters, bending effect and strain hardening, to de-
scribe the localized deformation mechanism. By numerical and experimental in-
vestigation they demonstrate that the deformation occurs not only in the contact
zone, but also in the neighboring wall which has been already formed in the vi-
cinity of the contact zone. And that the fracture tends to appear at the transitional
zone between the contact area and the formed wall. They conclude also that the
strain hardening has a positive effect on the material formability.
(Belchior et al. 2014) developed an approach applied in robotized ISF process,
this approach consists in coupling a finite element analysis of the ISF (using to
predict efforts in this process) with an elastic model of a robot structure where the
punch is mounted. The result of finite element simulation was used like an input
data for the elastic model at the aim to optimize the punch path and to make cor-
rection in robotized ISF when the punch makes deviation. The effect of the har-
dening law (Ludwick or Voce) has been evaluated.
Numerical Simulation of Incremental Sheet Metal Forming Process 491
(Kurra and Regalla 2014) analyzed the formability and thickness distribution in
ISF of extra-deep drawing steel sheet which is a widely used material in automo-
tive applications involving simple and complex parts requiring high formability.
(Seong et al. 2014) concluded after their research that the stress states in top,
middle and bottom surfaces exceed the forming limit curve at different times and
each layer has different stress state in terms of their deformation history to sup-
press necking.
(Behera et al. 2014) propose an analysis methodology using topological con-
ceptual graphs to capture the effects of different phenomena on the final accuracy
of a sheet metal part manufactured by SPIF. So they create an algorithm creating
partial tool paths when the shape of the sheet formed is to complex and demand
sum accuracy.
(Han et al. 2013) studied the springback at the aim to predict and control accu-
rately this factor to design an accurate tool path for ISF. A three-dimensional elas-
to-plastic finite element model was implemented to simulate ISF this model is
based on the particle swarm optimization neural network and it has shown good
results in the prediction of springback.
The present investigation is a contribution of work that started using an incre-
mental sheet metal forming (ISF) process by taking into account the anisotropic
effect presented in sheet metal combined with isotropic hardening behavior. Re-
sults including thickness distribution and forming forces can be estimated from the
numerical model.
ε = ε +ε
e p
- Partition of the total strain (1)
σ = D: ε
e
- Hooke’s law (2)
3
- Yield function f = ϕ (σ ) −σ p ≤ 0 (3)
2
ϕ (σ ) = σ Pσ
T
(4)
σ p =σY + R , R =Q 1− e ( − βκ
) (5)
492 L. Ben Said, M. Wali, and F. Dammak
∂f 3 1
ε = γ = γ n n= Pσ
p
- Flow rule , (6)
∂σ 2 ϕ
where ε denotes the elastic strain tensor and ε is the plastic strain tensor, σ
e p
is the stress tensor, D is the general elastic operator, Q and β are material pa-
rameters, γ is the plastic multiplier and P is a fourth order tensor which define
the yield criterion. This yield function includes the classical J2 plasticity yield
condition and the quadratic Hill criterion as special cases. The Hill yield criterion,
in three-dimensional cases, is obtained by taking
⎡H + G −H −G 0 0 0 ⎤
⎢ H+F −F 0 0 0⎥
⎢ ⎥
2 ⎢ F +G 0 0 0⎥
P= H , [H ]= ⎢ ⎥ (8)
2N 0 0
3 ⎢ ⎥
⎢ Sym 2M 0⎥
⎢ ⎥
⎣ 2L⎦
In this section, square box application of ISF was investigated and the results car-
ried out with the use of numerical treatment based on ABAQUS/Explicit code.
The material is an aluminum alloy (AA5754-O). The mechanical properties of the
AA5754-O material are given in Table 1. The isotropic hardening behavior is
modeled by the exponential law (Eq. 5), and the Hardening parameters of the
AA5754-O material are presented in Table 2.
Numerical Simulation of Incremental Sheet Metal Forming Process 493
The initial dimension of the sheet is 200 x 200 x 1.5 mm, the punch tool has a
diameter of 10 mm. The punch tool is assumed as a rigid body and no property is
assigned. The frustum of the square box has a depth of 10 mm and a base area of
80x80 mm. The tool path and the dimensions of the desired shape are shown in
Figure 2. The friction coefficient between punch tool and sheet metal is assumed
to be 0.1.
The sheet is meshed using 2728 triangular elements (S3R in Abaqus) with five
integration points through the thickness.
Fig. 2 Tool path and the final shape of sheet incremental formed (Robert et al. 2012)
494 L. Ben Said, M. Wali, and F. Dammak
Numerical predictions carried out using Abaqus are shown in Figures 3, 4 and
5. The thickness variation along the Y = 0 cut is presented in Fig. 3. According to
the thickness evolution, significant deformations of square box are located princi-
pally in the vicinity of punch path. The evolution of the vertical forming force dur-
ing the ISF process of the studied sheet is illustrated in Figure 4. The vertical
forming force due to the contact pressure reaches the maximum value in the last
pass of the punch when incremental forming is in a depth of 10 mm. the value of
this effort is in the same range of efforts recognized in the majority of articles
established by many research group.
The distribution of the von Mises stress in the deformed zone under the action
of the punch during the ISF process is as shown in Figure 5. According to this
Figure the major Von Mises stress is localized under the punch tool, located prin-
cipally along the punch path.
1.55
1.5
Thickness (mm)
1.45
1.4
1.35
1.3
1.25
0 0.05 0.1 0.15 0.2
X direction (m)
4 Conclusion
In this paper, finite element model is developed, based on Hill yield criterion and
isotropic hardening behavior, simulating an example of single point incremental
sheet forming. The material model is implemented in the commercial finite ele-
ment code ABAQUS/Explicit to predict ISF process. The results show significant
deformations have been located along the contour of the square box base but it
should be noted that there are also minor deformations on the borders near to the
clamped shape.
References
Azaouzi, M., Lebaal, N.: Tool path optimization for single point incremental sheet forming
using response surface method. Simulation Modelling Practice and Theory 24, 49–58
(2012)
Behera, A.K., Lauwers, B., Duflou, J.R.: Tool path generation framework for accurate
manufacture of complex 3D sheet metal parts using single point incremental forming.
Comput. Ind. 65, 563–584 (2014)
Belchior, J., Leotoing, L., Guines, D., Courteille, E., Maurine, P.: A Proc-ess/Machine
coupling approach: Application to Robotized Incremental Sheet Forming. J. Mater.
Process. Tech. 214, 1605–1616 (2014)
Ben Ayed, L., Robert, C., Delamézière, A., Nouari, M., Batoz, J.L.: Simplified nu-merical
approach for incremental sheet metal forming process. Eng. Struct. 62-63, 75–86 (2014)
Fang, Y., Lua, B., Chena, J., Xua, D.K., Oub, H.: Analytical and experimental inves-
tigations on deformation mechanism and fracture behavior in single point incremental
forming. J. Mater. Process. Tech. 214, 1503–1515 (2014)
Han, F., Mo, J., Qi, H., et al.: Springback prediction for incremental sheet forming based on
FEM-PSONN technology. Trans. Nonferrous Met. Soc. China 23, 1061–1071 (2013)
Kurra, S., Regalla, S.P.: Experimental and numerical studies on formability of extra-deep
drawing steel in incremental sheet metal forming. Journal of Materials Research and
Technology 3, 158–171 (2014)
496 L. Ben Said, M. Wali, and F. Dammak
Lemaitre, J., Chaboche, J.L.: Mechanics of Solid Materials. Cambridge University Press,
Cambridge (1990)
Robert, C., Delamézière, A., Dal Santoa, P., Batoz, J.L.: Comparison between in-cremental
deformation theory and flow rule to simulate sheet-metal forming processes. J. Mater.
Process. Tech. 212, 1123–1131 (2012)
Seong, D.Y., Haque, M.Z., Kimb, J.B., Stoughton, T.B., Yoon, J.W.: Suppression of neck-
ing in incremental sheet forming. Int. J. Solids Struct. 51, 2840–2849 (2014)
A Higher Order Shear Strain Enhanced Solid-
Shell Element for Laminated Composites
Structures Analysis
1
Mechanical Modelisation and Manufacturing Laboratory (LA2MP),
National Engineering School of Sfax, B.P W3038, Sfax, University of Sfax, Tunisia
2
Faculty of Engineering, King Abdulaziz University, North Jeddah, Saudi Arabia
abdhajlaoui@gmail.com, jarraya_abdessalem,mondherwali}@yahoo.fr,
fakhreddine.dammak@enis.rnu.tn
Abstract. This paper presents a free from locking higher order solid-shell element
based on the Enhanced Assumed Strain (EAS) for laminated composite structures
analysis. The transverse shear strain is divided into two parts: the first one is inde-
pendent of the thickness coordinate and formulated by the Assumed Natural Strain
(ANS) method; the second part is an enhancing part which ensures a quadratic dis-
tribution through the thickness. This permit to remove the shear correction factors
and improves the accuracy of transverse shear stresses. Also, volumetric locking is
completely avoided by using the optimal parameters in the EAS method. Compar-
isons of numerical results with those extracted from literature show the perfor-
mance of the developed finite element.
1 Introduction
In the last decades the use of composite materials is becoming more widely used
since they provide many advantages to structural designers. The stress and the
strain field in these structures is very complex, therefore numerical methods have
been developed to give an accurate prediction. Several methods have been pro-
posed in the literature namely: The Assumed Natural Strain method (ANS) used to
solve the shear locking problems in the case of thin plate and shell structures, we
can cite (MacNeal 1982), (Bathe and Dvorkin 1985), (Bucalem and Bathe 1993)
among many others. (Hauptmann and Schweizerhof 2000) assumed a linear distri-
bution of the strain in thickness direction and developed a solid-shell element for-
mulation. (Sze and Yao 2000) used this method to obtain a locking-free solid-shell
element.
The Enhanced Assumed Strain method (EAS) was proposed by (Simo and Rifai
1990). It allowed avoiding the thickness locking caused by coupling of the in-
plane and transverse normal stress and normal strain responses (Doll et al. 2000).
Based on this method, (Alves de Sousa et al. 2003) proposed a locking-free solid-
shell element formulation. The application of the EAS method in shell analysis
was presented in the work of (Andelfinger and Ramm 1993) with linear elastic
range. The extension to non-linear aspects was carried out by (Büchter et al. 1994)
and (Bischoff and Ramm 1997) among others.
The use of a single method cannot entirely solve the locking problems in solid-
shell elements. Therefore many studies have combined them for a better prediction
of the structure behavior. (Klinkel et al. 1999) and (Vu-Quoc and Tan 2003) de-
veloped three-dimensional locking free solid-shell elements by using the EAS and
ANS methods. (Hajlaoui et al. 2012) presented a solid-shell finite element based
on the classical ANS method to compute the shear locking effect and a nine pa-
rameters EAS method which avoided completely the volumetric locking to study
the bucking behavior of laminated composite plate with delamination.
In the literature, there is no solid-shell formulation with high-order transverse
shear enhancement. The only exception is the work of (Quy and Matzenmiller
2008), where the authors developed a finite element based on the theoretical foun-
dations of high-order shear deformation theories. In this works authors decom-
posed the transverse shear strains into two parts: the first one is compatible and
continuous, which can present a shear locking in thin limit structures, the second
part is based on the EAS with parabolic function in terms of natural thickness
coordinate. The membrane part in this work is not enhanced, which can present a
numerical locking in the case of in-plane bending and in the nearly incompressible
elasticity.
However, in the case of plates and classical shell, the higher order shear defor-
mation theory (HOSDT) has been widely used. (Liu and Reddy 1985), (Rastgaar
Aagaah et al. 2003) and (Reddy and Lee 2004), among others, used this theory to
improve the transverse shear strain and stress distribution. In a recent work, (Wali
et al. 2014) presents an efficiency three dimensional double directors shell element
for the functionally graded material shell structures analysis. The vanishing of
transverse shear strains on top and bottom faces is considered in a discrete form
inspired form the work of (Dammak et al. 2005). Thus, the third-order shear de-
formation plate theory (TSDT) is a particular case of the developed formulation.
Nevertheless, the treatment of delamination with this HOSDT is complicated
which is not the case with the solid-shell elements.
This paper is an improvement of the solid-shell formulation developed by (Haj-
laoui et al. 2012); the transverse shear strain is composed by two parts: a compati-
ble part based on the ANS formulation of (Bathe and Dvorkin 1985), and a second
part based on EAS method with parabolic function in terms of natural thickness
coordinate as proposed in (Quy and Matzenmiller 2008). In addition, the en-
hancements of membrane part and the transverse strain were used to avoid the
in-plane bending and volumetric locking.
A Higher Order Shear Strain Enhanced Solid-Shell Element 499
The strain field is enhanced with the introduction of internal variables (Simo and
Rifai, 1990), as following:
E = E + E
c
(1)
Where E and E are respectively the compatible part and the enhanced part
c
E =⎡
⎣ E11 , E22 , E33 , 2 E12 , 2 E13 , 2 E23 ⎤⎦
c c c c c c c T
(2)
c c
The compatible transverse shear strains E13 and E23 and the transverse thick-
c
ness strains E33 are developed on the base of ANS method proposed by (Bathe
and Dvorkin 1985). This formulation is detailed in the author’s previous work,
c c
Hajlaoui et al. (2012). The compatible transverse shear strains E13 and E23 are
evaluated at four midpoints of the reference element mid-surface A = (-1,0,0), B =
c
(0,-1,0), C = (1,0,0), D = (0,1,0) and the transverse thickness strains E33 is eva-
luated at the following four points of the reference element A1=(-1,-1,0), A2= (1,-
1,0), A3= (1,1,0), A4= (-1,1,0).
Then the compatible part of the Green-Lagrange strain tensor takes the
following form
500 A. Hajlaoui et al.
⎡ 1
2
( g11 − G11 ) ⎤
⎢ 1
( g 22 − G22 ) ⎥
⎢ 2
⎥
⎢ ⎥
4 (
1 + ξ A ξ )(1 + η Aη ) 2 ( g 33 − G33 ) ⎥
4
E =T
c −T ⎢ ∑ 1 1 A A
(5)
⎢ A =1 ⎥
⎢ ( g12 − G12 ) ⎥
⎢ 1 ⎡(1 − η ) ( g B − G B ) + (1 + η ) ( g D − G D ) ⎤ ⎥
⎢2⎣ 13 13 13 13 ⎦
⎥
⎢⎣ 2 ⎡⎣(1 − ξ ) ( g 23 − G23 ) + (1 + ξ ) ( g 23 − G23C ) ⎤⎦ ⎥⎦
1 A A C
δ E = B δUn Δ E = B ΔU n
c c
, (6)
⎡ T
g1 N I ,1 ⎤
⎢ T ⎥
g1 N I ,1
⎢ ⎥
⎢ ⎥
4 (
1 + ξ A ξ )(1 + η Aη ) g3 N I ,3
4
⎢ ∑ 1 T
⎥
BI = T ⎢ ⎥
−T A =1
(7)
⎢ g2 N I ,1 + g1 N I ,2
T T
⎥
⎢1 ⎥
( ) (
⎢ 2 ⎡⎣(1 − η ) g3 N I ,1 + g1 N I ,3 + (1 + η ) g3 N I ,1 + g1 N I ,3 ) ⎤⎥
T T T T
B B B B D D D D
⎦
⎢ ⎥
( ) (
⎢⎣ 12 ⎡⎣(1 − ξ ) g3A N IA,1 + g1A N IA,3 + (1 + ξ ) g3C N IC,1 + g1C N IC,3 ) ⎤⎥
T T T T
⎦⎦
E = M
α , δE = M
δα , ΔE = M
Δα (8)
A Higher Order Shear Strain Enhanced Solid-Shell Element 501
where M is the interpolation function matrix for the enhanced assumed strain
field. The interpolation is first defined in the natural coordinate and must be trans-
ferred to the global coordinates as follows
= det J 0 T − T M
M (9)
0 ξηζ
det J
Where the subscript ‘ 0 ’ means evaluation at the center of the element in the
natural coordinates, J = [G1 , G2 , G3 ] is the Jacobian matrix. The interpolation
, in equation (9), is expression in term of the parametric coordinates
matrix M ξηζ
( ξ ,η ,ζ ) .
Three choices of matrix M ξηζ
will be considered with, 7, 9 and 11
parameters.
⎛ξ 0 0 0 0 0 0⎞
⎜0 η 0 0 0 0 0
⎟
⎜ ⎟
⎜0 0 ζ 0 0 0 0 ⎟
⎜ ⎟
0 0 0 ξ η
=⎜
7 0 0 ⎟ (10.a)
M ξηζ
⎜ 1 ⎟
⎜0 0 0 0 0 −ζ 0 ⎟
2
⎜ 5 ⎟
⎜ 1 2 ⎟
⎜0 0 0 0 0 0 −ζ ⎟
⎝ 5 ⎠
⎛ξ 0 0 0 0 0 0 0 0 ⎞
⎜0 η 0 0 0 0 0 0 0
⎟
⎜ ⎟
⎜0 0 ζ 0 0 0 0 ξζ ηζ ⎟
⎜ ⎟
0 0 0 ξ η
=⎜
9 0 0 0 0 ⎟ (10.b)
M ξηζ
⎜ 1 ⎟
⎜0 0 0 0 0 −ζ 0 ⎟
2
0 0
⎜ 5 ⎟
⎜ 1 ⎟
⎜0 0 0 0 0 −ζ 0 ⎟
2
0 0
⎝ 5 ⎠
502 A. Hajlaoui et al.
⎛ξ 0 0 0 0 0 0 0 0 ξη 0 ⎞
⎜0 η 0 0 0 0 0 0 0 0 ξη ⎟
⎜ ⎟
⎜0 0 ζ 0 0 0 0 ξζ ηζ 0 0 ⎟
⎜ ⎟
ξ η
=⎜ ⎟
11 0 0 0 0 0 0 0 0 0
M ξηζ (10.c)
⎜ 1 ⎟
⎜0 −ζ 0 ⎟
2
0 0 0 0 0 0 0 0
⎜ 5
⎟
⎜ 1 ⎟
⎜0 0 0 0 0 0 −ζ
2
0 0 0 0 ⎟
⎝ 5 ⎠
Since the shear strain interpolation in thickness direction is parabolic, at least
2x2x3 Gauss quadrature integration rule is used. After including interpolation
functions for enhanced strain fields, we obtain the solid-shell element enhanced
with 7, 9 and 11 incompatible modes.
G( u , E ) = ∫
V
(
S : δ E + δ E dV −
c
) ∫ F .δ udV − ∫
V
V
∂Vf
FS .δ udA = 0 (11)
where S is the second Piola–Kirchhoff stress tensors and FS and FV are surface
and volumetric distributed load respectively. This equation will be solved by the
Newton-Raphson method and must be linearized. The strain parameters are then
eliminated at the element level which leads to the following tangent stiffness ma-
trix and the residual vector
−1 −1
KT = K e + K G − L H L , Re = L H h + f −f
T T ext int
(12)
int ext
where K e , L , H , f ,f and h are given by the following expressions
Ke = ∫ B B dVe , L = ∫ T B dV , H= ∫ T M dV
T
M e
M e
(13)
Ve Ve Ve
= ∫ = ∫ N FV dVe + ∫ N FS dA; h = ∫ T S dV
int T ext T T
f B S dVe ; f M e
(14)
Ve Ve ∂V fe Ve
4 amples
Numerical Exa
ments
Table 1 Listing of shell elem
Name Description
Q8A3E5
(Klinkel eet al. (1999), ANS with E33 , E13 E23 and 5 parameters
Q8A3E7
(Vu-Quocc and Tan 2003), ANS with E33 , E13 E23 and 7 parameters
Q8S12 (Quy and Matzenmiller 2008), with 12 parameters
C3D8C11 Present ellement with 11 parameters
1.8
1.6
Normalized displacement (x10-5)
1.4
1.2
1
Q8A3E5
0.8
Q8A3E7,C3D8C11
0.6
Q8S12
0.4
0.2
0
0 10 20 30
Number of Elements per Side
A short cylinder, witth two pinching vertical forces at the middle section,
ms at the end, is modeled using one octant with approo-
and two rigid diaphragm
priate symmetry boundarry conditions, (figure 1). The length of the cylinder is
504 A. Hajlaoui et al.
L=600mm, the radius is R=300mm, and the thickness is h=3mm. The material
properties are: E = 3.0 × 10 MPa, ν = 0.49 . The results are shown in Figure 1.
6
The material properties are: E11 = E22 = open , E22 = E33 , G23 = 0.5E22 ,
G12 = 0.5G13 = 0.6 E22 and ν 12 = ν 13 = ν 23 = 0.25 .
Figure 3 illustrates the variation of the buckling load versus the delamination
size for symmetric cross ply laminated square plates. It is shown that the largest
buckling load is obtained for the case of symmetric laminated [0/90/90//0], be-
cause the fiber orientation at the layer with the delamination is parallel to the
in-plane loads. From the plots it can be concluded that the influence of the dela-
mination is more important for the symmetric laminated [0/90/90//0].
A Higher Order Shear Strain Enhanced Solid-Shell Element 505
180
a/b=1 et b/h=20 E11/E22=30
160
5 Conclusions
In this study a new solid-shell finite element based on the partition of shear strain
is developed: a first part is independent of the thickness coordinate and formulated
by the ANS method which avoid the shear locking in thin limit structure; a second
part, enhancing part, ensures a quadratic distribution of the transverse shear strain.
This permit to remove shear correction factors and improves the accuracy of
transverse shear stresses. The finite element formulation is completely free from
the volumetric locking phenomena which occur in the treatment of nearly incom-
pressible elasticity. The developed finite element model is validated by comparing
numerical results with other results from literature. The buckling behavior of
composite laminates is investigated based on the use of solid-shell element.
References
Alves de Sousa, R.J., Natal Jorge, R.M., Valente, R.A.F., César de Sá, J.M.A.: A new vo-
lumetric and shear locking-free 3D enhanced strain element. Eng. Computation 20,
896–925 (2003)
Andelfinger, U., Ramm, E.: EAS-Elements for two-dimensional, three-dimensional, plate
and shell structures and their equivalence to HR-elements. Int. J. Numer. Meth. Eng. 36,
1311–1337 (1993)
Bathe, K.J., Dvorkin, E.N.: A four-node plate bending element based on Mindlin/Reissner
plate theory and a mixed interpolation. Int. J. Numer. Meth. Eng. 2, 367–383 (1985)
Bischoff, M., Ramm, E.: Shear Deformable shell elements for large strains and rotations.
Int. J. Numer. Meth. Eng. 40, 4427–4449 (1997)
Bucalem, M.L., Bathe, K.J.: Higher-order MITC general shell elements. Int. J. Numer.
Meth. Eng. 36, 3729–3754 (1993)
506 A. Hajlaoui et al.
Büchter, N., Ramm, E., Roehl, D.: Three-dimensional extension of non-linear shell formu-
lation based on the enhanced assumed strain concept. Int. J. Numer. Meth. Eng. 37,
2551–2568 (1994)
Dammak, F., Abid, S., Gakwaya, A., Dhatt, G.: A formulation of the non linear dis-crete
Kirchhoff quadrilateral shell element with finite rotations and enhanced strains. Euro-
pean Journal of Computational Mech. 14, 1–26 (2005)
Doll, S., Schweizerhof, K., Hauptmann, R., Freischlager, C.: On volumetric locking of low-
order solid and solid-shell elements for finite elastoviscoplastic deformations and selec-
tive reduced integration. Eng. Computation 17, 874–902 (2000)
Hajlaoui, A., Jarraya, A., Kallel-Kamoun, I., Dammak, F.: Buckling analysis of a laminated
composite plate with delaminations using the enhanced assumed strain solid shell ele-
ment. J. Mech. Sci. Technol. 26, 3213–3221 (2012)
Hauptmann, R., Schweizerhof, K., Doll, S.: Extension of the ‘solid-shell’ concept for appli-
cation to large elastic and large elastoplastic deformations. Int. J. Numer. Meth. Eng. 49,
1121–1141 (2000)
Klinkel, S., Gruttmann, F., Wagner, W.: A continuum based three-dimensional shell ele-
ment for laminated structures. Comput. Struct. 71, 43–62 (1999)
Liu, L.F., Reddy, I.N.: A higher-order shear deformation theory of laminated elastic shells.
Int. J. Engng. Sci. 23, 319–330 (1985)
MacNeal, R.H.: Derivation of element stiffness matrices by assumed strain distributions.
Nucl. Eng. Des. 70, 3–12 (1982)
Noor, A.K.: Stability of multilayered composite plates. Fibre Sci. Technol. 8, 81–88 (1975)
Quy, N.D., Matzenmiller, A.: A solid-shell element with enhanced assumed strains for
higher order shear deformations in laminates. Tech. Mech. 28, 334–355 (2008)
Rastgaar Aagaah, M., Mahinfalah, M., Nakhaie Jazar, G.: Linear static analysis and fi-nite
element modeling for laminated composite plates using third order shear theory. Com-
pos. Struct. 62, 27–39 (2003)
Reddy, J.N., Lee, S.J.: Vibration suppression of laminated shell structures investigated us-
ing higher order shear deformation theory. Smart Mater. Struct. 13, 1176–1194 (2004)
Simo, J.C., Rifai, M.S.: A class of mixed assumed strain methods and the method of in-
compatible modes. Int. J. Numer. Meth. Eng. 29, 1595–1638 (1990)
Sze, K.Y., Yao, L.Q.: A hybrid stress ANS solid-shell element and its generalization for
smart structure modeling. Part I—solid-shell element formulation. Int. J. Numer. Meth.
Eng. 48, 545–564 (2000)
Vu-Quoc, L., Tan, X.G.: Optimal solid shells for non-linear analyses of multilayer compo-
sites. I. Statics. Comput. Methods Appl. Mech. Engrg. 192, 975–1016 (2003)
Wali, M., Hajlaoui, A., Dammak, F.: Discrete double directors shell element for the func-
tionally graded material shell structures analysis. Comput. Methods Appl. Mech.
Engrg. 278, 388–403 (2014)
New Approch of High Cycle Fatigue Behaviour
of Defective Material under Multiaxial Loading
in 1045 Steel
1
LGM, université de Monastir ENIM, Avenue Ibn Eljazzar, Monastir 5019
hassine.wannes@gmail.com, chokri.bouraoui@enim.rnu.tn
2
LGM, EPEIM, Avenue Ibn Eljazzar, Monastir 5019
anouar.nasr @hotmail.fr
Abstract. This paper presents a new approach to predicting the HCF behaviour of
defective material submitted to Multiaxial loading .Defects resulting from the
casting processes are simplified to semi-spherical pore at surface of specimen.
Finite Element (FE) method was used to determine stress distribution around
defect. Papadoupolos criterion was used to evaluate stress equivalent around
defect. A definition of affected area was given, which is the area close to surface
defect where papadoupolos criterion was violated .The evolution of the affected
area, with the amplitude of loading and defect size, allows us to determine fatigue
limit for different defect sizes. Results of the new approach are in good agreement
with experimental results and show that affected area is a good parameter to
characterise the influence of a defect on fatigue behaviour.
1 Introduction
This study was carried on the 1045 carbon steel. It has long been used in aircraft
and automotive industry. An experimental database containing Multiaxial fatigue
results for 1045 steel with induced defects has been previously published [6].
3.1 F E Analysis
Defects are always the preferential sites for crack initiation .So we need to
characterise the stress distribution around the defect. Finite Element simulations
have been conducted for each size of defect and loading mode. Symmetry and
boundary conditions are applied . Plasticity occurs around defect so that all
simulations are conducted with an elastic- plastic Prandtl-Reuss type model with a
linear isotropic hardening law. Meshing is refined and optimized in the critical
zone around the defect. The cyclic loading is included in the cyclic stabilised
experimental stress-strain curve used to identify the model .
the defect in the maximum m shear plane [7] .And the macroscopic crack that leadds
to failure of the sample propagate
p in HLP. So that we will be interested in thhe
distribution of stress on th
his plane (Fig.5).
(a) (b)
ound defect on HLP (S22 maximal principal stress) (a) Tensioon
Fig. 1 Stress distribution aro
loading = 180MPa (b) to
orsion loading =150 MPa
A calculation of Papad
doupolos equivalent stress in the highest loadeed
plane HLP allows us to deeduce the following results:
− is almostt constant along the equator of defect (AB) with an erroor
that doesn’t exceed
d 4% .
− The distribution off Papadoupolos equivalent stress is independent of thhe
chosen direction.
As first approximation to
o interpolate the variation of in HPL we can propose thhe
following equations:
4.1 Methodology
From the results of numerical simulations, we can conclude that the variation of
is independent of the chosen direction. In fact, on a centred outline at the
centre of defect is constant .Our task is to study the variation of
in HLP.
In this plane, there is an area, close to the defect, where the Papadoupolos
criterion violated even with a loading less than the endurance limit of the defective
device. Consequently we are led to introduce the following definition:
Affected area
The affected area by the gradient of stress for a given defect and an applied
load is the area where the Papadoupolos equivalent stress is greater than or equal
to ( = =169MPa)
The approach that we suggest consists in following the evolution of the affected
area in relation to the loading and the size of the defect. When the loading grows
up, the affected area evolves.
To the loading of endurance limit corresponds an affected area limit. Thus, to a
given defect under a given loading; if affected area affected area limit, there is
no rupture before 10 cycles.
At the endurance limit we have: affected area = affected area limit.
New Approch of High Cycle Fatigue Behaviour of Defective Material 511
Using equations (1) and (2) the affected area as defined above is determined
by the following equations:
• Fully reversed tension
.
affected = ( D 1 √area (3)
Using experimental data base table 2.we note that at fatigue limit we have a
relationship between affected area limit and the fatigue limit as bellow:
* = ct = 24000 MPaμm in case of reversed
fully tension.
* =ct = 30000 MPaμm in case of fully
reversed torsion.
Affected area limit was calculated by equations (3) and (4) using the fatigue
limit value in each cases
Hypothesis
For fully reversed tension and torsion the product of fatigue limit by the square
root of is almost constant and independent of defect size .
The proposed approach consists on the following points:
♦At first we plot curves : = f( equations (3) and (4)
for many defect sizes.
♦ Secondly, from a medium size of a defect we deuce or
♦Then at the same graphic we plot the curve: =
or
♦Finally, the crossing points allow as to determinate the fatigue limit for each
defect size.
512 H. Wannes, A. Nasr, and C. Bouraoui
To summarize necessary steps for the proposed approach, we can use the
algorithm fig.3.
4.2 Applications
4.2.1 Alternative Tension
In alternative tension case the affected area was determined by (3). From an
experimental fatigue limit of a medium defect size we calculate the parameter
.In this study, a defect size √ = 400μm corresponding to a fatigue
strength = 152 MPa. Using hypothesis 2 we deduce =24100 MPa.um.
In the same graphic we plot the curves:
♦ Affected area versus loading amplitude, for many defect sizes.
♦ Affected area limit versus loading amplitude: Affected area = .
Affectedareaum2 40000
Affectedarealimit
30000
size:50um
H size:100um
20000 size:200um
size:300um
10000 size:500um
size:700um
size:900um
0
120 140 160 180 200 220 240
Loadingamplitude(MPa)
240
Fatiguelimt(MPa)
Experimentalresults
220
Proposedapproach
200
180
160
140
120
100
0 200 400 600 800 1000
Defectsize(um)
As shown, obtained results are in good agreement with experimental results and
the approach treats well the issue of small defects.
514 H. Wannes, A. Nasr, and C. Bouraoui
70000
Affectedareaum2
60000
50000 Affectedarealimit
size:170
40000
N size:200
30000 size:300
20000 size:400
size:600
10000
size:900
0
120 130 140 150 160 170 180
Loadingamplitude(Mpa)
Fatiguelimit(Mpa) 180
170 Experimentalresults
160
__ Proposedapproach
150
140
130
120
110
100
0 200 400 600 800 1000
dfectsize(um)
As shown, obtained results are in good agreement with experimental results but
the approach overestimates fatigue limit for small defects (less than 200μm).
5 Conclusions
References
[1] Murakami, Y., Endo, M.: The parameter model for small defects and non metallic
inclusion in fatigue strength: experimental evidences and application. In: Blom, A.F.,
Beevers, C.J. (eds.) Actes de “ Theoretical Concepts and Numerical Analysis of
Fatigue”, pp. 51–71. EMAS, Birmingham (1992)
[2] Endo, M.: Effects of small defects on the fatigue strength of steel and ductile iron
under combined axial/torsional loading. In: Ravichandran, K.S., Ritchie, R.O.,
Murakami, Y. (eds.) Proceeding of Small Fatigue Cracks: Mechanics, Mechanisms and
Applications, Hawaii, pp. 375–387 (1999)
[3] Susmel, L., Taylor, D.: Two methods for predicting the Multiaxial fatigue limits of
sharp notches. Far. Frac. Eng. Mater. Struct. 26, 821–833 (2003)
[4] Leopold, G., Nadot, Y.: Fatigue from an induced defect: experiments and application
of different Multiaxial fatigue approach. J. ASTM Int., 7 (2007)
[5] Nadot, Y., Billaudeau, T.: Multiaxial fatigue limit criterion for defective materials.
Engineering Fracture Mechanics (2005)
[6] Billaudeau, T., Nadot, Y., And Bezine, G.: Multiaxial fatigue limit for defective
materials: mechanisms and experiments. Acta Mater. 52, 3911–3920 (2004)
Determination of Stress Concentration for
Orthotropic and Isotropic Materials Using
Digital Image Correlation (DIC)
Abstract. The objective of this study was to perform a tensile test on isotropic and
orthotropic plates with holes and measured deformations around their hole by us-
ing the method of digital image correlation (DIC). The results of this measure
were used to determine the local stress at the edge of the hole and measured the
net and global stress concentration factor for isotropic and orthotropic materials
and see the difference between the two types of materials.
This paper presents an analysis of the net and global stress concentration factor
for different hole diameters and their influence on the isotropic and orthotropic
material using a camera DIC.
1 Introduction
The camera allows the use of in situ recording of a burst frame. These are used to
measure the displacement fields and local deformation. The principle of image
correlation is to compare local travel in the plane perpendicular to the camera be-
tween two different images. Comparing a said deformed image and the reference
image allows the measure of local movements (Figure 1). The calculations do not
take into account the movement of solid body. From these measurements, the local
deformations are deducted in the directions of the study design. To determine the
displacement at point of the specimen, it is divided into small images and is
squared. The grid represents the space between two thumbnail images, they may
intersect. The user can therefore play on two adjustable parameters are the size of
the thumbnail images and size of the grid. Although it appears to have the perfect
pattern thinnest and finest grid in order to gain accurate measurements, this confi-
guration is not always the most suitable. When the grid is too fine, the correlation
has a noise significant measurement.
This method is implemented on plate specimens with hole than is covered with
a white paint before the test. Is then carried out by spraying an artificial marking
paint with a speckle pattern on the surface of the material, via a black spray. The-
reafter, the image correlation is performed using the standard software
ICASOFT® [Icasoft] developed by Fabrice Morestin (Mguil-Touchal and al.1996)
[7] to LaMCoS laboratory of INSA Lyon (Example Icasoft image with Figure 2).
The principle is used to match two (original and distorted) images by comparison
pixel by pixel with using a criterion based on the grid level in order to evaluate the
displacement field and the corresponding strain.
Determination of Stress Concentration for Orthotropic and Isotropic Materials 519
.
Fig. 2 Example of an image with camera DIC
3 Experimental Study
The equipment used for the tests of uniaxial tension is: traction machine EZ20
with computer controlled device and image acquisition, including a lamp for light-
ing. The tests summers Films has captured with the camera positioned facing the
test, all using software (Icasoft, INSA Lyon, LAMCOS-MSE) to determine dis-
placement fields and strain. See Figure 3:
Traction
machine
Lamp
Camera DIC
PC
Software
Icasoft
Fig. 3 Diagram of the mechanical acquisition data and digital images with a uniaxial tensile
test.
Tensile tests were carried out in a universal testing range of the EASY TEST
Machine combine flexibility with exceptional usability, EZ20 model with trans-
verse hydraulically setting with constant velocity about 2 mm / min, with ambient
temperature.
Tensile tests were performed according to standard ASTM D3039 [8] (Figure
4), using a minimum of twelve specimens (The most current specimens to ASTM
D3039 has a constant rectangular cross-sectional dimensions: 250 mm long x 25
mm wide x 2 mm thick test .A typical speed for standard test samples of 2 mm /
min) for each family of laminate.
520 M.M. Makki and B. Chokri
The specimens were made from carbon / epoxy with 0.22 mm thick. The tapes
were strength carbon fibers (T700) high strength unidirectional of M21 Hexcel
epoxy resin (35% resin). The plate contains eight unidirectional layers [0/90] 2s
laminate with a total thickness of 1.76 mm. These tacking sequences have helped
to get an overview of the evaluation with different types of damage in notched
specimens under uniaxial loading using DIC.
The specimens were prepared by bonding the end of glass fiber / epoxy lami-
nate legs. This procedure resulted in the available 150 mm2 of the sample for the
tensile test surface. Figure below shows the laminated specimens with different
diameters:
4 Results
With an ambient temperature, it was two families of tests: steel and composite
plates with different hole in the center with the same size and speed of loading
(Figure 6) .on notice the break in the steel is still in hole and perpendicular to the
loading, against composite the breakage is around the hole but the orientation of
crack always followed the direction of tissue .In our example we have a orienta-
tion tissue [0.90]2s. So the crack almost perpendicular to the load.
Determination of Stress Concentration for Orthotropic and Isotropic Materials 521
Thereafter we can plot the stress-time curve for the composite plate (Figure7)
such that we can see the first damage and then the fiber break with the total damage:
Fiber
break
First dommage
400
ı d=3 400
ı d=4
300 d=5
300 d=5.5
d=7
200 200
d=9
d=7
100 100
0 İ 0 İ
0 0,01 0,02 0,03 0,04 0 0,005 0,01 0,015 0,02 0,025
Ͳ100
Fig. 8 Stress-Strain curve for a steel plate Fig. 9 Stress-Strain curve for a composite
with different diameters plate with different diameters
522 M.M. Makki and B. Chokri
failure stress
400
300
200
100
0
3 4 5 6 7 8 9 a
10
Fig. 11 Image of damaged Steel plate Fig. 12 Image of damaged steel plate with
camera DIC
Fig. 13 Image of damaged composite plate Fig. 14 Image of damaged composite plate
with camera DIC
Determination of Stress Concentration for Orthotropic and Isotropic Materials 523
İ0,12
11
0,1
0,08
0,06
Isotropic material 0,04
Orthotropic material
0,02
0 Y pixel
1,38E+01 1,58E+01 1,78E+01 1,98E+01 2,18E+01 2,38E+01
Ͳ0,02
Subsequently, the longitudinal stress around the hole can be calculated from the
deformation field (ε11 and ε22) obtained from DIC:
5 Theoretical Co
oefficient of Global Stress Concentration
Theoretical coefficient off stress concentration is defined as the ratio of the actuual
maximum stress in the zo one of discontinuity (notch, hole, for example) to the noo-
minal stress in the section
n:
Kt = σ /σ
σ : Is calculable by numerical
n methods such as finite element method or bby
analytical methods for siimple geometries. It is also measured by the techniccal
analysis of experimental constraints
c such as photo-elasticity or DIC methods.
σ : Is calculable using the formulas of strength materials, considering thhe
specimen like a bar or plaate without account the geometric discontinuity.
σ =
6 Evolution of th
he Global Stress Concentration with Hole
Diameter
In our case, the study focuses on an infinite orthotropic plate 3D is made oof
carbon /epoxy (HR [0/90] 2s) .According of Table 1 Kt equal to 5.24 for an inffi-
nite orthotropic plate.
According to the inveestigation of Heywood [1] on a rectangular orthotroppic
plate with center hole und
der the action of unidirectional axial load, the factor is inn-
troduced in the form of th
he following equation:
With
7 The Theoretica
al Value of the Net Stress Concentration
Another way to determinee the stress concentration is the application of the averagge
n section instead of the hole width of the plate (global
normal stress across the net
surface).
σ =
K T = K T (1-a/w)
526 M.M. Makki and B. Chokri
Table 1 Material Properties orthotropic carbon / epoxy (T700 / M21 [0 90] 2s)
(GPa)
148 7.8 4.5 0.35 2375 1465 95
A tensile load of 119 MPa at one and a lock on the opposite. The fineness of
the mesh is optimized in order to have a steady value of maximum stress. To re-
duce the computation time while maintaining good accuracy, only a quarter of the
plate is meshed with progressive mesh designed to produce a very fine mesh
around the hole and larger elsewhere (Figure 16).
For a given width, this ratio Kt equal to 3 for an infinitely thin plate with a cir-
cular hole and thereafter the global stress concentration factor Ktg increases if the
hole diameter increases against the net stress concentration factor Ktn decreas-
es.(Figure 17 and 18):
Ktg
11,5 DIC
10,5 FEA
9,5
Heywood
8,5
Howland
7,5
6,5
5,5
4,5
3,5
a/w a/w
2,5
0 0,2 0,4 0,6
Ktn
3,5 DIC
3 FEA
2,5 Heywood
2 Howland
1,5
0,5
0 a/w
0 0,2 0,4 0,6
For isotropic material we can notes that the results of finite element analysis are
very close to the results of DIC camera but for the stress concentration Ktg there is
a difference from a / w = 0.4 compared with Heywood and Howland.
Ktg
14 DIC
12
FEA
10
Heywood
8
0 a/w
0 0,2 0,4 0,6
On observed that a / w is increased when the edges of the plate is closer to the
hole. Consequently, the lines of force will be more compressed and Ktg increases.
6Ktn DIC
5
FEA
4
Heywood
3
0 a/w
0 0,2 0,4 0,6
Note that the net stress concentration decreases if the ratio a / w increases and
unlike different with global stress concentration factor Ktg.
It has almost the same variation of KTn respect to a/ w for isotropic and ortho-
tropic material.
9 Conclusion
In this work we have used the image correlation method with DIC camera to
measure the tensile stress concentration and which gives consistent results with
the finite element analysis and thereafter it is found that the DIC method given a
best results for a planar structure isotropic or orthotropic.
The results obtained show that the presence of a geometrical discontinuity in a flat
plate has a significant effect on the stress concentration to an isotropic and ortho-
tropic material. In addition we see the evolution of global and net stress concentra-
tion respect to different hole diameters for composite and steel.
Finally, the DIC approach don't provide good results for composite structures
because this method can be show only the surface damage against the composite
material having many damage inside the structure (delamination ,cracking ma-
trix….).
Acknowledgements. I want to thank deeply Mr. Jury members who have done me the hon-
or to judge my work and have improved my presentation with helpful comments.
References
[1] Heywood, R.B.: Designing by photoelasticity. Chapman and Hall (1952)
[2] Peterson, R.E.: Stress concentration factors. John Wiley & Sons (1974)
[3] Pilkey, W.D.: Peterson’s stress concentration factors, 2nd edn. John Wiley & Sons
Inc., New York (1997)
[4] Howland, R.C.J.: On the stresses in the neighborhood of circular hole in a strip under
tension. Phil. Trans. Roj. Soc. 229, 49–86 (1929)
[5] Lekhnitskii, S.G.: Theory of Elasticity of an Anisotropic Elastic Body, 1st edn., p. 4.
Holden-Day, San Francisco (1963)
[6] Tan, S.C.: Finite-width correction factors for anisotropic plate containing a central
opening. J. of Composite Materials 22, 1080–1097 (1988)
[7] Mguil-Touchal, S.: Une technique de corrélation directe d’images numériques. Thèse
INSA Lyon (1996)
[8] ASTM D3039/D3039M-00. Standard test method for tensile properties of polymer ma-
trix composite materi-als. American Society for Testing Materials [CD-ROM] (2004)
[9] Howland, R.: On the stresses in the neighborhood of a circular hole in a strip under
tension. Philosophical Transactions of the Royal Society A 229, 67 (1929-1930)
The Extended Finite Element Method for
Cracked Incompressible Hyperelastic
Structures Analysis
1
Mechanical Modelisation and Manufacturing Laboratory (LA2MP), National Engineering
School of Sfax, B.P W3038, Sfax, University of Sfax, Tunisia
{zaafourimehrez,mondherwali}@yahoo.fr, said.abid@issatgb.rnu.tn,
Fakhreddine.Dammak@enis.rnu.tn
2
Faculté des Sciences Ben M’Sik, Laboratoire dÊIngénierie et Matériaux (LIMAT),
Université Hassan II de Casablanca, B.P. 7955, Sidi Othman, Casablanca, Morocco
jamalmoham@gmail.com
Abstract. This paper aims to examine the contribution of the extended finite ele-
ment method (XFEM) in finite strain fracture mechanics problems. A generalized
neo-Hookean hyperelastic material is considered in an incompressible plane stress
approximation. The accuracy of the implementation is demonstrated by a series of
numerical tests.
1 Introduction
2 Constitutive Equation
∂φ
F = , J = det ( F ) (1)
∂X
C=F F
T
(2)
ψ = ψ ( I1 ,I 2 ,I3 ) (3)
(I )
1
I1 = tr ( C ) , I 2 = − trC I 3 = det ( C ) = J
2 2 2
1
, (4)
2
ψ = ψ ( I1 ,I 2 ) , J =1 (5)
2
∂I i
S = 2∑ψ ,i
−1
= α I +α C +α C
1 2 3
(6)
i =1 ∂C
534 M. Zaafouri et al.
−1
Where I , C and C are in plane tensors. However, the most general form
of the elasticity tensor in terms of the principal invariants is given by the fol-
lowing expression
∂S ∂ ψ ( I1 , I 2 )
2
=2 =4 (7)
∂C ∂C ∂C
= δ I ⊗ I +δ
1 2
(I ⊗C +C ⊗ I)+δ 3
(I ⊗C −1
+C ⊗I +
−1
)
(8)
δ C ⊗C +δ
4 5
(C ⊗ C −1
)
+ C ⊗ C + δ C ⊗ C + δ I C + δ I ijkl
−1 6 −1 −1 7
−1
8
(I ) (C )
1 1
C
−1
ijkl
=
−1
ik
C
−1
jl
+C
−1
il
C
−1
jk
, I ijkl = (δ ik
δ jl + δ il δ jk ) (9)
2 2
The numerical solution with the finite element method is based on the weak form
of equilibrium equations. The three dimensional form of the latter in the total La-
grangian formulation is given as
G= ∫ V
S : δ E dV − Gext = 0 (10)
where Gext is the virtual work of the external forces, S is the second Piola-
Kirchhoff stress tenser and δE represent the virtual Green-Lagrange strain
tensor.
1
E = (C − I ) (11)
2
( δF )
1 1
δE = δC = F + F δF
T T
(12)
2 2
The Extended Finite Element Method 535
φ= x = X +u (13)
F = [ x ,1 x ,2 ] , (14)
with
⎡ x ,1 ( .) ,1 ⎤
T
⎢ ⎥
δE = B δu B= x ,2 ( .) ,2
T
, (15)
⎢ ⎥
⎢⎣ x ,T1 ( .) ,2 + x ,T2 ( .) ,1 ⎥⎦
where ( .),i is a material derivative. With these matrix notations, the weak form
(10) can be written in the following form
∫ ( Bδ u ) S dV − Gext ( δ u ) = 0
T
G= (16)
V
In the optics of a resolution with the Newton method, its linearization is devel-
oped in two parts: material and geometrical
∫ ( B δ u ) ( B Δu ) dV + ∫ ∇ (δ u ) :∇ ( Δu ) SdV
T
DG ⋅ Δu = (17)
V V
Within the Xfem framework, the crack is taken into account by two different types
of enrichment: discontinuous enrichment and singular enrichment (Fig. 1). The
energy function of generalized neo-Hookean materials is
μ ⎛⎡ ⎤ ⎞
n
b
ψ = ⎜ 1 + ( I − 3) − 1 ⎟ (18)
⎢
2b ⎝ ⎣ n ⎥
⎦ ⎠
536 M. Zaafouri et al.
As in (Legrain at al. 2005), the singular enrichment function that will be used is
• Discontinuous enrichment
Singular enrichment
F= r { 1/ 2
sin (θ / 2 ) } (19)
X= ∑N X i
i
(20)
i∈I n
u= ∑ N u + ∑ N (H − H )a + ∑ N (F − F )b
i
i
i
i i
i
i i
(21)
i ∈I n i∈ I H i ∈I F
δ u = ∑ N δ ui + ∑ N ( H − H i ) δ ai + ∑ N ( F − Fi ) δ bi
i i i
(22)
i ∈I n i ∈I H i ∈I F
B = ⎡⎣ Bu Bb ⎦⎤
i i i i
Ba (23)
⎡ T
x ,1 N ,1
i
⎤ ⎡ T i
x ,1 M ,1 ⎤
⎢ ⎥ ⎢ ⎥
Bu = Ba =
i T i i T i
x ,2 N ,2 , x ,2 M ,2 (24)
⎢ ⎥ ⎢ ⎥
⎢⎣ x ,1T N ,i2 + x ,T2 N ,1i ⎥⎦ ⎢⎣ x ,1T M ,i2 + x ,T2 M ,i1 ⎥⎦
The Extended Finite Element Method 537
⎡ T i
x ,1 L,1 ⎤
⎢ ⎥
Bb =
i T i
x ,2 L,2 (25)
⎢ ⎥
⎢⎣ x ,1T Li,2 + x ,T2 Li,1 ⎥⎦
With M =N
i i
(H − H ) i
, L =N
i i
(F − F ) i
(26)
and ( ⋅) ,
, i = 1,3 are materials derivatives.
i
Finally, the element residual and material and geometric tangent operator are
expressed as:
∫ ∫ B B dV
T T
Gin = ΚM =
i i ij i j
B SdV , (27)
V V
⎡ f ( Ni ,N j ) I (
f N ,M )I i j
( )
f N ,L I ⎤
i j
⎢ ⎥
KG = ∫
⎢ f (M ,N ) I f ( M ,M ) I f ( M , L ) I dS
ij i j i j i j
S ⎥ (28)
⎢⎣ f ( Li , N j ) I f ( L ,M ) I
i j
f ( L , L ) I ⎥⎦
i j
where:
(
f N ,M
i j
) = N, (S i
1 11
j j
)
M ,1 + S12 M ,2 + N ,2 S 21 M ,1 + S 22 M ,2
i
( j j
) (29)
5 Numerical Example
In order to illustrate the accuracy and versatility of the extended finite element
method X-FEM in nonlinear fracture mechanical problems for large displacements
whose behavior is isotropic hyper-elastic kind, a numerical example is presented
in this section. The example is solved using both FEM and X-FEM techniques,
and the results are compared. For the purpose to perform a real comparison, the
number of elements in FEM and X-FEM meshes are taken almost equal. For X-
FEM a uniform meshes elements are used independent of the shape of crack.
While in FEM analyses, the crack is aligned with element edges and refined mesh
is applied near the crack region. All examples are simulated with the plane stress
approximation.
The example shown in Figure 2 illustrate a plate subjected to an imposed
displacement on its top edge equal to 4mm and its bottom edge is fixed.
538 M. Zaafouri et aal.
Plate geometry and maaterial properties are adopted from (Legrain et al, 20055).
The width and height of this plate are respectively w = 2 mm and h = 6 mm. A
crack of length a = 1mm is inserted at mid-height. The material is assumed to bbe
Neo-Hookean with μ = 0.4225.
The Extended Finite Element Method 539
The deformed configurations obtained by X-FEM and FEM methods are pre-
sented in figure 3. We notes that the XFEM field displacement discontinuity near
the crack is well represented and conformed to Abaqus results.
3,5 U2-XFEM
U2-Abaqus
U2 : Displacement
3
2,5
1,5
0,5
0 0,5 1 1,5
Distance X
Fig. 4 Comparison of crack U2-displacement
6 Conclusions
In this paper, the extended finite element method (XFEM) is used to analyze the
finite strain fracture mechanics problems with a neo-Hookean hyperelastic materi-
al with an incompressible plane stress approximation. In order to demonstrate the
efficiency of the X-FEM method, a numerical example was considered showing
displacement field near the crack and the deformed shape. A comparison of
XFEM and FEM results was carried showing the reliability of XFEM method to
nonlinear fracture mechanic.
Acknowledgment. The authors gratefully acknowledged and express their gratitude to the
CMPTM (Tunisian-Moroccan Joint Standing Committee) under reference 13/TM42 for
their support and technical assistance of this research.
References
Dammak, F., Regaieg, A., Kallel Kammoun, I., Dhiab, A.: Modeling the law of trans-
versely isotropic hyperelastic behavior of elastomers. European Journal of Computa-
tional Mechanics 16(1), 103–126 (2007)
540 M. Zaafouri et al.
Dolbow, J., Moes, N., Belytschko, T.: Modeling fracture in Mindlin-Reissner plates with
the eXtended finite element method. Int. J. Solids Structures 37, 7161–7183 (2000)
Dolbow, J.E., Devan, A.: Enrichment of enhanced assumed strain approximations for
representing strong discontinuities: Addressing volumetric incompressibility and the
discontinuous patch test. International Journal for Numerical Methods in Engineer-
ing 59, 47–67 (2004)
Fries, T.P., Belytschko, T.: The extended/generalized finite element method: An overview
of the method and its applications. Int. J. Numer. Meth. Engng. 84, 253–304 (2010)
Geubelle, P.H., Knauss, W.G.: Finite strains at the tip of a crack in a sheet of hyperelastic
material: I. Homogeneous case. J. Elasticity 35, 61–98 (1994)
Jarraya, A., Kallel-Kammoun, I., Dammak, F.: Theory and finite element implementa-tion
of orthotropic and transversely isotropic incompessible hyperelastic membrane. Multi-
discipline Modeling in Materials and Structures (MMMS) 7(4), 424–439 (2011)
Karoui, A., Mansouri, K., Renard, Y., Arfaoui, M.: The extended Finite Element Method
for cracked hyperelastic materials: a convergence study. Int. J. Numer. Meth.
Engng. 100(3), 222–242 (2014)
Knowles, J.K., Sternberg, E.: An asymptotic finite-deformation analysis of the elastostatic
field near the tip of a crack. Journal of Elasticity 3, 67–107 (1973)
Knowles, J.K., Sternberg, E.: Finite deformation analysis of the elastostatic field near the
tip of a crack: reconsideration of higher order results. Journal of Elasticity 4, 201–233
(1974)
Knowles, J.K., Sternberg, E.: Large deformation near the tip of an interface crack between
two neo-Hookean sheets. J. Elasticity 13, 257–293 (1983)
Legrain, G., Moës, N., Verron, E.: Stress analysis around crack tips in finite strain problems
using the eXtended finite element method. Int. J. Numer. Meth. Engng. 63, 290–314
(2005)
Long, R., Krishnan, V.R., Hui, C.Y.: Finite strain analysis of crack tip fields in incom-
pressible hyperelastic solids loaded in plane stress. Journal of the Mechanics and Phys-
ics of Solids 59, 672–695 (2011)
Möes, N., Dolbow, J., Belytschko, T.: A finite element method for crack growth without
remeshing. International Journal for Numerical Methods in Engineering 46, 131–150
(1999)
Mohammadi, S.: Extended finite element method. Blackwell Publishing (2008)
Stephenson, R.A.: The equilibrium field near the tip of a crack for finite plane strain of in-
compressible elastic materials. Journal of Elasticity 12, 65–99 (1982)
Displacement Influence on Frequencies
and Modal Deformations of a Sandwich Beam
1
Assistant to the ESIER: elkef road, 5 km Medjez elbâb, 9070, Beja
chenini.idris@yahoo.fr
2
Aassociate professor at the military academy, Address: Fondek ejjdid Grombalia, Nabeul
abdelliyoussef@gmail.com
3
Professor at the ENIT, Address: 37 B.P. Le Belvédère, 1002 Tunis
rachid.nasri@enit.rnu.tn
1, 2,3
Research laboratory on Applied Mechanics and Engineering (LR-MAI-ENIT)
Abstract. Composite structures are used in the aerospace, stars and automotive es-
pecially the structures made of composite sandwich panels which are subject to vi-
bration harmful sources of noise and mechanical failures. In this paper, we propose
to analyze the influence of the shapes of displacement fields on the vibration beha-
vior of a sandwich beam. For this energy method based on the minimum energy is
used to achieve the equations frequencies and a sandwich beam modes. The kinetic
and potential energies of the skins are, in turn, derived from the classical laminate
theory. Several fields of polynomial movements are tested. Other parts will study the
effects of rotational inertia, taking into account the bending energy of the body of
sandwich NIDA. We analyzed the frequencies and modes based on different para-
meters. The experimental data are obtained for recessed-free conditions limits excit-
ing near underrun using an impact hammer. The vibrational response is measured
with a laser vibrometer. The natural frequencies are obtained experimentally by
modal analysis. Numerical simulations complete this work for two types of sand-
wich Nomex paper and aluminum. The natural frequencies obtained from the
theoretical formulation for numerical solution of the system are compared with ex-
perimental results and the results of numerical simulation. The very good agreement
between the results shows that the model is correct.
1 Introduction
Energy methods are among the most important methods to study the vibration
behavior of beams. The problem with these methods is that they require the
2 Mathematical Formulation
Z
Y
c
X
b
a
2(1 − ν 2 ) ∫− fl ∫0
VBF = dz (ε xB ) 2 dx (3)
With the same method we determine the energy of elastic bending deformation
of the upper layer:
E c+ fu a
2 ∫c
V ' BF = dz ∫ (ε ' xB ) 2 dx (4)
2(1 − ν ) 0
To determine the tensile energy of one of the faces we must first determine the
normal force to this surface and its extension in the parallel axis to the face.
N FL : The normal Force at the lower face and N FU : the normal force at the
upper face
544 I. Chenini, Y. Abdelli, and R. Nasri
c + fu
N FU = ∫
0
N FL = ∫ σ f dz (5) And σ f dz (6)
− fl c
VNF : is the tensile energy of the lower layer of the normal force of this face.
ε xL : The elastic deformation in the x direction of the lower face.
ε xU : The elastic deformation in the x direction of the upper face.
1 a 1 a
2 ∫0 2 ∫0
V NF = N FL ε xL dx (12) V ' NF = (− N FL ).(−ε xL )dx (13)
The bending elastic deformation energy of the bottom layer per unit width is as
the following:
∂ 2W c
(14) and ε xU = − 1 (c + f u )( ∂ W c )
2
1
ε xL = (c + f l )( ) (15)
2 ∂x 2 2 ∂x 2
Substituting (10) and (14) in (12) we obtain:
E. f l .(c + f l ) 2 a ∂ 2W c 2
VNF =
8 ∫
0
(
∂x 2
) dx (16)
∂
With: γ XZ = 1 ⋅ τ XZ (19) and τ XZ dx = − N FL dx (20)
G XZ ∂x
The body bending energy is as following:
2
1 a ⎛ ∂ 2W ⎞ c 3 .b
VFC = ∫ Ec .I .⎜⎜ 2 c ⎟⎟ dx With I= (21)
2 0 ⎝ ∂x ⎠ 12
G XZ in Pa 5600.106 70.109
ν 0.33 0.33
546 I. Chenini, Y. Abdelli, and R. Nasri
Second, we take a beam where field checking the no translation and not the
non-rotation on the fixed end section x=0 is:
We present only the results where the differences between fields are observa-
ble. We mention that the fields are the following:
Field 1: order 2 polynomial Field 2: order 3 polynomial
Field 3: order 4 polynomial Field 4: order 5 polynomial
Field 5: order 6 polynomial
377 4000
sans prise en compte de l’inertie rotationnelle et de la flexion dans le corps sans prise en compte de l’inertie rotationnelle et de la flexion dans le corps
373
3000
371
2500
369
2000
367
365
1500
Champs 2 Champs 3 Champs 4 Champs 5
Champs 2 Champs 3 Champs 4 Champs 5
Fileds 1 Filds 2 Filds 3 Filds 4 Fileds 1 Filds 2 Filds 3 Filds 4
Fig. 2 First natural frequency depending Fig. 3 Second natural frequency depend-
of several parameters ing of several parameters
The first pulse from the first displacement fields (order 2 polynomial) with only
one variable is widely superior to other values which tend to the same region limit
370 rad/s if we consider both the effect of of rotational inertia and the bending
body. The value of the 2nd natural frequency tends to a limit close to 2300 rad/s
without the effect of the rotational inertia and converges to 2200 rad / s from the
third field (order 4 polynomial). The value of the third natural frequency con-
verges only from the fourth field (order 5 polynomial).
Fig. 6 Third modal deformation compari- Fig. 7 Fourth modal deformation compari-
son between different fields son between different fields
It can be concluded that the high order frequencies require more order to stabil-
ize and more the mode order rises more we need a higher order polynomial.
We notice that the first modal deformation obtained from the field (1) is not
close to other deformed because of the function that characterizes the displace-
ment field does not define the shear energy of the body (∂ 3W / ∂x 3 = 0) . Even the first
obtained natural frequency is clearly superior to the others results, therefore this
result will be eliminated from our comparative study.
When we increase the number of parameters that characterize the displacement
fields we notice that the natural frequencies and modal deformed curves converge
to a common solution.
Each time we increase the number of parameters that characterize the dis-
placement fields, the calculations become more and more difficult and sometimes
we can‘t solve the systems of the obtained equations where we find unusable
results (problems with the field (5)), that’ s why we will consider the results
obtained from the field (4) as a reference.
From figure 5, 6 and 7 we remark that the deformed reach their max and min in
the same x-axis and also get canceled at the same points.
4 Experimental Study
The test rig, shown in Figure 8, consists of a massive steel structure for embedding
beams and insulation from external vibrations. The beams are excited by an in-
strumented hammer connected to the acquisition system. The vibrational response
of the beam is itself measured with a laser vibrometer.
548 I. Chenini, Y. Abdelli, and R. Nasri
Performing a test of static bending with considering the effect of shear force
(the thickness of the specimen is relatively large) allows us to determine the static
stiffness and EIz. In these calculations the specimen is assumed to be elastic
homogeneous isotropic and linear.
The last two assumptions are verified in the used loading area.
The specimen which is fixed at one end and free on the other is suspended verti-
cally to avoid the effect of static deflection.
The excitation hammer (Figure 10) with rubber tip and incorporated force sen-
sor excites the bar horizontally with a transient load of small amplitude.
The bi- channels spectrum analyzer calculates the transfer function averaged
over several samples. The coherence function is checked at each test and is close
to one.
Fig. 11 Transfer function magnitude and Fig. 12 Transfer function magnitude and
phase of the Al beam phase of the Nomex
5 Numerical Simulation
Fig. 13 First modal deformation of flexion in Fig. 14 First modal deformation of flexion
plan (o,x,y) in plan (o,x,z)
550 I. Chenini, Y. Abdelli, and R. Nasri
Fig. 15 second modal deformation of flexion Fig. 16 First modal deformation of torsion
in plan (o,x,y)
6 Comparative Analysis
Comparing results among them, we see a slight variation between the natural
frequencies and modal deformation determined digital simulation compared with
analytical and experimental results this variation may be due to the not taken into
account the effect of the glue between the layers has a significant effect known
vibration damping.
7 Conclusion
The shape of the displacement field has a great influence on the natural frequency,
the right choice is essential to reach a common solution. The second value of the
first and the third natural frequency converge respectively from a displacement
field of 3, 4 and 5 order polynomial.
Displacement Influence on Frequencies and Modal Deformations 551
For each form of the displacement field we notice the relationship between the
coefficients of the matrix which may be written as follows
For n order polynomial:
i = 1....(n − 1)
M ij = K ij − L ij Avec
j = 1....(n − 1)
Where Lij can be in a general form.
1 2
L ij = ρω 2 a ( 3 + i + j)
2 ( 3 + i + j)
K 1 j = 4 (1 + j) a j ⋅ C K 2 j = N 2 j + (12 ⋅ j) ⋅ a ( j+1) ⋅ C
With relations between the N ij coefficients.
We can suggest to increase the degree of polynomial that characterizes the dis-
placement field or to modify its shape in order to reach at a general relationship.
We can also choose a field of displacement according to the three direction
(x, y, z) wherein for better vibration behavior of our beam while taking account of
the torsion and bending in both horizontal and vertical plane.
References
[1] Mindlin, R.D.: Influence of rotatory inertia and shear on flexural motions of isotropic,
elastic plates. J. Appl. Mech., 1831–1838 (1951)
[2] Yarlagadda, S., Lesieutre, G.: Fiber contribution to modal damping of polymer matrix
composite panels. Journal of Spacecraft and Rockets 32(5), 825–831 (1995)
[3] Wang Deducing, C.M.: thick plate solutions from classical thin plate solutions. Struct.
Eng. Mech. 11(1), 89–104 (2001)
[4] Meunier, M., Shenoi, R.A.: Dynamic analysis of composite sandwich plates with
damping modeled using higher-order shear deformation theory. Composite Struc-
tures 54, 243–254 (2001)
[5] Soula, M., Nasri, R., Ghazel, M.A., Chevalier, Y.: The effects of kine-matic model ap-
proximations on natural frequencies and modal damping of laminated composite
plates. Journal of Sound and Vibration 297, 315–328 (2006)
[6] Maheri, M.R., Adams, R.D.: Steady-state flexural vibration damping of honeycomb
sandwich beams. Composites Science and Technology 52(3), 333–347 (1994)
[7] Banerjee Frequency, J.R.: equation and modal deformation formulae for composite
Timoshenko beams. Composite Structures 51, 381–388 (2001)
[8] Nilsson, E., Nilsson Prediction, A.C.: Measurement of some dynamic properties of
sandwich structures with honeycomb and foam cores. Journal of Sound and Vibra-
tion 251(3), 409–430 (2002)
[9] Blevins, R.D.: Formulas for Natural Frequency and Modal deformation. Krieger Pub-
lishing Company, Florida (2001)
Author Index
Del Rincon, Alfonso Fernandez 459 Larbi, Ahmed Ben Cheikh 453
Deü, Jean-François 349 Larbi, Walid 349
Dhifelaoui, Hafedh 453 Loukarfi, Larbi 9
Driss, Dorra 101
Driss, Slah 131 Maaloul, Makram 101
Driss, Zied 91, 101, 111, 131, 209, Mabrouki, Ibrahim 359
359 Mahfoudi, Chawki 141
Dron, Jean Paul 69 Makki, Mhalla Mohamed 517
Mars, Jamel 469
Elaoud, Sami 339, 409 Masmoudi, Faouzi 289, 327
Meddour, Ikhlas 419
Fakhfakh, Hassen 369
Mejri, Sabrine 307
Felfel, Houssem 289
Merzoug, Mustapha 69
Fudym, Olivier 307
Meziani, S. 19
Ghanmi, Samir 37 Miloudi, Abdelhamid 59, 69, 151
Gloaguen, Jean-Michel 281 Mlayeh, Olfa 101
Gmati, Nabil 401 Mohamed, Anis 401
Graa, Mortadha 47 Mouloud, Guemana 379
Guedri, Mohamed 37 Moussaoui, M. 19
Guidara, Mohamed Amine 391 Mrabet, Elyes 37
Guizani, Amir 189 Mrad, Charfeddine 239, 259
M’Ziou, Nassima 269
Haddar, Mohamed 189, 459
Hadj-Taïeb, Ezzeddine 81, 339, 391, 409 Nasr, Anouar 507
Hafaifa, Ahmed 379 Nasr, Aymen 259
Hafsi, Zahreddine 409 Nasri, Rachid 219, 239, 259, 541
Hajlaoui, Abdessalem 497 Neder, Mahmoud 151
Hammadi, L. 27 Nejlaoui, Mohamed 47
Hammadi, Moncef 189, 249
Hammami, Ahmed 459 Ohayon, Roger 349
Hammami, Omar 299 Ouali, Nourdine 1, 169
Hariri, Said 391
Harras, Bilal 443 Patalano, Stanislao 199, 249
Hocine, Rachida 229
Houidi, Ajmi 47 Rabahi, Abderezak 317
Rachid, Belhadef 379
Issam Ziane, Mohamed 179 Renaud, Ruixian 249
Ichchou, Mohamed 37 Romdhane, Lotfi 47
Rueda, Fernando Viadero 459
Jamel, Mohammed 531
Jarraya, Abdessalem 497
Saidi, Mohamed 121
Kaddeche, Slim 121 Salah, Abid Mohamed 209
Kaffel, Ahmed 111 Samet, Ahmed 359
Kammoun, Imen Kallel 131 Santamaria, Miguel Iglesias 459
Karmed, Djamel 9 Schmitt, Christian 81, 391
Khlifi, Kaouthar 453 Sghaier, Jalila 307
Skander, Achraf 121
Lanteri, Stéphane 401 Soriano, Thierry 189
Lanzotti, Antonio 199 Soula, Mohamed 37
Author Index 555