You are on page 1of 17

A FINITE ELEMENT MODEL OF A SEMI-TRAILER LEAF SPRING INCLUDING FRICTION BETWEEN LEAVES.

Authors:

Jorge Rodriguez Matienzo Department of Applied Mechanics, CUJAE, Calle 127 s/n, Marianao, La Habana, Cuba. Tel + (537) 260-2267; fax + (537) 27-2964 e-mail matienzo@mecanica.cujae.edu.cu

Introduction.

Leaf springs have its main use in automotive suspensions. O. Elliot placed them first in a carriage in 1804; H. Ford was pioneer in using alloy steel in leaf springs in 1908. Since then, several efforts have been made in order to obtain a reliable design model. Nowadays devices like leaf springs are used too in orthopedic prosthesis, measurements systems, etc. The main problem in leaf spring theory is the determination of leaves shape under load: knowing the shape, the stresses can be determined. A theoretical formulation to determine the shape of several cantilevered metallic beams placed one onto another under bending, without considering the friction, is exposed by Osipenko et. al. [9]. Expressions for determining gap, press, displacement, etc. are shown. The author of this paper could probe these theoretical results with a FEM model and good agreement was obtained. A comparative study of a steel leaf spring and another of composites is made by Shokrie M. M. and Rezaei D [12], for automotive use. Limited results of 4 sheets leaf spring under static load by FEM model are given. These are compared with those using the procedure of SAE HS 788 and other experimental ones, showing good agreement. The causes of failure of leaf springs are studied in Mukhopadhyay N. K. et al. [8]. The most frequent premature failure is attributable to the fatigue. The author highlights the importance of controlling the hardening process in what is referred to the reduction of the superficial cracks and remarks the correct selection of the sheets thickness. For this, a well-suited leaf spring model, which takes into account contact and friction, is needed. The non-lineal character of the leaf springs is revealed in the simulations of the dynamics of vehicles, mainly in cases of heavy vehicles or articulates. In Potter C. [10] several ways for modeling a suspension based on leaf springs or air are discussed. In the case of the leaf spring is pointed out that has the biggest non-lineal character between automotive suspensions, manifested in the model with an error of 45%, using the UMTRI [13] method. To reduce the error the author intends to use a non-lineal parametric model combined with dynamic measurements in the vehicle.

The no-linearity of leaf spring suspensions and its implication in vehicle modeling is pointed out also in Gerdes J. C., et. al [2]. The ADAMS package was used, substituting leaf springs by other elements. The data on rigidity and friction force against load employed confirm the abovementioned, and in the same way the author stands out the difficulty of modeling the non-linearity correctly. In Winkler C. et al [14], according to the data of deformation against load on the suspension, is observed how the no-linearity diminishes with the load on the suspension, just as Potter C. says in [10]. The modeling of leaf springs considering the phenomena of the contact between sheets and the friction continues being a current problem. The study of the grade of no-linearity and its influence in the dynamics of the suspensions depends in great measure at this moment on experimental data, for lack of more representatives models. The treatment of the contact problem using FEM has received special attention for the investigators in recent years. Since Signorinis works between 1933 and 1959, a lot had advanced in the formulation. A excellent summary is made by Mijar A. R., Arora J. S. in [7], it is focused on the approaches using variational equalities or variational inequalities, optimization techniques and the formulation of general frictional contact problem; the penalty methods and augmented Lagrangian method are discussed. In general, a numerical method is necessary to solve the non-linear problem, and select a group of factors appropriately to ensure convergence, and at a reasonable time, avoiding an "ill-conditioned problem. The value of the penalty is of great importance; Gu R. J., Murty P., Zheng Q. in [4] presented a strategy to achieve the numerical convergence and a correct result in a general contact problem. They propose a variable penalty that ensures proper values during all the contact process. Structuring of the pattern in what concerns to discretizacin size, values of rigidity, adjustment of looseness, etc., is recommended by Johnson D. in [5]. The simulation of Coulomb friction law is usually made by an approach that considers static and kinematics friction without varying the friction coefficient, starting from the tangential stresses, which is valid if the sliding speed is low enough [7, 10].

It is useful to point out that in some contact problems the goal is not to the contact efforts itself (like in the of Hertz problem), but the behavior in another area outside of the contact zone. An example are gears, in which beside the contact stresses, the flexion stresses are needed in the tooth base, Refaat M. H., Meguid S. A [11 ]. The most of revised papers refers to contact problems and FEM modeling that include buckling and changes in relative position, etc. The study of bending dominated problems including contact and friction has been minor, and that is the case of a leaf springs. An appropriate model for a leaf spring should include the phenomenon of contact and friction and allow a solution at a reasonable time and not be "ill-conditioned". This will be a no-linear model, and should allow to determinate displacement, stresses, the amount of non-linearity and load range in which it is more significant.

The contact problem with friction.

Frictional contact problem emerge in engineering practice in many areas, but in most of the cases are treated in a simplified way, due to its inherent complexity. This arises due to several aspects, such as inherent nonlinearity. The contacts surfaces changes during contact process, also the boundary conditions (displacements and contact forces). The friction law is non-smooth and multivalued, requiring a special treatment. Friction can be static, quasi-static or kinematic. In some cases, a large deformation occurs, and the structure responds in a nonlinear way. The general approach to a contact problem is represented in Fig 1.

Fig. 1, Configuration before and after contact. The contact object has a volume and a surface , the target surface is , and movement of the contact object, prescribed by displacement vector u, has boundaries and occurs under body and surface forces, not represented. Two candidate contact surfaces on contact and target objects are defined: c and c , respectively. Particles of these candidate surfaces a1 and a2 will be in contact each with others. For each particle in both candidate surfaces a position vector is defined as x1 and x2, the distance between objects is minimized as:
1 2 1 2

do = min x1 x 2

(1)

This defines the initial gap between surfaces, expressed through the position of the most closed particles. The current gap during contact process is:

dn = do u n 0
1

(2)

where u correspond to the closest particle in c . Once the contact have been established, the contact object is deformed, a contact force tc arises; this force has two components, one in normal direction tn and the other in tangential direction tf . These forces are termed normal contact force and tangential contact force, respectively. The last is due to friction between surfaces. The condition of unilateral contact is characterized as:

gn = dn 0,

t n 0,

gn tn = 0

(3)

the first in (3) set that no penetration will occurs, the second that objects will pull each other, and the third that both bodies are or separated:

tn = 0, gn < 0
or pressed:

tn 0, gn = 0
The Coulombs classical friction law is used. Regarding its inability in handling dependence of friction coefficient on relative sliding, it is avoided by considering low relative velocities. No distinction is made between static and kinematic coefficient of friction. The bodies in contact either stick or slides depending on following conditions, is the friction coefficient:

t f tn < 0 ut = 0

Stick condition Sliding condition

(4) (5)

t f t n = 0 ut > 0

The virtual work equation formula for an elastostatic frictional contact problem is:
n f k k ij ijd = biuid + TiuidS + ti uidS + ti uidS + Fi ui s
2 c 2 c

(6)

Ti represents surface loads applied in 1 , bi represents body loads in , ui and I are virtual
displacements and deformations Fki are concentrated forces acting in k-th points. is a part of
2

c2 . The solution consists in finding a displacement field ui and contact forces tn and tf such that
all the boundary conditions, including contact and friction are satisfied. As contact surface and contact forces are unknown, an incremental solution is required, in which external load is applied gradually and Newton-Raphson method is employed. The contact condition is imposed treating contact force as additional unknown variable in Lagrange method, or as a penalty function in penalty regularization. The final solution can be represented as an equilibrium position of the system, which is achieved when the potential energy reaches a minimum. The expression for potential energy is, according to Bathe [1]:

1 k ijCijkl kld uib i d ui Ti dS uk i Fi 2 k


s

(7)

Cijkl is the stiffness of the system. The discretized form for (7) can be written as: (U) = UT K U UT F
(8)

where U is a global displacements vector, K is a global stiffness matrix and F is a global load vector. The requirements of minimal potential energy must be satisfied in conjunction with the boundary condition

min (u)
u

dn 0 .

The problem is treated as an optimization problem with constraints. The use of Lagrange multiplier method is expressed as in (9). A Lagrangian function is defined:

1 L = ku2 fu + fn (gn + s2 ) 2

(9)

fn is the Lagrange multiplier and s2 is a slack variable for the inequality constraints, satisfying
Karush-Kunh-Tucker (KKT) necessary conditions for gradients operators on L. Here fn is equivalent to a force required to impose the constraint: the contact (compressive) force. In the penalty method, a penalty function (10) is introduced as in (11):

p = r (g n ) 2 +
1 1 p = + p = ku2 fu + r (gn ) 2 + 2 2
where r is the penalty parameter and (gn)+ = max {0, gn} . When r

1 2

(10)

(11)

, the solution of (11) is

the solution of the original problem and [9] if the Lagrange multiplier is expressed as f n = rgn The penalty value r is like the stiffness of an artificial spring placed between contacting surfaces. Is easy to understand that in order to satisfy the no penetration boundary condition, r should be infinitely high, but it implies convergence problems in the numerical solution. Both methods: Lagrange and penalty are implemented in most of commercial FEM codes.

Modeling an automotive leaf spring.

The automotive leaf spring from a semi trailer was modeled using ANSYS 5.7. The general dimensions are shown below. No. of leaf 3 1 1 1 1 1 1 1 1 Large, mm 1060 770 660 600 510 440 380 320 220

Each leaf has 3/8 thickness and 90 mm width, made from ABNT 5160 steel, with limits YIELD =1010 MPa TENSILE =1200 MPa HB= 350 E=2.1 e 5 MPa = 0.3

The suspension carries a load of 40 kN with 100 mm of displacement, both extremes are free supported. One central bolt maintains all leaves jointed together in vertical direction, a pair of brackets hold leaves and eliminate lateral relative movement, at same time guarantying contact between superficies. A central bracket attaches leaf spring to the rear axle.

Fig. 2 Leaf spring front view. A 2D model was built using plane stress with thickness. In the model, only the central bolt and rear axle are implemented. The central bolt was introduced coupling UX and UY degrees of freedom of coincident nodes in the vertical symmetry axis. The axles external nodes are constrained in all degrees of freedom, except for the upper area. The contact surfaces were modeled with surface-to-surface contact elements. These elements were placed between leaves

and between the lower leaf and rear axle. Load was applied in the upper leaf, in two nodes, one in each extreme. Several values are necessary for contact elements. The most important is the penalty value, normal stiffness KN of each of the 13 contact sets. Assigning an enough high value ensures correct results, but convergence problems will arise without doubts. On the other side, lower values give a rapid convergence rate, but the risks of ill-conditioned problem exist, also with an excessive penetration between solids in contact. An approximate solution is to run the model, observe its convergence graphic, and analyze penetration. Then, correct in one or another direction and re-run the model. As a thumb of rule, if the displacement of the model for a given load is known, and the problem is bending dominated, the normal stiffness KN is calculated as:

KN =

load 1 displacement n

(12)

where n is the number of beams.

This is only an initial value; the numerical calculation will do the rest. Once this normal stiffness is given, another penalty is determined: the tangential stiffness KT,

KT = KN
Other important parameters should be established, like the amount of permissible penetration, the initial contact between surfaces and the behavior surfaces in contact, once it is reached. An amount of geometrical penetration is needed for numerical calculation. Using Lagrange method the penetration is controlled to be always below the limit, in the penalty method the final penetration is valued. The rigid body motion arise in many contact problems, it should be avoided in a static analysis in order to prevent zero or negative pivot, ought to an unconstrained model,

so, an initial contact between surfaces have to be done. Several types of contact surface behaviors are possible (no separation, no sliding, and standard), the proper for the problem would be selected. The model was built taking all this into account. A values of 36 MPa for KN and =0,3 were taken, KT was rounded in 12 MPa, with minimal permissible penetration of 0,47 mm (5% of thickness), closing the gap at the beginning of contact process (this implies an amount of initial penetration due to discretizacin) and added to the coupled degrees of freedom in coincident nodes forbids rigid body or zero pivot; contact between leaves will be maintained once it is established. Some results are shown below and in Fig. 3.

Fig. 3, Contour plots of displacements UY. The displacement under full load is 99.2 mm, a 0,8 % error related to nominal value. Maximum stresses by Von Misses are in node 809, 869 MPa, in the zone where 3rd leaf is supported by the 4th. Also significant values of stress are encountered in 1st and 2nd leaves. The amount of maximal penetration is 0,49 mm, located in the extremes of leaf 4th to 13th. The gap between leaves is in the range of 0,4 to 0,8 mm, mainly between leaves 4th and 5th. Pressure between leaves reaches 34-38 MPa in the zone of maximum penetration, between leaves 4th and 5th; the frictional stress is 5-7 MPa in the same zone. Being the yield stress of ABTN 5160 of 1010 MPa, this result looks a little high, but allowable; safety factor would be 1,17.

Plotting load values versus displacement in fig. 4 shows the non-linear character of a leaf spring, and also that this non-linearity diminishes with load. Referring to spring rate, it varies with displacements in a reduced band.

load
45000 40000 35000 30000 25000 20000 15000 10000 5000 0 1

load and spring rate vs displacement spring rate


402 400 398 396 394 392 390 388 386 2 4 6 9 14 22 32 43 53 63 73 83 93 101

disp.

Fig. 4 Load and spring rate vs displacement on a leaf spring. A modal analysis was made with the same boundary conditions, excluding friction. This is done since FEM model only consider viscous damping, but Coulomb friction is determining in this case. First six natural frequencies are in table below. I MODE II MODE III MODE IV MODE V MODE VI MODE The shape in the first four modes is in fig. 5 and 6. 0,992 Hz 1,351 Hz 4,023 Hz 5,385 Hz 7,631 Hz 13,794 Hz

Fig. 5, I and II modes

Fig. 6, III and IV modes It is a common practice to limit suspensions firsts natural frequencies between 1 and 2,5 Hz in order to avoid motion sickness or harsh ride. According to Gillespie [23] the whole suspension natural frequency is o :

o = 0.159
where:

(kss * k t ) (k ss + k t ) * Ms

o = suspension natural frequency, klf = leaf spring rate, kt = tire spring constant, Ms = sprung mass.

Here, klf is load dependent (so is kt ), so different values will be obtained for a load or unload vehicle. Optimum would be between 1,5 and 2 Hz, if the tires are included, the results would be in that range. The frequency response of the leaf spring in the range 0-20 Hz is showed in fig. 7. L1 correspond to one of the nodes in which load is applied, in leaf 1; L2 and L3 (almost coincident in FRF) to nodes directly below but in leaves 2 and 3. L4 and L5 correspond to nodes in the extreme of leaves 4 and 5, in all cases nodes moves in vertical direction. It is clearly noted that L1, L2 and L3 have resonance in the 1st, 3rd and 6th natural frequencies, with L2 and L3 with a peak in the first anti-resonance of L1. The others leaves have a different FRF shape but with resonance at 2nd and 5th natural frequencies too.

Fig. 7, FRF in 6 nodes, UY.

Conclusions A finite element model of a semi-trailer leaf spring has been made, considering contact and friction between leaves. A rule for estimate the penalty value normal stiffness of the contact pair for this kind of problem is given, so are another tips for modeling. The static analysis shows that stresses are below yield point, and are maximum in the principal leaves, being the most higher where the 3rd leaf is supported by the 4th. The model permits to evaluate the pressure and the friction between leaves. The modal analysis give natural frequencies of the leaf spring, in a frictionless case, without tires; the expected natural frequencies of whole suspension once the spring rate of tires is considered should be in the proper range. The FRF also illustrate the principal modes of vibration of the first leaves.

Bibliografia. [1] [2] Bathe, K. J.: Finite elements procedures. Prentice Hall, New Jersey, 1996. Gerdes J. C., et. al: Safety performance and robustness of heavy vehicles AVCS. California PATH program. Report MOU-390, 2002. [3] [4] Gillespie T. D.: Fundamentals of vehicle dynamics. SAE, Pennsylvania, 1992. Gu R. J., Murty P., Zheng Q.: Use of penalty variable in finite elements analysis of contacting objects. Computers and Structures, 80, pp 2449-2459, 2002, Pergamon Press. [5] [6] Jhonson D.: Principles of simulating contact between parts using ANSYS. (sin fecha). Lebon F.: Contact problems with friction: models and simulation. Simulation modeling Practice and Theory 11, pp 449-463, 2003. Elsevier. [7] Mijar A. R., Arora J. S.: Review of formulations for elastostatic frictional contact problem. Structures Multidisciplinary Optimization, No. 20, pp 167-189. Springler-Verlag, 2000. [8] Mukhopadhyay N. K. et al.: Premature failure of a leaf spring due to improper material processing. Engineering Failure Analysis. Vol 4, No. 3, 1997, pp 161-170, Elsevier [9] Osipenko, M. A. et al.: A contact problem in the theory of leaf spring bending. International Journal of Solids and Structures, 40, 2003; Pergamon Press. [10] Potter C.: Assessing the relative road damaging potential of HGV. Proceeding of International Symposium on Heavy Vehicles Weight and Dimensions, Michigan, 1995. [11] Refaat M. H., Meguid S. A.: On the contact stress analysis of spur gears using variational inequalities. Computers and Structures, Vol. 57, No. 5, pp 871-882, 1995, Pergamon Press.

[12]

Shokrie M. M. , Rezaei D.: Analisys and optimization of a composite leaf spring. Composite Structures, 60, 2003, pp 317-325, Elsevier.

[13]

Verheul C. and Sayfield Int.: Implementation of a leaf spring model in ADAMS/Rail. Sayfield International, Adams/Rail UC, 1999

[14]

Winkler C. et al: Parameter measurements of highway tractor and semi-trailer. Technical report UMTRI-95-47, Univ. of Michigan Transportation Research Institute, December 1995.

You might also like