EngOpt 2012 - International Conference on Engineering Optimization

Rio de Janeiro, Brazil, 1-5 July 2012.

Material identication of viscoelastic core materials in sandwich structures

IDMEC/IST - Instituto Superior Tcnico, Technical University of Lisbon, Portugal e e-mail: [email protected] COPPE/UFRJ - Federal University of Rio de Janeiro, Brazil e-mail: [email protected] Abstract

The inverse problem of estimating viscoelastic material parameters is addressed in this paper with application to sandwich plates with a viscoelastic core and laminated composite face layers. We use a sandwich plate nite element model, where the viscoelastic properties of the core layer are modeled using the complex modulus approach. Fractional derivative viscoelastic material models are used for the sandwich core frequency dependent properties due to the fact that these models use a minimal set of parameters, which are the ones to be estimated. The sandwich is modeled using a plate nite element model with a layerwise approach, where the laminated composite face layers are modeled with a rst order shear deformation theories and the viscoelastic core is modeled using a higher order shear deformation theory with cubic expansion of the in-plane displacements in the thickness coordinate. The nite element is an eight node serendipity plate element with 13 degrees of freedom per node. The inverse problem is solved using a gradient based approach with FAIPA - Feasible Arc Interior Point Algorithm. Parameter estimation is conducted using only natural frequencies and corresponding modal loss factors.

Keywords: Sandwich Structures, Viscoelasticity, Inverse Methods.

Introduction

This paper addresses viscoelastic parameter estimation of laminated sandwich composite structures with viscoelastic core. One of the most reliable experimental information that can be easily obtained and used for the identication of material parameters are natural frequencies of free vibration and the associated modal loss factors, as they are less prone to noise and measurement errors than other quantities. A recent survey on inverse techniques for parameter estimation of mechanical as well as piezoelectric properties on composite and adaptive structures was conducted by Arajo et al. [1], where the estimation of damping u parameters is also covered, essentially using hysteretic damping models. In this work we conduct estimation of frequency dependent viscoelastic material parameters in sandwich structures core materials, using only modal parameters (natural frequencies and modal loss factors). The inverse problem of identifying the viscoelastic material parameters is then formulated as a constrained optimization problem and solved using a gradient based method with analytic sensitivities.

Sandwich plate model

A layerwise nite element model is used to analyse sandwich laminated plates with a viscoelastic (v) core and laminated anisotropic face layers (e1 , e2 ), as shown in Figure 1. The basic assumptions in the development of the sandwich plate model are: all points on a normal to the plate have the same transverse displacement; no slip occurs at the interfaces between layers; the displacement is C 0 along the interfaces; elastic layers are modelled with rst order shear deformation theory (FSDT) and viscoelastic core with a higher order shear deformation theory (HSDT); all materials are homogeneous and orthotropic; elastic layers (e1 ) and (e2 ) are made of laminated composite materials; for the viscoelastic core (v), material properties are complex and frequency dependent. Applying displacement continuity conditions at the layer interfaces, allows us to retain 13 unknowns. 1

Figure 1: Sandwich plate As for the constitutive relations, we consider that bre-reinforced laminae in elastic multi-layers (e1 ) and (e2 ), and viscoelastic core (v) are characterized as orthotropic. For the viscoelastic core layer, the stiness coecients are complex and frequency dependent quantities, since the complex modulus approach was used in this work, using the elastic-viscoelastic correspondence principle. The equations of motion for the plate are obtained by applying Hamiltons principle, using an eight node serendipity plate element with 13 degrees of freedom per node. Assuming harmonic vibrations, the nal equilibrium equations are obtained for the free vibration problem: [K() M] un = 0 n where, the complex eigenvalue is written as: n = n (1 + in ) n (2) (1)

and n is the real part of the complex eigenvalue and n is the corresponding modal loss factor. One should note that the viscoelastic behaviour of the core translates into a complex and frequency dependent stiness matrix K() and the non-linear eigenvalue problem is solved iteratively.

Parameter estimation

The inverse eigenvalue problem is solved by minimising an error function, with respect to the frequency dependent damping material parameters, subject to design constraints based on thermodynamic requirements of isothermal linear viscoelasticity. The error function is built as the sum of two terms as follows:
I

=
n=1

w n

n n

+
n=1

w n

n n

(3)

where n are the real parts of the experimental eigenvalues, n are the corresponding real parts of the eigenvalues predicted by the numerical model, n and n are, respectively, experimental and numerical modal loss factors, wn and wn are weights used to express the condence level in each experimental eigenvalue and corresponding loss factor, respectively, and I is the total number of experimental eigenpairs under consideration. It is considered that the core is a linear isotropic viscoelastic material whose mechanical behavior can be described by means of a model based on the concept of fractional derivatives. Such model can be easily handled in order to express its complex shear modulus as a function of four model parameters [2]: G(j) = G0 2 1 + a(j) 1 + b(j) (4)

where a,b, , and G0 are parameters to be identied, with G0 the static shear modulus. It can be shown that this model is causal and satises the thermodynamic constraints on linear viscoelasticity [2], as long as the model parameters are non-negative. It is furthermore assumed that 0 1. The constrained minimisation problem is solved using FAIPA, the Feasible Arc Interior Point Algorithm with analytic sensitivities: = 2 wn xi n=1 where: n 1 = xi n ( ) ( ) n n n xi xi (6)
I

n n

1 n 2 wn n xi n=1

n n

1 n n xi

(5)

with ( ) and ( ) being the real and imaginary parts of the complex eigenvalue , respectively. n n n The derivative of the complex eigenvalue with respect to design variables is obtained, for elastic and viscoelastic design variables, using the following expression [5]:
uu uT Kuu (,xi ) Mxi(xi ) un n n xi n = xi uT Muu 2n Kuu (,xi ) un n n

(7)

where n = is the resonant frequency, and all derivatives are evaluated at = n . The derivatives of the system matrices are obtained analytically at the element level, where the usual nite element assembly procedure is used.

( ) n

Application

The inverse technique is applied to estimate viscoelastic material properties of a 3M damping polymer ISD-112 used as a core material in a sandwich panel with 2024-T6 aluminium faces, as described in Araujo et al. [3]. The in-plane dimensions of the panel are 300 mm 50 mm, the thickness of the layers are hv = 0.1016 mm for the viscoelastic core, and he1 = 1.2 mm and he2 = 0.8 mm for the elastic aluminium faces. For the 2024-T6 aluminium faces the following isotropic material properties are considered: E = 64 GPa, = 0.32, with specic mass = 2695 kg/m3 . For the ISD-112 core, Poissons ratio = 0.49 and material density = 1000 kg/m3 are considered known material properties. A 24 4 nite element mesh was used, corresponding to a total of 2832 degrees of freedom, associated to the 4 parameter fractional derivative model. The panel is clamped at one end and free on all the other three edges, in a cantilever beam conguration, and are presented in Table 1, along with the corresponding residuals obtained after identication, both by the present method associated to the 4 parameter fractional derivative model, and by the method reported by Barkanov et al. [4]. Residuals are dened as: fn fn n n 100 ; rn = 100 (8) n fn where fn and n are the natural frequencies and modal loss factors of the identied model. Figures 2 and 3 show the present identied material parameter frequency curves (identied parameters G0 = 0.041 MPa, a = 0.660, b = 3.35 104 and = 0.624) along with the ones reported by Barkanov et al. [4], using the curve-tted material model [4] of ISD-112 for the frequency range f = 5 . . . 1600 Hz: rfn = G = 4.759 0.9266/z + 2.405z 2 with z = 0.1918 + 0.0005148f G = E = 1.385 0.03673z 0.01342/z with z = 0.01 + 0.0006306f 3 [MPa] (9)

Table 1: Experimental natural frequencies and modal loss factors, and corresponding residuals after identication for 2024-T6 aluminium face sandwich with ISD-112 core Barkanov et al. [4] rfn [%] rn [%] 0.0 -1.0 1.1 2.8 2.9 -1.2 1.2 8.3 0.0 -3.9 7.7 -3.1 6.7 -2.9 Present [%] rn [%] 9.5 -0.2 -0.2 1.0 1.6 -2.3 -0.2 4.2 3.3 0.7 11.8 -4.3 4.3 -7.8

n 1 2 3 4 5 6 7

fn [Hz] 16 100 268 496 785 1180 1570

n [%] 12 22 26 26 32 30 34

rfn

Figure 2: Identied frequency curves for shear storage modulus The regenerated frequency response curves for the tip displacement of the beam, corresponding to an impulsive excitation force of 1 N applied also at the tip of the beam, are presented in Figure 4 for both the present identied material model (fractional derivative model with identied parameters) and the material model identied by Barkanov et al. [4], coresponding to Equation (9). The discrepancies found may be due to the dierent methods of identication, since the numerical models are substantially dierent and the identication technique reported by Barkanov et al. [4] makes use of response surfaces, which are themselves approximations to the numerical model. Also the clamped boundary conditions are not the most adequate for identication purposes.

Conclusions

Identication of frequency dependent properties of viscoelastic core materials in a composite sandwich conguration has been addressed, and results from an identication conducted with experimentally measured resonant frequencies and corresponding modal loss factors are presented and compared with alternative approaches [4], indicating that the inverse problem can be solved using only information from eigenfrequencies and modal loss factors, associated to fractional derivative viscoelastic behaviour models.

Figure 3: Identied frequency curves for shear loss factor

10-2

Tip displacement [m]

10-3

10-4

10-5

10-6

200

400

600

800 Frequency [Hz]

1000

1200

1400

1600

Figure 4: Regenerated frequency response curves after identication for the two material models

Acknowledgments
The authors thank the nancial support of FCT Project PTDC/EME-PME/120830/2010 (Portugal) and FCT-CAPES Proc. 4.4.1.00 CAPES (Brazil).

References

[1] Arajo A L, Mota Soares C M, Mota Soares C A. Inverse techniques for the characterization of u mechanical and piezoelectric properties on composite and adaptive structures: a survey. Computational Technology Reviews, 2010, 2, 103-123. Ed. B.H.V. Topping, J.M. Adam, F.J. Pallares, R. Bru, M.L. Romero. Saxe-Coburg Publications. [2] Pritz T. Analysis of four-parameter fractional derivative model of real solid materials. Journal of

Sound and Vibration, 1996, 195, 103-115. [3] Arajo A L, Mota Soares C M, Mota Soares C A, Herskovits J. Characterization by inverse techniques u of elastic, viscoelastic and piezoelectric properties of anisotropic sandwich adaptive structures. Applied Composite Materials, 2010, 17, 543-556. [4] Barkanov E, Skukis E, Petitjean B. Characterization of viscoelastic layers in sandwich panels via an inverse technique. Journal of Sound and Vibration, 2009, 327, 402-412. [5] Haftka R.T., Gurdal Z. Elements of Structural Optimization. Kluwer Academic Publishers, Dordrecht, 1992