Академический Документы
Профессиональный Документы
Культура Документы
96-110
Departamento de Engenharia Mec anica, Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias s/n, 4200-465 Porto, Portugal
e-mail: cmav@fe.up.pt and jdr@fe.up.pt
Departamento de Engenharia Mec anica, Universidade de Aveiro, Campus Santiago, 3810-193 Aveiro, Portugal
e-mail: rmoreira@ua.pt Submitted: 05/03/2010 Accepted: 18/04/2010 Appeared: 29/04/2010 c HyperSciences.Publisher
Abstract: This is the second of two companion articles addressing an integrated study on the mathematical modeling and assessment of the eciency of surface mounted or embedded viscoelastic damping treatments, typically used to reduce structural vibration and/or noise radiation from structures, incorporating the adequate use and development of viscoelastic (arbitrary frequency dependent) damping models, along with their nite element (FE) implementation, and the experimental identication of the constitutive behavior of viscoelastic materials. In the rst article [Vasques, C.M.A. et al., Viscoelastic damping technologiesPart I: Modeling and nite element implementation, Journal of Advanced Research in Mechanical Engineering 1(2): 76-95 (2010)] viscoelastic damping has been tackled from a mathematical point of view and the implementation, at the global FE model level, of time and frequency domain methods, namely the internal variables models, Golla-Hughes-McTavish (GHM) and anelastic displacement elds (ADF), and the complex modulus approach based ones, direct frequency response (DFR), iterative modal strain energy (IMSE) and an iterative complex eigensolution (ICE), respectively, were described and formulated. This second article is a natural extension of the rst one. It presents a generic methodology to identify the complex shear modulus of viscoelastic materials. In this case, the complex shear modulus of the well-known viscoelastic material 3M ISD112 is identied and up-to-date values for this material are used and curve-tted in order to obtain the modeling parameters of the GHM and ADF models. Afterward, a viscoelastic sandwich (three-layered) plate specimen and the correspondent FE model are considered numerically and experimentally. Measured and predicted frequency response functions (FRFs) are compared with the purpose of assessing the performance of the damping models presented in the companion article. The analysis allows to assess the validity of the methodology to determine the frequency dependent complex modulus, the GHM and ADF parameters identication procedure and the outcomes and drawbacks of the DFR, IMSE, ICE, GHM and ADF viscoelastic damping modeling strategies and their FE implementations, with the aim of assisting structural designers in the selection of the most appropriate viscoelastic damping modeling approach for their specic needs. Keywords: Damping, viscoelastic, experimental, complex modulus, internal variables, ISD112.
96
www.hypersciences.org
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2) Vasques et al. / Viscoelastic Damping TechnologiesPart II : Experimental / pp. 96-110
1. INTRODUCTION This is the second of two companion articles addressing an integrated study on the mathematical modeling and assessment of the eciency of viscoelastic damping treatments; it incorporates the adequate use and development of viscoelastic (arbitrary frequency dependent) damping models, along with their nite element (FE) implementation, and the experimental identication of the material properties of viscoelastic materials, which allows the use of more realistic material properties data. In the rst article Vasques et al. (2010) alternative approaches to the generation of representative viscoelastically damped structural FE models, e.g. beam, plate and shell structures with surface mounted or embedded viscoelastic damping layers, are discussed. Both time and frequency domainbased viscoelastic damping models are tackled from a mathematical point of view and their implementation at the global FE level is described. This second companion article is a natural extension of the rst one, addressing the experimental identication of viscoelastic material properties, the identication of the Golla-Hughes-McTavish (GHM) and anelastic displacement elds (ADF) model parameters, and the validation and assessment of the dierent viscoelastic damping models/approaches under consideration. The extensive use of passive or hybrid damping treatments using viscoelastic materials to reduce vibration and sound radiation from structures [Nashif et al. (1985), Mac e (1994), Sun and Lu (1995), Benjeddou (2001), Vasques et al. (2006), Vasques and Rodrigues (2008), Vasques and Dias Rodrigues (2008)] has motivated the integration of the viscoelastic damping models discussed in the companion article (Part I) [Vasques et al. (2010)] into commercial or in-house FE codes. As far as commercial FE codes are concerned, they often oer only solution methods based upon the use of direct frequency response and direct integration methods in the frequency and time domains, respectively, while the remainder models (internal variables models and the iterative complex modulus based approaches) proposed in the companion article are usually developed and implemented by the users and structural designers into in-house codes or adaptations of the commercial ones. The direct frequency response (DFR) approach is a very common strategy used by the major part of the structural designers for studying complex structures with viscoelastic damping treatments. This approach is based on the direct application of the complex modulus and allows directly obtaining frequency response functions (FRFs) with the shortcoming of, in principle, demanding a high computational cost. On the contrary, in general, time domain models represent better alternatives to frequency, or complex modulus approach (CMA), based models, allowing the reduction of the computational burden due to the recalculation of the stiness matrix during the frequency sweep and the study of the transient response in a more straightforward (direct) manner, even for highly damped
The joint funding scheme provided by the European Social Fund and Portuguese funds from MCTES through POPH/QREN/Tipologia 4.2 and project PTDC/EME-PME/66741/2006 are gratefully acknowledged by the authors.
structural systems. This latter approach may lead to a numerical analysis with lower computational cost. Therefore, in principle, time domain models would represent better alternatives than frequency domain ones, but this statement cannot be generalized since it strongly depends on parameters of the analysis such as the problem dimension, frequency range of the analysis, constitutive model detail level and damping properties of the materials. Among the time domain models, internal variables models are more interesting from the computational point of view and easiness of implementation into FE codes. Thus, the GHM and ADF models are two attractive alternative methods, used to model the damping behavior of viscoelastic materials in FE analysis, which yield a standard FE formulation (however with the addition of some nonphysical dissipative variables). In order to use them, one needs the GHM and ADF characteristic parameters which allow characterizing the complex (frequency dependent) constitutive behavior of the viscoelastic material being used. To this end, experimental procedures to measure the isotropic constitutive behavior (usually the shear modulus) may become necessary to be devised/performed and numerical identication procedures of the measured data need to be developed. In order to choose the most appropriate material for a specic application, the designer needs some information regarding the damping capabilities of these materials. The latter is usually obtained through the analysis of the complex (frequency dependent) constitutive behavior of viscoelastic materials at dierent temperatures. To that end, normalized and proprietary experimental procedures to the measurement and analysis of the dynamic constitutive behavior (usually isotropic) and numerical procedures for the identication and graphical representation of the complex moduli of viscoelastic materials have been proposed. There are numerous methods for evaluating the performance of damping materials. These methods can be roughly divided into two categories: those whose purpose is to rank the performance of damping materials on a dened structure [e.g. the SAE J1637 (2007) test of the Society of Automotive Engineers, which is based on the so-called Oberst Bar Test Method and is used to rank order materials used in passenger vehicle applications] and those whose purpose is to measure the properties of the damping material alone so that mathematical models can be used to predict its damping performance when applied to many dierent types of structures. As far as measuring the complex moduli of viscoelastic damping materials is concerned, there are various testing methodologies that have been devised through time and they can be divided in two distinct and somewhat complementary categories: (i) direct methods, which are based on dynamic measurements of bare viscoelastic material samples, and (ii) indirect methods, where the properties are inferred from dynamic measurements made on bars with surface mounted or sandwiched viscoelastic damping layers and from assumed underlying mathematical models which relate the composite bar damping behavior to raw viscoelastic material one. Both methods, though, rely on the time-temperature superposition principle to construct
97
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2) Vasques et al. / Viscoelastic Damping TechnologiesPart II : Experimental / pp. 96-110
master curves of the storage modulus and loss factor from data sets measured at several distinct temperatures, inherently assuming that the material is thermorheologically simple [Ferry (1980)]. Regarding the direct methods, they usually consider raw material samples in a single degree of freedom (DoF) system conguration, which is conceptually the simplest means for obtaining complex moduli data, where the damping material sample is deformed in a specic mode, such as shear or extension. These tests can be dynamic or pseudo-static and involve the measurement and acquisition of force and displacement time histories, which are later processed to obtain the complex modulus data. In the former case, the material is dynamically (usually harmonically) deformed and the frequency dependent complex modulus is directly obtained through transmissibility- or impedance-based approaches. Regarding the latter approach, usually a step deformation (relaxation test) or force (creep test), or other types of stimuli, is used to deform the material and it yields the complex moduli indirectly through time-frequency equivalences, based upon the identication of a representative time domain model by tting the measured response to the model. On the other hand, indirect methods consider test specimens where the damping material is surface mounted or embedded in beam structures, with the most popular conguration of the test specimen being the Oberst beam where the damping material is surface mounted only in one side of the beam being subjected mainly to extensional deformations, which was alleged rst proposed and studied by Li enard (1951) and Oberst (1952) in the early 1950s. Other congurations have also been used considering a symmetric Oberst beam with damping material on both sides (so-called Van Hoort beam ) or sandwiched beams [Nashif et al. (1985), Jones (2001), ASTM E756-05 (2005)], or even beams with constrained damping layers [e.g. Hambric et al. (2007)]. These test specimens are used to infer complex elastic moduli by resonance- or impedance-based methods, which use analytic expressions [e.g. Ross et al. (1959), Kerwin (1959), Liao and Wells (2008)] or approximated numerical models [e.g. Wojtowicki et al. (2004), Hambric et al. (2007), Castello et al. (2008)] of distributed parameters systems, that relate the complex modulus of the viscoelastic material to the resonances and loss factors or to the FRFs of the test specimen (beam and the damping material), respectively. The dierent experimental apparatus for measurement of the viscoelastic constitutive behavior by the aforementioned approaches (e.g. resonant beam tests, SDoF resonant and dynamic modulus testing, creep and relaxation tests), along with measurement and determination procedures and representation of viscoelastic materials properties are described and discussed, for example, in monographs and books [Drake and Soovere (1984), Nashif et al. (1985), Tschoegl (1989), Findley et al. (1989), Sun and Lu (1995), Jones (2001), Cremer et al. (2005)], articles [Paxson (1975), Ferguson (1984), Allen and Pinson (1991), Allen (1996), Willis et al. (2001), Etchessahar et al. (2005), Sorvari and Malinen (2007), Jaouen et al. (2008)] and standards [ISO 10112:1991 (1991), ASTM D5026-95a (1995), ANSI S2.21-1998 (R2007) (2007), ANSI S2.22-1998 (R2007)
(2007), ANSI S2.23-1998 (R2007) (2007), ANSI S2.242001 (R2006) (2006), ASTM D5023-01 (2001), ASTM E756-05 (2005), ASTM D5418-07 (2007), ASTM D502407 (2007)] and the references therein. Furthermore, some information of proprietary apparatus and test methods such as the Dynamic Mechanical Analyzer (DMA), Dynamic Mechanical Thermal Analyzer (DMTA), Rheovibron, Autovibron, Viscoanalyzer, RSA II, to name a few, from companies and laboratories such as CSA Engineering (Mountain View, CA, US), Polymer Laboratories (Amherst, MA, US) and Roush Technologies (Livonia, MI, US), can be found in [Allen (1996), Jones (2001), ASTM E756-05 (2005), Melo and Radford (2005), Price et al. (2008)] and the references therein. In the previous references, viscoelastic material master curves, tables and empirical and fractional derivative analytical expressions of the complex modulus and shift factor are presented for several types of viscoelastic materials. However, as referred by Jones (2001), that information is provided from various sources and is fairly accurate, in general, but it is by no means the best data which can be obtained. Better, trustworthy and up-to-date data may be obtained by additional testing or purchase of data from inhouse and proprietary data bases [e.g Drake and Terborg (1980), Drake (1988)]. The viscoelastic material considered in this work is the ISD112 from 3M (1993), which was chosen because of its widespread use, its commercial availability and because most of the related published work on the open literature is related to the application or characterization of this specic material. Regarding its constitutive behavior, an empirical explicit analytical denition of the complex modulus and shift factor obtained by curve tting master curves of 3M ISD112 is given by Drake and Soovere (1984). However, the accuracy of these expressions is not clear and, with time, manufacturers have changed the production process, with implications on the properties of the viscoelastic materials, making old data inaccurate. All these aspects strongly justify the development and usage of an experimental methodology to alternatively test and measure the viscoelastic material complex modulus, which is hereby proposed and validated for the 3M ISD112. However, the proposed methodology to identify the material properties can obviously be adapted and applied to any viscoelastic material and in fact the same experimental procedure and apparatus was also used to identify the frequency dependent material properties of cork compounds as described by Lopes et al. (2006). Regarding the structure of the article it is divided into three main sections: viscoelastic material characterization, identication of constitutive model parameters and experimental assessment and validation of the damping models. Therefore, the experimental test rig and methodology used to identify the complex shear modulus of the viscoelastic material 3M ISD112 are described rst. Next, the curvetted viscoelastic material data is compared with the measured one and the parameters of the GHM and ADF models are presented and used to represent the frequency dependent viscoelastic stiness and to model the viscoelastic damping behavior of the structure. The identied parameters are then utilized in a FE model and the analysis of a sandwich plate with a viscoelastic core and elastic
98
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2) Vasques et al. / Viscoelastic Damping TechnologiesPart II : Experimental / pp. 96-110
skins is performed. The measured and predicted FRFs are compared with the purpose of assessing the performance of all the damping models. The analysis allows to assess the validity of the methodology to determine the frequency dependent complex modulus, the GHM and ADF parameters identication and the outcomes and drawbacks of the direct frequency response (DFR), iterative modal strain energy (IMSE), iterative complex eigensolution (ICE), GHM and ADF viscoelastic damping modeling strategies and their FE implementations presented in the companion article [Vasques et al. (2010)]. The main contributions and novelties of this work are related to the fact that both measured and predicted results are utilized to validate the methodologies used to include viscoelastic damping into FE models. Therefore, the developed experimental procedure to determine the viscoelastic material properties is presented and discussed, and the characteristic material parameters of the GHM and ADF models of the 3M ISD112 are obtained by curvetting the measured shear storage modulus and loss factor of the actual 3M ISD112 material. It is important to point out, though, that the experimental data is usually taken by a manual graphical procedure from the plot of the nomogram given by the manufacturer of the viscoelastic material. A similar study is reported by Trindade et al. (2000) where the GHM, ADF and iterative MSE models are compared. However, such study does not comprise neither an eective assessment with measured experimental results nor a performance comparison with predicted results obtained by a DFR approach, which, being in some circumstances the most time consuming technique is, however, for sure, the most precise in the frequency domain. Additionally, the proposed ICE approach is assessed and compared with other damping modeling strategies. Therefore, to conclude, it is important to emphasize that the damping methodologies proposed here are assessed and compared in a more rigorous way. 2. EXPERIMENTAL IDENTIFICATION OF VISCOELASTIC MATERIALS 2.1 Underlying Analytical Model The proposed experimental methodology for the dynamic characterization of the complex modulus of viscoelastic materials is based on the direct identication of the complex equivalent stiness K (j ) of a discrete single degree of freedom (SDoF) system [Allen (1996)]. The complex stiness is physically materialized by a thin viscoelastic layer which is subjected to cyclic shear deformation imposed by a dynamic exciter. The dynamic response of the considered SDoF system is then used to evaluate the shear storage modulus and loss factor variation with frequency and temperature. The receptance FRF of a SDoF system, assuming stationary harmonic motion, is given by H (j ) = (j ) X 1 , = K (j ) 2 M F
The complex valued stiness K (j ) of the viscoelastic sample can be directly determined through the inverse of the receptance, the dynamic stiness function Z (j ) = H 1 (j ), as K (j ) = 2 M + Z (j ). (2) The viscoelastic material storage modulus G ( ) and loss factor ( ) functions can then be evaluated from G ( ) = h 2 M + Re [Z (j )] , AS Im [Z (j )] , ( ) = 2 M + Re [Z (j )]
(3a)
(3b)
where h is the thickness of the viscoelastic sample and AS its shear area. 2.2 Experimental Apparatus The test rig representing the dynamic SDoF system used to identify the complex shear modulus of viscoelastic materials is illustrated in Figure 1. The underlying analytical model of the SDoF system previously described comprises a moving mass M , represented by the upper bar (2), and a complex stiness K (j ), represented by the thin viscoelastic layer sample (4) introduced between two rigid blocks, as shown in Figure 2. The upper bar is guided by two thin lamina springs (3), which provide the restraint of the spurious DoFs, allowing the viscoelastic specimen to deform mainly in shear due to the relative horizontal motion between the moving bar (2) and the xed bar (1).
(1)
and X (j ) are the amplitudes of the dynamic where F force and displacement response and M is the active mass.
The dynamic excitation of the moving bar is provided by an electrodynamic shaker (Ling Dynamic Systems - model 201), using a thin stinger to minimize the rotation and lateral excitation, driven by a random signal generated by
99
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2) Vasques et al. / Viscoelastic Damping TechnologiesPart II : Experimental / pp. 96-110
the signal generator of a dynamic signal analyzer (Br uel & Kjr - model 2035) and amplied by a power amplier (Ling Dynamic Systems - PA25E). The applied excitation force is measured using a piezoelectric force transducer (Br uel & Kjr - model 8200). The acceleration response of the moving mass is measured with a piezoelectric accelerometer (Br uel & Kjr - model 4371) and the relative displacement response with a proximity probe (Philips PR6423). The signal conditioning and the frequency response functions are determined with the aforementioned dynamic signal analyzer. The test rig and the electrodynamic shaker are rigidly assembled onto an inertial block (Figure 3), which is supported by rubber pads in order to reduce the inuence of the rigid body modes of the assembly on the utile bandwidth. Furthermore, the experimental apparatus was introduced into a thermal chamber, providing nearly isothermal conditions between 0 and 35 C; the temperature range is limited by the shaker admissible range of operating temperature. The analysis at various temperatures allows the use of the temperature-frequency equivalence principle [Jones (2001)], extending the frequency range of the characterization and allowing data correlation. The temperature of the viscoelastic material is evaluated using two thermocouple probes located near the specimen, as depicted in Figure 2.
Viscoelastic damping is a property exhibited by a wide variety of materials such as polymeric materials, ranging from natural or synthetic rubber to industrial plastics, or even cork and cork compounds [Lopes et al. (2006)]. The class of polymer materials is extremely wide and many polymeric compounds, displaying somewhat dierent complex modulus properties, available from commercial manufacturers and other sources [see Nashif et al. (1985) and Jones (2001) for manufacturers] can be found. Several viscoelastic materials, specically tailored for passive damping treatments, are commercially available in the market, both as raw materials for OEM application or available as part of a damping solution service/product. The ISD series from 3MTM company (ISD110, 112 and 113), the DYAD series from Soundcoat R company (DYAD601, 606 and 609) and the HIP2 from the Heathcote Industrial Plastics company are just a few examples of such materials. In this work, a special attention is given to the 3M ISD112 material, which is experimentally characterized and applied as a damping treatment in the experimental specimen used to assess the dierent viscoelastic damping models and approaches. As stated before, the main reason for its choice is related to the fact that it is the viscoelastic material more often used in most of the numerical and experimental studies published in the open literature. Nevertheless, the experimental and numerical methodologies hereby implemented and applied may be straightforwardly applied and adapted to characterize other materials with a somewhat dierent viscoelastic constitutive behavior. The receptance and accelerance FRFs of a 0.57 mm 480 mm2 3M ISD112 viscoelastic material specimen, measured in a bandwidth of [0 400] Hz at nine dierent temperatures between 2.7 and 33.5 C are represented in Figure 4. The obtained FRFs, whether using directly the receptance or the accelerance FRFs, are the inputs for the complex modulus description calculated according to Equations (3). In order to identify and lter possible random errors due to the high frequency noise, rigid body modes or other error sources, a Wicket plot representation, which is proved to be a useful graphical representation to highlight erroneous data values [see Jones (2001) for further details], is applied to the identied complex modulus data. In this graphical representation the storage modulus calculated from the data at dierent temperatures is plotted against the corresponding loss modulus values (or loss factor values). Figure 5 (a) represents the Wicket plot of the identied complex modulus data of the 3M ISD112 specimen without any ltering of scattered data. From the analysis of this plot, the master curve of the complex modulus can be determined, making it possible to identify erroneous data sets and thus to eliminate the data not following the main tendency of the data distribution. After ltering the data, i.e. removing all scattered points, a reliable data set of complex modulus is obtained, as illustrated in Figure 5 (b), which can be used to characterize the viscoelastic material under analysis. The dierent sets of complex modulus functions, identied at dierent temperatures, should nally be correlated to identify the shift factor T (T ) distribution of the material by using the frequency-temperature equivalence principle.
Fig. 3. Experimental apparatus into the thermal chamber. 2.3 Analysis of Measured Complex Modulus Data The receptance and accelerance FRFs of the SDoF dynamic system described in the previous sections are measured using the transducer response signals (proximity probe, accelerometer and force transducer) for dierent temperatures. Since two dierent FRFs were measured using two dierent transducer responses (displacement and acceleration), the corresponding FRFs can be correlated in order to identify high frequency noise generated by the thin springs, the stinger and shaker trunnion (highly evidenced by the receptance results), and the rigid body modes eects at low frequencies (evidenced in the accelerance results). Therefore, the simultaneous use of both transducers allows to correlate the two response measurements in order to conveniently combine information from both measurements and to enlarge the utile frequency range.
100
Journal of Advanced Research in Mechanical Engineering (Vol.1-2010/Iss.2) Vasques et al. / Viscoelastic Damping TechnologiesPart II : Experimental / pp. 96-110
10
10