Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 49, 1, pp. 51-70, Warsaw 2011 APPLICATION OF ETR FOR DIAGNOSIS OF DAMAGE IN STEEL-CONCRETE COMPOSITE BEAMS Tomasz Wróblewski Małgorzata Jarosińska West Pomeranian University of Technology, Departament of Theory of Structure, Szczecin, Poland e-mail: wroblewski@zut.edu.pl; jarosinska@zut.edu.pl Stefan Berczyński West Pomeranian University of Technology, Institute of Manufacturing Engineering, Szczecin, Poland; e-mail: stefan.berczynski@zut.edu.pl The paper presents howETR (energy transfer ratio) changes for a steel- concrete composite beam depending on the degree of damage. A nume- rical model of composite beamwas constructed on the basis of research conducted on a real beam. The study presents how the frequency of na- tural vibrations, the damping ratio andETRalter due to changes in the structure of the beam.Theperformed studies showthatETR is themost sensitive parameter to damage taking place in the beam. Consequently, this parameter can be used to diagnose damage in this kind of beams. Key words: steel-concrete composite beams, nonproportional damping, energy transfer ratio (ETR) 1. Introduction Modal analysis is evermore often used to detect damage in engineering struc- tures.Anearlydetection andability to locate damage isparticularly important in bridge structures. Steel-concrete composite beams are often used as main carrying girders in bridge constructions. This kind of beams is composed of a steel girder and a reinforced concrete slab placed onto it. This connection is made by welding steel elements to the top flange of a girder. These elements will be later embedded in concrete thus creating a fixed connection between the reinforced concrete slab and steel girder. The observed increase of road transport intensity makes it necessary to monitor the technical condition of the existing bridge constructions. Particular 52 T. Wróblewski et al. attention should be paid to both static and dynamic properties of a structure. Theappliedmodal analysis canbeused to detect damage.Anobserved change of some selected modal parameters can be the evidence of damage which is developing in the structure. Early detection of defects allows one to undertake preventive steps, thus to avoiding risk of having to close off the construction in question. Traditional modal parameters, such as the frequency and modes of natural vibrations and their respective damping ratios, are not sensitive enough to some kinds of damage. The aim of the present study was to apply the Energy Transfer Ratio – ETR to diagnostics of defects in steel-concrete connections of composite be- ams. ETR is a modal parameter defined by Liang and Lee (1991). According to the authors, it is muchmore sensitive to appearing damage than the tradi- tionally defined modal parameters. The results of experiments conducted on a real composite beam have been used in this study to analyse and validate the effectiveness of ETR. 2. The state-of-the-art analysis Commonly appliedmodal analysis is a technique of testing dynamicproperties of a structure.Theconductedanalysis yields amodalmodel comprisinga set of frequencies andmodes of natural vibrations as well as damping ratios. Modal analysis is ever more often used to diagnose the technical condition of struc- tures. This approach can be seen in the studies into steel-concrete composite beams carried out by the teamof scientists fromtheUniversity ofUdine, Italy. Apaperpublished in 2parts byDilena andMorassi (2003) andbyMorassi and Rochetto (2003) presents the results of investigations conducted on composite beams with damage in the steel-concrete interface. The analysed beam had no supports and it was hung on four flexible ropes. The investigated damage was introduced at one of the free ends of the beams. The authors focused on the analysis of changes taking place in the frequency vibrations. Consequently, they obtained results, according to which longitudinal/axial vibrations of the beamwere not very sensitive to damage in the connections. In contrast, in the case of flexural vibrations, the differences amounted to 38%. Liang and Lee (1991) defined a new modal parameter – Energy Transfer Ratio –which describes the amount of energy transferred between variousmo- des of vibrations. Their definition of ETR says that it is the ratio of modal energy transferred during a cycle to the total energy stored in the structure prior to the cycle. Energy transferred between various modes of vibrations Application of ETR for diagnosis... 53 exists for non-proportionally damped systems, which are majority in the civil engineering, and for these kinds of systems the ETR is possible to be determi- ned. Details concerning this coefficient will be presented later in this paper. Theoretical analyses usually look into two kinds of damping: proportio- nal and non-proportional. Systems with proportional damping are extremely rare in the real world (it is an almost hypothetic case). However, scholars very often assume damping to be proportional. This is due to the fact that for proportionally damped systems and for those without damping vibration modes are identical and frequencies of natural vibrations are very similar. An assumption about proportional damping makes it possible to determine mo- dal properties of a structure by analysing a system without damping, which is quite a simplification. A system is proportionally damped if the following relation is met: – according to Caughey and O’Kelley (1965) CM −1 K=KM−1C (2.1) – according to Ewins (1995) C=βK (2.2) – according to Ewins (1995), Uhl (1997), Lee and Liang (1998) C=αM+βK (2.3) where: M is the mass matrix, C – damping matrix, K – stiffness matrix. α and β are proportionality coefficients. Details concerning the issue of pro- portional damping can be found in the above mentioned studies. In 1999 Liang and Lee were conducting research on a composite bridge model on a scale of 1:6, which consisted of a concrete slab resting on three single-span steel girders. They limited their study to two kinds of damage: 1) removal of the support at one end of the central girder, 2) crack of a stret- ched part of the girder in the middle span. On the basis of purely theoretical investigations and concrete assumptions, the authors determined a relation ac- cording towhichETRwas1000 timesmore sensitive todamage than frequency of natural vibrations. In order to validate their assumptions, they conducted experiments whose results are presented in Table 1. The results presented there clearly show that ETR is a good identifier of changes taking place in beams both for the first and second kind of damage. ETR is much more sensitive to changes in a structure than any other para- meter. Unfortunately, the experiment conducted by Liang and Lee did not confirm their assumption that ETR is 1000 times more sensitive to damage than frequency of natural vibrations. 54 T. Wróblewski et al. Table 1.Changes of modal parameters [%] Mode number Damage 1 Damage 2 Frequency Damping ETR Frequency Damping ETR ratio ratio 1 4.1 89.5 309 2.7 7.3 117 2 3.8 1.3 30 1.0 15.2 243 3 0.1 7.4 1892 7.8 129.0 236 Wang and Zong in their papers published in 2002 and 2003 analysed ETR as well. Themodel made on a scale of 1:6 consisted of a concrete slab resting on four steel girders. Two kinds of artificial damage were introduced into the model: 1) the first was a simulated lack of the support; 2) the second was buckling and cracking of the steel girder. The authors, similarly to Liang and Lee, introduceda relation according towhichETRis 1000 timesmore sensitive to damage than frequency of natural vibrations. However, also this time they failed to confirm their theoretical assumptions. The sensitivity analysis for the investigated modal parameters is presented in Table 2. Table 2. Sensitivity analysis for damages Mode number Damage 1 Damage 2 Frequency Damping ETR Frequency Damping ETR ratio ratio 1 5.7 38.4 592.5 7.5 1.3 1614.7 2 8.4 74.0 81.7 2.8 67.3 820.2 3 6.3 123.4 4524.0 3.5 18.9 68.2 Having analysed Table 2, it can be concluded that: • ETR ismuchmore sensitive to introduced damage than either frequency of natural vibrations or corresponding damping ratios. • The form of vibrations for which ETR showed the biggest changes is not the form for which the frequency vibrations and/or damping ratio reached their biggest differences both prior to and after the introduced damage, e.g. for the first kind of damage the biggest change in the fre- quency of natural vibrations occurred for the 2nd form, while for ETR the biggest change took place for the 3rd form. • ETR showsmuch bigger changes for damage resulting from themissing support than for damage such as cracking of the girder. Application of ETR for diagnosis... 55 As shownabove, procedures aiming at diagnosing damage in steel-concrete composite beamshave been applied for several years now. Scientists have been trying to determine how the analysed parameters change given defects recor- ded in their structure. The studies conducted so far seem to point out that ETR is theparameterwhich is themost sensitive to introduceddamage.There are, however, some discrepancies between relations that were introduced in a purely theoretical way and empirical tests. All these facts seem to suggest the need of conducting ETR analysis of steel-concrete composite beams. 3. Theory of complex damping In structural dynamics, the equilibrium of a vibrating n-DOF system can be given by a set of differential equations MX ′′+CX′+KX =F (3.1) where M,C,K are themass, damping and stiffness matrices (n×n), respec- tively, X′′, X′,X – acceleration, velocity and displacement vectors (n×1), F – vectot of the external forcing function. According to Lee and Liang (1998), the corresponding homogenous equ- ation of the system can be written as Y ′′+DY ′+Ω2Y =0 (3.2) where D=Q⊤M− 1 2CM − 1 2Q Y =Q⊤M 1 2X Ω2 =Q⊤M− 1 2KM − 1 2Q= diag(ω2ni) In the above equation, Q stands for the eigenvector of the generalized stif- fness matrix K=M− 1 2KM − 1 2 and ωni is the natural frequency of the system without damping. The above system may have n modes if the D matrix is diagonal, whichmeans that system is proportionally damped. If thematrix D cannot be diagonalized together with the matrix Ω2, the system cannot be decoupled into n isolatedmodes. There is some energy transfer between those ’modes’. In such a case, the system is non-proportionally damped.This type of damping exists inmost real structures. If a system is non-proportionally dam- ped, the Caughey criterion described in equation (2.1) will not be satisfied, and also the matrix Dwill not be diagonal. 56 T. Wróblewski et al. We can assume that the n-DOF system described by equation (3.2) has a mode shape matrix P, which is complex because of the non-proportional damping. It has to be noticed that the mode shape matrix P of the system is no equal to the eigenvector Q of the generalized stiffness matrix. The i-th eigenvalue of the system λi is also called the complex frequency λi =−ξiωi± j √ 1− ξ2iωi (3.3) where ξi is dampingratio and ωi isnatural frequencyof thenon-proportionally damped system. It should be noted that ωi 6=ωni. Now, we can rewrite equation (3.2) as follows Q ⊤ PΛ 2+Q⊤CPΛ+Ω2Q⊤P=0 (3.4) where C=M− 1 2CM − 1 2 Λ= diag(λi) The ii-th entry of each matrix of equation (3.4) becomes λ2iQ ⊤ i Pi+λiQ ⊤ i CPi+ω 2 niQ ⊤ i Pi =0 (3.5) The above equation can be rewritten to the form λ2i +λi [Q⊤i CPi Q⊤i Pi ] +ω2ni =0 (3.6) The term in the square brackets is called the generalized Rayleigh quotient. According to Lee and Liang (1998), we can notice that 1 2ωi [Q⊤i CPi Q⊤i Pi ] = ξi+ jζi (3.7) The term ζi is called to be the i-th energy transfer ratio – ETR. It can be also proven that ωi =ωniexp(ζi) (3.8) Detailed information and derivations of the above equations are presented elsewhere, see Liang and Lee (1991), Lee and Liang (1998). Application of ETR for diagnosis... 57 4. Composite beam The beam cross-section is presented in Fig.1. The beam measured 3200 mm in overall length (only the I-bar was 20mm longer at both ends). The beam consisted of a rolled steel I-bar IPE 160 made of S235JRG2 steel and a re- inforced concrete slab 60×600mm in section size made of C25/30 concrete. The reinforcement rods were placed in two layers in the way shown in Fig.1. Fig. 1. Cross-section of the composite beam Transverse reinforcement wasmade of rods, 6mm in diameter, whichwere placed every 150mm.The beamhad a stiff connection consisting of perforated steel slats. The perforated slats are a new generation of connecting elements used in the bridge engineering. The slats were made from a flat bar, 10mm thick, made from S235JRG2 steel. The distribution of connecting elements is shown in Fig.2. Fig. 2. Distribution of connecting elements 58 T. Wróblewski et al. The program of experimental tests included realisation of the following operations: (i) initial static load tests; (ii) tests to determine basic dynamic characteristics; (iii) other additional tests (material tests). During the initial static load tests, a simply supported beamwas assumed. Some static load was applied in order to test the correctness of workmanship of the beams and to prepare them for dynamic tests. The obtained results showed no anomalies. The tests whose aimwas to determine basic dynamic characteristics, were conducted for a free beam. This kind of a beam was achieved by suspending the tested beam on two steel frames by means of four steel cables, 3mm in diameter. A grid of measuring points and a system of coordinates used in the studies are presented in Fig.3. The acceleration of vibrations was measured as the system response. Each beamwas tested three times. In every test, the beam vibrations were excited in different points. An impulse excitation was used in the tests. The excitation points are marked in Fig.3 as A,B and C. At the points A and B, the excitation was applied in the y-direction, whereas at the point C, in the x-direction. Fig. 3. A grid of measuring points Toprocess the results obtained in the dynamic tests, TimeDomainMDOF ASMmodule implemented in CADA-X software package was used – LMS In- ternational (2000). Thismoduleuses auniversalmethodof estimatingparame- ters of amodalmodelLSCE/LSFD(Least SquareComplexExponential/Least Square Frequency Domain). The estimation of the model parameters in this method is global for all transfer functions. In order to determine frequencies of natural vibrations on the basis of the obtained transfer functions, SUM index was used. This index gives a normalized sum of amplitudes of the measured transfer functions at the selected measuring points and directions. Application of ETR for diagnosis... 59 Table 3 contains vibration frequencies and values of their corresponding modal damping ratios for the first five flexural vibrations modes of the beam. Table 3.Experimental natural vibration frequencies and their corresponding values of modal damping ratio Frequency 1flex 2flex 3flex 4flex 5flex fi [Hz] 77.75 186.65 304.72 417.57 531.13 ξi [%] 0.11 0.27 0.33 0.42 0.42 5. Rigid finite element model of the beam The numerical model of the beam has been created in the convention of rigid finite element method – RFE model. The central idea of the method is the division of a real system into rigid bodieswhich are called rigid finite elements (RFE),whichare then in turnconnectedbymeans of spring-dampingelements (SDE). For continuous parts of a structure it is customary to start creating a model with segmentation of a beam into equal or nearly equal segments. This segmentation is also called the primary division. A spring-damping element (SDE) isplaced in the centre of gravity of every segment.ThisSDE is supposed to concentrate all spring and damping properties of a given segment. The next step is to connect SDEs created in the primary division by means of RFE. This is the so-called secondary division. While modelling a composite beam, its steel and concrete parts should be treated separately. Figure 4a shows a composite beam with length L and divided into n segments of equal length ∆L (primary division). As it can be seen, two SDEs were placed in every segment of the beam resulting from the primary division: one concentrating the properties of the steel I-bar and ano- ther one concentrating the properties of the segment of a reinforced concrete slab. TheRFEsmodelling the beam (secondary division, Fig.4b) were placed between SDEs. The initial and final RFEs are half the size of indirect RFEs of length ∆L. The SDEs, modelling the connection are the last elements of the model. These SDEs connect RFEs modelling steel and concrete parts of the composite beam. EveryRFEof a given number i has its own independent coordinate system X̂ (i) 1 ,X̂ (i) 2 ,X̂ (i) 3 . The system is chosen so that it would overlap with the princi- pal central axes of inertia of the given RFE. Having applied this assumption, 60 T. Wróblewski et al. Fig. 4. RFE model: (a) segmentation of a beam into sections (primary division); (b) RFE system (secondary division) themass andmoments of inertia are the only parameters necessary to describe any RFE. These quantities can be given in a form of a diagonal mass matrix M (i) = diag[m(i),m(i),m(i),J (i) 1 ,J (i) 2 ,J (i) 3 ] (5.1) The first three terms of thematrix are equal to themass of anRFE, while the other three are the mass moments of inertia of RFE relative to the axes X̂ (i) 1 ,X̂ (i) 2 ,X̂ (i) 3 . Every SDEof the k number has its own independent co-ordinate systemof themain axes Ŷ (i) 1 , Ŷ (i) 2 , Ŷ (i) 3 . Themain axes of an SDE have a property that forces acting on the SDE in the direction compatible with these axes result in its translational deformations which take place only in the direction where these forces are applied. The main parameters that describe an SDE of the number k are coeffi- cients defining its spring and damping properties. The spring properties are described by means of two matrices: the matrix of translational stiffness co- efficients K (k) Y and the matrix of rotational stiffness coefficients K (k) ϕ . Both matrices are diagonal and they are 3×3 in dimension K (k) Y = diag[k (k) Y,1,k (k) Y,2,k (k) Y,3] K (k) ϕ = diag[k (k) ϕ,1,k (k) ϕ,2,k (k) ϕ,3] (5.2) While investigating thequestionofmotionofflat systems, theabovediscus- sedmatrices reduce their dimensions respectively to: M(i) (3×3), K (k) Y (2×2), Application of ETR for diagnosis... 61 K (k) ϕ (1×1).Practicalmethodsof creating the stiffnessmatrix Kon thebasis of translational K (k) Y and rotational K (k) ϕ stiffness coefficients as well asmethods of creating the inertiamatrix M on the basis of individualmassmatrices M(i) were described in details by Kruszewski et al. (1999) and by Wittbrodt et al. (2006). The relations on the basis of which the elements of matrices M(i), K (k) Y , K (k) ϕ are determined for RFE and SDE (spring-damping elements) modelling the constant part of the structure (the reinforced concrete slab and the steel I-bar), are throughly discussed in the literature on the subject (Kruszewski et al., 1999; Wittbrodt et al., 2006). These relations require knowing both the applied materials and parameters describing the cross-section of an element. Table 4 contains a list of all parameters which were possible to be determined onthebasis of thebeaminventoryor referencedata.Thetabledoesnotcontain values of the substitutemodulusof elasticity of the reinforced concrete slab Ec. Table 4.Parameters of the beammodel Reinforced concrete slab tc [m] 6.386E−02 Ac [m 2] 3.832E−02 Ic [m 4] 1.302E−05 ρc [kg/m 3] 2.447E+03 ec [m] 3.193E−02 Steel I-section As [m 2] 2.010E−03 Is [m 4] 8.690E−06 ρs [kg/m 3] 7.850E+03 Es [N/m 2] 2.050E+11 es [m] 8.000E−02 The modulus Ec is a substitute modulus which takes into account the longitudinal reinforcement of the concrete slab. It can be easily calculated when the moduli of elasticity for both steel and concrete are known. White the obtaining of the steel modulus is relatively easy, the concrete modulus is much more difficult to be found. Lee et al. (1987) as well as Memory et al. (1995) during their theoretical analyses aiming at determining dynamic characteristics of reinforced concrete andcomposite structuresmaintained that it was important to take into consideration the dynamicmodulus of elasticity of concrete Ed. According to Neville (1995), the dynamic modulus Ed can be determined by means of vibrations of a concrete specimen, with only a 62 T. Wróblewski et al. negligible stress being applied.Owing to a low stress level, there are nomicro- cracks and there is no creep. This is the reason that the dynamic modulus of elasticity is considered to be roughly equal to the initial tangent modulus defined in a static test. The dynamic modulus of elasticity is therefore much higher than the secant elasticitymodulus Ecm, which in a standard procedure is determined during application of a static load onto a sample. The matrix elements K (k) Y for SDEs modelling the connection reflect the stiffness of the connection. If we investigate a flat system, it is necessary to know the stiffness of connecting elements in the vertical direction (perpendi- cular to the connection plane) Kv and in the horizontal direction (parallel to the connection plane) Kh. The reference literature does not provide any information on this problem. The threemissing parameters defining themodel stiffness, i.e. Ec,Kv,Kh were determined on the basis of parametric identification. The best possible fit of natural vibration frequencies obtained in experimentally and numerical- ly was used as the identification criterion. Both flexural and axial vibration modes were analysed. Detailed data of the procedure can be found in Wró- blewski (2006). Table 5 contains a comparison of frequencies obtained during the investigations and on the basis of theRFEmethodmodel. It also contains values of the identified parameters of the model. Table 5.Parameters of the beammodel Model RFE i fiexp finum ∆ [Hz] [Hz] [%] 1flex 77.75 77.75 0.0 2flex 186.65 187.07 0.2 3flex 304.72 307.10 0.8 4flex 417.57 419.81 0.5 5flex 531.13 526.70 −0.8 Kh [N/m] 2.012E+09 Kv [N/m] 2.745E+08 Ec [N/m 2] 3.392E+10 The damping properties of SDEs are described bymeans of twomatrices: the matrix of translational damping coefficients C (k) Y and the matrix of rota- tional damping coefficients C(k)ϕ . In a general case, both matrices are 3× 3. For flat systems they reduce their dimensions, similarly to K (k) Y ,K(k)ϕ Application of ETR for diagnosis... 63 C (k) Y = diag[c (k) Y,1,c (k) Y,2,c (k) Y,3] C (k) ϕ = diag[c (k) ϕ,1,c (k) ϕ,2,c (k) ϕ,3] (5.3) For SDEs which replace the constant part of the structure made of a ma- terial with properties defined by theKelvin-Voigt model, the relation between the respective dampingand stiffness ratios of thematrix elements can be given by Kruszewski et al. (1999) c (k) Y,i = Q−1 ω k (k) Y,i c (k) ϕ,i = Q−1 ω k (k) ϕ,i i=1,2,3 (5.4) where Q−1 is the loss factor and ,ω is the vibration frequency. The loss fac- tor Q−1, just like the logarithmic damping decrement δ or the damping ra- tio ξ, is a parameter used for defining the damping.Assumping that the dam- ping is minimal, between the above given parameters, the following relations are to be observed (Marchelek, 1991) Q−1 = δ π ξ= δ 2π δ=2πξ (5.5) Values of the loss factor depend on the frequency, temperature and other factors. The higher value of the loss factor, the better damping properties of a material. According to Rao (2004), the loss factor for steel amounted to Q−1s ∈ 〈6 · 10 −4,2 · 10−4〉. Concrete has much better damping properties than steel and owing to this its loss factor values vary in the following range Q−1c ∈ 〈0.06,0.02〉 (De Silva, 2000). For interfaces of elements and connections, through analogy to loss factors, it is possible to derive a connection loss factor (Marchelek, 1991). In the in- vestigated case, the connection loss factor was introduced and it was defined as Q−1con. Matrices of damping ratios for modelling connections of SDEs were defined according to relation (5.4) by substituting coefficient Q−1con. It was decided that the loss factors Q−1s ,Q −1 c ,Q −1 con should be determined bymeans of fitting the frequency characteristics obtained fromtheRFEmodel to the characteristics obtained experimentally. The characteristic obtained on the assumption that the system input is force and its output is acceleration is called inertance. Such characteristics are fitted with each other. One of the optimization procedures of Optimization Toolbox/Matlab software package was used during the analysis. Loss factor values for which the frequency am- plitudes for some selected resonance frequencies obtained in the experimental and numerical investigations overlap were sought. An example of such values obtained both experimentally and numerically can be seen in Fig.5. For com- parison purposes, somemeasuring points located at the ends of the reinforced concrete slab (i.e. points 2 and 34, response direction y, point C excited)were 64 T. Wróblewski et al. chosen. The loss factors determined in the analysis are presented in Table 6. More detailed information on the loss factor procedure can be found elsewhere (Wróblewski, 2006). Table 6.Values of loss factors Q−1s 0.0003 Q−1c 0.0222 Q−1con 0.0058 Fig. 5. Comparison of experimental and numerical amplitude and frequency characteristics Thedevelopedmodel of the composite beamhas very low requirements for processing capacity. This fact is due to a limited number of degrees of freedom. The software used in the analysis had been prepared in Matlab environment. The fit of experimental and numerical results is very high, which can be seen in Table 6 and Fig.5. The thus determined model was used for numerical simulations of damage occurring in the beam connections. 6. Numerical simulation of damage The carried out numerical simulation of damage aimed at investigating the sensitivity of ETR to the beamdamage. Damage simulation in the connection was conducted by removing successive SDEs connections 1, 4, and 7, respecti- vely (compare Fig.4) at one end of the beam. The removal of SDE number 1 Application of ETR for diagnosis... 65 simulated the damage on a 50mm long segment. The removal of two adjacent SDEs number 1 and 4 simulated the damage on a 150mm long segment. In a similar way, subsequent segments from which SDEs were removed, can be determined (compare Fig.4). Themodal parameters, i.e. frequency of natural vibrations for a system without damping or with proportional damping fi and for a system with non-proportional damping fni, and the damping ra- tio together with ETR were determined for the first five modes of flexural vibrations. Changes of these parameters are presented in Tables 7, 8 and 9. Because the values of fni and fi were almost identical, only one frequency table is presented below – that for fi. Table 7.Changes in frequencies of natural vibrations fi Damaged SDEs 1 1, 4 1, 4, 7 i finum finum,d ∆ finum,d ∆ finum,d ∆ [Hz] [Hz] [%] [Hz] [%] [Hz] [%] 1flex 77.75 77.69 −0.1 77.37 −0.5 76.49 −1.6 2flex 187.08 186.36 −0.4 182.53 −2.4 172.02 −8.0 3flex 307.10 304.28 −0.9 289.51 −5.7 261.38 −14.9 4flex 419.82 413.77 −1.4 389.09 −7.3 365.67 −12.9 5flex 526.71 518.10 −1.6 493.44 −6.3 477.42 −9.4 Table 8.Changes in damping ratio ξi Damaged SDEs 1 1, 4 1, 4, 7 i ξinum ξinum,d ∆ ξinum,d ∆ ξinum,d ∆ [%] [%] [%] [%] [%] [%] [%] 1flex 0.26 0.26 0.0 0.26 1.1 0.28 5.8 2flex 0.28 0.28 0.5 0.30 7.9 0.39 40.8 3flex 0.33 0.33 1.6 0.40 21.0 0.50 53.4 4flex 0.41 0.42 2.7 0.49 18.5 0.46 12.5 5flex 0.52 0.54 2.6 0.55 6.1 0.49 −7.0 Ascanbeseen fromthepresenteddata, theETRhas thehighest sensitivity to damage, and the frequency of natural vibrations fi and fni have the lowest sensitivity. However, there is no unambiguous relation between the changes in ETR values and frequencies or damping ratios. Having analysed the values of damping ratio presented in Tables 3 and 8, their high consistency is to be stressed. It also confirms the correctness of the loss factor estimation. 66 T. Wróblewski et al. Table 9.Chnges of ETR ζi Damaged SDEs 1 1, 4 1, 4, 7 i ζinum ζinum,d ∆ ζinum,d ∆ ζinum,d ∆ [�]·10−2 [�]·10−2 [%] [�]·10−2 [%] [�]·10−2 [%] 1flex 2.55 2.54 −0.1 2.55 0.2 2.66 4.3 2flex 1.92 1.91 −0.5 2.06 7.2 3.02 57.3 3flex 1.39 1.38 −0.9 1.93 38.7 2.18 56.5 4flex 0.97 0.98 0.6 1.36 39.4 1.16 19.5 5flex 0.76 0.73 −3.9 0.76 −0.3 1.54 103.5 The obtained results do not confirm the relation derived by Liang and Lee (1999) and by Wang and Zong (2002, 2003), which claims that ETR is 1000 times more sensitive than the frequency of natural vibrations fi. The assumptions made by the researchers are presented in detail below. For these assumptions, the relation inquestionwas obtained.Let ”0”denoteundamaged structure and ”j” stand for changed conditions of the structure after intro- duction of the damage. Given the above, relation (3.8) can be expressed as ωi0 =ωni0exp(ζi0) ωij =ωnij exp(ζij) (6.1) Theauthorsassumedthat afterdamage thechanges in stiffnessof the structure are negligibly small, and therefore ωni0 =ωnij =ωni (6.2) Formulas (6.1) take the following form ωi0 =ωniexp(ζi0) ωij =ωniexp(ζij) (6.3) It can be observed that for ζi0 = 0.001 and for changes in natural vibration frequencies amounting to 0.1%, we obtain ωij −ωi0 ωi0 =0.1% → exp(ζij)− exp(ζi0) exp(ζi0) = 0.001 (6.4) ζij − ζi0 ζi0 =1(100%) For ζi0 = 0.001 and ∆ωi = 0.1% ETR changes 100%, from which it can be inferred that it is 1000 times more sensitive to changes in the structure than the frequencies ωi. However, it should be also noted that this result was Application of ETR for diagnosis... 67 obtained for the following concrete figures: ζi0 =0.001 and ∆ωi =0.1%.How doesETRcoefficient change for other values of ζi0 and ∆ωi?For lowdamping observed in steel and inconcrete structures,values ofETRamount to 0.00-0.01 (WangandZong, 2002). Consequently, the analysis focusedon changes ofETR for various values of ζi0 in the range of 0.001-0.005 for frequency variations ∆ωi =0.1% and ∆ωi =0.2%. The results are presented in Tables 10 and 11. Table 10.Changes of ETR depending on ζi0 for ∆ωi =0.1% ζi0 0.001 0.002 0.003 0.004 0.005 Change of ETR 100.0% 50.0% 33.3% 25.0% 20.0% Table 11.Changes of ETR depending on ζi0 for ∆ωi =0.2% ζi0 0.001 0.002 0.003 0.004 0.005 Change of ETR 200.0% 100.0% 66.7% 50.0% 40.0% As can be seen in the above presented tables, the general statement that ETR is 1000 times more sensitive to damage than frequency of natural vi- brations is not entirely correct. The statement is true only for ∆ωi = 0.1% and ζi0 =0.001. For other values of ∆ωi and ζi0 other sensitivity values are obtained, e.g. for ∆ωi =0.1% and ζi0 =0.005 ETR change amounts to 20%, so in this case ETR is ”only” 200 times more sensitive to changes than the frequency of natural vibrations. We should also remember about assumption (6.2). An assumption was made that changes occurring in the structure are so small that the structure stiffness does not change and, as a result, the stif- fness matrix K and frequency of natural vibrations ωni do not change either. However, our investigations showed that the frequency of vibrations ωni do- es change after the damage is introduced into the system, which is shown in Table 7. As can be seen, the biggest change of vibration ωni for 3 vibration modes and for damaged 3 SDEs amounts to 14.9%.Therefore, it is not correct to assume that ωni = const. 7. Summary The present study focuses on ETR sensitivity to damage in the connections of steel-concrete composite beams. The obtained results confirmed that this coefficient is markedly more sensitive to damage than frequencies of natural vibrations. Although this difference does not amount to 1000 times, as Liang 68 T. Wróblewski et al. and Lee (1999) as well as Wang and Zong (2002, 2003) claimed, it is never- theless a significant one. It was also observed that ETRwasmore sensitive to damage than the damping ratio. It should be noted that the simulation in the connection was conducted for one beam only, for one kind of damage located at one end of the beam. Further studies, both experimental and numerical, are planned. They will focus on a wider range of damage. It should also be stressed that the ETR analysis was conducted on the whole beam. However, detailed studies show that this coefficient can be determined locally, only for a part of the beam. Perhaps this way of determination ETR might prove to be a more effective method of damage detection. This issue is certain to be the focus of future studies. References 1. CaugheyT.K.,O’KellyM., 1965,Classical normalmodes in damped linear dynamic systems, Journal of Applied Mechanics, 32, 583-588 2. DeSilvaC.W., 2000,Vibration. Fundamentals andPractice,CRCPress,Boca Raton, FL 3. DilenaM.,Morassi A., 2003,Adamage analysis of steel-concrete composite beams via dynamicmethods: Part II. Analyticalmodels and damage detection, Journal of Vibration and Control, 9, 5, 529-565 4. EwinsD.J., 1995,Modal Testing: Theory andPractice, ResearchStudiesPress Ltd., England 5. Kruszewski J., Sawiak S., Wittbrodt E., 1999, Wspomaganie kompute- rowe CAD CAM, Metoda sztywnych elementów skończonych w dynamice kon- strukcji, Wydawnictwa Naukowo-Techniczne,Warszawa 6. LeeP.K.K.,HoD.,ChungH.W., 1987,Static anddynamic tests of concrete bridge, Journal of Structural Engineering, 113, 1, 61-73 7. Lee G.C., Liang Z., 1998,On cross effects of seismic responses of structures, Engineering Structures, 20, 4, 503-509 8. Lee G.C., Liang Z., 1999, Development of a bridge monitoring system,Pro- ceedings of the 2nd international Workshop on Structural Health Monitoring, Standford University, Standford, CA, 349-358 9. Liang Z., Lee G.C., 1991, Damping of structures, Part 1, National Center forEarthquakeEngineeringResearch,TechnicalReportNCEER-91-0004,State University of NewYork at Buffalo, Buffalo, NY Application of ETR for diagnosis... 69 10. Marchelek K., 1991, Dynamika obrabiarek, Wydawnictwa Naukowo- Techniczne,Warszawa 11. Memory T.J., Thambiratnam D.P., Brameld G.H., 1995, Free vibration analysis of bridges,Engineering Structures, 17, 10, 705-713 12. Morassi A., Rocchetto L., 2003, A damage analysis of steel-concrete com- posite beams via dynamic methods: Part I. Experimental results, Journal of Vibration and Control, 9, 5, 507-527 13. NevilleA.M., 1995,Properties ofConcrete, 4th ed., LongmanGroupLimited, London 14. Rao S.S., 2004, Mechanical Vibrations, Fourth Edition, Pearson Education International, New Jersey 15. UhlT., 1997,Wspomaganie komputerowe CADCAM,Komputerowo wspoma- gana identyfikacja modeli konstrukcji mechanicznych, WydawnictwaNaukowo- Techniczne,Warszawa 16. Wang T.L., Zong Z., 2002, Improvement of evaluation method for existing highway bridges (Final report No. FL/DOT/RMC/6672-818), FloridaDepart- ment of Transportation, FL 32399, USA 17. WangT.L., ZongZ.H., 2003, Structural damage identification byusing ener- gy transfer ratio, IMAC-XXI: A Conference on Structural Dynamics, Kissim- mee, Florida 18. WittbrodtE., Adamiec-Wójcik I.,Wojciech S., 2006,Dynamice of Fle- xibleMultibody Systems. Rigid Finite ElementMethod, Springer-Verlag,Berlin, Heidelberg, NewYork 19. Wróblewski T., 2006, Ocena właściwości dynamicznych belek zespolonych, Rozprawa doktorska, Politechnika Szczecińska,Wydział Budownictwa i Archi- tektury, Szczecin Wykorzystanie ETR do diagnostyki uszkodzeń w stalowo-betonowych belkach zespolonych Streszczenie Artykuł przedstawia, jak zmienia się współczynnik transferu energii ETRwyzna- czanydla stalowo-betonowychbelek zespolonychwzależności od stopnia ichuszkodze- nia. Model numeryczny belki opracowany został z wykorzystaniem wyników badań doświadczalnych przeprowadzonych na rzeczywistych belkach zespolonych. Artykuł 70 T. Wróblewski et al. prezentuje, jak zmieniają się częstotliwości drgań,współczynniki tłumienia orazETR wzależności od zmianw strukturze belki. Analizywykazały, że ETR jest parametrem wykazującymnajwiększąwrażliwość na symulowane uszkodzenia. Parametr tenmoże być użyty do detekcji uszkodzeń w tego typu belkach. Manuscript received November 6, 2009; accepted for print April 23, 2010