202 Belahcen Methods Magnetic and elastic fields The A−φ formulation of the magnetic field in two dimensions and the displacement based formulation of the elastic field are used. ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ S(A k+1 , U k+1 )+ ∂ S(A k+1 , U k+1 ) ∂ A A k+1 [D r ] T [LD s ] T ∂ S(A k+1 , U k+1 ) ∂U A k+1 D r C r 00 LD s 0 G s 0 − ∂ ˜ F k+1 ∂ A 00 ˜ K − ∂ ˜ F k+1 ∂U ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ A n k+1 u r n k+1 i s n k+1 U n k+1 ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ = ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ S (A k )A k + [D r ] T u r k + [D s ] T K T i s k − (S(A n k+1 )A n k+1 + [D r ] T u r n k+1 + [D s ] T K T i s n k+1 ) LD s A k − H s i s k −C s (V s k+1 + V s k )− (LD s A n k+1 − H s i s n k+1 −C s (V s k+1 + V s k )) D r A k −C r u r k − G r i r k − (D r A n k+1 −C r u r n k+1 − G r i r n k+1 ) ˜ F n k+1 − ˜ KU n k+1 ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (1) The 2D finite element (FE) equations for the magnetic, and elastic field are coupled through the displacements and the forces. These equations are solved together with the circuit equations of the windings of the machine as described in [9]. The system of equations to be solved at each iteration is written as (1), where A is the magnetic vector potential; C, D, G, and H are the coupling matrices between the magnetic vector potential and the electric parameters in the windings parts of the machine; L is a matrix for the connections of the stator windings; S and K are the magnetic and mechanical stiffness matrices; u and i are respectively voltages and currents and U is the nodal displacements vector. The superscripts r and s refer respectively to rotor and stator. The subscripts k and k + 1 refer respectively to previous andpresent stepsin the time stepping method. The sign ∼over the matrix K and the force F means that they are replaced by their dynamic counterparts. A full description of the above matrices is given in [9,11]. Magnetic and magnetostrictive forces The forces, which are the load for the elastic field, are separated into magnetic force (also called reluctance forces in some works [4]) and magnetostriction forces. The magnetic forces are calculated, at any iteration, from the calculated magnetic vector potential, based on the local application of the virtual work principle: F T =− ∂W ∂U φ = constant (2) II-6. Effect of Stress-Dependent Magnetostriction 203 With the energy W per element S e given by: W = S e B 0 H · dB dS e , (3) we obtain the contribution of one element to the nodal magnetic forces calculated on the reference element ˆ S e as: F T =− ˆ S e 1 μ e ∇A · ∂ ∂U (∇A)|J|+ B 0 H · dB ∂ ∂U (|J|) d ˆ S e (4) where,|J|isthe determinant of the Jacobianmatrixforthetransfor mation fromthereference element to the actual one. These individual contributions from elements to nodal forces are added to each other to obtain the global nodal magnetic forces. The nodal magnetostriction forces are calculatedalso at elementlevel and assembled in thesamemanner as themagnetic forces. The calculation of the magnetostrictive forces is based on an original method called the method of magnetostrictive stress. This method is explained hereafter. Let’s consider an element of iron in a magnetic field H. Due to magnetostriction of iron, this element will shrink orstretch depending on the signof itsmagnetostriction. This change in dimensions is described by a magnetostrictive strain tensor {ε ms }. Corresponding to this strain, a magnetostrictive stress tensor {σ ms } can be calculated using Hook’s law. The nodal magnetostrictive forces are the set of nodal forces due to this stress. The measurements presented in [10] give the component of magnetostrictive stress σ ms in a direction parallel to that of the magnetic field. The other component of magnetostric- tive stress orthogonal to the direction of the magnetic field H can be calculated within two assumptions. First,there is nomagnetostrictiveshear stress inthe frame definedby the direc- tion parallel to H and the one or thogonal to it. Second, there is no volume magnetostriction; which is a good assumption in the range of flux density occur ring in electrical machines. The latter assumption means that the magnetostriction strain in the direction orthogonal to that of the magnetic field is opposite and has half the amplitude compared to the strain parallel to the direction of the magnetic field. If σ ms⊥ is the magnetostrictive stress in the direction orthogonal to the magnetic field, using the first assumption, we can write: ⎡ ⎣ σ ms σ ms⊥ 0 ⎤ ⎦ = E ⎡ ⎣ ε ms ε ms⊥ 0 ⎤ ⎦ (5) where E is the stress-strain matrix. In the case of plane stress, use of the second assumption leads to: σ ms⊥ = 2v −1 2 − v σ ms (6) The magnetostrictive nodal forces are calculated for each element as follows. Let θ be the angle defined by the direction of the magnetic field and the x-axis. The projections of each edge of the element on the directions parallel and orthogonal to the magnetic field are respectively s = cos(θ)s x + sin(θ )s y (7) 204 Belahcen and s ⊥ =−sin(θ)s x + cos(θ )s y (8) where s x and s y are respectively the projection of the considered edge of the element on the x- and y-axis. The forces, per unit length, parallel and orthogonal to the direction of the magnetic field are respectively F ms = σ ms s (9) and F ms⊥ = σ ms⊥ s ⊥ (10) These forces are distributed equally between the two nodes of the given edge. The forces in the original Cartesian coordinate system are obtained as the projection of F ms and F ms⊥ on the axis of that system: F msx = cos(θ)F ms − sin(θ )F ms⊥ (11) and F msy = sin(θ)F ms + cos(θ )F ms⊥ (12) When the stress dependency of magnetostriction is taken into account, only the data of magnetostrictive stress are changed. All the rest is the same. Vibrations The solution of the magnetoelastic FE analysis produces among others the nodal displace- ments as a function of time. The displacements of a node on the outer surface of the stator core are transformed with Discrete Fourier Transform (DFT) and numerically differentiated to obtain the frequency components of the velocity of vibrations of the node considered. In the following calculations, a total of 3,000 time steps are calculated with 300 time steps per period of the line voltage (20 ms). This leads to a sampling frequency of 15 kHz, a frequency resolution of 5 Hz and a maximum frequency of 7.5 kHz. The amplitudes of these vibrations are the quantities under consideration in the result section. Results Validation An induction machine-like test device has been built to verify the presented model for magnetostriction. The test device is shown in Fig. 1. The test device is constructed in a way that simulates the flux path in an induction machine, meanwhile it minimizes the reluctance forces. The latter are due to the presence of the air gap while the test device has no air gap. The only magnetic forces in the test device are the magnetostrictive forces and the Lorentz forces. The effect of the latter ones on the vibrations of the test device can be neglected due to the low currents in the windings and also due to the high relative mass-ratio between the iron core of the device and its windings. Thus, the only cause of vibrations in the test device is the magnetostriction. Computations and measurements have been made for the test device. The simulated flux lines in the cross section of the device are shown in Fig. 2. The flux density in the back II-6. Effect of Stress-Dependent Magnetostriction 205 Figure 1. Picture of the test device. The search coil for measurement of the back iron magnetic flux density can be seen. Rubber tubes separate both mechanically and electrically the windings from the iron core. iron core of the test device has also been measured by a search coil. This flux density is compared tothe simulatedone inFig. 3. The vibrations of the outer surface ofthe testdevice have been measured with a laser vibrometer at different points. The same values have been calculated. The simulated and measured displacements at a point on the outer surface are shown in Fig. 4. Application Themethoddevelopedinthisworkisappliedtoasmallsize(37kW)inductionmachine.Dif- ferent computational approaches are used to establish both the effect of stress-independent and stress-dependent magnetostriction on the vibrations of such a machine. The parameters of the simulation machine are given in Table 1. Figure 2. Plot of the calculated flux lines in the test device. 206 Belahcen Table 1. Parameters of the induction machine Rated power 37 kW Rated voltage 380 V Connection Star Slip 1.7% Frequency 50 Hz Number of phases 3 Number of poles 4 Stator outer diameter 310 mm Stator inner diameter 200 mm Stack length 289 mm Number of stator slots 48 Rotor outer diameter 198.4 mm Number of rotor slots 40 1.5 1 0.5 0 –0.5 –1 –1.5 1.5 1 0.5 0 –0.5 –1 –1.5 0.3 0.32 0.34 0.36 0.38 0.4 Time (s) Time (s) Flux density (T) Flux density (T) 0.1 0.12 0.14 0.16 0.18 0.2 Figure 3. Measured and simulated flux density in the back iron core of the test device. (a) Measured. (b) Simulated. 0.3 0.32 0.34 0.36 0.38 0.4 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5 x 10 −8 Time (s) Displacement (m) 0.1 0.12 0.14 0.16 0.18 0.2 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 x 10 −8 Time (s) Displacement (m) Figure 4. Measured and simulated displacements at the surface of the test device (in measurement the DC-component is omitted). (a) Measured. (b) Simulated. II-6. Effect of Stress-Dependent Magnetostriction 207 −0.2 −0.1 0 0.1 0.2 −0.2 −0.1 0 0.1 0.2 X-coordinate (m) Y-coordinate (m) − 0.2 − 0.1 0 0.1 0.2 − 0.2 − 0.1 0 0.1 0.2 X-coordinate (m) Y-coordinate (m) original deformed Figure 5. Calculated magnetostrictive force(normalized to 50,870 N/m) and deformation (magnified 20,000 times) of the stator core of the induction machine. (a) Forces. (b) Deformation. The calculated magnetostrictive forces acting on the stator core of the induction machine at thelast timestep and the corresponding deformation are shownin Fig. 5(a,b)respectively. Theseareforces and deformations calculated with stress-independentmagnetostriction.The reluctance forces are not shown. Calculations with no magnetostriction and these with stress-dependent magnetostriction have also been undertaken. The velocity of vibrations of a node on the outer surface of the stator core of the machine (point P in Fig. 5(b)) calculated with different approaches are compared. Fig. 6 shows the relative difference in the amplitudes of velocity between the cases no magnetostriction and stress-independent magnetostriction. Fig. 7 shows the relative 0 1000 2000 3000 4000 5000 6000 7000 8000 −10 −8 −6 −4 −2 0 2 Frequency (Hz) Relative difference in amplitude Figure 6. Calculatedrelative difference in the amplitude of velocities of point P (Fig. 5b). Differences between the cases no magnetostriction and stress-independent magnetostriction. 208 Belahcen 0 200 400 600 800 1000 −70 −60 −50 −40 −30 −20 −10 0 10 Frequency (Hz) Relative difference in amplitude Figure 7. Calculatedrelative difference in the amplitude of velocities of point P (Fig. 5b). Differences between the cases stress-independent and stress-dependent magnetostriction. difference in the amplitudes of velocity between the cases stress-independent and stress- dependent magnetostriction. The relative differences are calculated as (|v 1 |−|v 2 |)/|v 1 |. Analysis and discussion Validation The measured displacements from the test device are slightly higher than the simulated ones. This difference can be seen also in the measured and simulated flux densities. They are duemainlyto differences in the magnetic properties of the materials used in simulations. Indeed, the manufacturing process of the test device slightly deteriorated the magnetic properties of the iron sheets. However, the correspondences between measured quantities and these simulated with the presented model for magnetostriction are rather good from both the amplitudes and wave forms points of view. In the future, better magnetic properties of the manufactured machine can be introduced into the simulation software for better results. Simulations Thevibrationsoftheinductionmachineareaffectbythemagnetostriction.The amplitudesof most of thefrequencycomponents are increased.The most increasedfrequencycomponents are these at 490 Hz (840%), 40 Hz (400%), 60 Hz (270%), and 50 Hz (320%). Some other frequency components are damped due to the magnetostriction. Among these the ones at 1,190 and 1,470 Hz damped respectively 80% and 70%. The 100 Hz component is damped only by 7%. II-6. Effect of Stress-Dependent Magnetostriction 209 The stress dependency of magnetostriction adds to the effect of stress-independent mag- netostriction so that the amplitudes of almost all the frequencies are increased. The increase reaches some 7,000% for the frequency component at 1,595 Hz e.g. However, the accuracy of the simulations with stress-dependent magnetostriction cannot be established due to the effect of magnetostrictive stress. Indeed, in the FE iteration process, the stress from mag- netostrictive forces cannot be separated from the stress due to other forces (reluctance and Lorentz forces). Thus the stress state of the material is not accurately estimated, leading to inaccuracies in calculation of the stress-dependent magnetostriction. Although, we can say that both stress-independent and stress-dependent affect the vibrations of rotating electrical machines. Conclusions A model for the magnetoelastic coupling is presented and used in the simulations of an induction machine. The goal of these simulations is to establish the effect of the magne- tostriction on the vibrations of rotating electrical machines. For this purpose an original method for the calculation of magnetostrictive forces is presented. It is shown that the magnetostriction affects the vibrations of rotating electrical machines by increasing or decreasing the amplitudes of velocities measured at the outer surface of the stator core of the machine. These velocity are the ones responsible for acoustic noise. Furthermore, The stress dependency of the magnetostriction adds to the increase of the above amplitudes. The modeling of vibrations and noise of electrical machines should take into account the effect of magnetostriction and its stress dependency. References [1] P. Witczak, Calculation of force densities distribution in electrical machinery by means of magnetic stress tensor, Arch. Electr. Eng., Vol. XLV, No. 1, pp. 67–81, 1996. [2] L. L˚aftman, “The Contribution to Noise from Magnetostriction and PWM Inverter in an Induction Machine”, Doctoral thesis, IEA Lund Institute of Technology, Sweden, 94 p, 1995. [3] K. Delaere, “Computational and Experimental Analysis of Electrical Machine Vibrations Caused By Magnetic Forces and Magnetostriction”, Doctoral thesis, Katholieke Universiteit Leuven, Belgium, 224 p, 2002. [4] Z. Ren, B. Ionescu, M. Besbes, A. Razek, Calculation of mechanical deformation of mag- netic material in electromagnetic devices, IEEE Trans. Magn. Vol. 31, No. 3, pp. 1873–1876, 1995. [5] O.Mohammed, T.Calvert,R.McConnell, “A ModelforMagnetostriction inCoupledNonlinear Finite Element Magneto-elastic Problems in Electrical Machines”, International Conference on Electric Machines and Drives IEMD ’99, Seattle, Washington, USA, pp. 728–735, May 1999. [6] F. Ishibashi, S. Noda, M. Mochizuki, “Numerical simulation of electromagnetic vibration of smallinductionmotor”,IEEProc.Electr. PowerAppl.,Vol. 145, No.6,pp.528–534,November 1998. [7] G.H. Jang, D.K. Lieu, “The effect of magnetic geometry on electric motor vibration”, IEEE Trans. Magn., Vol. 27, No. 6, pp. 5202–5204, November 1991. 210 Belahcen [8] C.G. Neves, R. Carlson, N. Sadowski, J.P.A. Bastos, N.S. Soeiro, “Forced Vibrations Calcu- lation in Switched Reluctance Motor Taking into Account Viscous Damping”, International Conference on Electric Machines and Drives IEMD’99, May 1999. [9] A. Belahcen, “Magnetoelastic Coupling in Rotating Electrical Machines”, IEEE Tran. Mag., Vol. 41, No. 5, pp. 1624–1627, May 2005. [10] A. Belahcen, M. El Amri, “Measurement of Stress-Dependent Magnetisation and Magne- tostriction of Electrical Steel Sheets”, Internation Conference on Electrical Machines, Cracow, Poland, CD-ROM Paper No. 258, September 5–8, 2004. [11] A. Arkkio, “Analysis of Induction Motors Based on the Numerical Solution of the Mag- netic Field and Circuit Equations”, Doctoralthesis, Acta Polytechnica Scandinavica, Electrical Engineering Series No. 59, 97 p. Available at http://lib.hut.fi/Diss/198X/isbn951226076X/. II-7. COMPARISON OF STATOR- AND ROTOR-FORCE EXCITATION FOR THE ACOUSTIC SIMULATION OF AN INDUCTION MACHINE WITH SQUIRREL-CAGE ROTOR C. Schlensok and G. Henneberger Institute of Electrical Machines (IEM), RWTH Aachen University, Schinkelstraße 4, D-52056 Aachen, Germany christoph.schlensok@iem.rwth-aachen.de, henneberger@iem.rwth-aachen.de Abstract. In this paper the structure- and air-borne noise of an induction machine with squirrel-cage rotor are estimated. For these, different types of surface-force excitations and rotational directions are regarded for the first time. The comparison of the different excitations shows, that it is necessary to take the rotor excitation into account, and that the direction of the rotation has a significant effect on the noise generation. Introduction Thedriversofpassengercarsnowadaysmakegreatdemandsonthe acoustics of the technical equipment such as the electrical power steering. Therefore, it is of high interest to estimate the audible noise radiation of these components. The induction machine with squirrel-cage rotor used as power-steering drive is computed in three steps: coupled to the casing caps by the bearings. Forthis, therotor excitation hasto be takeninto account as wellfor comparison reasons. 1. electromagnetic simulation, 2. structural-dynamic computation, and 3. acoustic estimation. The theory is briefly described in [1] and therefore not repeated. In the case of an induction machine with skewed squirrel-cage rotor the location of the maximum force excitation of the stator teeth depends on the rotational direction. So far, only stator-teeth excitation has been regarded in literature [2–4].Further ontheimpact of theforceexcitingtherotor is taken into account. Therefore, four different cases of electromagnetic surface-force excitation are compared and discussed in this paper as listed in Table 1. Since the rotor of the induction machine is skewed (skewing angle = 10 ◦ ) the stator teeth are excited very asymmetrically. The location of the maximal tooth excitation de- pends on the direction of rotation. In case of right-hand rotation the highest excitation S. Wiak, M. Dems, K. Kom ˛ eza (eds.), Recent Developments of Electrical Drives, 211–223. C 2006 Springer. . location of the maximal tooth excitation de- pends on the direction of rotation. In case of right-hand rotation the highest excitation S. Wiak, M. Dems, K. Kom ˛ eza (eds.), Recent Developments of Electrical. by 7%. II-6. Effect of Stress-Dependent Magnetostriction 209 The stress dependency of magnetostriction adds to the effect of stress-independent mag- netostriction so that the amplitudes of almost. henneberger@iem.rwth-aachen.de Abstract. In this paper the structure- and air-borne noise of an induction machine with squirrel-cage rotor are estimated. For these, different types of surface-force excitations