Comparison of Different FE Calculation Methods for the Electromagnetic Torque of PM Machines Dieter Gerling Institute for Electrical Drives, University of Federal Defense Munich Neubiberg, Germany Summary: Generally, the torque of electrical machines can be calculated applying different methods. In this paper, the Maxwell’s stress tensor method, the magnetic co-energy method, and the lumpedparameter method are investigated. As an example, a PM machine with surface mounted magnets in the rotor is analyzed by means of the FE-software package ANSYS. For the lumped-parameter method, the dq-parameters of the electric machine are derived with the Fixed Permeability Method (FPM). With this method the parameters of the PM machine can be calculated with high precision. The obtained results for the electromagnetic torque applying the different calculation methods are compared concerning accuracy, ease of use and computing time. Keywords: Torque Calculation, Finite Element Method, ANSYS, PM Machine, Fixed Permeability Method (FPM) NAFEMS Seminar: „Numerical Simulation of Electromechanical Systems“ 1 October 26 - 27, 2005 Wiesbaden, Germany 1. Introduction The basic task of any electric machine is to generate torque to accelerate and drive a load over a specific range of speeds. Thus, torque is a very important consideration for both analysis and design of electrical machines. The finite element method provides an accurate approach to torque evaluation from the derivation of electromagnetic field distribution. This calculation method allows precise determination of machine parameters through the magnetic field solutions as it takes into account the actual distribution of windings, details of geometry, and non-linearity of magnetic materials of an electrical machine. In this paper, a PM machine with surface mounted magnets in the rotor is analysed with ANSYS. Figure 1 shows the geometry of the studied PM machine. The main geometrical data are presented in the Table 1. TABLE 1- GEOMETRY DATA Rotor radius [mm] Air gap length [mm] Thickness of magnet [mm] Magnet pole arc [deg] Number of poles [ -- ] Stator outer radius [mm] Stack length [mm] 21.6 0.9 4 127.8 4 40 35 Fig. 1: Geometry of the studied PM machine Different methods based on finite element solutions have been used for the calculation of the electromagnetic torque of this machine. The methods analysed here are Maxwell’s stress tensor-, dqmodel of PM machine-, and magnetic co-energy method. The use of Maxwell’s stress tensor is probably the simplest method, since it requires only the local flux density distribution along a specific contour around the air-gap of the machine. The accuracy of torque calculation with this method relies on model discretization and on contour selection. According to the virtual work principle, the electromagnetic torque equals the derivative of the magnetic co-energy with respect to angular position at constant current. With this method at least two finite element solutions are required to obtain the co-energy change due to an incremental displacement, and this inevitably increases the computing time. Calculation of the electromagnetic torque with the third method is based on the dq-mathematical model of the PM machine. The dq-parameters of the machine are derived accurately using fixed permeability method (FPM) [2-4]. This method ensures that the saturation effect which occurs in the load model is not ignored during calculation of the • • dq-axes inductances (single current excitation of the stator windings), back emf (single magnet excitation). The calculation algorithm for the dq-parameters of the PM machine with the fixed permeability method is presented in the fourth section of this paper. NAFEMS Seminar: „Numerical Simulation of Electromechanical Systems“ 2 October 26 - 27, 2005 Wiesbaden, Germany 2. Electromagnetic Torque 2.1 Maxwell’s Stress Tensor Method The Maxwell’s stress tensor method is probably the most commonly used since it is uncomplicated to apply and needs a relatively small calculation time. But, on the other hand, the accuracy of this method is obviously dependent on the type of the problem to be solved, on the model discretization and on the selection of the integration line or contour. If a Maxwell’s stress tensor is used, it is important to force at least three layers of elements in the small air-gap and to calculate the torque in the middle element layer. The electromagnetic torque Te in the motor air-gap, on a closed surface of radius r, can be calculated by integrating the Maxwell’s stress tensor. For two-dimensional electromagnetic field models: Te = 2π L µ0 ∫r 2 Br Bθ dθ (1) 0 where Br and Bθ are radial and tangential components of the flux density, and L is the active length of the machine. Taking advantage of the specific periodicity of electrical machines, the integral can be performed from 0 to 2π electrical radians and the result multiplied by the number of pole pairs p. 2.2 Co-Energy Method The second most frequently used method is based on the stored magnetic co-energy change or the virtual work with a small displacement. According to the virtual work principle, the electromagnetic torque Te equals the derivative of the magnetic co-energy Wco-eng with respect to angular position at constant current, Te = ∆Wco-eng ∆θ (2) i = const In co-energy method the calculation time is doubled, because with this method the field is solved twice to get an energy difference. 2.3 Electromagnetic Torque Based on the dq-Mathematical Model of PM Machine The PM motor is usually modeled and analyzed by means of the space-vector theory, which is well known and effective. The motor equations, valid also when the iron is saturated, are reported in the following. The electromagnetic torque based on the dq -formulation is Te = 3 p (ψ d ⋅ iq − ψ q ⋅ id ) 2 (3) while the voltage dynamic equations are d ψ d − ω ⋅ψ q dt d uq = R ⋅ iq + ψ q + ω ⋅ψ d dt ud = R ⋅ id + (4) where, in general, the flux linkages depend on both current components, according to the following magnetic model ψ d = ψ d ( id , iq ) (5) ψ q = ψ q ( id , iq ) NAFEMS Seminar: „Numerical Simulation of Electromechanical Systems“ 3 October 26 - 27, 2005 Wiesbaden, Germany If the motor works in the magnetic linear region (low current), the d- and q-axis flux linkages vary linearly with the corresponding d- and q-axis current components, i. e. ψ d = Ld ⋅ id + ψ dm ψ q = Lq ⋅ iq (6) where, ψ dm is the d-axis flux-linkage due to magnets. For the non-saturation case the total PM fluxlinkage flows through the d-axis. When the motor operates with high torque, hence high current, saturation effects are not negligible and the dq flux linkages become involved functions of both d- and q-axis current components. The complete magnetic model for a saturated motor can be thus expressed as ψ d ( id , iq ) = Ld ( id , iq ) ⋅ id + ψ dm ( id , iq ) (7) ψ q ( id , iq ) = Lq ( id , iq ) ⋅ iq + ψ qm ( id , iq ) Compared with the equ. (6), an additional q-axis PM flux-linkage component is present in the expression of the total flux-linkage. Like Ld and Lq , when the saturation occurs the ψ dm and ψ qm also depend on the d- and q-axis current components. In an analysis made in [1] for a PM machine with inset magnets in the rotor (IPM machine), it is shown that the ψ dm increases with id , while it decreases with iq components. In the same way, when the saturation occurs, the magnitude of ψ qm (which is negative for the motor operation condition of the PM machine) increases with iq , while it decreases with id ; its variation reaches up to 24% of the ψ dm . This is the main effect of the saturation. Using the equ. (3) and (7), the following expression for the electromagnetic torque is obtained Te = { } 3 p Ld (id , iq ) − Lq (id , iq ) ⋅ id ⋅ iq + ψ dm (id , iq ) ⋅ iq −ψ qm (id , iq ) ⋅ id 2 (8) The above expression is used to calculate the electromagnetic torque with the third FE method. The dq-parameters of the PM machine are calculated using the fixed permeability method (FPM) [2, 3, 4]. The main point of this method is that it transforms non-linear problems into linear ones by storing the permeabilities from the non-linear analysis. For each operation point the magnetic permeability of each element obtained from the previous FE simulation under the double excitation is fixed and stored for the further analysis. With this fixed permeability, a second calculation is carried out either with current or magnet single excitation. This ensures that the saturation effect which occurs in the load model is not ignored during calculation of the dq-parameters of the PM machine. 3. Finite Element Analysis Results Based on the above calculation FE methods, in the following the electromagnetic torque is calculated for I = 2.56 A (peak value) and δ = 0° operation condition. δ is the electrical angle between total flux linkage due to the stator currents and the rotor q-axis (in the literature this angle is known as torque angle or load angle). The flux line distribution and the flux density along the surface of a PM pole pair for this operation condition are shown in the figure 2. It is shown that the rotor surface is subjected to a relatively large flux density fluctuation due to the varying reluctance as teeth are passed. All following finite element calculations have been performed using the software package ANSYS. NAFEMS Seminar: „Numerical Simulation of Electromechanical Systems“ 4 October 26 - 27, 2005 Wiesbaden, Germany B [T] Fig. 2: Flux line distribution and the flux density along the surface of a PM pole pair. Firstly, the electromagnetic torque is evaluated using the Maxwell’s stress tensor and the magnetic coenergy method for a fixed current excitation in the stator and for different incremental rotor positions. Figure 3 compares the results obtained with these methods. It is shown that the results obtained from the Maxwell’s stress tensor method are in good agreement with the results obtained with the magnetic co-energy method. The torque curves vary sinusoidally with the load angle (for a fixed current in the stator, the load angle is equivalent with the rotor position). As is known the PM machine with surface mounted magnets in the rotor gives the maximum torque for the case when the load angle is zero ( δ = 0° , I = I q ). Due to the slotting effect some harmonics (torque pulsations) are added in the toque curve of this type of the machine. This torque component is known as cogging torque. The cogging torque is often the largest component of torque pulsation in permanent magnet motors. It is caused by the interaction between the magnets and the stator teeth. This effect is normally not depending on the stator current. Only if the teeth are saturated due to stator current, usually the cogging torque increases because of the wider effective slot openings. I=2.559 A; Torque-load angle 1,2 1 0,8 T_maxw 0,6 T_Coen T [Nm] 0,4 0,2 0 -0,2 -0,4 -0,6 -0,8 -1 -1,2 0 20 40 60 80 100 120 140 160 180 delta [elec. degree] Fig. 3: Electromagnetic torque versus load angle obtained with Maxwell’s stress tensor- and the magnetic co-energy method. NAFEMS Seminar: „Numerical Simulation of Electromechanical Systems“ 5 October 26 - 27, 2005 Wiesbaden, Germany For the same operation condition as before in the following the electromagnetic torque versus rotor position is evaluated using Maxwell’s stress tensor method and the dq mathematical model of the PM machine. The fixed permeability method is used to derive the motor parameters. Using this method the saturation condition of the machine under the double excitation is fixed and stored for the further analysis. The dq-inductances and flux linkages due to magnets are derived under fixed permeability from the double excitation condition. The dq-parameters of the motor are: iq = 2.559 A, id = 0 A, L q = 6.483 mH, L d = 6.483 mH, ψ dm = 1.531 ⋅10−1 Vs, ψ qm = −7.663 ⋅10−4 Vs . Figure 4 shows the saturation condition of the iron parts of the PM machine for the given operation condition. Three regions of the machine can be seen to have low relative permeability values: the stator teeth (in the main flux path), the stator yoke (in the main flux path), and the rotor yoke (in the main flux path). Fig. 4: The saturation condition (relative permeability) of the PM machine at I = 2.56 A and δ = 0° . The electromagnetic torque results versus rotor position obtained with the Maxwell’s stress tensor and from the dq mathematical model of the PM machine are presented and compared in the figure 5. Also here the obtained results show a good agreement between these methods. delta=0°, I=2.559 A: Torque-rotor position 1,4 1,2 T [Nm] 1 0,8 T_maxw T_dq 0,6 0,4 0,2 240 228 216 204 192 180 168 156 144 132 120 108 96 84 72 60 48 36 24 12 0 0 theta [elec. degree] Fig. 5: Electromagnetic torque versus rotor position obtained with Maxwell’s stress tensor- method and the dq-mathematical model of PM machine. NAFEMS Seminar: „Numerical Simulation of Electromechanical Systems“ 6 October 26 - 27, 2005 Wiesbaden, Germany As the calculation of the electromagnetic torque with the Maxwell’s stress tensor method gives the total torque components in the air-gap, the dq model gives only the average components of the electromagnetic torque. The dq-theory is established based on the assumption that both the winding flux-linkage and the currents are sinusoidal. The air-gap harmonics are ignored using this model. 4. Computing time As is discussed in the previous sections, the Maxwell’s stress tensor method is probably the most commonly used method for the torque calculation since it is uncomplicated to apply and needs a relatively small calculation time. Only one simulation for each operation point is needed with this method. In co-energy method the calculation time is doubled, because with this method the field is solved twice to get an energy difference. Also, the torque calculation with dq-mathematical model of the PM machine parameters of which are derived with fixed permeability method needs two simulations for each operation point. The first simulation should be done under the double excitation, from which the permeability of each element is fixed and stored for the further analysis. With this fixed permeability, a second calculation is carried out either with current or magnet single excitation. The calculation time for the second simulation with the “frozen” permeability condition is faster because the electromagnetic problem is linearized. Figure 6 shows the calculation procedure of the dq-parameters with fixed permeability method. If the saturation effect occurs it is required that the calculation procedure to be repeated for different rotor positions, therefore to derive the average values for these parameters. For this case this method is very time consuming. start Iˆ, δ dq-transformation theory θ = 0 : ∆θ : τ P Lq , Ld , ∗ψ qm , ∗ψ dm Double excitation condition Torque, T Total flux linkage: ψ T τ P − Pole pitch θn < τ P Fixed Permeability method Current flux-linkage: ψ Average values i Lq , Ld , ∗ψ qm , ∗ψ dm , T Magnet flux-linkage: end ψ m = ψ T − ∗ψ i ∗ Fig. 6: Calculation flow-chart of the dq-parameters of the PM machine. NAFEMS Seminar: „Numerical Simulation of Electromechanical Systems“ 7 October 26 - 27, 2005 Wiesbaden, Germany 5. Conclusions The torque is a very important parameter for both analysis and design of electrical machines. The finite element method provides an accurate approach to torque evaluation from the derivation of electromagnetic field distribution. In this paper, three calculation methods based on the results of electromagnetic field FE analysis are used to derive the electromagnetic torque of a PM machine; the Maxwell’s stress tensor, the co-energy method and the dq mathematical model for the PM machine. Using these methods the electromagnetic torque versus load angle and rotor position for a given operation point is calculated. The obtained results show a good agreement between these methods. References [1] [2] [3] [4] Gerling D., Dajaku G.: “Finite element analysis and torque calculation of the IPM2910 Machine”, Institute for Electrical Drives, University of Federal Defense Munich, July 2005 Kwak S. J., Kim K. J., Jung H. K.: “The characteristics of the magnetic saturation in the interior permanent magnet synchronous motor”, International Conference on Electrical Machines, 2004, Cracow, Poland (ICEM 2004) Bianchi N., Bolognani S.: “Magnetic models of saturated interior permanent magnet motors based on finite element analysis”, IEEE IAS Annual Meeting, 1998, St. Louis, USA Jianhui H., Jibin Z., Weiyan L.: “Finite element calculation of the saturation dq-axes inductance for a direct-drive PM synchronous motor considering cross-magnetization”, International Conference on Power Electronics and Drive Systems, 2003, Singapore (PEDS 2003) NAFEMS Seminar: „Numerical Simulation of Electromechanical Systems“ 8 October 26 - 27, 2005 Wiesbaden, Germany