You are on page 1of 6

MAGNETIC FIELD CALCULATION OF CLAW POLE PERMANENT MAGNET MACHINES USING MAGNETIC NETWORK METHOD Y.G.

Guo and J.G. Zhu Faculty of Engineering, University of Technology, Sydney P.O. Box 123 Broadway NSW 2007 Abstract This paper reports the three-dimensional magnetic field analysis using the magnetic network method for evaluation of parameters and performance of a claw pole permanent magnet machine. The magnetic network model is built based on the principle of equivalent magnetic flux pipes and the network equations are derived to solve the nodal magnetic scalar potentials, which are used to calculate the magnetic flux flowing through different cross-sectional areas. The results are validated by 3D finite element method. The magnetic network technique has the advantage of combining the speed of the conventional magnetic circuit method and the computational accuracy of finite element analysis. 1. INTRODUCTION Claw pole machines are commonly used as automobile, wind and hydro generators due to their simplicity and low manufacturing cost. The unique armature construction with one circumferentially wound global excitation coil for all poles in one claw pole stack implies a peculiar armature pole structure, which results in a truly three-dimensional (3D) magnetic field in the machine. In contrast to conventional types of electrical machines, where analysis can be conducted in a two-dimensional (2D) plane perpendicular to the machine shaft, the claw pole machine model must take into account all three field components: radial, tangential, and axial. Some parts of the core structure cannot be modelled to an equivalent 2D frame because they are not invariant along one of the axes and hence only 3D analysis can provide accurate results. For this reason, it is usually impossible or very difficult to accurately predict the parameters and performance of such complex structured devices by the conventional magnetic circuit (CMC) method. The flux through one part is considered to flow in the direction of one axis only, which is not true for the 3D flux electromagnetic devices. In addition, the consideration for the effect of saturation and leakage fields relies heavily on experience or experimental data. Therefore, the conventional empirical method cannot provide reliable results in situation where complex structures or completely new designs are being developed although it may be very useful for analysing small changes or rapidly determining the machine dimensions for given specifications. Theoretically, the 3D finite element method (FEM) is capable of accurately analysing magnetic field distributions in any devices since it can include all the structure, material and excitation details. In order to obtain acceptable accuracy of finite element analysis, the machine needs to be split into a huge number of geometrical elements. The field values are assumed to be a simple function of position within these elements, enabling interpolation of results. The time required for calculating the field distribution, however, could be very long, especially in design optimisation stage, which needs a great number of iterations to reach an optimum design. This is only possible with high-speed computers with large RAMs and it suggests that the finite element analysis (FEA) may not be the most effective method for solving the everyday problems in electrical machine design. In order to overcome the shortcomings of the FEM and the CMC, the magnetic network method (MNM), based on the principle of equivalent magnetic flux pipes [1,2] is used in this paper for calculating the magnetic field distribution in a claw pole permanent magnet machine. Similar to the FEM, the solution region can be chosen according to the symmetry of the structure and the whole solution region is divided into a number of elements with quasi-regular shape and relatively uniform field distribution. The equivalent magnetic permeance (or reluctance) of an element can be evaluated by the magnetic Ohms law. The elements are connected with nodes. The nodal magnetic scalar potential, the magnetic flux flowing through the elements and various parameters are computed based on the similarity between the magnetic and electrical networks. Compared with the FEM, the MNM is capable of analysing electrical machines of any structure with an acceptable accuracy with a much smaller number of elements (allowing rapid iteration of computation). The compromise between the computational accuracy and the CPU time is good enough to be appealing. Accuracy can be improved by concentrating elements in critical and saturated parts of the machine, e.g. in the main air gap, the claw pole parts, and the sub-air

gap between claw pole pieces [3]. Opposite to the triangle (2D) and tetrahedra (3D) commonly used in the FEM, each permeance in the magnetic network has distinct physical meaning and has direct influence on the output results. The flux in the magnetic network can flow through a magnetic permeance in the defined direction only, whereas in the FEM there is no restriction on the direction of flux through any element. In a magnetic network, resistive components represent magnetic reluctances or permeances. The values depend on the material and geometry, and for ferromagnetic parts, the flux density in the region as well. To compute all the three components of the magnetic field at one point, the magnetic property of the element with a block shape, in which that point is located, can be modelled by 6 reluctances and 7 nodes, as shown in Fig.1. Any irregular hexahedron used for the solution region meshing can be transformed to an regular equivalent block.

x = y = z

(4)

This paper introduces how a 3D magnetic network model of a claw pole permanent magnet machine is built, considering all the peculiarities of machine construction and material properties. The MNM is validated by the 3D finite element analysis with ANSYS 5.6, a powerful commercial finite element software package. The computation shows that the MNM can be a very effective design tool for electromagnetic devices with complex structures. 2. THE MNM MODEL 2.1 The Claw Pole Machine Structure Fig.2 shows the structure of a claw pole permanent magnet (PM) machine prototype. This machine has an outer rotor structure, which comprises 20 surfacemounted NdFeB PMs and a mild steel cylinder. The stator consists of two claw-pole pieces mounted on a shaft. The stator winding is located between the two claw pole pieces and not shown in the figure for clarity.

Fig.1 3D reluctance model of a solid element In Fig.1, Rx, Ry, and Rz are the equivalent reluctances on the X, Y, and Z axes, respectively, and can be calculated as following (a)

Rx =

X 2 x 0 YZ Y Ry = 2 y 0 XZ Z Rz = 2 z 0 XY

(1) (2) (3)

where x, y, and z are the relative permeabilities on the X, Y, and Z axes, 0 is the permeability of air, X, Y, and Z are the side lengths of the solid element. The soft magnetic composite material and the mild steel, used as the stator core and the rotor core respectively, are magnetically isotropic and hence

(b) Fig.2 Magnetically relevant parts of (a) rotor and (b) stator of a claw pole PM machine The rotor is similar to that of a common PM ac machine, but the stator structure is more complex and

its magnetic field is truly 3D. Therefore, the magnetic network model must be 3D as well. Only 3D magnetic circuit network can correctly estimate the influence of the stator iron parts (particularly the tapered claws) and the air gaps with irregular shapes on the machine performance. Each magnetic flux path in the machine is represented by a magnetic permeance or reluctance in the magnetic network, which takes into account the geometry of the path and magnetic properties of the material. For the convenience of analysis, the claw pole stator and rotor iron cores are divided into several segments along both axial and circumferential directions, as shown in Fig.3. Each segment has a reasonably regular shape. The magnetic field path in any segment is assumed to be independent and connected to the others by non-linear magnetic reluctance.

estimated by the magnetic Ohms law as in the following equation when the tube has a regular shape: Rm = L r 0 A
(8)

where A is the equivalent or computational area, 0 the permeability of air, and r the relative permeability of the tube depending on the material, and for ferromagnetic parts, the flux flowing through it. When the flux tube is made of the permanent magnet, it is modelled as a combination of a magnetomotive force and a magnetic reluctance in series. The magnetomotive forces (mmf) of the magnet is calculated by Fm = H c Lm (9)

where Hc is the coercive force of the magnet and Lm the length along the field direction. The demagnetisation curve of the NdFeB magnets can be modelled as a straight line. The magnet reluctance is calculated by (8) where the magnetisation is equal to zero and the reluctivity is a constant. The stator and rotor reluctances are also calculated with constant reluctivities. The saturation phenomena of the ferromagnetic materials are taken into account as a variation of the relative reluctivity, which is a function of the flux flowing through them based on the given B-H curve. When the rotor rotates, the magnetic field in the machine changes periodically. This is taken into account by the main air gap reluctance, which is a function of the rotor position. Fig.4 illustrates one magnetic network model of the claw pole PM machine prototype at the original rotor position. The so-called original rotor position (0o) is where the major path of the flux produced by the PMs has the highest permeance, namely the rotor magnets and the stator claw poles share the same radial axes, respectively. In Fig.4, Rry, Rrm, Rml, Rg, Rcp, Rcl, Rsy and Rsh are the magnetic reluctances of one rotor yoke segment, the magnet, the leakage flux path between the rotor magnets, the main air gap flux path from the magnet to one claw-pole segment, the claw-pole segment, the leakage flux paths between claw-pole segments, the stator yoke, and the shaft, respectively. Frm is the mmf of the PM and Fs the mmf per pole of the stator winding. Fig.4 shows the basic structure of the magnetic network, but the practical model used in the magnetic field calculation may have much more nodes and elements. For example, the claw pole segment is a sub-network consisting of a number of radial, axial and circumferential reluctances, based on the accuracy requirement.

Fig.3 Claw-pole surface segments along both axial and circumferential directions 2.2 The Magnetic Network The magnetic network model is built with flux tubes described by magnetic reluctance or permeance values, and magnetomotive forces or flux sources. A flux tube is a 3D solid in which the flux is constrained to flow in one direction and may be identified by a finite element analysis [4,5]. The network topology is chosen according to the geometrical consideration based on the general knowledge of the direction of the flux tubes. The three dimensional magnetostatic finite element analyses can be used, assisting to achieve this goal, especially for the main air gap and the leakage reluctances. In the case of 3D magnetic field analysis using the magnetic scalar potential:

m = B dA
A

Fm = H dl = FmA FmB
l

(5) (6)
(7)

Rm =

Fm m

where Fm is the magnetomotive force applied along the length L of the tube, or the magnetic scalar potential difference between the two equipotential surfaces at the extremities, m the flux though the tube, and Rm the equivalent reluctance, which can be also

nodal magnetic potentials in the network equations. Assume m(k) to be the kth solutions after k times iterations, the (k+1)th solutions are achieved by the following linear equations:
[ J ]( k ) { m }( k ) = { s }( k )

(12)

where [J] is the Jacobian matrix and


{ m }( k ) = { m }( k +1) { m }( k ) { s }
(k )

(13) (14)

= { s } [G ] { m }
(k )

(k )

After a number of iterations, the right-hand side vector of (12) should approach zero if the iteration is convergent. The solution reaches the required accuracy when the following condition is satisfied: Fig.4 Equivalent reluctance network 3. FIELD ANALYSIS WITH MNM MODEL 3.1 Network Equations When the magnet permeances (or reluctances) and mmfs are known, the nodal magnetic scalar potentials can be calculated by the following network equations: [G ]{ m } = { s } (10) where [G] is the magnetic permeance matrix, {m} the nodal magnetic scalar potential vector, and {s} the flux source vector. Then, the magnetic flux, the flux density and field strength in each element, which is required for the calculation of the electrical equivalent circuit parameters and performance of the electrical machine, can be determined from the nodal magnetic potentials. Because the permeance matrix [G] includes some nonlinear elements Gi (i=1,2,,n), these equations are solved by the Newton-Raphson iteration method. These elements form a non-linear network and the values can be determined by the magnetic potential differences across them. The network equations are solved by the following steps: (1) A series of nodal magnetic potentials are assumed as the initial values under a certain mmf Fm (e.g. the magnet mmf for no-load analysis), and the potential differences Fi across the two terminals of the equivalent permeance network Gi are considered as the equivalent magnetic potential sources. When the total flux i flowing through Gi are obtained by solving the modified network equations, where Gi are replaced by the equivalent magnetic potential sources Fi , the Gi elements can be determined by 3.2 Flux Calculation When the nodal magnetic potentials are solved, the flux flowing through the branches can be obtained with the magnetic Ohms law:

[{ s [G ]( k ){ m }( k ) }]2
where is a small enough positive quantity.

(15)

m =

F m Fm1 Fm 2 Fs = Rm Rm

(16)

where Fm1 and Fm2 are the magnetic scalar potentials at the two terminal nodes, Fs is the mmf produced by the permanent magnets and/or the stator current, along the direction from node 2 to node 1, and Rm is the equivalent reluctance of the branch. The magnetic reluctance is flux-dependent when the branch consists of ferromagnetic materials. In this case, the numerical iteration is needed as discussed in Section 3.1. The flux flowing through several typical crosssections at no-load operation by the magnetic network model in Fig.4 are listed in Table 1. The results obtained by the FEA, which is to be presented in the next section, are also listed in the table for comparison. Table 1. Calculated magnetic fluxes through various cross-sections (in mWb) Cross-sections MNM FEM Stator yoke and shaft 0.503 0.489 Rotor yoke 0.0480 0.0468 One rotor magnet 0.123 0.115 Main airgap per pole-pitch 0.118 0.114 3.3 Calculation of Stator EMF Waveform When the rotor rotates, the flux linking the stator winding varies and an electromotive force (emf) is induced. The emf frequency depends on the rotor speed while the emf waveform is determined by the waveform of the flux. The total flux flowing though

Gi =

Fi

(11)

(2) The Newton-Raphson method is used to solve the

the stator yoke and shaft can be considered as the total flux linking the stator coil. At the different rotor positions, the air gap reluctances have different values and the magnetic network model may need to be modified. For example, when the rotor moves 180o (electrical), the magnetic network model is the same as that shown in Fig.4, but that the direction of the mmfs produced by the rotor magnets changes. The flux 1 linking the stator winding at different rotor angles from 0o to 180o (electrical) are calculated and the corresponding emf magnitude induced can be determined by
E1 = 2fN 11

periodic boundary conditions such that the flux density on the two half periodic boundary planes obey the following equations:
Br (r , , z ) = B r (r , , z ) B (r , , z ) = B (r , , z ) B z (r , , z ) = B z (r , , z )

(19) (20) (21)

and the degrees of freedom or the scalar magnetic potentials have to satisfy

m (r , , z ) = m (r , , z )

(22)

(17)

where N1 is the number of turns of the stator coil and f the emf frequency determined by the rotor speed, viz.

f =

pn 60

(18)

where p is the number of pole pairs of the machine and n the rotor speed in rev/min. When the machine runs at 300 rev/min, the emf frequency is 50 Hz. The emf waveform of the stator winding is plotted in Fig.5. It is almost sinusoidal versus the rotor position (in electrical degrees). It should be noted that the flux linking the stator winding and the induced emf have the same shape of waveforms, but shifted by 90o (electrical).

Fig.6 illustrates the magnetic flux density vector plots at the original rotor position when the machine operates at no-load. It can be found that the major path for the magnetic flux produced by the permanent magnets is along one of the permanent magnets, the main air gap, one of the claw pole stator pole pieces, the stator yoke and shaft, the other claw pole stator piece, the main air gap, another permanent magnet, and then the stator yoke to form a closed loop. It can also be seen that a large amount of leakage flux exists between the claw poles and the magnets. This plot proves the general knowledge about the flux directions used in the magnetic network method. The flux flowing through different cross-sectional areas are calculated and compared with those obtained by the MNM, as shown in Table 1. The results by two methods have a reasonably good agreement.

Fig.5 EMF waveform of the stator winding 4. FIELD ANALYSIS WITH 3D FEM To validate the MNM method, 3D finite element analysis has been performed to calculate the magnetic field distribution in the claw pole permanent magnet machine prototype as shown in Fig.2. ANSYS 5.6, a powerful commercial finite element software package, was employed for the calculation. Considering the symmetrical nature of the structure, it is only required to analyse the magnetic field in one pole pitch [6]. To determine the 3D magnetic field distribution, the scalar magnetic potential was employed with half

Fig.6 Magnetic flux density vectors at no-load The theory applied to the calculation of the flux passing through one cross-sectional area is

m = B dA
A

(23)

For numerical calculation, the cross-section is divided into a number of small sub-areas and the flux density in the middle of each sub-area can be considered as the value through all the sub-area if the division is fine enough. Then, equation (23) can be rewritten as

REFERENCES [1] Ostovic, V., Dynamics of Saturated Electric Machines, New Yoke, Springer-Verlag, 1989 [2] Ostovic, V., Miller, J.M., Garg, V.K., Schultz, R.D., and Swales, S.H., A Magnetic-EquivalentCircuit-Based Performance Computation of a Lundell Alternator, IEEE Transactions on Industry Applications, Vol. 35, No. 4, July/August 1999, pp825-830 [3] Rasmussen, C.B., and Ritchie, E., A Magnetic Equivalent Circuit Approach for Predicting PM Motor Performance, Proceeding of IEEE IAS Annual Meeting, New Orleans, Louisiana, 5-8 October 1997, pp10-17 [4] Roisse, H., Hecquet, M., and Brochet, Simulations of Synchronous Machines Using an Electric-Magnetic Coupled Network Model, IEEE Transactions on Magnetics, Vol. 34, No. 5, September 1998, pp3636-3659 [5] Hecquet, M. and Brochet, P. Time Variation of Forces in a Synchronous Machines Using Electric Coupled Network Model, IEEE Transactions on Magnetics, Vol. 34, No. 5, September 1998, pp3214-33217 [6] Guo, Y.G., Zhu, J.G., and Ramsden, V.S., Calculation of Cogging Torque in Claw Pole Permanent Magnet Motor, Proceeding of Australasian Universities Power Engineering Conference, Darwin, Northern Territory, Australia, 26-29 September, 1999, pp343-347

m = Bi Ai
i

(24)

The flux flowing through the stator yoke and shaft is calculated as the effective flux of the stator winding. When the rotor rotates, the flux varies and an emf is induced in the winding. The flux waveform was calculated by rotating the rotor magnets for one pole pitch in 20 steps. The corresponding emf waveform at 300 rev/min, computed from (17) and (18), is plotted in Fig.5. The results by two methods have reasonably good agreement. 5. CONCLUSIONS The magnetic network method is in fact a distributed magnetic circuit method. The MNM can achieve comparable computational accuracy to the FEM with much fewer elements and shorter CPU time. The fact that the nodes are concentrated in the crucial regions such as the air gap enables accurate calculation of magnetic fields. On the other hand, owing to its distributed nature, the MNM is far more accurate than the conventional magnetic circuit method when the electrical machine structure is very complex. Therefore, the MNM can be a practical and useful method for the optimum design of electromagnetic devices with complex structures.

You might also like