IEEE TRANSACTIONS ON MAGNETICS, VOL. 48, NO. 2, FEBRUARY 2012 915

Performance Comparison of Longitudinal Flux and Transverse FluxPermanent Magnet Machines for Turret Applications With Large Diameter

Ji-Young Lee�, Do-Kwan Hong�, Byung-Chul Woo�, Doo-Hwan Park�, and Byoung-Uk Nam�

Korea Electrotechnology Research Institute, Changwon, Gyeongnam, 641-120, KoreaDoosan Corporation Mottrol, Seongsan-Gu, Changwon, Gyeongnam, 456-3, Korea

Agency for Defense Development, Yuseong, Daejeon, 305-600, Korea

The aim of this paper is to provide performance comparisons between conventional longitudinal flux and transverse flux permanent-magnet machines using three-dimensional finite element analysis considering both electro-magnetic and mechanical fields. For a turretapplication with 2 meters in diameter size, not only electro-magnetic performance but also mechanical structural strength is important.Therefore, the comparison is focused on the torque density, machine efficiency, time constant, magnetic forces, and mechanical equivalentstresses occurred by the magnetic forces. For a specific turret application without an external cooling system, the obtained results providean indication on which type of the machines is best suited regarding performance and size.

Index Terms—Finite element analysis, longitudinal flux, transverse flux, turret.

I. INTRODUCTION

S EVERAL papers have presented direct-drive machine de-signs, and the direct-drive machines from [1]–[3] are made

for high torque and low torque ripple application, and they arebased on surface mounted permanent-magnet (PM) machines.Most of these are about conventional longitudinal flux machines(LFMs). Some papers have insisted that transverse flux ma-chines (TFMs) are good for direct-drive applications becauseof high torque density [3]–[5], but the TFMs have high torqueripple. And the TFMs with high torque mostly possess interiorPM rotor or long axial length which are good components to in-crease torque even in LFMs. Moreover it is hard to find a paperregarding performance comparisons between LFM and TFM fordirect drive applications under the fair and various conditions.One good example is in [3], but the conditions are focused ondownhole applications which have limited outer diameter andrelatively long axial length, and there is no consideration formechanical construction.

In this paper, PM excited LFM and TFM are investigated forturret applications where the machine inner and outer diame-ters are limited to 1 and 2 m respectively, and the axial lengthis also limited to less than 100 mm by bearing thickness. It isa totally opposite configuration from the downhole applicationshape in [3]. First, the configuration and constraints of the LFMand TFM are introduced. Next, it is explained that the analysismodels and computation methods for electro-magnetic and me-chanical parameters based on three-dimensional finite elementanalysis (3-D FEA). And then, the comparison results are pre-sented. The comparisons are focused on the torque density andmachine efficiency for output performances, and resistance andinductance for precision controllability, and magnetic forces andmechanical equivalent stress for structural stability.

Manuscript received July 07, 2011; revised September 29, 2011; acceptedOctober 21, 2011. Date of current version January 25, 2012. Correspondingauthor: J.-Y. Lee (e-mail: jylee@keri.re.kr).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TMAG.2011.2174038

II. CONFIGURATIONS AND CONSTRAINTS

A. Configurations of LFM and TFM

According to the relative plane direction of the magnetic fluxloop to the direction of motion, electric machines can be catego-rized into LFMs and TFMs [6]. In LFMs the loop of the usefulflux lies in longitudinal or axial planes to the direction of motion.These machines are the conventional types, and they have gen-erally distributed or concentrated windings. In TFMs the loopsof the working flux lie in planes transverse to the direction ofmotion, and they have generally torus or ring shaped windings.Each of them has many construction variations, depending onspecific applications. In turret applications, machine construc-tion is chosen based upon the following considerations.

1) Cylindrical shape and radial directional air-gap: The turretsystem requires cylindrical pan-cake shaped machines.Since strong axial-directional exterior impacts are ex-pected in a special environment, radial directional air-gapis selected.

2) External-rotor machines: Normally, with the same dimen-sion, external-rotor machines could provide higher torquedensity than internal-rotor machines because the formercan have greater air-gap radius [3].

3) The shortest coils: Considering the limited machinevolume, concentrated winding is considered for LFM andring shaped winding is considered for TFM.

Fig. 1 shows the selected LFM and TFM configurations.

B. Constraints of LFM and TFM

To fairly perform comparisons between the two machines,some constraints have to be given, as listed in the following.

1) The inner and outer diameters and axial length are fixedbecause of the system volume limit.

2) The great number of poles is good to increase torque andto reduce time constant in both LFM and TFM [3]. Andin the case of using incremental encoder, the choice of 2multipliers is good for this number of poles to eliminatethe accumulated position error. The number of PM in TFMwas the same as that of LFM in the beginning. For the easyfabrication, PM was divided and shifted by a pole-pitch as

0018-9464/$31.00 © 2012 IEEE

916 IEEE TRANSACTIONS ON MAGNETICS, VOL. 48, NO. 2, FEBRUARY 2012

Fig. 1. Machines for turret applications. (a) LFM, (b) TFM.

TABLE ICOMMON CONSTRAINTS FOR THE DESIGN

Fig. 2. Meshed 3-D FEA model for magnetic field analysis. (a) LFM, (b) TFM.

shown in Fig. 1(b) instead of stator core twisting as shownin [5]. Therefore the number of poles looks physically dif-ferent, but the same number of poles is magnetically con-sidered between the two machines.

3) Number of slots and phases: The variable pole and slotcombinations can reduce torque ripple in LFM [2], butTFM has a limited combination because of separated coreshapes and independent phase arrangement. Since torqueripple can be reduced by several ways such as in [1] and[5], the variable pole and slot combinations are not consid-ered in this paper. Phase is also fixed for the same standardcontrol systems.

The assumed common constraints for the design are listedin Table I. The specific design have been done first for LFMunder the condition that winding fill factor and current densityare 70% and 4 respectively in addition to the commonconstraints in Table I because the turret system is totally en-closed. In the TFM design, the same rotor is used and only statoris redesigned. The axial length of stator tooth in TFM is the sameas that of LFM to have the same outer volume, and the width ofstator tooth is selected to have the highest torque. The numberof winding is selected for the same electro-motive force (EMF)at no load condition and the maximum speed.

III. ANALYSIS MODELS AND COMPUTATION METHODS

A. Analysis Models

3-D FEA is used to compute the two motor performances andthe non-linear magnetic characteristics of materials are consid-ered. Fig. 2 shows the meshed models using second order tetra-hedral elements for magnetic field analysis. Considering peri-odic boundary condition, 1/128 model is used for the analysismodel of LFM. In TFM only one phase among 1/128 model ismodeled for magnetic field analysis due to the decoupling ofmagnetic flux paths and armature coils [7].

When the magnetic forces are calculated in the rotor model,the force calculation regions should be divided according to thenumber of PM, as well as the configuration of PM segmentsinstead of calculating over entire rotor model. It is because localforces can cause mechanical stress although the vector sum iszero.

B. Analysis and Computation Methods

As mentioned before, torque density, machine efficiency,electric impedance, magnetic forces and mechanical equivalentstress are calculated. Except the equivalent stress, all param-eters are calculated with the electro-magnetic field analysisresults. And the equivalent stress is calculated by mechanicalfield analysis with the magnetic force as a source.

The each parameter was calculated as follows.1) Torque density: Torque is determined using the principle

of Coulomb virtual work (CVW), and the torque, , on amoving object about the axis of rotation is given by thefollowing relationship [8], [9]:

(1)

where is the magnetic coenergy of the system, isthe rotor position, and is the phase current. The currentis held constant. Unlike the classical virtual work method,the moving object is not actually rotated during the forcecomputation. Instead, only the tetrahedral that lie along theoutside surface of the object are virtually distorted [9]. Thetorque density is derived from dividing the torque by totalweight of the electro-magnetic circuit consisting of statorand rotor cores, coils, and PMs.

2) Machine efficiency: The machine efficiency, , can be cal-culated by

(2)

where is the torque, is the rotational speed, is thecopper loss, is the mechanical loss, and is core loss.The denominator part is input power.The mechanical loss is the sum of the windage and frictionlosses, and frequently the loss can be ignored compared tocore loss. It is because windage loss is trivial amount at lowspeed and friction loss is also negligible due to bearings.Therefore, mechanical loss is ignored in this paper.The core loss includes static hysteresis loss, eddy currentloss, and excess loss. The eddy current loss can be caused

LEE et al.: PERFORMANCE COMPARISON OF LONGITUDINAL FLUX AND TRANSVERSE FLUX PERMANENT MAGNET MACHINES 917

by the both normal and tangential component of flux den-sity [10]. When LFM or TFM has separated laminationcores, the eddy current loss caused by the normal compo-nent of flux density is very small amount [11], [12], andthe selected model between LFM and TFM will be fabri-cated with the separated lamination core. Therefore, onlyeddy current loss caused by the tangential components offlux density, hysteresis loss, and excess loss are calculatedhere with the general formulas consisting of each loss coef-ficient, frequency, and flux density amplitude as presentedin [10].

3) Electric impedance: The electric impedance consists ofstator winding resistance and inductance. A coil resistancehas nonlinear property according to its temperature [13],but here it was assumed that the coil temperature of thetwo machines is the same as room temperature.For the inductance calculation, differential inductancevalue was calculated by the flux linkage method as follows[14]:

(3)

where is the total flux linking the th winding, andis the current circulating in winding . Thus, this is aboutthe mutual inductance between circuit and , and it canbe self inductance when p equals q. Since the inductancewaveform is sinusoidal, the average values of the induc-tance are used when time constant is calculated.

4) Magnetic force: For the local force calculation, the CVWmethod is used like the torque computation. The force, ,on the moving object in the direction of the displacement,, is given by the following relationship [8], [9]:

(4)

where is the magnetic coenergy of the system,is the magnetic flux density, and is the magnetic fieldintensity. Here the current is also held constant.

5) Mechanical equivalent stress: The magnets themselves ofsurface mounted PM machine bear the possibility to bebroken or shifted from the original position by stress suchas electromagnetic exciting force. Therefore the electro-magnetic forces on each PM module consisting of PMand its back yoke are computed first, and then mechanicalequivalent stress by the forces is calculated.As the mechanical equivalent stress, the von Mises stressis computed. A physical interpretation of von Mises crite-rion is that yielding begins when the elastic energy of dis-tortion reaches a critical value. The von Mises criterion isalso known as the maximum distortion strain energy crite-rion [15]. The von Mises stress, , is computed with theprincipal stresses as follows [15], [16]:

(5)

IV. PERFORMANCE COMPARISONS

After computing all above-mentioned parameters for theLFM and TFM, the results are compared in Table II.

TABLE IIDESIGNED AND CALCULATED PARAMETERS

Although the total volume is the same for each machine, thetotal weight is different. This is because the two machines havedifferent stator core configurations due to different phase ar-rangement. In the LFM, the phase is distributed in radial di-rection, so the stator cores become almost full in the assignedvolume for stator except coil volume as shown in Fig. 1(a). Inthe TFM, however, the phase is distributed in axial direction,and the phase coil has a ring shape around axis as shown inFig. 1(b). So, there are vacant spaces between each stator tooth.In conclusion, the different weight in the same volume can meanthe different ability to make the best use of the volume. Addi-tionally, this difference affects the effective spaces where forcesoccur. The torque of TFM is less than the torque of LFM. Thisalso applies for the total weight, therefore the torque densitiesare almost the same. Effectively, the torque density of TFM isslightly higher than that of LFM.

The number of turns per phase in LFM is much more thanthat in TFM, but, the number of turns per slot in LFM is muchless. This is because of the different winding patterns, whichcause different impedances. The resistance in TFM is 28 timesas much as that in LFM, and the inductance is 133 times. As therelationship of the resistance and inductance, the time constantin TFM is 5 times bigger. As a consequence, the copper lossin TFM is extremely more than that in LFM because of biggerresistance, and the efficiency becomes considerably less. It isnoticeable that the core loss is not much different compared withcopper loss. This result shows that the flux density or saturationlevel in each core is similar.

Fig. 3 shows the force calculation results for one rotormodule in the two machines. In LFM the theta-directional forceis caused by composite magneto-motive force (MMF) of threephases, but in TFM it is caused by MMF of only one phase.If the theta-directional forces are summed for three phases inTFM, the difference of the total amount of theta-directionalforces of the two machines is the same as the difference oftorque. The r- and z-directional forces are not effective, andthey cause mechanical stress or vibration instead. In LFM theaverage r-directional force is much bigger than that in TFM,

918 IEEE TRANSACTIONS ON MAGNETICS, VOL. 48, NO. 2, FEBRUARY 2012

Fig. 3. Force variations according to time in one PM and its back yoke of anal-ysis models. (a) LFM, (b) TFM.

Fig. 4. Equivalent stress in the rotor. (a) LFM, (b) TFM.

but the ripple component of the force in LFM is relativelynegligible. If uniform forces are distributed in circumferentialdirection, ideally there is no movement, and so is the situationby the r-directional force in LFM. On the contrary, the r-di-rection force in TFM is variable according to rotating, and theforce variation cannot be cancelled out by sum of three phaseforces. For the z-direction force in TFM, the ripple becomesmuch bigger.

Fig. 4 shows the equivalent stresses occurred by the three di-rectional forces on the rotor as shown in Fig. 3 at 6.5 millisec-onds. The equivalent stress in the rotor of LFM which is influ-enced by the composite forces by three phases is much smallerthan the stress in the rotor of TFM which is influenced by eachphase force.

V. CONCLUSION

For a turret application, the two machines—LFM with con-centrated winding and TFM with ring shaped winding—aredesigned under the limited volume having pan-cake shape. Andthen, electro-magnetic and mechanical parameters are com-puted based on 3-D FEA. After the performance comparison,the following conclusions are obtained.

1) The weight of TFM is lighter and the torque density issimilar to that of LFM.

2) In the LFM, the machine efficiency is higher, and the timeconstant and the mechanical stress are smaller.

Therefore, LFM is recommended for the turret application withlarge diameter and short axial length.

Table III is the list of calculated and measured results fora fabricated LFM after optimum design process under slightly

TABLE IIICALCULATED AND MEASURED RESULTS FOR A FABRICATED LFM

changed conditions such as increased requirement torque over6000 Nm. This comparison shows the accuracy of analysis re-sults presented in this paper.

ACKNOWLEDGMENT

This work was supported by the Research Project for Di-rect Drive & High Efficiency Electric System of DefenseAcquisition Program Administration and Agency for DefenseDevelopment.

REFERENCES

[1] W. Fei and P. C. K. Luk, “A new technique of cogging torque sup-pression in direct-drive permanent-magnet brushless machines,” IEEETrans. Ind. Appl., vol. 46, no. 4, pp. 1332–1340, 2010.

[2] R. Wrobel and P. H. Mellor, “Design considerations of a direct drivebrushless machine with concentrated windings,” IEEE Trans. EnergyConvers., vol. 23, no. 1, pp. 1–8, 2008.

[3] A. Chen, R. Nilssen, and A. Nysveen, “Performance comparisonsamong radial-flux, multistage axial-flux, and three-phase trans-verse-flux PM machines for downhole applications,” IEEE Trans. Ind.Appl., vol. 46, no. 2, pp. 779–789, 2010.

[4] Y. G. Guo, J. G. Zhu, P. A. Watterson, and W. Wu, “Development ofa PM transverse flux motor with soft magnetic composite core,” IEEETrans. Energy Convers., vol. 21, no. 2, pp. 426–434, 2006.

[5] H. T. Ahn, G. H. Jang, J. H. Chang, S. U. Chung, and D. H. Kang,“Reduction of the torque ripple and magnetic force of a rotary two-phase transverse flux machine using herringbone teeth,” IEEE Trans.Magn., vol. 44, no. 11, pp. 4066–4069, 2008.

[6] S. A. Nasar and I. Bolder, Linear Motion Electric Machines. NewYork: Wiley, 1976, pp. 220–261.

[7] E. Schmidt, “3-D finite element analysis of the cogging torque ofa transverse flux machine,” IEEE Trans. Magn., vol. 41, no. 5, pp.1836–1839, 2005.

[8] A. Benhama, A. C. Williamson, and A. B. J. Reece, “Force and torquecomputation from 2-D and 3-D finite element field solutions,” IEEProc.-Electr. Power Appl., vol. 146, no. 1, pp. 25–31, 1999.

[9] “Maxwell 3D technical notes,” Maxwell Online Help Ansys Inc., 2010.[10] D. Lin, P. Zhou, and Q. M. Chen, “The effect of steel lamination core

losses on transient magnetic field using T- � method,” presented atthe IEEE Vehicle Power and Propulsion Conf. (VPPC), Harbin, China,2008.

[11] B. C. Mecrow and A. G. Jack, “The modeling of segmented laminationsin three dimensional eddy current calculation,” IEEE Trans. Magn., vol.28, no. 2, pp. 1122–1125, 1992.

[12] J. Y. Lee, S. R. Moon, D. H. Koo, D. H. Kang, G. H. Lee, and J. P.Hong, “Comparative study of stator core composition in transverse fluxrotary machine,” J. Elect. Engrg. Technol., vol. 6, no. 3, pp. 350–355,2011.

[13] S. M. Cho, J. K. Kim, H. K. Jung, and C. G. Lee, “Stress and thermalanalysis coupled with field analysis of multilayer buried magnet syn-chronous machine with a wide speed range,” IEEE Trans. Magn., vol.41, no. 5, pp. 1632–1635, 2005.

[14] R. Escarela-Perez, E. Campero-Littlewood, M. A. Arjona-Lopez, andA. Laureano-Cruces, “Comparison of two techniques for two-dimen-sional finite-element inductance computation of electrical machines,”IEE Proc.-Electr. Power Appl., vol. 152, no. 4, pp. 855–861, 2005.

[15] [Online]. Available: http://en.wikipedia.org/wiki/Von_Mises_yield_criterion

[16] “Theory reference,” Ansys Help System Ansys Inc., 2009.