Lumped-Parameter Thermal Network Model and Experimental Research of Interior PMSM for Electric Vehicle

CES RANSACIONS ON ELECRICAL MACHINES AND SYSEMS, VOL. 1, NO. 3, DECEMBER 017 367 Lumped-Parameter hermal Network Model and Experimental Research of Interior PMSM for Electric Vehicle Qixu Chen, Zhongyue Zou, and Binggang Cao Abstract A 5kW interior permanent magnet synchronous machine (IPMSM) applied to the electric vehicle is introduced in the paper. A lumped-parameter thermal network model is presented for IPMSM temperature rise calculation. Furthermore, a 3D liquid-solid coupling model considering the assembly clearance is compared with the D lumped-parameter thermal network model. Finally, a dynamometer platform for temperature rise measurement is established to verify the above-mentioned methods, which obtains the measured efficiency map at rated load case and overload case. At the same time, the measured no-load back electromotive Force (EMF), load line input voltage and load current are gathered. hermocouple PC100 is used to measure the temperature of the stator winding and iron core, and the FLUKE infrared thermal imager is applied to measure the surface temperature of PMSM and controller. esting result shows that the lumped-parameter thermal network have a high accuracy to predict each part temperature. Index erms Interior permanent magnet synchronous machine, lumped-parameter thermal network, liquid-solid coupling, thermal resistance, thermal conductance. I. INRODUCION NERIOR permanent magnet synchronous machine I(IPMSM) is applied to the field of electric vehicle (EV) and hybrid electric vehicle (HEV) for its higher power density and torque density. he massive heat bring challenge for safe operation of PMSM. herefore, accurate temperature rise calculation becoming increasingly more and more important. [1] here are three basic analysis methods: 1) lumped-parameter thermal model, ) the finite element method (FEM) and 3) computed fluid dynamics (CFD). Mellor et al [] studied an induction motor using the lumped parameter thermal model to calculate the average temperatures at different parts of the motor. A lumped-parameter thermal network model for radial flux PMSM [3 6] and axial flux PMSM [7 8] is used to calculate the average temperatures for different parts of PMSM. Fluid flowing characteristics including gap air and coolant are studied [9 10]. Coupled electromagnetic and thermal analysis his work was supported in part by the new energy source motor and its control laboratory of Suda Electric Vehicle co. ltd. he authors are with electric vehicle and system control institute, Xi an jiaotong University.( e-mail:xianjiaotong014@163.com; moon4715@163.com; cbg@mail.xjtu.edu.cn). of PMSM is carried out in the paper [11 14] and a testing platform was established by temperature sensor and infrared thermal imager. FEM simulations can be accurate to compute temperature distributions of motor parts, but convection heat transfer coefficients generally use empirical formula, which lead to the uncertainty of results. In this paper, an interior radial-flux PMSM is introduced in the paper. Forced liquid (50%water+50% glycol) cooling in shell jacket is adopted. A lumped-parameter thermal network models for thermal design and analysis of IPMSM is presented. his method that is divided into two stages combines both analytical method and CFD simulations. In the first stage, fluid can be modeled using CFD tools to compute the convective heat transfer coefficients between solid surface and fluid surface, which are necessary for FEM and lumped-parameter thermal network model. In the second stage, convection heat transfer coefficients obtained by CFD simulations can be used to amend the empirical formula. In this way, a simplified model considering the symmetry and period reduces the computation time obviously. Finally, a temperature rise testing platform taking the rated load and overload into account is built to verify the above-mentioned lumped-parameter thermal network result and CFD simulation result. Furthermore, a measured efficiency map and phase current wave are obtained at operation region. he temperature measurements of PMSM are carried out using thermocouple PC100 and an FLUKE infrared thermal camera device. II. LUMPED-PARAMEER HERMAL NEWORK A lumped-parameter thermal network model applied to IPMSM is presented in the paper. he main parameters of the IPMSM in the paper is listed in able I. Its geometry model and equivalent heat source model are shown in Fig.1 and Fig. respectively. Each cylindrical component can be descripted by lumped-parameter thermal network composed of radial resistance, axial resistance and heat source. he definition of thermal resistances in lumped-parameter thermal network is related to thermal conductivity of material and its dimensional information. In the paper, the heat transfer mechanisms of IPMSM, which are generally classified as thermal conduction and thermal convection, are discussed in detail.

368 CES RANSACIONS ON ELECRICAL MACHINES AND SYSEMS, VOL. 1, NO. 3, DECEMBER 017 ABLE I IPMSM DIMENSION AND PARAMEERS Parameter Value Rated power P [kw] 5 Rated speed n [rpm] 3300 inner diameter D si [mm] 148 outer diameter D so [mm] 08 iron length L ef [mm] 10 Number of stator parallel branches a Number of turns per coil N s 10 Length of air gap δ [mm] 1.5 Permanent magnet thickness h pm [mm] 7 Number of stator slots Q 36 Number of pole pairs p 3 Winding connection Y Coolant inlet flow velocity V [L/min] 5 Outlet iron coil PM Rotor iron Fig. 1. IPMSM geometry model (half model). Fig.. Equivalent model of heat source (half model). A. hermal Resistance Definition Inlet Shell Water channel A general cylindrical component and its lumped-parameter thermal network that is derived from the solution of the heat conduction equations are shown in Fig.3. he corresponding thermal resistance definition is given in able II [5-7], where k r and k a are the thermal conductivities in the radial and axial directions respectively; l is the axial length and r and r 1 are the outer and inner radius of the cylindrical component; 3 and 4 are the unknown temperatures on the inner and outer surfaces, and 1 and are the unknown temperatures at two end face; m represents the average temperature of the component; P and C represent the corresponding internal loss and thermal capacitance. hermal capacitance in steady state analysis is here under no consideration. 1 hermal Capacitance 3 r1 4 Ra1 Axial C Radial Rr1 network network r m 3 l Ra3 Rr3 Ra 1 P Rr Heat source (a) (b) 4 Fig. 3. (a) Cylindrical component (b) Radial and axial thermal network. ABLE II HERMAL RESISANCE DEFINIION OF CYLINDRICAL COMPONEN R Radial direction Axial direction R 1 1 r ln ( r r ) R R 3 1 π kl r r 4 r 1 π kl r r 4 r 1 1 l π k ( r r ) a 1 r1 ln ( r1 r) l 1 1 π k a ( r1 r ) 4rr 1 ln ( r1 r) r l r1 r 6π k a 8π kl r ( r1 r ) ( r r ) 1 r + 1 able III gives some typical values for the thermal conductivities of solid materials applied to IPMSM [6]. ABLE III PMSM HERMAL PARAMEERS hermal conductivity Value Silicon steel sheet λ SiW/(K.m)](x,y,z) (4.5,45,45) Copper λ Cu [W/(K.m)] 386 Aluminum alloy λ Al [W/(K.m)] 01 Winding insulation λ Insu [W/(K.m)] 0.6 PM λ pm [W/(K.m)] 8 Steel 4CrMo λ St [W/(K.m)] 45 Heat conduction and thermal convection including liquid coolant and air are considered in the calculation of the IPMSM temperature rise. B. Heat Conduction Calculation General formula of heat conduction is shown as λ Amat Gmat = (1) Lmat he equivalent thermal conductance with or without assembly clearance is considered as follows. 1 Gmat1, = 1 Gmat1+ 1 Gair + 1 Gmat () 1 Gmat1, = 1 Gmat1+ 1 Gmat C. Heat Convection Calculation Most of heat generated by IPMSM could be taken away by circulating liquid coolant in shell jacket. Shell jacket includes S-type channel filled with liquid coolant, and its pressure loss is shown in Fig.4. Outlet Fig. 4. Pressure loss along the path. Inlet

CHEN et al.: LUMPED-PARAMEER HERMAL NEWORK MODEL AND EXPERIMENAL RESEARCH OF INERIOR PMSM 369 FOR ELECRIC VEHICLE he classical Bertotti loss separation model [15] is used to calculate core loss, which breaks the total core loss into static hysteresis loss, classical eddy current loss and extra loss. Core loss Pv is obtained by a group of flux density loss BP curve at different sinusoidal work frequency 100Hz, 00Hz, 400Hz, 1000Hz. Pv = Ph + Pc + Pe (3) α 1.5 1.5 = Kh fbm + Kc f Bm + Ke f Bm where P v iron core loss; P h hysteresis losses; P c eddy current losses; P e extra eddy current loss; B m AC component magnitude of flux density; f frequency; K h,α coefficient of hysteresis losses (α=); K c coefficient of the eddy current losses ;K e coefficient of the extra eddy current loss. hree iron loss coefficients K h, K c, K e is obtained by 1.5 v 1 m m P = KB + K B (4) 1.5 where K1 = Kh f + Kc f ; K = Ke f Coefficient of the classical eddy current losses K c is calculated by Kc σ d 6 = π (5) where σ electrical conductivity; d thickness of a slice of silicon sheet. In order to satisfy the minimum of the quadratic form, the parameter K 1, K can be computed by 1.5 1 = vi 1 mi + mi = (6) f( K, K ) P ( KB K B ) min where P vi,b mi the i th data point (P vi,b mi) measured on the iron core loss curve Loss coefficient K h,k e can be described as 1.5 h ( 1 c 0 )/ 0 ; e / 0 K = K K f f K = K f (7) where f 0 measured frequency of loss curve Due to the stator slot opening and harmonic current, rotor iron core and PM are exposed to high-order time harmonic of current and armature magnetic-motive-force (MMF) space harmonic in the air-gap field, which rotate at different speed relative to the rotor speed. he induced eddy current loss [16], [17] in PM and rotor iron core is calculated by 1 1 P E dvdt J σdvdt edd _ PM = σ = 0 0 z v v (8) where P r total rotor loss; σ electrical conductivity; E electric field intensity; J z eddy current density; V space integration volume of loss. he induced eddy current density J z can be represent by J z Az = σ t he eddy current losses P edd_pm can be determined directly from the magnetic vector potential as (9) P edd _ PM 1 Az = σ 0 t V he above equation (10) in discrete form is dvdt (10) N step edd _ PM = σ step z, k + 1 z, k k = 1 ( ) P p f N V A A (11) he total loss is the sum of the losses of all parts as shown in P = P + P + P + P + P + P + P = QC losses Cu Fe edd _ pm ro air be ex (1) Q v = (13) A where P losses is total loss of PMSM, P Cu is copper loss, P Fe is stator iron loss, P pm is eddy current loss of PM, P ro is eddy current loss of rotor, P air is friction loss of air, P be is bearing loss, P ex is extra loss, V is the velocity of the liquid coolant, Q is the flow of liquid coolant, A is the cross-sectional area of water channel. hermal conductance between inner wall of shell jacket and liquid coolant is shown by 1/3 0.14 1/3 1/3 d e v Nu 1.86 Rewa Pr Re<00 wa = L υwa 0.8 0.4 0.03Rewa Pr Re<00 (14) de ab Rewa = v = v υwa υwa ( a+ b) (15) λ h Nu wa wa = D (16) Gwa = hwa Awa (17) where λ wa is the thermal conductivity of liquid coolant, Nu wa is the Nusselt number of liquid coolant, Re wa is the Rayleigh number of liquid coolant., Pr is the Prandtl number, D e is the equivalent diameter, a is the width of water channel, b is height of water channel, υ wa is kinematic viscosity coefficient of liquid coolant, h wa is convective heat transfer coefficient of liquid coolant, L is the path length of water channel. hermal conductance G air1 between rotor end surface and gap air is shown by (18) (1).Where the heat transfer coefficient h air1 of air can be obtained by classical Nusselt number. Nu λ h air1 = (18) δ for a<1700 0.367 4 Nu = 0.18a for 1700<a<1e (19) 0.41 4 7 0.409 a for 1e <a<1e where aylor number a and Reynolds number Re is shown as Re δ a = (0) r v δ Re = (1) υ e

370 CES RANSACIONS ON ELECRICAL MACHINES AND SYSEMS, VOL. 1, NO. 3, DECEMBER 017 where υ is kinematic viscosity of the air (m /s) and v is linear speed of the rotor (m/s); hermal conductance at air gap is obtained by G = h A = h π DL () air1 air1 air1 air o Start Establish -type heat network model Compute loss Establish heat equilibrium equation Compute node temperature rise Heat transfer coefficient h air and thermal conductance G air between the rotating shaft and internal air can be written as G = h A = h π DL (3) air1 air1 air1 air o G = h A (4) air air air Heat transfer coefficient h air3 and thermal conductance G air3 at the location of end winding surface can be written as illustrated by hair3 = 14. (1 + k v) (5) G = h A (6) air3 air3 air3 where k is the coefficient of the air blowing rate he thermal conductance G air4 and external surface heat transfer coefficient h air4 between frame and ambient air can be evaluated as ( ) shell hair 4 = 14 1+ 0.5 v 5 air 4 air 4 air 4 1/3 (7) G = h A (8) where v is wind speed (that is linear speed of shell surface. unit: m/s); shell is external surface temperature of shell (unit:k).a air3 is surface area of shell (unit:m ) D. hermal Conductance and Lumped Parameter hermal Network For steady-state thermal analysis, the temperature rise of each node of lumped-parameter thermal network is calculated as shown in equation (9) (31), where the dimension of thermal conductance matrix G nxn is n=34. G = W (30) n (, ) (, ) (,1) (,) G i i = G i j = G i + G i + j= 1, j i (, 1 ) (, 1 ) (, ) + G i i + G i i+ + + G i n (31) Gauss elimination method, Gauss-seidel iterative method, Jacobi-cholesky elimination method, or Conjugate gradient method can be adopted to solve thermal conductance matrix G. able IV gives loss values for the heat generation applied to IPMSM. ABLE IV IPMSM PAR LOSS VALUE Parameter Value tooth losses P Fet [W] 110 Slot winding copper losses P Cu1 [W] 76 End winding copper losses P Cu [W] 141 yoke losses P Fej [W] 05 PM eddy losses P pm [W] 8 Rotor iron losses P ro [W] 41 Air friction losses P air [W] 10 Bearing losses P be [W] 4 Extra losses P ex [W] 13 Losses values at 5kW power and speed of 3300rpm Compute the coefficients of heat conductance and heat transfer Compute node heat resistance Generate heat conductance matrix No Fig. 5. Nodal temperature rise calculation flow chart. Satisfy temperature rise condition? Yes Save results Nodal temperature rise data at rated case are obtained in table V, which follow the flow chart of nodal temperature rise calculation using lumped-parameter thermal network model is shown in Fig.5. It is very important to figure out the heat generation, heat conductance, heat convection, and heat dissipation. herefore, the temperature node distribution and the equivalent lumped-parameter thermal network model in Fig.6 and Fig.7 are established in the thermal analysis. he power losses defined in able IV are injected into the specified thermal nodes of the parts. In the thermal model, the geometry of the IPMSM is divided into the following parts: 1), ) internal air, 3), 4) shell, 6) stator yoke, 8), 1) end-winding, 10) slotting winding, 14) stator teeth, 16) gap air, 18) rotor shoe, 1) magnet, 4) rotor yoke, 7), 31) bearing, 9) shaft, 33), 34) end cap. Fig. 6. emperature node distribution of IPMSM. Fig. 7. Lumped-parameter thermal network model. o validate the above-mentioned methods, a liquid-solid End

CHEN et al.: LUMPED-PARAMEER HERMAL NEWORK MODEL AND EXPERIMENAL RESEARCH OF INERIOR PMSM 371 FOR ELECRIC VEHICLE coupling model is built, which considers assembly clearance, equivalent insulation, gap air and internal air in Fig.8.According heat source and heat dissipation, the parts temperature distribution at rated load is reasonable as shown in Fig.9. ABLE V NONE EMPERAURE RISE DAA Part Node Δ Part Node Δ 8 5.5 3 4.5 Rotor 9 50.3 4 44. yoke 10 50.7 5 4.3 winding 11 50. 0 43.5 1 5.3 PM 1 44.6 yoke tooth Rotor shoe Δ (Unit: K) Shell Assembly clearance 5 36.1 43.8 6 36.5 6(30) 18.6 7 36. Bearing 7(31) 0. 13 41. 8(3) 18.4 14 4.5 End Cap 33(34) 7.5 15 41.4 Shaft 6 10.6 17 44. 3 3.4 Shell 18 46.6 4 3.6 19 44.7 Convective wall Magnet Rotor Rotor iron thermocouple Fig. 10. Rotor and stator of PMSM. Oscilloscope Load motor controller Power analyzer Fig. 11. Dynamometer platform. iron Electric parameter sensor box Motor controller Current clamp PMSM winding Constant temperature cooling tank Load motor Upper computer console iron core coil Magnet Rotor iron core Symmetric wall Fig. 8. Liquid-solid coupling model. Periodic wall End cap Bearing Internal air Gap air Shaft Cooling-water machine DC power supply Fig. 1. Power supply and upper computer console. Measured phase current wave at rated load case is obtained by three current clamps in Fig.13. Fig. 9. emperature distribution of PMSM. III. PROOYPE ESING In order to verify the lumped-parameter thermal network model and liquid-solid coupling model of IPMSM, A load testing platform is established. Rotor and stator of IPMSM are shown in Fig.10. Dynamometer system platform mainly include DC power supply, load motor controller, load motor, electric parameter sensor box, power analyzer, AC current clamp, oscilloscope, upper computer console, IPMSM, controller, constant temperature cooling tank and cooling-water machine as shown in Fig.11-Fig.1. Fig. 13. Phase current wave at rated load. Hall current sensors and voltage sample probe are mounted on the bus bar. hree voltage probes are used to gather the phase voltage. In addition, another two voltage probes are used to gather the bus voltage. wo hall current sensor are located on the 3 phase AC bus bar in order to gather the phase current data. At the same time, a hall current sensor located on the DC bus bar is used for obtaining the DC bus current in Fig.14 (L). For the aspect of output power, a torque-speed data acquisition analyzer gathers data of torque and speed, which transmit data to the power analyzer in Fig.14 (R). Finally, IPMSM efficiency

96.18 37 CES RANSACIONS ON ELECRICAL MACHINES AND SYSEMS, VOL. 1, NO. 3, DECEMBER 017 map in both the constant torque and the constant power operating regions is measured as shown in Fig.17. 3 Pin = 3 1e Uline _ rmsipha _ rms cosϕ U 3 AB _ rms + UBC _ rms + UCA_ rms IA _ rms + I (3) B _ rms = 3 1e cosϕ 3 n Pout Pout Pout = ; η = = (33) 9549 Pin Pout + Ploss where P in, P out are the input power and output power of IPMSM respectively; U line_rms is the root-mean-square value of line voltage; I pha_rms is the root-mean-square value of phase current; cosφ is the power factor; is torque; n is speed (rpm);η is the efficiency; P loss is the sum of all the losses. Voltage probe 3ph AC busbar Hall current sensor DC busbar Power analyzer IPMSM Fig. 14. Electric parameter sensor box and power analyzer. No-load back electromotive force (EMF) at 3500rpm is shown in Fig.15.Furthermore, line voltage at load case 3500rpm, 10N.m is obtained in Fig.16. he efficiency of IPMSM in the region of low speed and high torque is relatively low. Because the motor armature current is large, and the in-duced voltage is obviously lower than stator voltage, which leads to the small DC bus voltage utilization ratio and the low power factor. he output power of the motor compared with the rated operation condition is still relatively low. However, winding copper loss is larger with re-spect to the rated operation condition in the proportion of the input power. herefore, effi-ciency is still low. he efficiency is relatively low in the constant power region, especially at high speed and low torque. On the one hand, high-frequency skin effect in armature winding leads to increase of the phase winding resistance, which consequently result in increase of the high frequency copper loss. At the same time, a large percentage of stator current is used to produce negative d-axis demagnetization current, copper loss also increases significantly in the weak magnetic area. On the other hand, the hysteresis loss, eddy cur-rent loss in the stator core and eddy current loss in the rotor core and PM are proportional to the square of the frequency. In addition, extra loss is proportional to the 1.5th power frequency as described in equation (3) and (11), which result in loss of stator core and rotor losses increase obviously in the proportion of total input power, herefore, efficiency in the region of the constant power relative to rated condition fallen remarkably. orque(n.m) 75 70 60 50 40 30 81.73 79.3 86 55 76.91 60.04 55. 50.40 45.59 40.77 35.95 6.31 64.86 31.13 69.68 74.50 57.63 5.81 47.99 43.18 38.36 33.54 67.7 6.45 8.7 7.09 3.90 81.73 79.3 96.18 76.91 5.81 47.99 43.18 6.45 57.63 38.36 33.54 8.7 67.7 7.09 55. 50.40 45.59 35.95 60.04 40.77 31.13 6.31 64.86 69.68 74.50 3.90 96.18 96.18 81.73 79.3 5kW rated load efficiency map 76.91 57.63 38.36 5.81 47.99 43.18 67.7 6.45 33.54 8.7 81.73 55. 50.40 45.59 35.95 60.04 40.77 6.31 31.13 64.86 74.50 69.68 7.09 3.90 79.3 76.91 57.63 5.81 47.99 43.18 38.36 33.54 67.7 6.45 8.7 3.90 55. 50.40 45.59 35.95 40.77 60.04 31.13 6.31 64.86 90 80 70 60 50 Fig. 15. No-load emf wave at 3500rpm. 0 10 84 14 4 00 7 57 67 6 5 47 43 38 33 809 76 63 7 45 81 99 18 36 54 791 3 90 81 73 69 64 60 55 50 45 40 35 31 6 7468 86 04 40 59 77 95 13 31 50 96.18 81.73 79.3 3 90 74 50 7 09 76.91 6 60 55 50 40 35 31 645 04 40 59 77 95 13 31 57 5 47 43 38 33 8 63 81 99 18 36 54 7 64 86 7.09 5 81 67.7 64.86 6.45 57 60.04 63 67 7 1000 000 3000 4000 5000 6000 7000 8000 Speed(rpm) 69 68 Fig. 17. Efficiency map at rated load case. he temperature measurements of IPMSM are carried out using thermocouple PC100 and an FLUKE infrared thermal camera imager. hermocouple temperature sensors PC100 are located into each phase slot windings for measuring winding temperature and iron core temperature. he temperature measurement of outer surface including PMSM and its controller is measured by using a FLUKE infrared thermal camera imager in Fig.18-Fig.19. 81.73 79.3 76.91 74.50 69.68 55 0 40 40 30 Fig. 16. Line voltage wave at 3500rpm, 10N.m. he efficiency of IPMSM at the rated speed of 3300rpm, rated torque 75Nm can reach 96%.he region area of efficiency greater than 80% is about 85%, which meets the perfor-mance requirements of the IPMSM. Fig. 18. PMSM infrared thermal image at =76N.m, n=3300rpm, Ia=130.5A.

CHEN et al.: LUMPED-PARAMEER HERMAL NEWORK MODEL AND EXPERIMENAL RESEARCH OF INERIOR PMSM 373 FOR ELECRIC VEHICLE Fig. 19. Motor controller infrared thermal image at =76N.m, n=3300rpm, Ia=130.5A. Similarly, efficiency map and IPMSM phase current at overload case are obtained in Fig.0-Fig.1. Infrared thermal image of IPMSM and its controller at overload case are shown in Fig.-Fig.3. orque(n.m) 180 160 140 10 100 80 60 40 0 65.77 70.05 74.34 7.0 76.48 78.6 80.76 8.91 74.34 76.48 50.78 80.76 5.9 33.64 55.06 35.78 31.50 78.6 93.6 8.91 8 91 93.6 37.93 40.07 59.34 57.0 61.49 50.78 5.9 35.78 35.78 37.93 40.07 44.35 70.05 46.49 48.64 50.78 5.9 55.06 59.34 61.49 65.77 63.63 67.91 7.0 78.6 80.76 35.78 37.93 40.07 44.35 46.49 48.64 78.6 80.76 50.78 5.9 55.06 59.34 61.49 70.05 65.77 63.63 67.91 44.35 70.05 65.77 63.63 67.91 46.49 48.64 76.48 78.6 80.76 4.1 7.0 74.34 93.6 8.91 93.6 33.64 31.50 57.0 80.76 4.1 8.91 78.6 7.0 93.6 7.0 74.34 76.48 8.91 76.48 74.34 67.91 65.77 70.05 93.6 33.64 31.50 61.49 59.34 57.0 57.0 4.1 80.76 55.065.9 8.91 74.34 76.48 78.6 76.48 8.91 74.34 63.63 33.64 31.50 55.06 7.0 67.91 65.77 70.05 48.64 44.35 3 00 1000 000 3000 4000 5000 6000 7000 8000 Speed(rpm) Fig. 0. Efficiency map at overload case. Fig. 1. Phase current wave at overload case. 5kW overload efficiency map 46.49 90 80 70 60 50 40 he results of the lumped-parameter thermal network method, fluid-solid coupling CFD method and experimental validation method are listed in table V. Due to using many empirical formulas in computing the coefficient of heat conductance, heat convection and heat source. he methods of the lumped-parameter thermal network and the fluid-solid coupling CFD need to amend by experiment validation. By comparing results among three methods in table VI, the methods of the lumped-parameter thermal network and the fluid-solid coupling CFD can better predict the parts temperature of the IPMSM. ABLE VI CALCULAED AND MEASURED EMPERAURE A RAED LOAD Parameter Method A Method B Method C Slot winding 74 76 75 iron 60 6 64 Rotor iron 66 68 69 PM 66 67 70 Bearing 4 38 41 Shell 7 9 5 End cap 31 33 30 Shaft extension 34 36 3 A: Lumped-parameter thermal network; B: fluid-solid coupling CFD; C: Measured (5kW, 3300rpm, 130.5A) [ C IV. CONCLUSION In this paper, convective heat transfer coefficients of fluid can be accurately computed using CFD simulation, which can be used to amend the empirical formula. herefore, a simplified model considering the symmetry and period is built to reduce the computation time obviously. Assembly clearance between shell and stator iron core is taken into account. By comparing lumped-parameter thermal network model with 3-D liquid-solid coupling FEA model, lumped-parameter thermal network method is more effective and accurate than traditional thermal network. Meanwhile, temperature rise testing platform including prototype manufacture is built to validate the above two methods using FLUKE infrared thermal imager and thermocouple PC100. Efficiency map, no-load EMF, line voltage and phase current wave are tested on dynamometer. he agreement between simulation and experimental results shows that proposed lumped- parameter thermal network model is quite convincible. he method will be important for practical engineering application. Fig.. Infrared thermal image of PMSM at =144N.m, n=3300rpm,i a=51a. Fig. 3. Infrared thermal image of motor controller at =144N.m, n=3300rpm,i a=51a. REFERENCES [1] A. Boglietti, A. Cavagnino, D. Staton, M. Shanel, M. Mueller, and C. Mejuto, Evolution and modern approaches for thermal analysis of electrical machines, IEEE rans. Ind. Electron., vol. 56, no. 3, pp. 871 88, Mar. 009 [] P. D. Mellor, D. Roberts, and D. R. urner, Lumped parameter thermal model for electrical machines of EFC design, Proc. Inst. Electr. Eng., B, vol. 138, no. 5, pp. 05 18, Sep. 1991. [3] B. H. Lee, K. S. Kim, J. W. Jung, emperature Estimation of IPMSM Using hermal Equivalent Circuit, IEEE rans. Magn., vol. 48, no. 11, Nov. 01 [4] J. Nerg, M. Rilla, and J. Pyrhönen, hermal analysis of radial-flux electrical machines with a high power density, IEEE rans. Ind. Electron., vol. 55, no. 10, pp. 3543 3554, Oct. 008. [5] C. Jungreuthmayer,. Bauml, O. Winter, M. Ganchev, H. Kapeller,A. Haumer, and C. Kral, A detailed heat and fluid flow analysis of an internal permanent magnet synchronous machine by means of

374 CES RANSACIONS ON ELECRICAL MACHINES AND SYSEMS, VOL. 1, NO. 3, DECEMBER 017 computational fluid dynamics, IEEE rans. Ind. Electron., vol. 59, no. 1,pp. 4568 4578, Dec. 01. [6] A. M. EL-Refaie, N. C. Harris,. M. Jahns, and K. M. Rahman, hermal analysis of multibarrier interior PM synchronous machine using lumped parameter model, IEEE rans. Energy Convers., vol. 19, no.,pp. 303 309, Jun. 004. [7] S.. Scowby, R.. Dobson, and M. J. Kamper, hermal modeling of an axial-flux permanent magnet machine, Appl. hermal Eng., vol.4, pp. 193 07, 004. [8] N. Rostami, M.R. Feyzi, J. Pyrhönen, A. Parviainen, and M. Niemela, Lumped-Parameter hermal Model for Axial Flux Permanent Magnet Machines, IEEE rans. Magn., vol. 49, no. 3, pp.1178 1184, Mar 013 [9] D. A. Staton and A. Cavagnino, Convection heat transfer and flow calculations suitable for electric machines thermal models, IEEE rans.ind. Electron., vol. 55, no. 10, pp. 359 3516, Oct. 008. [10] D. A. Howey, P. R. N. Childs, and A. S. Holmes, Air-gap convection in rotating electrical machines, IEEE rans. Ind. Electron., vol. 59, no. 3,pp. 1367 1375, Mar. 01. [11] F. Marignetti,, V. D. Colli, and Y. Coia, Design of Axial Flux PM Synchronous Machines hrough 3-D Coupled Electromagnetic hermal and Fluid-Dynamical Finite-Element Analysis, IEEE rans. Ind. Electron., vol. 55,no. 10, pp. 3591 3601, Oct. 008. [1] F. Marignetti and V. D.Colli, hermal analysis of an axial flux permanent-magnet synchronous machine, IEEE rans. Magn., vol. 45, no. 7, Jul. 009 [13] G. Zhang, H.Wei, M. Cheng, B. F. Zhang, and X.B. Guo, Coupled Magnetic-hermal Fields Analysis of Water Cooling Flux-Switching Permanent Magnet Motors by an Axially Segmented Model, IEEE rans. Magn., vol. 53, no. 6, 810654, Jun 017. [14] H. Vansompel, A. Rasekh, A. Hemeida, J. Vierendeels, and P. Sergeant, Coupled Electromagnetic and hermal Analysis of an Axial Flux PM Machine, IEEE rans. Magn., vol. 51, no.11, 8108104, Nov 017. [15] Lin.D, Zhou.P, Fu.W.N, Badics.Z and Cendes Z.J. A Dynamic Core Loss Model for Soft Ferromagnetic and Power Ferrite Materials in ransient Finite Element Analysis. IEEE rans. Magn., vol. 40, no., pp: 1318-131, Mar 004. [16] Shigematsu K,Oyama J,Higuchi,et al. he study of eddy current in rotor and circuit coupling analysis for small size and ultrahigh speed motor[c]// Wang Zhaoan. Proceedings of 4th International Power Electronics and Motion Control Conference. Xi'an: Xi'an Jiaotong University Press,004,1: 75-79 [17] G. Dajaku1, D, Gerling, Eddy current loss minimization in rotor magnets of PM machines using high-efficiency 1-teeth/10-slots winding topology, International Conference on Electrical Machines and Systems (ICEMS 011), 0-3 Aug. 011, Beijing, China. Zhongyue Zou He received the B.S. degree from Shandong University of echnology and the M.S. degree from Shandong University of Science and echnology. He is currently working toward the Ph.D. degree in mechanical engineering from Xi an Jiaotong University, Xi an,china. His research interests include energy management and energy management. Binggang Cao He received the Ph.D. degree from Xi an Jiaotong University, Xi an, China. He is a Professor and the Director of the Institute for Electric Vehicle and System Control with Xi an Jiaotong University, where he was previously the Dean of the School of Mechanical Engineering and the Director of the Research Center for Electric Vehicles. His research interests include system optimization, mechatronic systems, electric vehicles, and electric drive systems. Qixu Chen was born in Shandong, China in 198. He received bachelor s degree in mechatronic engineering from North University of China, aiyuan, China, in 007, and master s degree in mechatronic engineering from Xidian University, Xi an, China, in 010. He is currently working toward a doctor's degree in electric engineering from Xi'an Jiaotong University, Xi an, Shaanxi, China, His research interests include AFPMSM design and drive of the electric vehicle.