2010 ANSYS, Inc. All rights reserved. 11 ANSYS, Inc. Proprietary

Transcription

1 Integrated Simulation Environment for Electric Machine Design Scott Stanton Technical Director Advanced Technology Initiatives ANSYS Inc ANSYS, Inc. All rights reserved. 11 ANSYS, Inc. Proprietary

2 Design Automation Example: RMxprt - Maxwell

3 UDP: User Defined Primitives 2009 ANSYS, Inc. All rights reserved. ANSYS, Inc. Proprietary

4 External Circuit Coupling: Maxwell - Simplorer Thevenin equivalent (impedance matrix, source voltages) 2D/3D FEA Lumped field parameters (inductances, induced Internal voltages) System Simulator

5 System/Circuit - FEA Coupling: Simplorer - Maxwell Differentiating Feature Example: Axial-Disk PM Motor Control Current Control Loop Position Control Loop

6 External Circuit Coupling Differentiating Feature 3ph Line Current Profile Flux Linkage Profile Axial Motor Speed & Torque Profile Magnetic Flux Density

7 Nonlinear Lamination and Nonlinear Anisotropy Differentiating Feature Nonlinear lamination is extensively used in low frequency electromagnetic devices for significant reduction of eddy current loss. Nonlinear anisotropy is widely used in magnetic recording, power transformers and large size electrical machines. Oriented electrical steels have high induction but with much lower core loss in the rolling direction.

8 Nonlinear Lamination and Nonlinear Anisotropy Differentiating Feature Example: Reluctance Motor Application Y X Rotor lamination is defined along r direction in the local cylindrical coordinate system The rotor local coordinate system is attached to the moving rotor

9 PM Characteristic to 2 nd & 3rd Quadrant Differentiating Feature Expand the existing algorithm to the 3rd quadrant for demag computation Base on the actual user-input B-H curve in the 3rd quadrant Load line without other sources Load line with other sources B Hc after demagnetization Initial Br Br after demag 0 H Demagnetization point

10 Generator Fault Example 550 W PM generator 4 Pole 3 Phase, 50HZ AC Ceramic 8D PM Rated Speed, Open-Circuit to Short-Circuit Fault

11 Magnet 2 nd quadrant demagnetization (demag) Spatially dependent demag due to fault Initial Radial Magnetization

12 Short-Circuit Analysis Short circuit at 15.2ms: Phase A peak Ansoft LLC Volts [V] D_EMF_save_demag Maxwell3DDesign2 ANSOFT Curve Info InducedVoltage(PhaseA) Setup1 : Transient InducedVoltage(PhaseB) Setup1 : Transient InducedVoltage(PhaseC) Setup1 : Transient Time [ms] Ansoft LLC B [T] BH_Data_Points_Initial_Demag Material BH Curve Maxwell3DDesign E E E E+000 H [A/m] ANSOFT Operating Points Bus short for all phases

13 Short-Circuit Analysis Subsequent use of the magnet results in reduced performance Ansoft LLC D_EMF_demaged Maxwell3DDesign3 ANSOFT Curve Info InducedVoltage(PhaseA) Setup1 : Transient InducedVoltage(PhaseB) Setup1 : Transient InducedVoltage(PhaseC) Setup1 : Transient Ansoft LLC BH_Data_Points_Demag Maxwell3DDesign3 ANSOFT Volts [V] B [T] Time [ms] 0.05 Operating Points E E E E+000 H [A/m] Weak Back EMF Addt l short for all phases

14 Short-Circuit Analysis Leading edge is weakened significantly Ansoft LLC D_EMF_save_demag Maxwell3DDesign2 ANSOFT Original Volts [V] Time [ms] Ansoft LLC D_EMF_demaged Maxwell3DDesign3 ANSOFT Fault Volts [V] Time [ms]

15 Robust Mesher Mesh a higher percentage Higher mesh quality Effective on imported geometries Matching boundary more robust Fewer total elements Smoother and more uniform element transition Automatically healing and repair High Quality Mesh Using Minimal Settings Special Attention given to Air Gap Region

16 Induced Eddy Currents in Magnets Symmetry Boundary A n s o ft C o rp o ra t i o n N o r m a liz e d P o w e r L o s s C u r v e NIn fo os p li t _ S M N o r m a liz e d P o w e r L o s s Im p o r t e d N o r m a liz e d P o w e r L o s s End of Rotor Normalized Power Loss [W] T i m e ( m s ) [s ]

17 PM Loss Reduction Add cuts: Reduce Eddy Currents Reduce Loss Up to 32 Magnet Segments

18 Power Loss in Magnet vs. Number of Segments Normalized Loss with Carrier Harmonics Loss= n mag J σ 2 dv

19 PMSM Core Loss Calculation Core Loss is Expressed as the Sum of: Hysteresis P h Classical Eddy Current P c Excess Loss P e P Fe = Ph + Pc + Pe 2 2 P Fe= khfbm+ kc( fbm) + ke( fbm) Minimizing the Error: error( k h, k c, k e ) = m n i i= 1 j= 1 [ p vij 1.5 ( k h m: number of loss curves ni: number of points of the i-th loss curve Pvij = f(fi, Bmij): two dimensional lookup table for multi-frequency loss curves f i B 2 mij + k c f 2 i B 2 mij + k e f 1.5 i B 1.5 mij )] 2

20 Integrated EM Field and Core Loss Analysis B z Core Loss Effects on Input Power and Force/Torque z y z y x x J e d J e d B x Eddy current produced by B n Eddy current produced B t Reference: D. Lin, P. Zhou and Q. M. Chen, The Effects of Steel Lamination Core Losses on Transient Magnetic Fields Using T-Ω Method, IEEE VPPC, September 3-5, 2008, Harbin, China

21 Core Loss Ansoft Corporation Curve Curve Info Curve Info XY Info Plot 1 avg avg avg Sinusoidal Excitation, 60Hz Curve Sine plus 1kHz triangular wave, 60Hz Curve Sine plus 1kHz triangular wave, multiple CL Curves Test_tran CoreLoss [kw] Time [ms]

22 Core Loss Effect on Torque Ansoft Corporation XY Plot 53 Test_MagnetLoss Curve Info Curve Info Torque Not Including Torque Difference CL Effect Torque Including CL Effect Moving1.Torque Torque [knewtonmeter] Time (ms) [ms][s]

23 3 Phase Induction Motor Losses Core loss Rotor copper loss at no-load condition Due to flux pulsations in the air gap Difficult to measure Are present in every squirrel cage induction motor Two induction motors in the 200 kw range that only differ in the slot design Both motors have 48 rotor slots Stators have 60 and 36 slot for motors A and B respectively Reference: J.Germishuizen, S.Stanton No Load Loss and Component Separation for Induction Machines, Proceedings ICEM 2008, Paper ID 1144

24 No Load Losses Input power: 36 Stator slots Why the big difference between the two designs? Input power: 60 Stator slots Stator Winding Loss Noticeable difference between input power and copper losses with different stator slot number. Reference: J.Germishuizen, S.Stanton No Load Loss and Component Separation for Induction Machines, Proceedings ICEM 2008, Paper ID 1144

25 FEA Rotor Copper Losses Where: J z is the current density ρ r is the resistivity of the rotor bar l is the stack length n is the total number of bars A n is the cross-sectional area of a bar Reference: J.Germishuizen, S.Stanton No Load Loss and Component Separation for Induction Machines, Proceedings ICEM 2008, Paper ID 1144

26 Measurement vs. Simulation Maxwell Transient Results 60 Slots 36 Slots U LL V I s A P Cu kw s P Cu kw r P Fe kw Measured Results Motor with its cage removed. The no load loss without the cage requires a special measurement setup to rotate the rotor with the same speed as the stator rotating field. 36 slot with cage 36 slot without cage 60 slot with cage Maxwell Results Reference: J.Germishuizen, S.Stanton No Load Loss and Component Separation for Induction Machines, Proceedings ICEM 2008, Paper ID 1144

27 Manufacturing Factor for Iron Loss Iron core losses are calculated based on k h, k c and k e using near perfect samples of laminated materials. Other factors that contribute to core loss include: Lamination Punching, Stress and Heat Other Losses such as intra-lamination currents 60 Slots 36 Slots 60 Slots 36 Slots Reference: J.Germishuizen, S.Stanton No Load Loss and Component Separation for Induction Machines, Proceedings ICEM 2008, Paper ID 1144

28 Workflow : Coupled Electromagnetic and Thermal Analysis for Electric Machines ANSYS Mechanical/ CFD Model Temperature Geometry Losses Maxwell Model Mapped Losses

29 PMDC Motor

30

31 Integrated Design Flow Torque Current Voltage Customer Requirements Scale N, L Create Initial Design. Map of solution domain T, ψ d, ψ q = f(i d, i q, Θ ) using Static Solver Scale N, L Integrating FEM in an everyday design environment to accurately calculate the performance of IPM motors, J.Germishuizen, S.Stanton and V. Delafosse ISEF XIV International Symposium on Electromagnetic Fields in Mechatronics, Electrical and Electronic Engineering Arras, France, September 10-12, 2009

32 Integrated Solution: Verify with Transient Torque Current Voltage Customer Requirements Scale N, L Create Initial Design. Map of solution domain T, ψ d, ψ q = f(i d, i q, Θ ) using Static Solver Scale N, L Integrating FEM in an everyday design environment to accurately calculate the performance of IPM motors, J.Germishuizen, S.Stanton and V. Delafosse ISEF XIV International Symposium on Electromagnetic Fields in Mechatronics, Electrical and Electronic Engineering Arras, France, September 10-12, 2009

33 Cogging Torque Optimization Motivation: Primary Contributor to: Torque Ripple Mechanical Vibration Acoustic Noise Drive System Instability Thus: Lower Efficiency Goals: Reduce Cogging Torque Maintain Machine Performance

34 Method Ansoft Corporation 3.00 Torque Maxwell2DDesign1 PM Shape Modification Pole Embrace Pole Arc Offset Moving1.Torque [NewtonMeter] Curve Info Ansoft Corporation Moving1.Torque XY Plot 2 PMSM_CT Setup1 : Transient Curve Info Bradial Setup1 : Transient Time='0ns' Magnet Thickness Transient FEA Time [s] Norm alizeddistance Genetic Optimization Algorithm Minimize Peak Cogging Torque Maintain Average Air Gap Flux Density Bradial

35 Nominal Setup PM Geometric Parameters Pole Embrace = 0.75 Pole Arc Offset = 0 mm Magnet Thickness = 3.5 mm Calculations Torque: Virtual Work Method Average Radial B in Air Gap: B rad = B x *cos(ϕ) + B y *sin(ϕ)

36 Nominal Solution Ansoft Corporation 1.20 XY Plot 2 PMSM_CT Curve Info Ansoft Corporation 3.00 Torque Maxwell2DDesign1 Bradial 1.00 Setup1 : Transient Time='0ns' N-m Curve Info Moving1.Torque Setup1 : Transient Bradial Avg = 0.76 T Moving1.Torque [NewtonMeter] Norm alizeddistance Time [s] Magnetic Field Eqn. Virtual Work Method T dw( Θ, i) dθ = i= const = d ( dθ ( B v 0 H dh) dv)

37 Goals Cogging Torque Peak Nominal Value = 2.2 N-m Optimal Goal = 0.2 N-m G1 = 1 + (max(abs(torque)) 0.2) * 9 / 5.3 When Cogging Torque = 0.2 N-m, then G1 = 1.0 Nominal Bavg Value = 0.76 Tesla Optimal Goal = 0.76 Tesla G2 = 1 + (Brad_Avg 0.5) * 9 / 0.31 When Air Gap Flux Density = 0.76 T, then G2 = 8.55 Magnet Area Minimize G3 = 1 + (Mag_area 220) * 9 / 290 When Magnet Area = 220 mm 2, then G3 = 1.0

38 Results A nsoft Corporation C og g in g Torq u e PMSM_CT_V erify Curve Info Optimized Design Setup1 : Transient Mov ing1.torque Imported Nominal Des ign Y1 [NewtonMeter] Nominal (0.76 T) --- Optimized (0.73 T) Ansoft Corporation 1.20 Air Gap Flux Density PMSM_CT_Verify Curve Info Bradial Setup1 : Transient Time='0ns' Tim e [s] M X 1 : M X 2 : Bradial Imported Nominal Design --- Nominal (2.2 N-m peak) --- Optimized (0.4 N-m peak) Bradial NormalizedDistance

39 Summary