Development of Active Magnetic Bearings with Sin-cos Winding for Application to Medium Voltage Machines

（Calnetix Technologies）

Abstract：Calnetix has recently completed the design and development of the next generation medium voltage,megawatt high speed motor drive system using 10 kVSilicon Carbide power devices. A1.6MW medium voltage motor supported by active magnetic bearings has been designed as a test platform for the system.The heteropolar radial magnetic bearings feature a winding topology with separate bias,cos(x-control)and sin(y-control)coils on each pole.This approach is shown to produce a more uniform load capacity versus angle than the conventional bi-quadrant winding scheme.This paper describes the design and motivation for the machine with primary emphasis on the magnetic bearing system.Preliminary bench test results of the actuator are described and compared to predictions.Future applications for this type machine in wastewater recovery, exhaust gas remediation, and natural gas compression are discussed.

Keywords：Magnetic Bearing，High Speed Motor

0 Introduction

There is an increasing demand on energy efficiency in the large machine industry,considering that electric motors use almost 40% of all electrical energy in the US[1].Variable speed drives(VSD)have met some of these goals in the last few decades,but the US Department of Energy(DOE)recently identified performance targets for the next significant step in electric drive energy efficiency that will require newer technologies than those currently in use.The challenge is to improve the efficiency with a combination of new technologies-high speed direct drive machine,high efficiency power electronics[2]and an enabling active magnetic bearing system.

Medium Voltage(MV)high power motor drives are commonly used in natural gas production and distribution,wastewater recovery,offshore platforms,mines,and metal processing plants,as they operate at higher supply voltages to obtain lower losses and use smaller cables to improve overall drive efficiency.This results in lower system cost,particularly when the VFD and motor are separated by a long distance due to environmental and application requirements.Existing MV drives rely on Si based devices as they are a widely available and mature technology.However,high switching losses of Si devices,such as insulated gate bipolar transistors(IGBT),prevent their use in very high-speed motor drive applications,which require higher switching frequencies.

Power switching devices using SiC technology have recently been used in some high-performance motor drives due to their high breakdown voltage,high operating temperature,and high switching frequency,leading to designs with high efficiency and power density[3].However,there are many technical challenges to overcome when designing MV drive systems using high voltage SiC devices.Additionally,the cost of a SiC MOSFET device is considerably higher than its Si IGBT counterpart[4].This led to the DOE funding development of SiC based MV drives and compatible MV motors to demonstrate the technology and to encourage adoption of the technology to drive down costs.

The key metrics defined by the DOE for the next generation direct drive MV MW system are shown in Table 1.The VSD requirements were defined based on the state of the art in wide band gap(WBG)power devices and high-speed powerdense drives.

Tab.1 Estimated target performance metrics of WBG based direct drive system at the 1 MW level

The nearest term technical approach to achieve these goals requires:1)significant improvement to the electromagnetic power density of the high speed motor,2)use of wide band gap(WBG)Silicon Carbide(SiC)devices in the VSD,and the use of active magnetic bearing systems that naturally would eliminate conventional oil-fed bearings and associated auxiliary systems altogether.Another motivation for using magnetic bearings is their non-contact operation will mitigate the effects of common mode voltages and currents due to the high dv/dt(rate of change of voltage)of the drive[5].

This paper briefly describes the new SiC based MV drive developed for this program.This is followed by a more detailed description of the demonstration motor with a focus on the magnetic bearing system.The heteropolar magnetic bearings were designed with separate bias coils and a sincos winding scheme for the orthogonal radial control coils.This Bias-Sin-Cos winding scheme results in a more uniform load capacity versus load angle compared to the bi-quadrant winding scheme typically used in heteropolar radial magnetic bearings.Using separate bias coils results in control coils with fewer turns and lower inductance compared to the typical design where the bias and control flux are generated by the same coil.

1 Vsd Design

The VSD developed for the program uses a 3 phase,4 160VAC input to deliver up to 1MW power at 500Hz.Additional basic specifications for the VSD are given in Table 2.The inverter uses a passive-active configuration in a proven 2-level topology with 10kV SiC WBG devices for the motor side inverter(Figure 1).The VSD design effort was focused on the development of the SiC inverter technology while maintaining a low risk on all other components using past experience in the industry[3].The design considerations and early development of this drive were discussed in[4].The grid side passive rectifier design uses a standard 12 pulse transformer input with filtering to meet the grid side THD requirements.Suitable isolation between the grid and the drive ground paths is essential to the interference-free operation of the system and this has been taken into consideration in the design of the ground paths which are critical.The passive rectifiers deliver 6 500V into the DC bus.This configuration is standard in all MV drives of the present technology,field proven,and warranted no change to this development.

The key considerations driving the design were:

1）Developing the motor inverter power platform with high density 10kV SiC modules,thermal management,bus interface design and interfacing the new gate driver into the existing controls

2）Optimization of the motor side filtering to meet the insulation life requirements while taking maximum advantage of the high switching capability of the SiC devices

3）Optimization for meeting the grid side power quality requirements,keeping the total harmonic distortion of current into the grid below 7% while addressing cost and the use of existing technology

4）Using field proven control electronics architecture and firmware that are pre-qualified and certified by agencies like UL and CE in the low voltage drives to de-risk the control problem.

Tab.2 MV VSD Specification

Fig.1 1 MW 4 160 V 2-level VSD Module

The motor drive has currently been tested up to 4 800 VDC,70 A peak current into an inductive load to evaluate its performance.Figure 2 shows 14.3kHz switching waveforms for this case.Figure 2a shows:1)gate drive voltage,2)high side drain-to-source voltage (1/2 scale),3)low side drain-to-source voltage(1/2)scale,and 4)load current.Figure 2b uses a smaller time scale to show the turnon switching waveform.The delta voltage is 2.36*2 kv,the switching time is 220 nsec giving a dv/dt of 21.5 kv/µsec.The performance is normal and as expected.

Fig.2 Load test 4 800V,70Arms inductive load at 14.3 kHz,with Ch1:Gate drive voltage,Ch2:high side drain-to-source voltage(1/2 scale),Ch3:low side drain-to-source voltage(1/2 scale),Ch4:Load current amps

2 Machine Design

A cross-section of the 1.6 MW,4 160 V demonstration machine is shown in Figure 3.The motor will be tested to 1.0 MW with the SiC based MV MW VSD,then later used in a higher power application with two paralleled VSDs.The motor is supported by a five-axis magnetic bearing system.There is a two-axis,heteropolar radial magnetic bearing on each end of the machine and a magnetic thrust bearing on the drive end.Calnetix standard reluctance,radial position sensors are located outboard of the radial bearings.There is a constant-flux edge sensor[6]incorporated into the radial sensor housing of the drive(thrust)end of the machine.An edge sensor detects rotor axial displacement from a radial surface which makes the overall machine design simpler and easier to assembly.The constant flux feature reduces sensor sensitivity to leakage fields.The backup(touchdown)bearings are supported in resilient mounts and are located outboard of the position sensors on both ends of the machine.The design of the backup bearing system is very similar to the one used in[7].The output shaft is designed to accept a double flexible disk coupling to transmit torque to a driven machine or dynamometer for testing.Magnetic bearings were selected for this machine for oil free operation and low losses.Additionally,the non-contact operation of the magnetic bearings will mitigate effects of common mode voltages from the high dv/dt of the drive[5].

Fig.3 1.6 MW 15 000 RPM 4 160V High Speed PMSM on magnetic bearings.

2.1 Motor Rotor

The motor is a 4-pole surface permanent magnet(SPM)design with a rated speed of 15 000r/min(500Hz electrical frequency).The rotor,Figure 4,comes from a family of proven SPM technologies that has been developed,tested and proven in the field using proprietary composite retention technologies over standard 32MGOe Samarium Cobalt magnets.While the composite retention sleeve provides for a very efficient electromagnetic design,the harmonic losses concentrated in the magnets poses unique problems for high speed machines due to the poor thermal conductivity of composites.Additional enhanced technologies to minimize the rotor harmonic and eddy current losses were developed to mitigate the rotor thermal management challenge.The result was a compact high-speed rotor that also has a high level of rotordynamic stiffness to enable sub-critical operation of the system.

2.2 Motor Stator

One of the main challenges faced in the stator design was the high dv/dt caused by the high switching capability of the SiC devices at medium voltage.SiC devices have a dv/dt of 20kV/µsec as opposed to the fast switching IGBTs that currently can switch about 600V/µsec.Existing insulation technologies are designed to meet the present generation of IGBT switching devices with some filtering.Hitherto,no medium voltage systems have been designed with the full switching capability of the SiC devices.The medium voltage insulation system for the demonstration machine was specially designed to withstand the higher dv/dt.The insulation system included supplemental turn insulation,a multi-layer overwrap of Mica tape and a final overlap of Armor tape.Additionally,the motor side filters and reactors were optimized to soften the leading edge of the switching transition such that desired motor life and operating thermal class for the stator could be achieved.

Fig4.Picture of the rotor for the demonstration machine

3 Magnetic Bearing Design

The motor is supported by a five-axis magnetic bearing system.There is a two-axis,heteropolar radial magnetic bearing on each end of the machine and a magnetic thrust bearing on the drive end.Key design characteristics of the bearings are given in Table 3.The listed load capacities are for the linear portion of the force vs.current curve.The peak load capacities are approximately 20% higher as noted in a following section.

Tab.3 Basic magnetic bearing minimal design parameters

4 Design and Analysis of the Radial Actuator

The radial AMBs used in this machine are electromagnetic(EM)biased 8-pole heteropolar actuators with a Bias-Sin-Cos winding configuration;the structure and outline dimensions are in Figure 5.

Fig.5 Structure and outline dimensions(mm)of the EM biased 8-pole heteropolar radial actuator with Bias-Sin-Cos winding configuration.

Except for the Bias-Sin-Cos winding configuration,this is the most common type of radial actuator used in industrial AMBs.The Bias-Sin-Cos winding configuration,while known for years,to the author’s knowledge has not been used in industrial applications,whereas the simpler bi-quadrant scheme is the most topology.However,use of the Bias-Sin-Cos winding configuration may often bring enough benefits to offset additional complexity.To understand the tradeoffs of the Bias-Sin-Cos winding configuration,it is helpful to first review the conventional bi-quadrant winding configuration.

Heteropolar bi-quadrant winding.The construction and operation of a conventional 8-pole heteropolar radial actuator using the bi-quadrant winding configuration are illustrated in Figure 6.In the most typical implementation,each pole is equipped with a single coil,with the following pole coils connected in series in order to produce flux circulation between the corresponding poles:0 and 7,1 and 2,3 and 4,5 and 6.Each of these pole pairs is then energized with a unipolar transconductance amplifier.In the absence of external dynamic loading,the amplifiers produce only a steady cur-rent in the control coils.In Figure 6,the resulting bias fluxes are shown with blue arrows and the resulting bias polarities of the poles follow the sequence S-N-S-N-S-N-S-N going from pole 0 to pole 7.Another commonly used sequence is S-S-N-N-S-S-N-N[8].When a force must be produced,for example in the positive X-direction as shown in Figure 6,the current in the coil pair 0-7 is increased,whereas the current in the pole pair 3-4 is decreased,resulting in a stronger magnetic pull towards pole pair 0-7.Functionally,the current flowing through each coil is the sum of two components:(1)a steady bias current which produces a bias field and(2)a steady control current which produces its own magnetic field.The control currents in diametrically opposite coils have equal magnitudes and opposite signs,thus causing the control field to subtract from the bias field in one coil and add to the bias field in the opposite coil.With this configuration,a net force is produced along the axis of the coils.For example in Figure 6 the control flux adds to the bias flux in the pole pair 0-7,but subtracts from the bias flux in the pole pair 3-4,resulting in a stronger magnetic field in poles 0 and 7 and a net magnetic force pulling the rotor towards these poles.

Fig.6 Bi-quadrant control of a radial heteropolar 8-pole actuator

In a similar way,a force can be produced in theY-direction by utilizing pole pairs 1-2 and 5-6 or a force in an arbitrary direction can be generated as a superposition ofXandYforces.

Heteropolar bi-quadrant winding with bias coils.In another implementation of the bi-quadrant control,each pole coil is split into two:bias coil and control coil.As the names suggest,the bias coil is used to produce the bias field and the control coil is used to produce the control field.All the bias coils then can be connected in series and energized with a dedicated bias amplifier,whereas every two pairs of coils associated with diametrically opposite pole pair(for example 0-7 and 3-4)can be also connected in series so that the control field they produce would add to the bias field in one pole pair(0-7 in Figure 6)and subtract from it in the other(3-4).The main advantages of this approach are:1)a smaller control coil with less inductance can be used which yields a higher slew rate capability,2)fewer amplifiers are required(3 instead of 4)and fewer wires must exit the actuator(3 pairs instead of 4).The control coil amplifiers must be bi-polar(PWM with H-bridge configuration is common)as they need to supply both positive and negative current.

Heteropolar bias-sin-cos winding.One of the disadvantages of the bi-quadrant scheme is that its load capacity varies depending on the load angle with the flux-carrying capacities of the actuator poles being underutilized when loads are applied in any direction but in between theXandYaxes.For example,poles 1 and 6 in Figure 6 also could contribute to generation of the force Fxin the x-direction,but they are not used.The most complete pole utilization can be achieved if each pole coil is controlled independently,or each pair of diametrically opposite poles is controlled independently in a“bias/control,two-coil-per-pole”configuration.This,however,would require either 8 amplifiers(in the first case)or 5 amplifiers(in the second case)and the matching number of wire pairs.

The Bias-Sin-Cos control scheme offers a good compromisebetweencomplexityand performance.With this scheme each pole has three windings:Bias,Cos(or Ctrl-X)and Sin(or Ctrl-Y).The number of turns in all bias coils are the same,but the polarities alternate.The number of turns in Cos and Sin coils change approximately proportional to the cos and sin of the angle between the pole axis and theX-axis respectively.This is shown in Table 4 where minus signs indicate a coil has opposite polarity to the coils with plus signs.All bias coils for one radial actuator are connected in series and driven by one amplifier.The Cos(Ctrl-X)coils are connected in series and driven by one bi-polar amplifier.Similarly,the Sin(Ctrl-Y)coils are series connected and driven by one bi-polar amplifier.The bias-sin-cos winding scheme can also be applied in a similar way to larger bearings where more than 8 poles per quadrant are used(12or16 poles for example).

Tab.4 Number of turns for each coil in bias-sin-cos actuator

Comparision of bi-quadrant and biassin-cos winding.Figure 7 shows a comparison of the magnetic field distributions when generating radial force in theX-direction with bi-quadrant winding scheme(a)and Bias-Sin-Cos winding scheme(b).TheX-control control current is 10A in both cases.It can be seen that in Figure 7(a)only pole pairs 0-7 and 3-4 are contributing to the force generation,whereas in Figure 7(b)pole pairs 1-2 and 5-6 are also involved.As a result,the force produced with the Bias-Sin-Cos winding configuration is 15% higher than with the Bi-quadrant configuration.

Figure 8 shows magnetic field distributions produced when a force is generated in betweenXandYaxes(45 degrees with respect to theXaxis).In this case,forces produced with the two winding configurations are about the same(strictly speaking the force obtained with bi-quadrant configuration is 1% higher).Noticeably,the field distribution for the Bias-Sin-Cos case in Figure 8b is the same as the field distribution in Fig 7(b)rotated by 45 degrees,whereas the field distribution for the Bi-quadrant case in Figure 8(a)is very different from the field distribution in Fig 7(a).

Figure 9 shows a comparison of the force vs.current curves obtained with bi-quadrant configuration and with Bias-Sin-Cos configuration when the load is directed along theX-axis(a)and between theXandYaxes(b).Clearly the Bias-Sin-Cos winding more fully utilizes the actuator force capability if the load vector direction changes or is unknow a priori.As a counterpoint we note that for heavy horizontal machines standard industry practice is to orient the actuator such that the weight load is shared between poles(as in Figure 9(a)).However,machines with a vertical spin axis,such as many wastewater blowers,the more uniform load capacity characteristic of the Bias-Sin-Cos winding can be beneficial.

Table 5 summarizes key qualitative differences between the three different winding configurations.Using a separate bias coil results in a smaller control coil and subsequently a higher slew rate limit for the same bus voltage.The bipolar amplifer needed is already commonly used in Calnetix’s permanent magnet bias magnetic bearings.The bias-sin-cos version was selected to take advantage of the slew rate benefits and the increased load density while accepting somewhat higher power consumption for the additonal coil turns needed.

Fig.7 Magnetic flux distributions when generating radial force in the X-direction with 10Acontrol current

Fig.8 Magnetic flux distributions when generating radial force between axes X and Y with 10Aradial control current Ir(geometric sum of Ixand Iy)

Fig.9 Comparison of the radial force vs current curves with the load direction along the X-axis

Tab.5 Comparison of the actuator winding schemes

4.1 Static Bench Tests on the Radial Actuator

Prior to the integration into machine,coil polarities and number of turns in the coils were verified using a bench test fixture shown in Figure 10.The test fixture employs a stationary rotor mockup with a laminated actuator target to mimic actuator operation inside a machine.The Bias,Cos and Sin windings were individually energized with+3A DC currents(windings are sized for 10A DC continuous currents),then the magnetic fields in the air gaps between the poles and the actuator target are measured with a 0.5 mm(0.02 inch)hall effect flux probe.A sample of the test results for one actuator are compared to predicted values in Table 4.The maximum percent error between the expected and measured values was less than 10%.Much of the residual error was found to be caused primarily by the centering accuracy of the actuator target(rotor mockup)relative to the actuator stator.

Note that with the ControlXcurrent set to+3ADC,control flux in pole pairs 0-7 add to the bias flux magnitude,and the control flux in pole pairs 3-4 subtract from the bias flux magnitude.This is also the usual action in a standard bi-quadrant winding scheme.The source of extra on-pole load capacity from the Bias-Sin-Cos winding scheme can be understood by examining the Bias and Control-Xfluxes for poles 1-2,wherepole1 increase the totalflux magnitude（-0.24-0.09=-0.337）and pole 2 decreases the magnitude（+0.24-0.09=+0.157）.The action for poles 5-6 is similar.The relation of the ControlYpoles to the bias is similar to ControlX,except with a 90 degree phase shift.

Fig.10 Radial actuator test setup

Tab.6 Radial actuator static flux-comparison of prediction to measurement*

Next,a number of frequency response measurements were performed on the radial actuator test fixture to verify design characteristics of the actuators.These tests included control/bias transfer functions to identify coupling due to mutual inductance,current/voltage transfer functions to measure inductance,and flux/current transfer functions.The frequency response of the bias coil to dynamic control voltage is shown in Figure 11 for a centered rotor.The gain ranges from-38.9dB to-38.7dB which represents about 1.0% to 1.2%.The coupling will get larger if the rotor is offset significantly in the air gap,however,with adequate power amplifier bandwidth this should be easily rejected.

Fig.11 Transfer function from control coils to bias coil with rotor centered in stator

4.2 Design and Analysis of the Axial Actuator

The thrust actuator used in this machine is of a conventional,EM biased type.Its structure and envelope dimensions are shown in Figure 12.

Two bias coils are connected in series and generate bias magnetic fluxes in the axial air gaps on two sides of the thrust disk.Two control coils are also wired in series in such a way that the magnetic fluxes they produce add to the bias flux on one side from the thrust disk and subtract from it on the other side resulting in a net magnetic force pulling the ro-tor towards the side with the larger magnetic field.Reversing the control current changes the direction of the force.Figure 13 shows force vs current curve for the thrust actuator.

Fig.12 Structure and outline dimensions(mm)of the EM biased thrust actuator.

Fig.13 Thrust actuator force vs current curve

5 Future Applications/Commercialization

There is a significant interest in MV MW level high-speed,high efficiency drivers in three specific industrial use areas:

1)Natural gas movement and compression.The need for efficient movement with low environmental impact has led to sustained interest in electric drives for large pipeline and process compression.While technologies hitherto did not provide enough of a step increase in efficiencies necessary to offset the higher initial capital costs,SiC technology shows promise.A major US gas turbine company is already invested in this technology and product development is being pursued.The need for machines in the 8～10MW range has already been established by trade studies.

2)Wastewater treatment.The case is even more pressing.High population density urban areas must operate large water recycling systems where the greatest operating costs are electric energy and maintenance.The higher efficiencies of the new SiC drive systems have potential to have impact this application.Aeration blowers in the 1 to 1.5MW size have been shown to be more cost-effective than smaller parallel blower arrangements.The major barrier here for the SiC technology is the initial capital cost which must be reduced to gain market penetration.

3)Offshore applications.Higher power densities,lower footprint and lower operating costs enable offshore operations to be scaled up within the same platform size and increase production.Large offshore operators have already specified or are evaluating this technology in the 6～10MW range for both power generation and compression on the offshore units.

As of today,the SiC MOSFET device cost is considerably higher than its Si IGBT counterpart as the SiC devices are adopted in various applications that the device cost will reduce significantly and allow for full adoption of SiC devices and replace Si devices

6 Conclusion

The new high-speed,medium voltage megawatt drive based on SiC MOSFET switching devices has been designed to meet the performance targets of 0.79m2/MW footprint and 1.510m3/MW inverse volumetric density.Testing is underway and results to date indicate the VSD can meet the target performance including a SiC inverter efficiency of over 98.5%.This is significantly higher than the conventional drives and the present IGBT based drive technology.Switching rates of up to 21.5kV/µsec have been demonstrated.It is expected that technology maturation of the devices and greater application volume will bring the costs of the devices in line with economic feasibility,as it was with the current IGBT technology.

The drive is to be tested on a 1.6MW motor supported by 8-pole,heteropolar radial magnetic bearings and a conventional EM-bias axial bearing.The radial heteropolar bearings employ a unique bias-sin-cos winding configuration which reduces the dependence of the load-carrying capacity on load direction.The radial bearings were constructed and tested prior to being integrated into the machine to characterize the static and dynamic characteristics.The results showed good agreement with predictions.

Calnetix plans to continue to test and apply the MV drive and 1.6MW motor in future applications including natural gas pipeline compressors,wastewater treatment,and offshore applications.These technologies are expected to play a key role as several industries continue to advance towards large-scale electrification.

7 Acknowledgements

This development work is supported by Department of Energy(DOE)under Award No.DE-EE0007251,where Calnetix is in process of developing a＞1 MW at＞15 000 RPM motor and VSD that will demonstrate the commercial viability of wide band gap power electronics for use with MW class motors.Cree Fayetteville Inc.developed the 10 kV SiC module and corresponding gate drive board.