Innovative and Highly Integrated Modular Electric Drivetrain

A highly integrated electric drivetrain module with 157 kW peak power is presented, which incorporates novel technologies in the field of electric machines, power electronics and transmissions: 1. High-speed electric machine with six phases and injection mould polymer-bonded magnets; 2. High-ratio dual-speed transmission with double planetary gear set (Ravigneaux gear set); 3. Gallium nitride (GaN) power electronics with winding reconfiguration feature.The combination of these components in one single housing makes the drive module flexible to integrate and to combine with conventional or alternative propulsion technologies, thus allowing various hybrid and electric drivetrain topologies. All technologies are selected in accordance with mass production potential and can therefore have a high impact on the automotive market in the future. Currently, the drive module is under development; the first models will be assembled in winter 2019. The integration into a demonstrator vehicle in 2020 will prove the potential of many new technologies and the suitability for the automotive market.


Introduction
Electrification of passenger cars and light-duty vehicles -which are still responsible for around 20% of Europe's greenhouse gas emissions [1] -will have an enormous effect on the reduction of greenhouse gas emissions. Additionally, pollutant-free air in dense urban areas is possible with electrified propulsion. However, the maturity of electric drives needs a final push for better performance and efficiency so that battery capacity can be saved and driving range can be increased. Combined with a cost reduction, a massive adoption of such transport in Europe and worldwide can be generated. This paper presents a new generation of a modular electric drive module, which can be used for various configurations of Full-and Hybrid Electric Vehicles. Focus is on cost reduction, low environmental impact, efficiency, and mass manufacturing readiness. The multiphase electric machine based on permanent magnets will be combined with the latest Gallium Nitride (GaN) power switches, advanced control with higher fault tolerance, and cooling features with reduced dimensions and higher efficiency. It will be linked with a performant two-speed transmission, adapting also new regenerative braking strategies. The development takes industrial-and user requirements into account as well as environmental impacts through life cycle analysis, to gear the activities towards a full vehicle design approach and the realization of each component and the complete powertrain. The project gathers nine cutting-edge partners from the fields of automotive, power electronics, transmissions and electric motors including three research centres, three Tier-1 suppliers and "Small and Medium Entrepreneurs".

Machine Design
In order to achieve very high power densities even with low available current and voltage, sophisticated electromagnetic designs are required. Additionally, the automotive market has strong requirements, adding additional effort to the design of a motor which is safe and strong at the same time. To meet these challenges, the Hybrid Synchronous Motor (HSM) was invented a few years ago which is an evolution of the internal permanent magnet motor (IPM). It is called Hybrid, because it utilizes the combination of reluctance-and electromagnetic torque. The torque production by these two effects is the reason, why it is difficult to design such machine type. Small changes in the geometry or materials have influence on both effects and therefore can have significant impact on performance, torque ripple, efficiency, power factor, controllability, demagnetisation and more.
The main goal for the EM development was to save more than 50 % magnet mass compared to existing PMSM with similar power, to reduce the active weight and to improve the efficiency. Many iterations were needed to satisfy all of the previously defined drive requirements.
The resulting electric machine is a 6 pole, 6 phase HSM for high speeds up to 22500 rpm. Since power is proportional to both speed and torque, the torque demand for the peak power of 101 kW @ 6600 rpm is as low as 145 Nm. For decent accelerations from 0 to 50 km/h of a C-class vehicle with the given gear ratio of 21.65, a max. torque of 150 Nm is required. With this gear ratio and motor torque the vehicle can accelerate from 0 to 50 km/h in around 3 seconds.
As the required motor torque is fairly low, the required phase current is also low, in this case only 225 A (rms) per phase. A challenge for high speed motors are the centrifugal forces at the rotor which require mechanical stabilization and special rotor geometries to hold the permanent magnets in place.
The stator windings are made from a special wire type called "Formed Litz Wire" (FLW, fig. 1). This wire type provides low DC resistance (similar to massive copper bars) and low AC resistance (similar to conventional Litz wires) at the same time [2]. FLW technology combined with the high speed motor concept leads to a very low phase resistance and high continuous power, because a small motor has also short wires. Additionally, iron losses are reduced because of the low flux harmonic content, even if the motor speed is very high.  [2] The project target torque of 150 Nm is surpassed from the simulations by 156 Nm at a current of 225 A (rms) and 320 VDC. Since the nominal speed will also be higher than expected (8650 rpm instead of 6600 rpm) the power at the new nominal speed rises to 141 kW. Above the nominal speed the power does not drop, it rises again to more than 157 kW (213 hp) between 12000 and 16000 rpm. This is very impressive if we consider the low remanence flux density (Br) of only 0.75 T @ 25°C, compared to sintered magnets which are available with more than 1.4 T @ 25°C. To achieve these high levels of power with only 320 VDC and 225 A (rms), perfect magnetisation, optimal motor control and also very good mechanical design such as low friction bearings, a stiff shaft and many other intelligent features are needed.
At the end of the day not only a strong and light motor is strived for, it is no less important to have a good efficiency. Simulations showed that also with this high-speed drive, peak efficiency of more than 97 % can be realized ( fig. 2). However, the key to save energy and battery size is mainly not peak efficiency, but a high drive cycle efficiency. Since the majority of drive cycles demand only low power, the good efficiency of this HSM design in the low-load areas helps to improve the overall energy consumption and to save battery capacity. With an active weight of less than 30 kg the developed motor leads to a powertrain with a high power density and very low amount of rare earths.

Magnet Injection Moulding
Injection moulding of polymer-bonded magnets in the rotor cavities instead of insertion of sintered magnets was identified as one of the possible ways to further reduce the content of critical raw materials in the motor. Unfortunately, bonded magnets have less remanence flux density (Br) than sintered magnets but they contain no heavy rare-earth (which is more critical than Nd and whose price might raise more severely if a supply crisis were to occur). They also allow complete freedom in the shape of the magnets, yielding some magnetic optimization potential of the rotor. Furthermore, all of the 66 magnets of a rotor slice can be injected in one single step which lowers production effort.
As the mechanical strength of bonded magnets is lower than of sintered magnets, centrifugal forces can be an issue. Hence, a mechanically stable rotor geometry has been developed for which the calculated power and efficiency of this injected magnet rotor is comparable to the sintered magnet one. Magnetostatic finiteelements calculations are now being performed in order to design a magnetizer tool to produce the proper magnetic field in the rotor cavities during the injection process ( fig.3, left). With regards to materials, several grades of composites are available; the most suited one has been identified and injected to parallelepipedic samples ( fig.3, right). The mould allows a magnetic field tuning in order to precisely reveal the fielddependence of magnetic properties and to allow further optimization of the injection mould.

Advantages of Gallium Nitride Devices
The low voltage Gallium Nitride (GaN) power device technology (650 V) provides a lot of advantages compared to its Silicon (Si) and Silicon Carbide (SiC) competitors [3,4]. In terms of performance, the Gallium Nitride devices have:  an unrivalled switching performance with switching on/off in less than 10 ns,  a low package inductance,  a low on-resistance  and reduced switching losses: 5 times better than Si devices and 2 times better than SiC devices.
With these advantages, the switching frequency can be raised and hence the size of passive components can be reduced: the inverter becomes smaller and more efficient at the same time. In terms of costs, fig. 4 shows that a Gallium(III) trioxide (Ga2O3) wafer is 3 times less expensive than a SiC one and it can be further reduced [5].  [5] The GaN technology is young and under ongoing development. As an example, the current capacity per die is still low: 60 A in 2017 and 120 A in 2019. In the framework of this research project, in order to fulfil power requirements, extensive work has been put in developing a reliable parallelization of dies.

GaN Power Board Conception
To obtain the required torque of 150 Nm in the 6-phase high speed motor, a current of 225 A (rms) per phase is requested. Each phase is independently connected to a full bridge in which each of two switches is realized with four parallelized GaN devices as specified in [6]. The topology of the full bridge is depicted in fig. 5 (left) with each switch (K1,2,3,4) realized by four GaN devices in parallel. A first power board prototype of one leg of the bridge is shown in fig. 5 (right). Hardware development started with the parallelization of two devices. The experimental waveforms in fig. 6 show the results for two parallelized GaN devices with 300 at double pulse test with an inductive load ( 330 , 200 ). As can be shown, the switching of the two devices is synchronous which is essential to maintain an equal current in the two of them. Also, oscillations on the DC bus at a frequency of approximately 1.2 MHz could be observed, which are induced by an exchange of reactive energy between the required decoupling capacitors.

Series-independent Winding Reconfiguration
In this section, a six-winding EM is used to explain the winding reconfiguration principle. Its windings are independent from each other, meaning there is no star point, which enables a free choice of the machine winding configuration. Fig. 9 shows the implementation for one of the three phases, i.e. for two windings. The two windings are connected to the supply by means of a switch matrix, comprising two full bridges and two series switches, depicted as contactors, and each machine winding is represented by a single inductor. The matrix is able to interconnect the windings in series, or to connect each winding independently to the supply. In the series configurations, either series switch 1 or 2 is closed. Hence, the current in the seriesconnected windings can be controlled by means of two phase legs only, which is advantageous with respect to switching losses in the phase legs.
In the independent configurations, the winding currents are controlled individually. Nevertheless, due to the symmetric construction of the motor, both currents will in principle be controlled to the same value.
In a first implementation of this concept [7], a single series switch was used, and the reconfiguration was applied depending on the actual speed of the motor. However, this way of operating has to include a certain margin for varying supply voltage, stator impedance, and magnetic field strength in the machine. Additionally, once the independent configuration is used, it applies during a complete cycle of the voltage wave of the machine phase, although in the sections close to a zero crossing the series configuration could still be used. Finally, with only a single series switch available, the loading of the four phase legs is asymmetric, as always the same two legs are disabled in series operation. Therefore, in the present work it was decided to extend the reconfiguration concept: when possible, a series configuration is used, and only where necessary (starting at the peak of the voltage wave applied to a machine phase) the windings are controlled independently. Also, in order to spread the losses more evenly over the phase legs, a second series switch was introduced, as already shown in fig. 9. As an ultimate refinement, the PWM patterns in the independent configuration are shifted 180 degrees with respect to each other, thereby minimizing the net torque ripple in the machine. Due to the reconfiguration, the system order changes during operation: in the series mode only one current per phase needs to be controlled, but in the independent mode two currents have to be regulated. The resulting control method for one phase is depicted in fig. 10. As shown, the switching on/off of the Difference Current controller (by definition zero in the series configuration) is governed by a binary signal derived from the output of the Sum Current controller (the desired winding current): once the demanded voltage for the series connection becomes too large in view of the DC link voltage, the configuration is changed to independent, the Difference Current controller is enabled, and the controller gain halved in order to avoid a transient in its output due to the changed load configuration.
For controlling all three phases, the sum currents will be transformed to the d-q-0 frame by means of the Park transformation. For brevity, this approach, which is quite standard for the control of three-phase machines, will not be detailed further in this paper.

Simulation Results of Winding Reconfiguration
A first result of the control concept discussed above is shown in fig. 11. Here, the circuit is generating a sine wave voltage across one winding pair. As shown, once the modulation index for the series connection reaches 1 or -1, the configuration is changed to independent and correspondingly the modulation index used to generate the PWM signals is halved. Once the resulting index drops below 0.5, the series configuration is restored. The shape of the individual currents in a similar situation is shown in fig. 12. It can be seen that each current follows the set-point, and the ripple of the sum current remains low in both the series and independent configurations.

Industrialization Potential of GaN Devices
GaN is a young and emerging technology, which has a lot of advantages and great potential for automotive applications [3,4]. With increased switching frequency and low losses, it allows to build very efficient and compact power converters and inverters. Soon the cost of such a device will be similar to Si and much cheaper than SiC [5]. Moreover, as it is a non-mature technology, technical improvements are to be expected, whether for the device performance itself (e.g. current capability and reliability) or the manufacturing process. The prediction of CEA is that SiC devices will stay at the foreground for high voltage (> 1 kV) applications. However, GaN technology will in a few years dethrone low voltage (< 1 kV) technologies which have been installed for quite some time in power electronics such as IGBTs, Cool-MOS etc.
Interestingly it was noticed during the power board development, that auxiliary components such as passive components, drivers, measurement probes etc. are not always adequate for the use with GaN devices and progress has to be made also in those fields, without which, the use and development of GaN-based inverters remains difficult.

Transmission
For the transmission two options were considered in the project. The first option is a fixed ratio, single speed transmission. The second option is a dual ratio, switchable transmission. Several aspects were considered to help making the decision: EVS32 8  Overall powertrain efficiency during type approval cycles: no clear winner between the two options  A dual ratio transmission with adapted ratios provides higher launch torque in first gear and higher top speed for the same motor speed in second gear: this benefits the projects demonstrator that has limited motor torque due to the current limitations of the available GaN devices.  Market demand: some customers prefer a dual ratio transmission for EV powertrains for particular reasons (e.g. gradeability, large difference between empty vehicle mass and fully loaded vehicle mass).
As for the projects targets advantages of dual ratio transmission surpass disadvantages, it was decided to develop a dual ratio, switchable transmission. The ratios of the transmission are 21.65 for the first gear and 12.18 for the second gear which leads to a ratio step of 1.78 The development process is described in the following: With the help of the gear synthesis software PlanGear from ZG, different gear configurations could be elaborated on the level of stick diagrams. Out of 34 possible configurations which fulfil the drivetrain's requirements, the four configurations with the highest efficiencies were depicted for further investigation. These configurations were then evaluated with respect to expected costs and overall size. Finally, the combination of a Ravigneaux planetary system and a pair of cylindrical gears was identified as most promising solution. In a next step, gear and bearing dimensioning was elaborated with the use of the transmission calculation software MASTA by SMT. Shaft and bearing loads were based on highly demanding drive cycles. Fig. 13 illustrates the design process of the gear and bearing configuration. Several configurations were reviewed whilst taking the manufacturing methods for parts and the way of assembly of the powertrain into account. Subsequently, detailed gear and bearing lifetime calculations based on highly demanding drive cycles were performed. Load-capacity, efficiency, weight and costs of possible bearing arrangements were evaluated and compared. Where possible, ball bearings are applied to reduce transmission losses. Once the transmission layout was known, the oil routing was developed to assure all bearings and gears are lubricated sufficiently.

Housing and Cooling
One of the ambitions of this project is to obtain a highly integrated, compact electric powertrain. The integration of motor, inverter and transmission into a single unit simplifies the powertrain assembly in the vehicle. Furthermore, the integration reduces the cost of interfacing by eliminating connectors, cables and covers. At the same time a modular approach should allow exchanging (a part of) the motor, inverter or transmission for repair.
From different layouts the consortium chose an arrangement where the inverter boards are placed around the drive shaft as can be seen in fig. 14 (left). This choice provides a compact unit with the best balance of length (L), width (W) and height (H).

Figure 14: Cross section of chosen layout (left) and main dimensions of sample powertrain (right)
An important feature of the integrated housing is the cooling of both motor and inverter. Prior experience has learned that a too high back pressure in the cooling system requires a fairly high coolant pump power for the required flow. For this reason the cooling subsystem puts the cooling of motor and inverter in parallel. The motor cooling itself uses several parallel channels.

Holistic Design Approach
The best electric machine, the best inverter and the best transmission do not necessarily combine to the best drive module. This is because properties of fully electric drive modules result from multi-physical interaction of the powertrain components (electrical, mechanical and thermal). A design of the components with regard to the performance on vehicle level is necessary in order to find a global optimum for a specified vehicle and fixed driving requirements [8]. Furthermore, the control system design for the drive module is important to consider because it can increase the interaction between different subsystems but also be a remedy for unwanted interaction.
Parallel to the engineering of the drive module, a design process is being developed which aims to identify the best combination of components for the demonstrator vehicle. "Best" herein means the ideal compromise between costs, efficiency, performance, weight, volume and complexity. The result is an iterative and computational intensive process, which involves detailed component modelling of transmission, electric machine and inverter, as well as vehicle longitudinal dynamics simulation in order to evaluate the components on system level ( fig. 15). In the first step, the requirements on vehicle level (e.g. acceleration from 0 to 60 km/h in 3.9 seconds) are broken down to component level (1 & 2). This results in a set of component requirements such as torque, power, gear ratio, etc. (3). The component level calculation blocks layout and design components to fulfil the requirements and return the results (4 & 5). The components are virtually assembled and the performance on system level is evaluated through numerical and analytical vehicle simulations. By weighting of different design criteria such as costs, performance, weight, etc., the best set of components for the specified target vehicle is found (6). As this study is ongoing and targeted to complete in Summer 2019, no resulting drive module specification can be shown yet. However, the outcome of this design study is not only a suitable specification for drive modules for the Volvo C30e, but also to give insights into the component interaction between electric machine, inverter and transmission. The following findings have been gained until now during the development and usage of the presented design environment: -Electric Machine: High speed EM generate power over high speeds and therefore do need less winding current than low speed EM. Ohmic losses in the windings are reduced while the iron losses slightly rise. However, the savings in winding losses surpass the additional iron losses and the machine is more efficient. -Inverter: Classical Si-based inverters are still a good choice for slow to medium switching frequencies. At high switching frequencies, GaN and SiC provide a suitable characteristic. The latter option leads to very compact Inverters, as the size of the DC-link capacitor can be reduced with higher switching frequencies. -Transmission: With higher maximum EM speeds, higher reduction ratios are required and the transmission gets bigger and less efficient. For high reduction ratios, more complex topologies show advantages compared to simple spur gear designs. -Module Level: A module based on a high speed electric machine as presented above is more efficient and compact compared to conventional modules, even though the transmission itself is not. The drawbacks on the transmission side are overcompensated by the advantages on the machine and inverter side. However due to high switching frequencies, conventional IGBTs produce high switching losses and thus wide bandgap materials (GaN and SiC) should be used. Based on the current cost structures the use of these materials results in a higher price on module level.