Digital Collaborative Development of a High Reliable Auxiliary Electric Drive System for eTransportation: From Dual Three-Phase PMSM to Control Algorithm

For electrified transportation (eTransportation) systems, multi-phase motors can provide higher performance and reliability than three-phase ones, but also bring more challenges in their optimal design and control. In this article, a set of high reliable electric drive system based on dual three-phase permanent magnet synchronous motor (DTP PMSM) is developed for auxiliary systems in eTransportation field. A digital collaborative develop process is proposed with the support of multiple software tools. Design, manufacture, and bench testing stages of the DTP PMSM, the two-level six-phase inverter, and the control algorithm are efficiently incorporated. A prototype of the multi-phase electric drive system is fabricated and tested. Comparison of the simulation analysis and experimental results confirms the effectiveness of the collaborative develop progress. Control algorithms based on dual d-q model and vector space decomposition model are both verified and compared via the bench test. Operation mode switching from six-phase mode to three-phase mode is also realized with the prototype system, verifying its capability in fault-tolerant and potential in efficiency optimizing.


I. INTRODUCTION
Transportation electrification is bringing a worldwide technical revolution during the past decades, benefitting from its environmental friendliness, high energy efficiency, and good propulsion performance [1], [2]. Electric transportation systems such as electric vehicles (EVs), more/all electric aircrafts and electric ships, are gaining more and more attentions from engineers and researchers all over the world [3]. In general, electric transportation systems desire that their electric drive systems satisfy the following requirements: good energy efficiency and power performance, low vibration The associate editor coordinating the review of this manuscript and approving it for publication was Sudhakar Babu Thanikanti . and noise, reduced volume and weight, great robustness and fault-tolerance ability [4]. In fact, multi-phase electric drive systems, compared with classic three-phase ones, may provide a more effective solution meeting all the aforementioned challenging requirements [4], [5].
There have been various applications of multi-phase motors in the field of electrified transportation [6]. A dual three-phase motor was designed and compared with the Nissan leaf three-phase machine, and shown improved torque density and torque quality without compromising efficiency [7]. Authors in [8] combined hybrid energy storage system with dual three-phase permanent magnet synchronous motor (DTP PMSM), achieving simplified powertrain and good performance for an EV. A nine-phase machine was VOLUME 8, 2020 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ integrated into the on-board battery charger of EVs, with a fuzzy logic controller designed for charging control [9]. Besides electric propulsion systems, the electrified auxiliary systems, such as electric actuators and electric pumps, also prefer multi-phase configurations. Six-phase motors have been widely employed by safety critical systems such as electric power steering systems (EPS) [10], [11] and electric braking systems (EBS) [12] of electrified vehicles. Due to the application demands of multi-phase motors, there have been plentiful studies and achievements in their design, modeling and control. Among various types of multiphase motors, a very interesting option is the dual threephase machine consisting of two sets of three-phase windings with isolated neutral points. DTP PMSM with 24 slots have been widely studied and optimized for better torque performance and low harmonic [13], [14]. Effect of phase shift angle on radial force and vibration behavior in DTP PMSM was studied in [15]. Authors in [16] provided analytical model of N × 3-phase PMSM for control-oriented purpose. In [17], a control algorithm was specially designed to drive a dual three-phase induction motor with a five-leg inverter. High quality current control [18], [19] and fault-tolerant control [20], [21] are also research hotspots of DTP motors.
More phases bring more novel configurations and design innovations as well. In a six-phase hybrid-excitation motor, three phases were used as excitation winding while the other three phases worked as armature winding [22]. In [23], novel toroidal windings were proposed for a six-phase direct-drive PMSM. In [24], windings of two six-phase induction motors were connected in series, and independently controlled from a single six-phase inverter. A six-phase transverse flux tubular permanent magnet linear motor was designed with a faulttolerant control strategy in [25]. Authors in [26] optimized the torque profile of a six-phase PMSM for a direct-drive EV with supervised machine learning. In [27], multi-level inverter was optimally designed with two sets of three-phase windings for better fault-tolerant and reduced torque ripple of an induction motor.
As previously reviewed, multi-phase configurations can provide more optimizing potentials and flexibility to electricdrive systems. However, there are also some challenges in the development process of multi-phase systems for electrified transportation. Transportation facilities such as electrified vehicles are systems with multiple requirements, constraints, and trade-offs. Most of existing studies mainly focus on individual research point of multi-phase systems, such as motor design, inverter configuration, or control algorithm. Collaborative development of the whole multi-phase systems is urgently desired for their application in eTransportation. The multi-phase motor, the inverter, and the control algorithm should be synergistically researched and developed, to achieve comprehensive optimum scheme of the whole system, and meet the rigorous requirements in power, efficiency, weight, volume, and reliability.
Facing those aforementioned challenges, some work has been carried out in this article. The main contributions of this work are twofold. First, a multi-software based digital collaborative development process is proposed and practiced in this work, achieving higher development efficiency and optimized performance of the multi-phase electric drive system. Design, manufacture and validation of the DTP PMSM, the six-phase two-level inverter, and the control algorithm are efficiently coordinated and iterated in the design loop. Second, two kinds of control algorithms for the DTP PMSM are designed with simulation, applied to actual system, and compared via experiments. Torque ripple suppression and fault-tolerant capabilities of the whole system have been adequately validated via bench tests.
The remaining sections of this article are organized as follows. In Section II, the design problem of the multiphase electric drive system is formulated, and the multisoftware based digital collaborative development process is introduced. In Section III, the design, optimization and simulation analysis of the DTP PMSM is elaborated. Section IV presents the design of a six-phase two level inverter as the controller of the DTP PMSM. In Section V, control-oriented models of the DTP PMSM are built, and control algorithm is developed. Section VI accomplishes experimental validation of the motor prototype and the controller, and gives some discussions. Conclusions of this article are drawn in Section VII.

II. DIGITAL COLLABORATIVE DEVELOPMENT PROCESS OF THE AUXILIARY ELECTRIC DRIVE SYSTEM A. PROBLEM FORMULATION
Oil pumps are key auxiliary components in vehicle powertrain systems, and electric-drive oil pumps are especially meaningful for hybrid powertrains [28], [29]. In this work, a high reliable 6kW DTP PMSM system is developed for oil pumps of electrified vehicles, and also can be applied to auxiliary systems of other electrified transportation facilities.
It should be noted that, limited by the on-board power distribution scheme, a DC bus voltage scheme is determined as 260V-280V which is also consistent with other auxiliary voltage systems of aviation and marine transportation. Primary requirements of the system are listed in Table 1. For the DTP PMSM system, two aspects should be taken into account: Firstly, a high power-density motor with good performance should be designed with multi-factor optimizing. The slot-pole combination and the rotor magnetic circuit should be optimized so as to reduce the cogging torque and leakage inductance of the motor, and improve the torque output and efficiency at the same time. Secondly, a high efficiency motor controller is expected to maximize the efficiency of the motor, make full use of DC bus voltage, enhance the system response, and guarantee the stability and robustness of the whole system. The design process should be rapidly iterated to improve the development efficiency.
Therefore, strong requirements for the collaborative design process of the motor and controller have been put forward, along with the development of high-performance auxiliary electric drive system.

B. DIGITAL COLLABORATIVE DEVELOPMENT PROCESS OF THE ELECTRIC DRIVE SYSTEM
In order to satisfy the aforementioned requirements, an efficient development method needs to be used. Multiple engineering tools, such as EDA and CAD software, have evolved in the last decades and different solutions for electrical, mechanical and software engineering enable a high efficiency digital design process of the electric drive system [30]- [32]. On the one hand, the digital design process has been conducive to shorten the development cycle and lower the economic cost, because requirements from one phase can be directly taken into account in other phases, and meanwhile, independent work can be done parallelly. On the other hand, product data management (PDM) has been widely employed in the industry, and data exchanging between engineering tools and the database is more convenient in this way. Fig.1 shows the main engineering tools helping enhance the development efficiency. The digital development process with multi-software collaboration is depicted in Fig. 2. According to demands, design tasks can be decomposed into different sub processes including motor design, controller hardware design, and software design. To better understand data exchanging between different phases, the detailed processes have been summarized in Fig.3.
For motor design, multi parameters optimization is essential to improve motor performance. With the help of motor design software, a preliminary scheme can be directly selected from the abundant alternative motor libraries including brushless permanent magnet motors (BPM), induction machines (IM) and so on, which has a low demand of engineering experience for designers. And finite element analysis (FEA) can be utilized to carry out the simulation results of motor performance for parameters optimization design such as winding cogging scheme and magnetic circuit structure. Due to switching frequency limitation and deadtime increasing the harmonic component in output voltage, it is necessary to consider the reduction of high frequency loss in motor design. The influence of modulation voltage utilization on the motor should also be considered. The detailed design and optimization results will be introduced in Section III.
In hardware design, constraint parameters including voltage and current level of each phase and DC bus are obtained from motor FEA results. To improve efficiency of development, EDA software can be used for circuit design and simulation, at the same time for electro-thermal analysis. Combined with the control algorithm suppressing harmonics, the value of the support capacitor can be reduced in the design. Moreover, by adopting automatic coding tools, the control algorithm can be easily converted from simulation algorithm to embedded code, which is much more efficient than conventional handwriting code.
During control software design, the motor parameters, such as inductance and permanent flux, can be obtained from the FEA results, and after that a high-precision mathematical model of the motor can be built for the model in the loop (MIL) analysis. In addition, the control algorithm can be optimized through collaborative development. For example, the copper loss and iron loss at different speeds calculated by FEA can be utilized to improve the control algorithm based on the minimum energy loss. For high performance control, the time delay and dead-time of inverter should be taken into account as well.
In summary, developing process based on multi-software digital collaborative design integrates the advantages of various software to improve development efficiency and flexibility. It should be pointed out that the design scheme will be optimized according to the prototype test in the form of correction factor.

III. OPTIMAL DESIGN OF THE DUAL THREE-PHASE PMSM A. MAIN PARAMETERS DESIGN OF THE DTP PMSM
The DTP PMSM involves in two sets of Y-type three-phase stator windings with separated neutral points. Two sets of stator windings that indicated by (A, B, C) and (X , Y , Z ) lay in the stator but not fracture with each other, as shown in Fig. 4. There are two kinds of winding topologies in common use: one is that the two sets of windings are completely coincident in electrical angle, namely, the windings are diverted by 0 electrical degrees, and the other is that the windings are spatially shifted by 30 electrical degrees. For the latter winding structure, due to the sixth harmonic torque generated by the two sets of windings is equal with 180 • phase difference, the lowest order of the torque ripple is 12 th . Whereas fault tolerance and torque ripple should be considered, the scheme  with winding shifted by 30 • shown in Fig. 4 (b) is employed in this article.

1) OPTIMIZATION OF THE SLOT-POLE COMBINATION
Owing to the influence of the controller switch frequency range on the system performance and hardware cost, the motor polar numbers should be considered first. In the following design, the schemes of 6,8,10 poles integral and fractional slot winding are compared.
The three schemes mentioned above can be arranged in dual-three phase stator windings with conventional integer slot scheme. The pole distance (number of slots per pole) is set to 6, and the pole slot matching schemes are respectively 6-pole/36-slot, 8-pole/48-slot and 10-pole/60-slot with the coil pitch being 5. Under the condition that other parameters and electromagnetic load of the motor are the same, the cogging torque waves for the three schemes in the range of 12 mechanical degrees at the rated speed is illustrated in Fig. 5.
The cogging torque of electrical machine with integral slot winding is large, which can be reduced by fractional slot winding [13], [14], as shown in Fig.5. After FEA, 10-pole/ 12-slot fractional-slot double-layer winding scheme is adopted, by which the peak value of cogging torque is less than 0.45N·m, and the distortion rate of line back EMF waveform is small.

2) OPTIMAL DESIGN OF ROTOR MAGNETIC CIRCUIT
The rotor permanent magnet of DTP PMSM is usually fixed in two ways: surface mounted pole structure and interior pole structure. The surface mounted permanent magnet is located on the outer surface of the rotor core. In order to avoid the permanent magnet being thrown out during the rotation of the motor, the outer surface of the permanent magnet is usually covered with a non-magnetic sleeve or tied with a fiberglass tape to protect it. Due to the poor mechanical performance of the surface mounted structure, which is not suitable for highspeed motors, the interior structure is adopted as shown in Fig.6(a). Since the permanent magnet size and the pole arc coefficient are constant, the measure of magnetic isolation becomes the decisive factor of motor performance, containing the rib (the rib between adjacent magnet poles) and the bridge (the bridge between outer web corners and rotor surface).
The optimal design is carried out through changing the bridge thickness since the rib thickness mainly affects mechanical strength rather than electromagnetic performance of the motor. The bridge thickness varies from 1.0 mm to 2.0 mm in a scale of 0.1 mm, and Fig.6(b) presents the electromagnetic performance obtained by FEA. For the motor, the cogging torque and airgap flux density shall be small, and the back EMF shall be close to and slightly smaller than the power supply voltage. Besides, in order to improve sufficient mechanical strength, the bridge thickness shall not be too small.
Considering comprehensively, we believed that the scheme with 1.5 mm bridge thickness was the best option. The parameters of the DTP PMSM are determined as shown in Table 2.

B. ELECTROMAGNETIC SIMULATION AND ANALYSIS OF THE OPTIMAL SCHEME
According to the proposed optimal design scheme of the DTP PMSM, the performance was evaluated under the no-load and rated operation conditions in terms of the flux density, back EMF, current and torque. The results were represented in Figs.7(a)-(e).
The magnetic field distribution in the motor shows that the flux density of stator teeth is under 1.8 T while the flux VOLUME 8, 2020 density of rotor bridge is locally over 2.0 T with no load, which is beneficial for the flux to go into the stator core from the teeth after passing through the air gap. Therefore, it can be considered that, theoretically, the magnetic circuit design of the motor is reasonable and provides good material utilization.
The peak value of no-load cogging torque is about 0.29 Nm, accounting for 1.52% of the rated torque (19.1 Nm), which is quite small. At the rated speed 3000 rpm, the lineline back EMF waveform is sinusoidal and its RMS is 178 V.
According to FFT, fundamental back EMF amplitude is 245 V with THD under 1.80%. Taking the existence of friction resistance and stray loss into account, the electromagnetic torque at the rated load is 19.6 Nm slightly higher than the rated value, and the steady-state current is 11.5 A, having sinusoidal waveform and low harmonic components. Hence, from a mathematical point of view, the design of winding meets the requirement under either set of conditions.

IV. DESIGN OF THE SIX-PHASE TWO-LEVEL INVERTER FOR THE DTP PMSM SYSTEM A. TECHNICAL SCHEME OF THE CONTROL SYSTEM
Based on the demand for auxiliary electric drive system, the following general technical solutions are adopted. The DC bus voltage provided by the on-board battery is inputted into the driving circuit after filtering by support capacitors. The DTP PMSM is fed by a six-phase two-level voltage source inverter (2L-VSI). The digital signal processing (DSP) chip is used to generate PWM signals and drive 2L-VSI to output appropriate AC voltage. The DTP PMSM operates in the control pattern of rotational speed, and is able to switch over between single three-phase (STP mode) and double three-phase operation modes (DTP mode) automatically and smoothly, so as to deal with the failure of a certain phase and remain running for a short time. Upon over-current fault or over-voltage fault occurring, the control system can cut off the AC voltage instantly to stop the motor.
Six-phase VSI is deemed to combine two three-phase VSIs, hence two three-phase H-bridge intelligent power modules (IPMs) can be utilized to form the driving circuit. IPMs, integrating IGBTs and circuit of logic, control, monitoring and protection, have the advantages of compact structure and high reliability. Moreover, owing to the low power and load current of oil pump motor, the scheme by using IPMs is suitable enough for design requirements.
The main controller chip is a 32-bit single-precision floating-point DSP, TMS320F28335, produced by Texas Instruments (TI) (Dallas, TX, USA). This processor is used to start the DTP PMSM; control the speed of the motor; switch over among two operation modes; provide protection for over-current, over-voltage and overheat of the motor and the controller; and also provide data communication interface of RS485 and CAN.

B. HARDWARE DESIGN OF THE HIGH-VOLTAGE MAIN POWER CIRCUIT
As shown in Fig. 8(a), the main power loop for the DTP PMSM is composed of the on-board battery; large DC filter capacitors (DC-link support capacitors); two IPMs; sampling of current, voltage, temperature and speed signals; communication interface; and system controller.
The basic working process of the main power circuit is as follows: the on-board battery provides 260-280V DC voltage, and through 540µF/450V capacitors (two 270µF/450V capacitors in parallel) and a 2µF/400V non-inductive absorption (high-frequency) capacitor, the DC bus voltage becomes relatively stable and the ripple is greatly improved. And then, the DC bus voltage is input into two IPMs. After modulation with control algorithms, the six-phase (ABCXYZ) high voltage is generated by IPMs, which is directly applied to two sets of three-phase windings of the DTP PMSM. The internal circuit of IPMs is illustrated in Fig. 8 (b). In addition to realizing the function of the DC/AC inverter, the IPMs can send the fault signals to the DSP as well, which is vital for protecting the equipment.
The DSP controller samples real-time signals of DC bus voltage, stator current, and rotor position and speed, and then sends 12-channel PWM signals generated by closed-loop control algorithm to drive the IPMs. Besides, the controller can realize software protection by analyzing the fault signal from peripherals, and it can send out a protection signal in the case of faults. The level-shifting transceiver not only translates PWM signals from 3.3V to 5V for asynchronous communication between data buses, but also disables PWM signals translation when receiving DSP's protection signal, so as to respond to software protection command.

C. HARDWARE DESIGN OF THE DIGITAL CIRCUIT
The hardware part of the DTP PMSM controller, shown in Fig.9, includes the DSP minimum system, power supply circuitry, optocoupler isolation circuitry, signals detection and conditioning circuitry, resolver interface circuitry, hardware protection circuitry, and communication circuitry. The minimum system of the DSP, TMS320F28335, consists of the chip power supply (3.3 V, 1.8 V), 30 MHz external crystal oscillator (the maximum working frequency of the chip is 150 MHz through PLL frequency modulation), 14-pin JTAG interface for program debugging, as well as requisite bias resistors and bypass capacitors.
Power supply circuitry is DC/DC conversion circuit, providing stable voltage source for the DSP and peripheral circuits. Optocoupler isolation module is an essential part of IPMs drive circuit, isolating high voltage from low voltage. The analog signals from sensors before being sent to AD module will be filtered and denoised by signal detection and conditioning circuitry so as to improve the sampling accuracy. Furthermore, resolver as a position sensor needs the resolver interface circuitry, including the exciting signal circuit and VOLUME 8, 2020 resolver/digital converter (RDC), to detect the rotor position and speed information, and output digital signals of rotor position and speed to external interface (XINTF) data lines of DSP through parallel ports.
Hardware protection circuitry automates protecting for over-current, DC bus over-voltage or under-voltage, and overtemperature realized by voltage comparator circuit, which is a supplement for software protection. As shown in Fig.8(c), protection signal is generated after fault signals and protection command passing through a NAND gate. Trip-zone (TZ) submodule of DSP connecting to protection signal is programmed to respond when faults occur. For example, upon over-current detected, protection signal will be forced to shift level ignoring software command, and TZ will disable PWM signals output from DSP, and meanwhile, the level-shifting transceiver in Fig.8(a) will be disabled as well. Other fault signals are processed in the same way.

V. CONTORL SOFTWARE DESIGN OF THE DTP PMSM A. CONTROL-ORIENTED MODELS OF THE DTP PMSM
Two analysis methods for the DTP PMSM in different coordinate systems are introduced in this section, corresponding to different mathematical models: dual d-q transformation model and vector space decomposition (VSD) model. The dual d-q transformation regards each set of three-phase windings as a basic unit, and applies the conventional three-phase motor modeling method to each set of three-phase windings separately, which can be regarded as two three-phase motors with coupling relationship. According to VSD method, the overall machine model, such as fundamental components, harmonic components and zero sequence components, is transformed and decomposed into three decoupled subspaces, corresponding to electromechanical energy conversion and harmonic loss respectively [18], [24].
In the dual d-q model, six variables of the DTP PMSM are transformed into two sets of reference variables [33], which can be expressed in stationary frame as d 1 -q 1 components and d 2 -q 2 components respectively by: where: θ is the electrical angle between d 1 axis and phase A axis, O is a zero matrix of 2 rows × 3 columns.
Stator voltage equation and flux equation in the dual d-q model are expressed respectively as follows: where: R s , L aad , L aaq , and L aal are the stator resistance and inductances. ω is the electric angular frequency. ψ fd is amplitude of rotor flux. Electromagnetic torque equation of the motor with pole pairs p 0 in the dual d-q model can be obtained: The electromagnetic torque equation indicates that the total torque of the DTP PMSM is a superposition of the torque generated by two sets of windings. Due to mutual inductance between the two sets of windings, this kind of torque superposition is not linear, nevertheless, it is obvious that the DTP PMSM has larger torque peak than the three-phase motor, under the same current limits and electrical parameters of d and q axes. The relationship between the VSD variables and the motor variables is given in the following formula [34]: where: Stator voltage equation and flux equation in VSD model are expressed respectively as follows: where Electromagnetic torque equation of the motor with pole pairs p 0 in VSD model can be obtained: In fact, the control method with dual d-q control is very similar to VSD method. The former is easy to be understood and realized, but with more complex voltage decoupling problems. Although the latter decouples the fundamental component and harmonics completely, the physical meaning of variables in VSD control is not so clear compared with the double d-q control. In this article, both of the above two mathematical models are used for control algorithm design: VSD model for steady-state control and double d-q model for fault-tolerant control.
For STP mode, only one set of three-phase windings is working under fault-tolerant control. When phase windings XYZ are cut off, for instance, u d2 = 0 and u q2 = 0 in (2), which is exactly the same as the voltage equation of threephase motor. As is shown in (1), transformation matrix of dual-three phase motor can be directly transformed into that of three-phase motor without extra computation in STP mode. Therefore, the dual d-q model instead of VSD model will be used for further STP algorithm design.

B. CONTROL ALGORITHM DESIGN OF THE DTP PMSM
The control algorithm of the DTP PMSM controller is developed by model-based design (MBD) method in MATLAB/Simulink (MathWorks Corp., USA), and the embedded source code is generated by Simulink Coder toolbox. With this rapid development method, the initialization of hardware and software is automatically completed by Simulink without manual setup, and the main control program is executed in enhanced PWM (ePWM) interrupt of DSP. The control software consists of the following five functional modules: closed-loop control algorithm, operational status display, AD sampling processing, communication module and fault status monitoring.
Field orientation vector control (FOC) is adopted in the closed-loop speed-regulating strategy to control the DTP PMSM to start and reach the target speed. Whereas the operating speed of the oil pump is under the motor rated speed, the flux-weakening control is not involved here. Through orienting the flux on the d-axis and setting d-axis reference current to 0, (4) and (8) can be written respectively as Thus, FOC can control the electromagnetic torque output of the motor by adjusting the q-axis current, in order to minimize the dependence of control system on motor parameters. Fig.10(a) is a principle block diagram of the operation control strategy, which adopts double closed-loop vector control of speed and current.
The reference value of q-axis current (i * q ) is calculated by the automatic speed regulator (ASR) taking the speed error through comparing the given value (n * ) with the sampling value (n) as the input. Detecting the current from the six-phase windings of the motor and using the coordinate transformation, according to (1) and (5), the current used by the regulator are obtained, as shown in Figs. 10 (b) and (c). These values are compared with the given values, and the voltage reference values of each axis (u ref ) are calculated by the respective PI regulators. In Fig. 10 (b), the strategy based on the double d-q model treats the DTP PMSM as two threephase motors, and sets d-axis current reference (i * d1 , i * d2 ) value to zero while q-axis current reference (i * q ) value given by ASR. In Fig.10(c), the harmonic current in the z1-z2 subspace (i z1 , i z2 ) is obtained by VSD, despite independent of the energy conversion of the motor, and will contribute to the copper loss, therefore, the reference value of the harmonic current (i * z1 , i * z2 ) is given to zero, while the current of the d-q subspace is the same as the former method.
The saturation algorithm of the inverter current and voltage is designed. Reference values of d and q axis current (or voltages) given by ASR (or ACR) are q ref and d ref , and the output value of the saturation algorithm q * and d * can be calculated by where and DQ limit is the maximum magnitude of the stator phase current (or voltage) that the inverter can supply to the motor. Finally, by applying SVPWM, the converted PWM drive pulse (S A -S Z ) is inputted into the six-phase VSI to generate six-phase voltage driving the motor to achieve control of the DTP PMSM.
A flowchart for control software is shown in Fig. 11. After hardware initialization and software initialization in order, the software enables interrupt and enters the circular waiting. When ePWM interrupt set as the main control program entry point occurs, the program enters the interrupt subprogram. First, the current, voltage, temperature and rotor position and speed values in real-time are obtained by AD sampling processing and XINTF to determine whether faults (such as over-current, over-voltage, over temperature and other faults) have occurred. The software protection program will be executed if faults occur, sending out protection directive signal to close PWM signals output, and the corresponding fault information will be displayed by fault status monitoring module. Only after troubleshooting can the control program continue. When no fault occurs, the PWM signals of each phase is generated according to the control strategies in Figs.10(b)-(c), output through ePWM of DSP.
On account of two sets of three-phase windings of the DTP PMSM physically independent of each other, the motor can operate in STP mode, which is very necessary for the oil pump safety. Particularly, in case of one set of windings broken down, the other set of windings can continue working to reduce mechanical impact and protect equipment. In addition, when the load is small requiring lower driving power, the DTP PMSM can be switched to STP operation mode driven by one set of three-phase windings, so as to reduce the switching times of IGBTs and the energy loss of the controller.

VI. EXPERIMENTAL VERIFICATION AND DISCUSSIONS
A prototype of the DTP PMSM and the controller was fabricated according to the design scheme, and several experiments were carried out on the experimental setup as shown in Fig. 12. The test DTP PMSM was under the control pattern of rotational speed driven by the DTP PMSM controller. The induction motor working in torque mode could change the load torque of the DTP PMSM.

A. VALIDATION OF STEADY-STATE PERFORMANCE
In order to evaluate the steady-state performance of the motor, the back EMF and current waveforms of the motor at 3000 rpm with no load and rated load were recorded and presented in Figs. 13 (a)-(c). Fig. 13 (a) shows the line back EMF of the designed DTP PMSM at the rated speed. Intuitively, the voltage waveform is sinusoidal and has less harmonics. The P-P (peak to peak) value of the line voltage is 519 V and the RMS is 175 V, which are very close to the electromagnetic analysis results in Fig. 7. The reason for the difference between experiment and simulation results is that the actual parameters of the permanent magnet will change due to the influence of the environment temperature.
We compared the steady-state performance of the two control algorithms, ACR based on the dual d-q model and VSD, under rated conditions (3000 rpm, 19.1 Nm). The stator current waveform and FFT results are shown in Figs. 13 (b)-(c). For the results based on the dual d-q control, the RMS of phase A current (i A ) is 13.01 A and that of phase X (i X ) is 12.38A. As for the VSD control, the RMS of i A is 11.73 A and that of i X is 13.67 A. The experimental results of steady-state current are consistent with the simulation results (RMS 11.5A) obtained by electromagnetic analysis, but the test bench has large mechanical moment of inertia and friction resistance, hence, the larger current is needed to reach the rated torque. Through FFT analysis, the 1 st to 12 th order current are illustrated, and as can be seen in Figs. 13 (b)-(c), the harmonic components are mainly 5 th and 7 th order accounting for a small proportion. In addition, the current THD for VSD control strategy (under 8.0%) is lower than that for dual d-q control strategy (over 9.0%). It can be seen from Fig.10(c) that the harmonic current references (i * z1 , i * z2 ) are set to zero based on decoupled model, while the dual d-q control in Fig.10(b) do not suppress harmonic components because of the coupled currents. Thus, VSD control strategy can reduce current harmonics which is consistent with the experimental results. It should be noted that THD of practical current will never be zero due to the existence of inverter delay and measurement errors.
To further compare the effect of the control algorithms on the DTP PMSM system efficiency, a series of loading tests at the rated speed were performed. A DC power supply provides the input DC power to the inverter, and the voltage (U DC ) as well as the current (I DC ) is measured. According to the experimental results, the overall efficiency of the proposed DTP PMSM system (η) is obtained by (12):  where P out is the mechanical output power of the motor, P in is the input electrical power, and T L and n are the load torque (Nm) and speed (rpm) of the motor respectively. The efficiency and power curves are shown in Fig. 14. At a higher load torque up to the power limit (6 kW), the output power increases and so does the efficiency. The efficiency of the motor implemented with the VSD algorithm is higher than that with dual d-q algorithm, and the maximum efficiency is about 85% achieved at the rated load. Compared with dual d-q algorithm, because of lower current harmonics leading to less harmonic loss [18], the VSD algorithm can increase the fundamental current promoting the efficiency under the same load. In summary, VSD algorithm is superior to the other one for the motor steady-state control to reduce system power loss.

B. PERFORMANCE IN OPERATION MODE SWITCHING
An experiment for switching operation mode was devised to measure the fault-tolerant performance of the motor system. In the experiment, a switching step signal was injected into the algorithm to cut off the voltage output of phase XYZ of the controller to simulate phase failure. The controller was able to automatically switch from DTP mode to STP mode after the failure. With a view to the reduction of maximum torque in STP mode, we selected a typical working condition (2000 rpm, 10 Nm) instead of the rated condition to prevent the motor from over current. Fig. 15 (a) shows the stator current waveforms during switching as well as the fluctuation of the speed and torque. From the phase current point of view, i X goes to zero after switching, but the other set of windings (ABC) is unable to provide enough torque immediately to balance with the load, so the speed drops rapidly. After the speed decreasing, the ASR adjusts the current reference rapidly to increase the electromagnetic torque of the motor to exceed the load torque, and the motor starts to accelerate. When the speed returns to the target, i A is reduced and the electromagnetic torque is decreased as well to reach the load torque. In the process, the motor speed is in a 6% fluctuation (120 rpm) and the torque is in an 8.6% fluctuation (0.86 Nm). And it takes 2.6 seconds for the controller to track the speed reference under the given load torque, which show a good performance.
Before switching, i A and i X are about 6.0 A (RMS) in the steady state as the controller is operating in DTP mode as shown in Fig. 15 (b). After switching to STP mode, the steady-state value of i A increases to 12.1 A that is twice as large as before as seen in Fig.15 (c). It can be seen that, the winding fault affects the phase current significantly, since the electromagnetic torque of the DTP PMSM is generated by two sets of windings according to (4). Assuming that the current RMS value of the two sets of windings is completely equal, then i q1 = i q2 , i d1 = i d2 = 0. And the current of STP mode will be doubled so that it can achieve the same T e as that of DTP mode. Therefore, in the process of the fault-tolerant control strategy design, the phase current must be reasonably restricted to avoid over-current fault.

VII. CONCLUSION
In this article, a 6kW DTP PMSM system has been developed for driving auxiliary systems in eTransportation application. Digital collaborative develop process was proposed with the support of multiple software tools. The DTP PMSM, the six-phase two-level inverter and the control algorithm were synergistically designed and iteratively optimized. Complete development details including motor scheme optimizing, electromagnetic analysis, modeling and control design of the system were reported. The prototype testing and experimental results indicate that the 6kW DTP PMSM system is qualified for the actual application requirements, and the proposed development process has practical value in engineering.
Since the control strategy using the VSD algorithm in steady-state control and the dual d-q control algorithm in fault tolerance can improve the performance with both superior steady-state control and fault tolerance, a robust switching algorithm realizing smooth excessive will be studied in our future work. DTP PMSMs are gaining wide attention in the field of electrified transportation. This article makes some preliminary attempts in their high-efficiency development, and more practical work aiming at eTransportation demands will be launched in the future.