A Novel Concept of Ribless Synchronous Reluctance Motor for Enhanced Torque Capability

The rotor structure of synchronous reluctance machines (SynRel) is conventionally retained mechanically by iron ribs. In this paper, a novel structure for high-speed synchronous reluctance rotor is presented. The novelty of this paper is the proof of a concept of SynRel machine without iron ribs. Structurally, the rotor iron lamination segments are embedded in an adhesive resin material with high temperature resistance and mechanical strength. Three four-pole SynRel machines have been designed with the target of improving motor torque, and compared for different ribs configuration. It has been shown that the proposed motor performs enhanced torque, power factor, and efficiency with respect to conventional SynRel with iron ribs. An extensive sensitivity analysis of the ribless rotor geometry is carried out, followed by both mechanical analysis and experimental over speed test to guarantee its robustness above the operating speed range. The manufacturing procedure of this novel rotor is introduced. Finally, the experimental results on both SynRel prototypes are presented, showing the increase in torque, power factor, and efficiency of the proposed solution. This paper is a first step toward the definition of a viable and novel solution of SynRel machines with improved performance.


I. INTRODUCTION
S YNCHRONOUS Reluctance (SynRel) machines have gained increasing interests in several applications thanks to their advantages with respect to induction machines [1]- [4]. It has been shown that they exhibit good torque capability, Manuscript  wide operating speed range and high efficiency because of the rotor anisotropy [5]. Their use is especially beneficial when a low-cost machine is required and if the use of permanent magnets (PMs) is not desired. Due to the absence of PMs, their application field is limited to low power density range, as they cannot compete with Permanent Magnet Synchronous Machines (PMSM) [6]. A number of research activities have demonstrated that with an optimal rotor design or skewing techniques, the torque ripple can be significantly reduced [7]- [10]. Another aspect that can be considered a downside of Syn-Rel motors is the relatively poor power factor [11]. The low power factor is depending on two main factors: the absence of any excitation in the rotor structure and the presence of the iron ribs, which are normally necessary to guarantee mechanical robustness of the rotor against centrifugal forces [12]. In the past other types of rotors like axially laminated anisotropy (ALA) rotors have been prototyped and tested. However, due to their manufacturing complexity and high rotor iron loss they have hardly found any further development [13]. In order to exhibit a good torque, the conventional SynRel is characterized by a small airgap and a high anisotropic rotor as shown in Fig. 1. Several rotor flux barriers force the flux to flow through given iron paths. The function of iron ribs is purely structural. However, as part of the rotor magnetic circuit, they affect the electromagnetic performance of the machine because they act as a short circuit for the flux between the iron islands. In order to saturate those iron ribs, a "magnetizing" current from the stator winding is required. As a consequence, this current is not contributing to the torque production and it results in a generation of flux leakage that is closing through the rotor iron ribs [14]. One of the alternative options is including PMs inset within the flux barriers for the same purpose of saturating the iron ribs [15], [16].
For those applications where the PMs are not allowed, the only option to enhance torque production is to optimise the rotor geometry for the highest saliency ratio and with the minimum iron ribs thickness [17]. The latter is defined by the size of the rotor, mass and rotational speed to guarantee a safety margin for mechanical robustness. In most of the designs, the minimum thickness of the ribs is limited not only by mechanical reasons but by the minimum manufacturing tolerances [18]. Normally, ribs thickness below 0.5mm are not recommended no matter whether the laminations are manufactured by punching, laser or electro erosion cutting methods [6].
In this work, a novel concept of ribless machine is proposed aiming to demonstrate that an ideal SynRel machine designed without iron ribs could lead to an improved torque, power factor and efficiency with respect to conventional SynRel motors. The goal has been achieved through the design and manufacturing of a novel rotor structure filled with an epoxy resin.
This paper is organised in four main parts. In section II, the geometry of three different rotors is introduced. Each machine is analysed via Finite Element (FE) demonstrating the advantages of the novel concept. Section III presents a detailed torque sensitivity analysis of the novel rotor considering different parameters. In section IV, a structural analysis is carried out to understand the stresses on the rotor structure. These are supported by over speed tests. Finally, the experimental results are shown in section V to prove the feasibility and potential of the proposed concept.

A. Reference stator
In order to guarantee the highest saliency ratio an Integral Slot Distributed Winding (ISDW) has been chosen. Since SynRel machines present a high rotor anisotropy, it has been demonstrated that this winding arrangement is the most effective solution compared to Fractional Slot Concentrated winding (FSCW) [19], [20]. The stator and winding configurations of the three machines refer to the benchmark SynRel in [6]. They share the same stator with ISDW as well as the same air-gap length 0.35 mm. A portion of the stator geometry is shown in Fig. 2. The parameters shown in Fig. 2 are identified in Table I

B. Reference rotor geometries
All rotors have four poles and the barrier distribution is symmetrical. Fig. 3 and Fig. 4 are showing the rotors of the three machines considered and compared in this section: • M1: with both tangential and radial ribs; • M2: with only tangential ribs; • M3: novel resin-aided ribless rotor. It is characterized by two flux barriers per pole and one cutoff region. All the edges of flux barriers are drawn based on the N. E. Joukowski airfoil potential function [21] according to the 'natural flux lines' method proposed in [5]. Two main differences can be noticed between the conventional rotors (M1 and M2) with respect to the M3. The barriers are filled with resin and the tangential and radial iron ribs are completely removed thanks to the structural support of the introduced material.  In addition, the first barrier of M3 has been replaced by a cut-off region, which will also serve as a slot for the resin (see Fig. 4).
The main geometrical parameters considered are summarised below: • D re and D ri are the rotor outer and inner Diameters, respectively and L stk is the lamination stack length; • ϑ b1 , ϑ b2 , and ϑ b3 are flux barrier position angles; • W r1 , W r2 and W r3 are the radial ribs thickness in M1; • W tan is tangential rib thickness in M1 and M2; • ∆b2 and ∆b3 are two barrier end opening angles in M3; • C cut is the length from the rotor center to the center of cut-off region on q-axis in M3.
In Table II, the dimensions of the three machines are summarised. The optimized parameters of M1, M2 and M3 are obtained through a genetic algorithm (MOGA-II) optimization via modeFrontier [22], which is deeply described in [16] and [23]. The two objectives are to maximize the average torque and to minimize the torque ripple. Among the Pareto front solutions of M3, shown in Fig   the epoxy resin in both barriers and cut-off regions is holding the iron lamination segments to form a mechanically robust structure. A 3D rotor sketch is shown in Fig. 6 with a clearer view of the two resin end-caps. All the transversally laminated iron islands are held together by the adhesive epoxy resin. Since SynRel machines saturation is strongly dependent by the thickness of the flux barriers, it is common to define an insulation coefficient k air [24] which is a ratio between L air (length of air along q-axis for conventional M1 and M2, L resin in M3 case) and D re -D ri (difference between rotor outer and inner diameters). Each flux barrier's thickness and the cut-off region of M3, which is modified by a cubic function depending on two parameters (ϑ b1 , C cut ), have been analysed by means of finite element simulations in order to obtain a desired saturation level in the iron magnetic paths.  The thickness of the iron bridges of M1 and M2 have been chosen according to the sizing method presented in [25] where a detailed analysis on the SynRel speed capability is given. For example, referring to Fig. 1, both the radial and tangential iron ribs have to guarantee a robust structure and resist to the centrifugal forces insisting on rotor parts. In general, the tangential iron ribs are less mechanically stressed with respect to the radial ones [26]. However, if the central ribs are removed (M2), it will induce a higher stress to the tangential ribs. Consequently, the maximum rotating speed of the machine is decreased (in this case 8000rpm with a safety factor of 1.5).

C. Motor characteristics comparison
In this section, the three SynRel motors (M1, M2 and M3), presented in section II-B, with optimized rotor geometry for torque production have been selected for comparison.
The static torque as a function of the current angle has been determined by means of FE simulations (with MagNet 2D software) for all motors. The results are shown in Fig. 7 under different current loadings. It can be noted that for the same current the average torque of M3 is the largest with respect to M1 and M2. This confirms that the absence of iron ribs can significantly improve the torque performance. Always referring to Fig. 7, for the same current loading the MTPA locus of M3 is given by lower current angles as highlighted with the continuous line. To give a clearer idea of what torque improvement can be ideally achieved by removing the iron ribs, a comparison is given in Fig. 8 for the same phase current of 4.62Apk. It is worth noticing that M3 with the novel rotor structure proposed can achieve an average torque of 1.395Nm, which is 21.2% and 10% higher with respect to M1 and M2 correspondingly.
A summary of the main output characteristics of the three motors, at the rated speed of 6000rpm, is given in Table III. This clearly shows that for the same torque demand, i.e. 1.15Nm, the novel M3 present some benefits: • the phase peak current is reduced by about 11.1% compared to the original 4.62Apk in M1; This type of machine presents only reluctance torque. Its production depends on the d and q axis inductance difference (L d −L q ), or more commonly saliency ratio (L d /L q ), which are affected by saturation and cross saturation effects. Therefore, the analysis of the inductance behaviour, of M1 and M3 motors, is offered in Fig. 9, Fig. 10 and Fig. 11. For both machines, the L q inductances are presented as a function of I q , for different I d values in Fig. 9 and Fig. 10 respectively. The cross saturation of M3 is relatively small, as the variation of L q is less than 1mH when the I d is varying from 0Apk to 6Apk, for I q current value, as shown in Fig. 10. However, the variation of L q in M1 is relatively large especially at low I q as shown in Fig. 9. In M1, the L q decreases from 48mH to 25mH when I d varies from 0Apk to 6Apk, for an I q of 0.5Apk. As a result, it is clear that M3 presents less cross saturation effect with respect to M1. This is due to the poor saturation of the iron ribs in M1 at low current condition. The poor saturation  also results in relatively large L q in M1 compared to L q in M3 at low currents. The L q in M1 is almost comparable to its L d value as shown in Fig. 11 for currents below 1Apk whereas the L q in M3 appears almost constant in a small value around 8mH with respect to both I d and I q . In M3, L q is much smaller compared to M1 and the inductance difference is particularly large under low current load condition. It can be concluded that the enhanced torque of M3 is determined by the difference of L q , which is represented by the area highlighted in Fig. 11, between inductance profiles of M1 and M3, respectively. M1 and M3 are compared in detail since they represent the benchmark machine and the proposed novel rotor respectively. A summary of their main parameters is given in Table IV. From the simulation results reported, M3 is capable of delivering higher average torque, power, efficiency and power factor for the same input voltage thanks to the novel rotor with higher saliency ratio.

III. TORQUE SENSITIVITY ANALYSIS OF THE NOVEL RIBLESS SYNREL
All SynRel motors, if are not well designed, can present a very high torque ripple [10], which is due to the interaction between the spatial harmonics of Magnetic Motive Force (MMF) from the stator currents and the rotor geometry [7]. As a result, besides investigating the motor behaviour, it is also essential to establish the influence of the various design parameters of the novel ribless SynRel rotor. Before starting the torque sensitivity analysis, it is necessary to recall the optimized parameters of M3 shown in Table. II. The motor has an average torque of 1.35Nm and 16% torque ripple at the rated current condition (in MTPA mode). The torque behaviour is calculated on 60 electrical degrees, corresponding to a complete cycle for a three-phase machine and it is computed by means of Maxwell stress tensor.
In Fig. 12, the average torque, expressed in pu, is represented as a function of the barrier position angles ϑ b2 and ϑ b3 . The optimal solution is represented by the red dot (corresponding to ϑ b3 =39.18 deg, ϑ b2 =27.36 deg). It appears that the average torque remains almost the same (the variation is less than 2%).   Fig. 13. Average torque as a function of Ccut and ϑ b1 (M3).
The influence of the cut-off region has also been taken into account. As shown in Fig. 13, the average torque is not affected by the angle ϑ b1 and for each C cut value it remains almost constant. Besides, the position of cut-off region on q-axis C cut has little impact on average torque (only 1.4% difference for a variation of 2mm).
Similar analysis is conducted on the torque ripple to understand the effects of the geometry. As shown in Fig. 14, the torque ripple is very sensitive with respect to ϑ b3 variation. When ϑ b3 varies from the optimal value 39.18 deg to 40 deg, with ϑ b2 equal to 27.36 deg, the torque ripple increases from 16% to more than 48%. Meanwhile, ϑ b2 has less influence on torque ripple compared to ϑ b3 . Moving the angle from 27.36 deg to 28 deg, with ϑ b3 constant to 39.18 deg, the torque ripple increases from 16% to about 26%.
As the novel rotor is completely ribless, it is worth to consider the torque ripple sensitivity of barrier end opening angles ∆b2 and ∆b3. In Fig. 15, it can be seen that the torque ripple is optimal when the two angles are kept as a combination of relatively small values (i.e. 0.72 deg for ∆b2 and 0. 8      barrier position angles ϑ b2 and ϑ b3 , they are still essential for torque ripple reduction. Similar torque ripple sensitivity analysis can be done on C cut and ϑ b1 , as shown in Fig. 16. Starting from the optimal solution, it is interesting to notice that the torque ripple decreases as C cut increases, with ϑ b1 kept constant to its optimal value of 11.26 deg. It exhibits an opposite behaviour with respect to the average torque. To conclude, the above sensitivity analysis confirms the importance of selecting the rotor parameters in order to reduce the torque ripple. Also, its oscillations are more sensitive to the geometry parameters with respect to the average torque. Similar behaviour has been identified on the conventional SynRel motor.

IV. STRUCTURAL ANALYSIS AND MANUFACTURABILITY
Conventional SynRel motor rotors have iron ribs (tangential and central bridges) to maintain the mechanical stability against centrifugal force. The novel rotor instead only relies on the adhesive resin and its mechanical properties to hold the structure. In this section a summary of the structural and manufacturing considerations applied is given.

A. Structural Analysis
Since the rotor maximum speed is designed as 16000rpm, a mechanical finite element simulation is carried out to verify its integrity against centrifugal forces. The analysis is carried out by simulating one rotor symmetry with ANSYS software. The injected material is an industrial grade epoxy resin with high adhesive properties, good mechanical strength and temperature resistance when it is fully cured. Basically, it is an epoxy material with two components with the mechanical properties shown in Table V.
Since there is no load along the rotor axis, its iron parts are modeled as a consistent isotropic material along the axis to reduce the model complexity. All contacts between resin and iron are set as bonded, meaning that sliding or separation between two materials is not allowed. It is necessary to determine the Von-Mises stress on the rotor epoxy regions. As shown in Fig. 17, the small barrier end opening regions is the one showing highest Von-Mises stress. However, its value is 16.5MPa at 16000rpm, which is less than half of the material limit. The deformation status of the rotor at 16000rpm is displayed in Fig. 18. The maximum deformation occurs in correspondence to the cut-off region and is less than 0.01mm, which is considered acceptable.
A number of over-speed tests have been carried out to ensure that the rotor can maintain mechanical integrity. The rig for speed test is shown in Fig. 19. In order to simulate critical temperature scenarios, the rotor, together with the bearing supports and base shown in Fig. 19, is first heated up in the oven for one hour to a certain temperature. After that, the rotor diameter is measured as D2 and compared to the initial rotor diameter D1 to check thermal expansion. Then the rotor is placed onto the speed test rig to spin for 30 minutes, driven by the master motor through the belt. At the end of the test, the rotor diameter is measured again as D3 to check if any deformation has occurred. The tests have been performed at 60 • C, 80 • C, 100 • C and 120 • C. The qualitative results have shown that the deformations of the rotor are always less than 0.01mm compared to D1, which proves that the prototype is capable of operating at temperature up to 120 • C. Additionally, there is a 0.35mm gap between the rotor outer surface and the bottom plate resembling the air gap in real motor, so that any deformation will be observed if the deformation of the rotor becomes too large during the tests. However, no destructive experiments have been performed so far to identify the exact breaking point. The limitations of the materials, from both thermal and mechanical point of view need to be further explored. However, to the purpose of validating the feasibility of the novel concept, proposed in this paper, a number of experimental results are offered in section V.

B. Manufacturability
The manufacturing process of the M3 rotor protype is described as below: • Step 1: A rotor core with ribs (initial laminations with ribs in Fig. 20(a)) is manufactured with conventional process by wire cutting laminations bolted together. The width of the ribs is designed to be thin and consistent to the diameter of the rotor, which is marginally larger than the final product to provide allowance for follow-up machining.  part2) used to positioning the rotor at the center and prevent resin from entering the shaft hole in the rotor core. The resin is poured into the mold first. Then the rotor core is pressed down to the bottom of the mold, pushing the resin to fill up the flux barriers and the cavity between the rotor core and the mold. After the resin cured, the rotor core and the resin become a whole part and is removed from the mold by a hydraulic press for further processing as shown in Fig. 20(b). • Step 3: The outer layer of rotor is machined away to remove the ribs and excess resin so that final dimensions are achieved. The iron islands are mechanically separated while being held together solely by resin. A picture of the final manufacutred novel rotor with shaft is shown in Fig. 21.

V. EXPERIMENTAL VALIDATION
In order to verify the electromagnetic torque characteristics of the proposed SynRel, M1 and M3 have been manufactured and tested. The experimental measurements are performed on the machine mechanically coupled with a Magtrol setup (Dynamometer, Hysteresis Brake and Torque Sensor). The Magtrol controller has a minimum torque resolution of 0.001Nm. The motor is driven by a smart power module STK5Q4U362J and controlled by the 32-bit Toshiba TMPM375 microcontroller. The rotor position is measured by means of an encoder.  A picture of the Magtrol rig is shown in Fig. 22, highlighting the main test rig components. The experimental torque results are shown in Fig. 23. For both prototypes, the torque has been measured for different current values (from 2Apk to 6Apk). The measurements have been compared with the simulation results carried out by means of FEA. The torque versus current characteristics are showing a very good agreement with FEA results for both machines. Considering the current 4.6Apk, the average torque of M3 is 1.38Nm, which is 22.1% larger with respect to the 1.13Nm in M1.
A summary of the torque increment given by the novel motor design is given in Table VI. It is clearly showing that the torque is enhanced for any current loading. Furthermore, it confirms the prediction estimated by means of finite element simulations that has been shown Fig. 8, giving an improvement of the same order (about 22.1%). This means that the proposed motor can achieve the same torque with a reduced current   (more than 10% reduction), with a consequent improvement of the copper losses. Further experimental results are presented in Fig. 24 and Fig. 25. The first is showing the experimental voltages captured at 6000rpm at different current values. The results are in good agreement with the ones carried out via finite element and their difference is within 2%. In Fig. 24(b), the phase peak voltage V ph and V q of motor M3 have similar values since V d is relatively small. The measured phase voltage is 189Vpk in M1 and 188.3Vpk in M3 at rated condition.
Power factor and efficiency are shown in Fig. 25 for both M1 and M3, at the rated speed of 6000rpm. FEA predictions are validated with a difference within 3% and 1% for power factor and efficiency respectively. Compared to M1, for different set points of current (2Apk, 3Apk and 4Apk), both power factor and efficiency are showing higher values in motor M3.

VI. CONCLUSIONS
This paper has proposed a novel structural and electromagnetic concept of SynRel motor completely without iron ribs (M3). A detailed analysis of the novel SynRel on torque production is carried out, showing the impact level of each rotor geometry parameter and an optimal range for opening the rotor barrier angle. The analysis shows how the dq-axis inductances are changing with respect to the conventional SynRel motor (M1). First, the simulations results showing the advantages of the novel SynRel motor proposed are analysed.
The high speed capability of the novel SynRel is verified both via mechanical FEA analysis followed by an over speed validation test at high temperature. The manufacturing process has been described in detail. Finally, a number of experimental results have been shown and it has been validated that the proposed novel rotor design can lead to improve torque, power factor and efficiency. These are justified by the enhanced saliency ratio and minimised cross saturation effects given by the absence of the iron ribs. The results are showing an enhanced capability of novel SynRel for any current loading compared to conventional SynRel.