3D Printing as a Technology Enabler for Electrical Machines: Manufacturing and Testing of a Salient Pole Rotor for SRM

Among the various technology enablers for modern electrical machines, additive manufacturing plays a key role. The advantage of having a precise control of the shape of ferromagnetic structures, whilst achieving good electromagnetic performance, fits well with the design requirements of rotating electrical machines. To a certain extent, some of the physical properties of the material can be “tuned”, allowing for quick trade-off studies (i.e., prototyping), as opposed to conventional manufacturing techniques. Despite being considered an enabling technology, 3D printing of soft magnetic materials for electric motors is still at an embryonic stage. This work, thus, aims in providing an initial proof of concept. For the purpose, a switched reluctance machine is chosen as a case study. Its rotor core is additively manufactured through selective laser melting. Its performances are compared to those of an identical commercial motor featuring a laminated rotor core, via in-depth experimental tests. Initial results show that the 3D printed machine can actually develop the rated power, but with an efficiency reduction.


I. INTRODUCTION
In the last two decades, the development and commercialization of high performance magnetic and insulating materials, together with the introduction of unconventional cooling systems, have allowed electrical machines (EMs) to achieve outstanding power density and efficiency levels [1]. The ever-increasing request for high power and low weight EMs, mainly pushed by the transportation industry, is breaking new grounds for socalled enabling technologies. From a system level point of view, the widespread development of fast-switching power electronics converters has contributed in achieving very-high speed machines, with an excellent power to weight ratio [2]. At the same time, the concept of physics of failure [3][4][5][6][7], when applied to EMs' insulation systems and power electronics converters, is seen as a viable methodology for better exploiting the thermal and electrical capabilities of electrical insulating systems, leading to compact designs [8].
The scientific community [9,10] alongside with a number of international bodies / agencies [11][12][13], seem to agree that future electrification roadmaps can only be achieved by relying on a number of "unconventional" technology enablers including superconductivity [14], high energy density electrochemical storage devices [15], additive manufacturing etc.. Among these, metal additive manufacturing (AM) and more specifically selective laser melting (SLM), presents undoubtedly the highest technology readiness level (TRL), when applied to EM design [16][17][18].
AM allows to precisely manufacture virtually any object with high geometrical complexity, by progressively depositing single layers of material, with the limits thus being a) the maximum build for a single part (i.e., 3D printer "size") and b) the minimum resolution. As opposed to more conventional manufacturing techniques, such as laser / plasma cutting and other machining methods, AM is considerably more sustainable, as the wasted material during manufacturing is minimum [17]. Another perceived advantage is the possibility for rapid prototyping, before mass-production. Indeed, AM does not require any stamping (i.e., pressing) die for manufacturing a new part / component.
The application of AM to EM design is still at an early stage, although various recent works have proposed 3D printed solutions for both magnetic [17,18] and nonmagnetic EM parts [16,19]. A suitable manufacturing method for 3D printing metallic part is the so-called SLM [20], which has also recently been used for processing soft magnetic materials. In SLM, various layers are stacked one on top of the other by melting a metallic powder through a high-intensity laser [21]. Such procedure is followed by a rapid melting / cooling process leading to a fine grain microstructure [22].
In [18], the viability of the SLM technique for 3D printing the rotor of a synchronous reluctance machine is proven. Various manufacturing details, as well as the material's magnetization curve, are provided, without, however, including experimental results on the actual machine. In [23], finite element (FE) simulations are used for quantifying the efficiency improvement arising from AM applied to a synchronous reluctance machine. In [24], a 3D printed rotor for a permanent magnet machine is developed. Its magnetic anisotropy is increased thanks to AM, leading to an improved senseless control capability.
This paper aims at providing a comprehensive, experiment-based analysis of an additively manufactured rotor for a switched reluctance machine (SRM). For doing so, a commercial machine, with laminated rotor core is _____________________________________ initially selected. Its rotor is replaced with a 3D printed one, whose composition features a 5%w.t. silicon content. The material's mechanical and magnetic properties are initially characterized, before moving to the rotor manufacturing stage. Once the 3D printed rotor is manufactured and integrated in the commercial SRM stator, its electromagnetic performance is evaluated though experimental tests and apprised to the "benchmark" machine. As a main result it is found that the 3D printed SRM is able to actually develop the rated torque despite an efficiency decrement, caused by the increased losses in the solid rotor structure.

II. COMMERCIAL SRM
This Section gives a brief introduction on the theory and operations of SRMs. The benchmark machine is then presented, by providing its main geometrical dimensions and properties.

A. Introduction to SRMs
SRMs have been known since the nineteenth century [27], but their requirement for non-standard control, as opposed to e.g. DC or induction machines, has prevented their widespread adoption/use. With the availability of modern power electronics converters, there is renewed interest in SRMs and they are being proposed for various applications, including automotive [28] and aerospace [29].
SRMs are characterized by salient poles on both the stator and rotor. When a stator phase is excited, torque is produced by the tendency of the rotor aligning to the minimum reluctance position. Therefore, the stator phases are switched "on and off" according to the rotor position, so that a continuous motion is obtained [30]. For doing so, a suitable power electronics converter and control algorithm are necessary. For a 3-phase SRM, the basic converter topology counts 6 switches and as many freewheeling diodes, although their configuration differs from that for conventional AC machines. The main advantages of SRMs are the absence of permanent magnets and the ruggedness and robust structure [31]. Thus, they are particularly suited for harsh operating environments.

B. The benchmark machine
The benchmark motor is a commercial § 1.1 kW rated power, three-phase SRM with a 12/8 configuration (i.e., 12 stator teeth and 8 rotor teeth). Its main geometrical dimensions and parameters are listed in Table I. The cross-sectional view of the machine and its winding layout are shown in Fig. 1. Each phase has a total of 4 coils, and each coils features 220 turns. A standard, commercial converter implementing a soft-switching strategy is used for controlling the machine. The controller receives the rotor position feedback from a simple optical position sensor based on three photodiodes and three light emitting devices. Based on the rotor position, the control algorithm generates and sends the control signals to the switching devices. A digital hysteresis control algorithm is used for maintaining a flat-topped current when the SRM operates at low speed (i.e., the phases voltage is chopped for maintaining the current within the hysteresis band). The turn-on and turn-off angles of each phase are automatically selected, according to the instantaneous torque request, for minimizing the required current (i.e., maximum torque per ampere algorithm).

III. MATERIAL DEVELOPMENT AND ROTOR PRINTING
Silicon-based steels (i.e. Fe-Si) are conventionally adopted for manufacturing standard soft magnetic materials for EM laminations [32]. The higher silicon content helps in reducing the electrical conductivity of the alloy, and thus acts in directly reducing eddy-current losses. Commercial machines, including the case study SRM, adopt silicon content of approximately 3%w.t., while in high performance aerospace motors cores, up-to c.a. 6% silicon content in weight can be found. In this study a silicon steel powder with 5% silicon content (Fe-5.0%w.t. Si) is produced, by mixing a pre-alloyed high silicon steel powder with high purity Fe powder. The powder blend results in an average particle diameter of 36.2ȝm.
The SLM process employs a Renishaw AM125 SLM machine equipped with a 200 W D-Series redPOWER ytterbium fibre continuous wavelength laser. Carrying out the process in an argon atmosphere minimizes oxidation. The configuration of the build chamber is shown in Fig. 2. Samples with different shapes are produced for determining physical, magnetic and mechanical properties of the developed material. In particular, cylindrical samples are used for determining the magnetic performance of the material with the use of a vibrating sample magnetometer (VSM), whilst dog-bones are employed for performing tensile tests. The resulting magnetization curve is reported in Fig. 3 where the magnetic moment per unit volume M is plotted against the magnetic field magnitude.  After having tuned and optimized the processing parameters of the AM process, the SRM rotor is printed. Its geometry is identical to that of the benchmark EM, although its construction is solid. Due to build height restrictions in the SLM machine, the AM motor is axially divided in three blocks that are shown in Fig. 4. Clearly, the larger the number of axial segments, the lower the eddy current losses in the magnetic structure, since each segment is electrically insulated through a non-conductive coating.
The three segments of the SRM rotor were subjected to annealing at 1100 o C for 1 hour and then they were removed from the substrates via Electrical Discharge Machining (EDM). The SRM shaft has been manufactured via conventional machining techniques and has been assembled on it as shown in Fig. 5. Finally, high precision bearings have also been fitted on the shaft.

IV. FE ANALYSIS
A 2D FE analysis has been carried out for the benchmark machine. Such a preliminary analysis serves the purpose of a) determining the iron loss distribution and b) checking if there is any rotational speed characterized by excessive torque ripple. Point a) is necessary for establishing the rotor's share of iron losses, so that the results from the experimental tests can be better analyzed. In other words, the knowledge of the rotor loss magnitude for the benchmark machine alongside with the measured total losses for the 3D printed machine, allows to quantify the rotor loss variation (i.e., laminated/benchmark vs 3D printed) according to (1), where Prot are the rotor losses, Pel is the electrical input power, PJoule are the winding Joule losses, Pstat are the stator iron losses, T is the shaft torque and Ȧm is the mechanical speed.
Since the EM operates at relatively low speed, windage losses are neglected. Point b) on the other hand, needs to be verified mainly for practical and safety purposes. Indeed, excessive torque ripple might damage mechanical couplings, and thus should be avoided. Fig. 6 shows the flux map and magnetic field lines of the laminated machine operating at base speed and delivering the rated (average) torque of 18.7 Nm, whilst Fig. 7 reports the iron loss distribution (per unit mass), under the same operating condition. Fig. 6. Flux density map and magnetic field lines for the benchmark/laminated SRM operating at rated condition.
Despite the visible loss concentration in the rotor teeth tips, the rotor losses share over the total iron losses is less than 20%, as reported in Table II, where the loss results obtained from the FE simulation are summarized. These include also the Joule losses, which are one order of magnitude higher than iron losses. A speed-sweep analysis has been carried out for determining if there is any particular operating speed range in which the SRM develops excessive torque ripple. As a result it has been found that within the range 800 -1200 rpm an anomalous ripple is detected. This is visible in Fig. 8, where the instantaneous torque developed at 1000 rpm is plotted. Its torque ripple is higher than 120%. Therefore, experimental tests with the machine operating within the aforementioned speed range should be avoided.

V. EXPERIMENTAL TESTS
This Section describes the experimental test procedure and reports the main results for both the benchmark and the 3D printed SRMs.

A. Test-bed description
In order to test the electromechanical performances of the manufactured 3D printed SRM and to compare them with the benchmark machine, the test-bed shown in Fig. 9 has been used. The description of each component / instrument is reported in Table III. The SRM is flange-mounted to an "L" plate. Its shaft is mechanically coupled through a Magtrol ® torquemeter to a 70 kW, variable-speed Oswald ® induction motor. The load induction machine is torque controlled via an Emerson Unidrive ® three-phase inverter.
Phase current and voltage are measured through a halleffect current clamp and a differential probe respectively. These are connected to a Lecroy ® Wavetouch oscilloscope for instantaneous time-domain analysis and logging. In addition, the instantaneous electric power analysis is carried out using a Rode & Shwarz power analyser, which records and processes the phase current and voltage. Fig. 9. Experimental test-bed (the description is provided in Table III). Fig. 9 Description

B. Tests results
Both machines (i.e., benchmark SRM and 3D printed one) have been tested at base speed delivering the maximum torque and at 2.5 times rated speed. At the rated speed (i.e., 600 rpm), the SRM fitted with the 3D printed rotor produces 13% less torque compared to the benchmark SRM, as can be seen in Table IV, where the measured electrical and mechanical quantities at rated speed are tabulated. When the operating speed is increased up to 1500 rpm, the 3D SRM's performance improves both in terms of maximum mechanical power as well as in terms of efficiency, as can be observed in Table V, where the electric and mechanical power quantities measured at 1500 rpm and maximum torque are listed. The next sub-section will provide a more detailed discussion on the obtained results.

C. Discussion on the obtained results
In order to provide a more insightful discussion on the experimental results, it is worthy reporting the measured electric quantities in the time domain for both SRMs, as done in Figs. 10 to 13. In particular, in Figs. 10 and 11 the instantaneous electric quantities for the benchmark SRM running at base speed are plotted. Similarly, Figs. 12 and 13 report the measured electric quantities for the SRM with the 3D printed rotor operating at 600 rpm.  A comment can be made on the voltage waveforms for both machines. For the benchmark EM (i.e., Fig. 10), it is possible to observe the soft-chopping action, necessary to maintain the current within the hysteresis band. On the contrary, a flat-topped voltage waveform is recorded for the 3D printed rotor SRM (i.e., Fig. 12). This indicates that the power electronics converter has actually reached its current limit. Such an observation is also confirmed by the timeperiod of current / voltage waveforms. Indeed, for the benchmark EM, two electrical periods last 25ms, whilst they last c.a. 27ms for the 3D printed rotor SRM. Accordingly, by calculating the mechanical speed as in (2), where fel is the electrical frequency, it is possible to verify that the 3D SRM is actually slowing down by c.a. 45 rpm.
Further considerations can be made relying on the instantaneous power waveforms (Figs. 11 and 13), as well as the average power values reported in Tables IV and V. First of all, by calculating the ratio between active and apparent power, it is possible to obtain the power factor. At base speed, the latter, is equal to 0.52 for both machines. This interesting result indicates that, in principle, the volt-ampere rating of the power converter for a 3D printed rotor SRM can be identical to a standard laminated one (with the same rated power).  However, for the SRM at hand, because of the increased rotor losses caused by the solid rotor structure, the power converter results slightly underrated. In fact, the RMS input power for the benchmark SRM is 3.83 kVA, whilst it is equal to 4.24 kVA for the 3D printed rotor SRM. The larger demanded apparent power is needed for partially compensating the efficiency reduction, which can be calculated as the ratio between average mechanical power and average electric power. In particular, the efficiency reduction (at base speed) is c.a. 22 %.
As the operating speed is increased to 1500, the efficiency rises to 83% for the benchmark SRM and 65 % for the 3D printed one. Such a conclusion could be also reached by calculating the rotor loss share for the 3D printed machine as in (1). This is quite high when the machine operates at 600 rpm, as it peaks at c.a. 500 W, whilst is negligible for the benchmark machine. It keep being quite elevated, even when the machine operates at higher speed, although, in this case, the overall efficiency increases because of the Joule loss reduction (i.e., the RMS phase current decreases as the speed increases for constant / decreasing power operations).

VI. CONCLUSIONS
Additive manufacturing is considered one of the technology enablers for achieving near-future power density targets in electric drives. Although there is a wide ongoing research on the topic, few are the examples of operational 3D printed active parts for electrical machines. The main objective of this work is to prove the concept that additive manufacturing for electric motors is actually viable and feasible from a manufacturing and technological point of view. Therefore, the focus of the paper is mainly in demonstrating how from the raw materials (i.e., metallic powder), it is possible to use AM/SLM in order to manufacture and spin an actual rotor for an electrical machine. Clearly, because the rotor is a relatively simple solid block, its losses are considerably higher with respect to a standard laminated machine. However, thanks to the strengths of additive manufacturing, a more lightweight complex geometrical rotor can be built for optimizing the overall machine's efficiency. Thus, future work will investigate through a finite element optimization, how the magnetic structure can be modified in order to mitigate the effect of eddy current losses.
VII. REFERENCES

VIII. BIOGRAPHIES
Leonidas Gargalis is a Research Assistant at the Centre for Additive Manufacturing (CfAM) working towards the completion of his PhD. After he received a Marie Curie Doctoral Fellowship in 2016, he joined CfAM and the Institute for Aerospace Technology at the University of Nottingham. His research has focused on developing soft magnetic alloys using selective laser melting for the design and fabrication of rotating electrical machines. His research interests in metal AM span between process optimisation, such as parametric studies for materials qualification, and characterisation of 3D printed materials such as the microstructural, mechanical and electromagnetic properties.