Improved Thermal Modeling and Experimental Validation of Oil-Flooded High-Performance Machines With Slot-Channel Cooling

Thermal management is often considered a bottleneck in the pursuit of the next-generation electrical machines for electrified transportation with a step change in power density. Slot-channel cooling is considered to be an effective cooling technique, either as an independent method or as a secondary heat transfer path, which compliments traditional cooling systems. The slot-channel specific geometry and position effects on the thermal benefits are not thoroughly investigated in the literature, while previous work focuses on passing fluid through the unused space left in between coils forming concentrated windings. In this article, slot-channel cooling is implemented within an oil-flooded cooling system for a high power density motor that is used as a pump. A flexible and detailed lumped parameter thermal network (LPTN) is proposed for the cooling system, with the LPTN used to optimize the slot-channel dimensions and location for obtaining maximum thermal benefits. Finally, a surface-mount permanent magnet (SPM) machine with the optimized slot channel geometry is built and tested to validate the thermal model, experimentally achieving an armature continuous current density in excess of 30 A/mm2.


Improved Thermal Modeling and Experimental
Validation of Oil-Flooded High-Performance Machines With Slot-Channel Cooling transportation technologies at an unprecedented rate. High power density (kW/kg or kW/L) electrical machines are at the heart of the electrified concepts [1]. The design of such electrical machines is a multidomain challenge, which melds the electromagnetic aspect, the mechanical aspect, and thermal management [2], [3].
Liquid cooling provides a more efficient cooling system for electrical machines with respect to natural convection, with the latter often resulting in bulky and heavier electrical machines [4]. A liquid cooling jacket is a mature liquid cooling method used reliably in many applications, where the heat generated within the machines is removed to the coolant located in a jacket surrounding the stator core pack. The thermal performance can be furtherly improved with enhanced cooling techniques implemented in the end-winding region or the stator slot region, which plays a critical role in determining the hot spot temperature of the electrical machines [4]- [6]. Potting the end-winding with high thermal conductivity materials [7], [8] and using higher thermal conductivity materials inside the slot, such as back-iron extension [9], are proven to shorten the heat transfer path between the windings and the coolant jacket and, thus, effectively reduce the peak temperature. Thermal techniques that combine the thermal improvements applied to both the end-winding and the stator active parts are investigated in [10], where the end-winding is potted with a ceramic-based compound, while copper bars are inserted in the stator tooth to further enhance indirect stator cooling.
Compared to traditional housing jacket cooling, direct cooling, whereby the fluid is in physical contact with the windings, is highly effective in removing heat, and different approaches have been investigated. In [15], liquid coolant is injected into the stator to directly cool the windings, while a semiflooded cooling system is proposed in [11] for a high-speed application to limit the airgap losses. Spray cooling in the endwinding region [12], [13] and fully flooded cooling systems [4] are effective to cool machines with short stack lengths. Fully flooded cooling arrangements are commonly applied in the systems where the oil pump and drive motor are fully integrated. In such an arrangement, the oil from the pumping system can be used to cool the machine [14]. In a fully flooded oil cooling system (FFOCS), the oil enters from one side of the machine unit, flows along the gap between the stator and the rotor, and, finally, leaves from the other side of the machine [4]. However, in some circumstances, this intensive fully flooded cooling system is not sufficient to remove the massive heat generated by the windings in a high power density electrical machine, and thus, combinations of different cooling techniques may be required.
Recently, slot channel cooling, being direct cooling or indirect cooling, has been receiving increased attention [2], [15]- [20]. In such an arrangement, the heat transfer path between the winding and the slot is significantly reduced since the coolant is located in the slot and directly touches the windings. In [15], utilizing the space left between coils in a concentrated winding, a designed "T"-shaped slot channel is proposed to achieve a current density of 20.4 A/mm 2 , while the hot spot temperature is 148 • C. Similarly, cooling channels derived from the unused space are proposed in [2]. However, the correlations between temperature and slot channel parameters, such as size and location, remain uninvestigated, while they also play critical roles in the holistic process of electrical machine design. Meanwhile, the modeling technique for the aforementioned slot channel cooling thermal analysis presented in the existing literature is CFD, which is a very time demanding simulation approach [15], [21] and, therefore, not straightforward to be used within a multidomain optimization environment. Computationally effective thermal models with CFD and conjugate heat transfer (CHT) model implementation are investigated in [2] and [22]- [25] for time-effective thermal performance of the machine, where the copper loss increase with temperature is considered. Compared to CFD, the lumped parameter thermal network (LPTN) is flexible and provides thermal results very quickly. Its accuracy can be furtherly improved with careful design, and the temperature distribution can be obtained as well [26], [27].
In this article, LPTN is used for the design and thermal analysis of an FFOCS for a high power density electrical machine. After an initial thermal analysis, FFOCS alone is found not able to remove all the heat generated and allow the electrical machine to operate within the wire temperature limit (210 • C). The slot channel cooling system (SCCS) is then proposed to be implemented together with the FFOCS. The slot cooling channel (SCC), including its geometry and location, is optimized in a custom-designed thermal model, used within a multidomain environment.
The main contributions of this article are given as follows.
1) The combined cooling system of FFOCS and SCCS is investigated with thermal analysis, and its effectiveness is experimentally validated. The presented cooling technique is applicable to high-performance electrical machines.
2) The slot channel geometry and location are optimized based on the balance between the increased copper losses due to reduced copper fill factor and the corresponding thermal benefits that have not been studied in the previous literature. The methodology is applicable to various machine types, including those employing concentrated or distributed windings.
3) The proposed and experimentally validated thermal modeling technique is a thermal approach to obtain accurate temperature distributions, including prediction of hot spots, while benefitting from the LPTN's high computation speed. 4) The proposed thermal modeling is accounting for novel cooling methodologies, i.e., slot channel cooling in this article. With careful design, the developed thermal model can be extended to more complex geometries, such as those having multiple slot channels or irregular slot channel shapes. The structure of this article is given as follows. In Section II, the designed surface-mount permanent magnet (SPM) machine is presented, together with the proposed LPTN thermal model for the cooling system, which incorporates both the FFOCS and the SCCS. The SCC, including both its geometry and location in the slot, is optimized in Section III. In Section IV, an SPM machine prototype with the optimized SCCS and FFOCS is built and tested to validate the thermal modeling of Section III. Finally, conclusions from this research are summarized in Section V.

II. THERMAL NETWORK FOR THE MACHINE
In this section, the SPM machine is first introduced in Section II-A, with the losses calculation, including copper and windage losses presented. An LPTN, considering both the SCCS and FFOCS, is proposed and detailed in Section II-B.

A. SPM Machine Introduction
The requirement for the machine is 9.6 kW at the rated speed of 8700 rpm, corresponding to a 10.6-Nm rated torque. At the maximum speed of 19 000 rpm, the corresponding torque is 5 Nm [28]. The machine parameters are listed in Table I, together with the nine-slot, eight-pole permanent magnet prototype. SPM topology with a Halbach array and a hollow rotor shaft was selected for this application, taking into account the available converter kVA, together with the gravimetric power density (kW/kg) and rotor inertia requirements. Other IPM motors with different rotor topologies could also be considered; however, to minimize the inertia of the rotor, the advantages of an SPM Halbach and the very thin rotor back iron did represent the best choice for the case in hand.
As this article investigates the thermal performance of the SPM machine with SCCS and FFOCS, losses play a critical part in determining the thermal conditions of the electrical machine. The losses considered in this study are: 1) winding losses in the stator slot, including dc and ac losses; 2) windage losses due to the fully flooded cooling system and the high rotor speed; 3) iron losses in both stator and rotor cores; and 4) magnet losses.
As with most high power density machines, the majority of power losses are generated in the stator winding assembly, which consists of dc and ac losses [29]. The ac losses are caused by skin effects, proximity effects, circulating effects, and slot leakage effects. Skin conductor effects are due to the current's tendency to flow on the periphery when the frequency increases, while proximity effects are due to the eddy currents caused by the changing magnetic flux produced by neighboring conductors. For high-frequency electrical machines, ac resistance is significant and cannot be ignored [30]. The ac losses in (1) are calculated based on the ratio of ac resistance to dc resistance k ac , using standard 3-D FEA techniques, where the conductors within the slot are modeled at the strand level. From this analysis, k ac at the base speed of 8700 rpm is 1.04, while, at the maximum speed of 19 000 rpm, it is 1.14. The copper dc electrical resistance R dc increases with temperature, with the rate described by the temperature coefficient of resistivity, α = 0.003862. Equation (2) is used for copper dc losses calculation Q dc with phase current I . In (2), R ph0 is the phase winding electrical resistance at the reference temperature T 0 Therefore, copper losses Q total calculation is linked with the winding temperature and cannot be determined at this stage. The copper losses will be presented together with the winding temperature derived from the thermal modeling at the end of Section II.
Bertotti's formulation method is used within a 2-D FEA model to compute iron losses directly P Fe , which are composed of three components: hysteresis, classical, and excess loss, as described by formula (3) [31], [32]. The stator iron losses are 208.7 W when the motor is operating at 8700 rpm where f and B denote the frequency and magnetic flux density, respectively. Furthermore, k h , k eddy , and k exc are the hysteresis, eddy current, and excess loss coefficients, respectively. In this article, for the electrical steel grade considered (VACOFLUX 48 0.1 mm), k h = 0.007097, k eddy = 1.268 × 10 −6 , and k exc = 0.001735, while α = 5. Due to the friction effects caused by the relative motion between the rotor and the stator, windage loss is generated and contributes to the temperature rise in the electrical machine. Formula (4) is applied for windage loss calculation, where C m is the loss coefficient, ω is the rotor speed in [rad/s], ρ is the fluid mass density in [kg/m 3 ], l ro is the rotor length in [m], r is the rotor radius in [m], δ is the airgap thickness in [m], and ν is the fluid (oil) kinematic viscosity in [m 2 /s] [33]. The oil's kinematic viscosity for an inlet temperature of 90 • C is 3.19 × 10 −6 m 2 /s. The windage loss is 139 W at a rotor speed of 8700 rpm, half of which is applied to the stator, while the other half is applied to the rotor [34] when Ta ≤ 400 when Ta > 400

B. Thermal Modeling
For the SPM machine described in Section II-A, FFOCS is initially intended as the sole cooling technique, which is found as being not sufficient to remove all the heat generated within the machine. SCCS is then proposed to enhance the cooling of the machine to operate within the temperature limit of the constituent materials. This part will present the thermal modeling for the electrical machine with both FFOCS and SCCS, which can also be intuitively modified to simulate the thermal performance of the machine with FFOCS only.
As can be seen from Fig. 1, the cooling system in the electrical machine consists mainly of two elements: 1) flooding oil in the end-winding region and in the air gap and 2) forced oil convection in the SCC.
Considering heat transfer in both the radial and axial directions, a 3-D LPTN with a self-tuned number of nodes is built for the machine with SCCS and FFOCS, as shown in Fig. 2. The definition of "self-tuned" here refers to the fact that the nodes' number for each part of the machine can be adjusted based on the thermal model accuracy and the computational efficiency requirements. The calculation time is ranging from seconds to few minutes with nodes number increasing to significant large values. Exploiting the machine symmetries, half slot, and the corresponding rotor parts is simulated in the thermal network. In Fig. 2(a), it can be seen that the half slot is divided into five regions: "Region-I," "Region-II," "Region-III," "Region-IV," and "Region-V," where the widths of "Region-I" and "Region-V" are equal to the width of half the SCC, while the height of "Region-III" is equal to that of the SCC. The number of nodes of "Region-I" and "Region-II" in the "y"-direction is independent, while they share the same number of nodes in the "x"-direction. Similarly, the nodes' numbers of "Region-IV" and "Region-V" are the same in the "x"-direction, while they can be different in the "y"-direction. The numbers of nodes of "Region-II," "Region-III," and "Region-IV" regions are the same in the "y"-direction, while they are independent in the "x"-direction. For the nodes' geometry determination, the maximum slot node width (and/or height) is set to 0.2 mm in this article, which gives 1254 nodes in the half slot for one axial plane. The number of nodes in the "x"-direction "NM x " is first determined using (9) for the aforementioned slot regions. "DS x " is the maximum section length in the "x"-direction (mm). The mathematical symbol "" is used to indicate rounding up. The numbers of nodes in the "y"-direction are determined using a similar process Similarly, the stator tooth is divided into three regions, as shown in Fig. 2(a): "Tooth-I," "Tooth-II," and "Tooth-III," which shares the same number of nodes in the "x"-direction with "Region-II," "Region-III," and "Region-IV" in the slot, respectively. There are three regions for stator backiron. In the "y"-direction, "Back-iron-I," "Back-iron-II," and "Back-iron-III" share the same number of nodes with "Region-I," "Region-II," and "Tooth" regions, respectively.
For the sake of clarity, repetitive figures are avoided, and only four sections are presented in the axial direction ("z"-direction) in Fig. 2(b), covering the three types of nodal connections: 1) between two end-winding sections, i.e., "end-winding section a" and "end-winding section b"; 2) between stator core and end-region section, i.e., "endwinding section b" and "core machine section c"; and 3) between stator core sections, i.e., "core machine d" and "core machine e." In the thermal model, three end-winding sections for each end-winding region and 11 sections within the active length are selected to obtain time-efficient and accurate results. Fig. 2(c) shows the nodes' connection to the flooding oil in the end-winding region, where all the windings are assumed to be extended from the slot and corresponding to "Region-I" to "Region-V." The thermal model for the machine without SCCS is derived from that with slot channel cooling, where the slot channel (i.e., "rectangular blue region" in Fig. 2(a) and the corresponding region in Fig. 2(b) and (c) for both "x"-and "y"-directions) width and depth are set to infinitesimally small values.
Different nodes in Fig. 2 are linked with (10) for transient state calculation, where c i and m i are node j heat capacity and mass, respectively, T i (t) and T i (t − 1) are the node i temperatures at time t and t − 1. q i (t − 1) is the heat generated within node i at time t − 1, T j (t − 1) is node j temperature at time t − 1, and t is the time step between iteration time t and t − 1, which is 0.0001 s in this article. R i, j is the thermal resistance between nodes i and j For q i in (10), the calculated losses in Section II-A are evenly distributed into each node based on the volume. Copper losses are updated with temperature, while other losses, including iron losses, are assumed constant. For wire node i , the node loss q i (t) is calculated with formula (11), where α is a copper temperature coefficient, which is equal to 0.003862. q i (0) is the heat generated within node i at the reference temperature Equations (10) and (11) are used to compute the system transient state temperature, while, for thermal steady state, (12) is used Resistances R m, j in (10)-(12) consist of conduction, convection, and radiation thermal resistances. These correspond to conduction, convection, and radiation heat transfer, respectively. Radiation heat transfer is ignored in this article as it is comparatively small compared to the other two types of heat transfer. Equation (13) is applied for the conduction thermal resistance calculation, where l is the length in [m], A cond is the node cross-sectional area in [m 2 ], and λ is the node thermal conductivity in [W/(m·K)], which is determined by the material properties Equation (14) is used to calculate the convection heat transfer occurring in the following situations: 1) in the airgap region; 2) in the end-winding region; and 3) in the SCC. A conv is the contact area between the fluid and the solid, while the heat transfer coefficient h is calculated using formulas from (15) to (24) In the case of convection heat transfer in the air-gap region [35], [36], with a rotor speed of 8700 rpm, h is obtained as 2562 W/m 2 K, and the Nusselt number Nu = 13.17 with formulas (15)- (18), where D a is the equivalent diameter, δ is the air-gap thickness in [m], v a is the fluid axial flow velocity in [m/s] calculated with the flow rate and the air-gap area, v p is the rotor peripheral velocity in [m/s], v e is the effective velocity in [m/s]; r s and r r are the stator inner radius and the rotor outer radius, L is the motor length, and ν, λ, and Pr are the oil's kinematic viscosity, thermal conductivity, and Prandtl number, respectively, . (17) In the end-winding region, particular attention should be paid to the Nusselt number Nu determination [29]. Different formulas are used to calculate Nu corresponding to different parts in the end-winding region [37], [38]: near the stator, near the stator winding, and near the rotor. Equation (18) is applied to the heat transfer between the stator and the flooding oil, while (21) and (22) Nu = 0.076Re 0.7 Pr 0.36 (20) where k 1 , k 2 , and k 3 are the curve fit coefficients, while vel is the rotor peripheral velocity.
The heat transfer coefficients in the air gap and the endwinding region are mainly dependent on the rotor speed and less sensitive to the oil flow rate. The convection heat transfer in the SCC, however, is changing with the oil flow rate, with the variation plotted in Fig. 3, for a fixed slot channel geometry and location in the slot at the rated speed of 8700 rpm.
With the presented thermal model, at the rated power and speed, considering an oil inlet temperature of 90 • C, the theoretical highest temperature for the machine with FFOCS only is 269 • C. The total copper loss obtained is 1109 W. It can be deduced that the FFOCS alone is not sufficient to cool the machine below the wire thermal limit of 210 • C, for which thermal limit the maximum continuous current density achieved is 22.1 A/mm 2 .
SCCS is, therefore, incorporated with the FFOCS in this article in order to help dissipate the heat generated within the machine, especially in the slot, to the flooding oil. However, the cooling channel occupies space inside the slot, which would, otherwise, be available for the copper and, thus, for a given reference temperature, increases the resistance of the winding. From the foregoing discussion, the slot channel geometry is an important element that needs to be carefully investigated to achieve maximum thermal benefits.

III. THERMAL IMPROVEMENT WITH SLOT CHANNEL COOLING
In this section, based on the thermal modeling proposed in Section II, the SCC geometry and its location within the slot are investigated for maximum thermal benefits.
Three main variables are used to describe the shape and position of the slot channel geometry, as shown in Fig. 4, namely: 1) slot channel width; 2) slot channel depth; and 3) slot channel location. Section III-A investigates the slot channel width and depth [i.e., 1) and 2)] effects on the peak temperature of the machine, while Section III-B focuses on the cooling channel location. The three variables are scanned within defined ranges, as detailed in Sections III-A-III-C, to provide a clear picture of the temperature tendency.

A. Slot Channel Width and Depth Sensitivity Analysis on Temperature
For a comprehensive view of the slot channel geometry effects on winding temperature and manufacturing flexibility consideration, the selected SCC width dimensions are discrete values, starting from 1 mm and increasing at a step size change of 1 mm, up to a maximum width of 4 mm (corresponding to ∼62% of the top slot width), as listed in Table II and described in Fig. 4. On the other hand, a finer resolution is considered for the SCC depth, which varies from 0.5 to 4.9 mm (∼60% of the winding depth), with a step change of 0.4 mm. In the first analysis, only the SCC width and depth are varying, while the SCC center is fixed at the middle point of the slot (i.e., the slot channel center location "H " = 4.25 mm in Fig. 4).
The copper fill factor, defined as the ratio of copper area to the slot area, is first studied and plotted in Fig. 5. It can be seen that the copper fill factor decreases linearly with SCC depth for all the four SCC widths considered from the original value of 0.47 due to the fact that the SCC occupies space, which would, otherwise, be available for winding. The smaller the copper fill factor is, the higher the copper losses are for a given current and winding temperature. In practice, the copper losses are updated with SCC geometries to the developed 3-D thermal modeling in Section II. Using formulas (2) and (11), it can be seen that the losses vary since the channel affects the coil temperature and its resistance. In this case, the copper losses are affected by two factors: 1) the winding temperature and 2) the increased electrical resistance due to the reduced copper fill factor in Fig. 5. The copper losses taking into account the temperature effects are plotted in Fig. 6.
In Fig. 7, the peak temperature is plotted for all the combinations listed in Table II. As mentioned earlier in Section II, the original motor temperature is 269 • C without the SCC.   The original temperature is now plotted in Fig. 7 for a clear view of the temperature trends when considering slot channel cooling.
It can be seen from Fig. 7 that all the proposed SCCs provide temperature reductions compared to the original motor. The temperature is decreasing throughout the studied SCC depths for "SCC-i," which is the narrowest channel considered, 1-mm wide. For "SCC-i," the cooling benefits with the slot channel outweigh the increased copper losses due to the reduced copper fill factor. The significant temperature reduction achieved is also the main reason that the copper losses for "SCC-i" are lower than those for the original motor (1109 W from Section II-B), as shown in Fig. 6, even with a lower copper fill factor.
For wider SCCs ("SCC-ii," "SCC-iii," and "SCC-iv"), the peak temperature tendency is to decrease first and then increase due to the higher copper losses coming from the reduced copper fill factor.
As shown in Fig. 7, both SCC widths of 3 and 4 mm provide promising thermal benefits with 2-mm SCC depths. Further increase in SCC depth does not guarantee significant temperature reduction and reduces the flexibility of slot channel location optimization while complicating the manufacturing process.
Therefore, for further slot channel location improvement, two types of slot channel geometries are selected: 1) slot channel "A" with 3-mm SCC width and 2-mm SCC depth and 2) slot channel "B" with 4-mm SCC width and 2-mm SCC depth.

B. Slot Channel Location Analysis
In this section, the two selected slot channel types "A" (i.e., slot channel width of 3 mm and depth of 2 mm) and "B" (i.e., the slot channel width of 4 mm and depth of 2 mm) are compared, while the center of the slot channel location "H " in Fig. 4 is varying from 2.25 to 5.25 mm for both channel types. Of note, when the slot channel center is at the center of the slot, "H " = 4.25 mm.
It can be seen from Fig. 8 that the temperature drops and then increases with the value of "H ," while the optimized location for slot channel A is "H " = 3.6 mm, and it is "H " = 3.7 mm for slot channel B. With the optimized slot channel location, the temperature for slot channel A is 191 • C, while it is 184 • C for slot channel B.
For a clear view of the temperature distribution in the slot, the temperature contour is plotted in Fig. 9, for slot channel A  located at 2.25, 3.25, and 5.25 mm. When the slot channel A center is at 2.25 mm ("H " = 2.25 mm), the hot spot is near the slot opening, as shown in Fig. 9(a). This is because significant thermal barriers are resulted due to low thermal conductivity wedge placed in between the winding and the flooding oil. Conversely, the hot spot is pushed to the corner of the slot, near the back-iron and the tooth, when the slot channel is near the slot opening, as shown in the case of Fig. 9(c).
In this article, considering the manufacturing feasibility, slot channel A with 3-mm width and 2-mm depth is selected to be applied with the flooding oil cooling system. The slot channel center location "H " = 3.6 mm is selected as being optimal for minimizing the coil temperature, as shown in Fig. 8.

C. Thermal Results
With the optimized slot channel A dimensions (3-mm SCC width and 2-mm SCC depth) updated into the thermal model, Fig. 10 is obtained for the 26.84-A/mm 2 current density. As can be observed from Fig. 10(b), the temperature is reduced largely with the SCCS, with respect to the original machine of Fig. 10(a). Meanwhile, with the channel, the hot spot moves from the center of the slot to the corner of the slot bottom.   Fig. 11 presents the temperature distribution in the axial direction, for the different slot regions described earlier in Section II. It can be clearly seen that, with the proposed thermal management, the peak temperature occurs within the core part in "Region-II," which concurs well with Fig. 10.

D. Generalization Aspect
Sections III-A-III-C presented the thermal benefits with the optimized slot channel geometry and location. To derive the relationship between the machine dimensions and the SCC geometry for best thermal performance, different motors with scaling ratios of 0.5 to 5 are studied. In this exercise, the slot channel width is selected as being 50% of the slot top width "w 2 " for manufacturing reasons, as shown in Fig. 12, while the slot channel depth is varying from 10% to 80% of the winding depth. It is found out that the slot channel depth for best thermal benefits is 40%-60% of the winding depth. With fixed slot channel width as 0.5w 2 and the slot channel depth as 0.5h w , the optimized slot channel center location is around 0.4h w , slightly above the slot center toward the slot bottom. However, for a specific machine design, mechanical restrictions should also be taken into consideration during the SCC optimization process.

IV. EXPERIMENTAL RESULTS
An SPM machine prototype with the proposed slot channel geometry is built and tested in this section. First, a dummy stator with windings is used to test the slot channel feasibility, followed by the manufacturing and testing on a full motor. Finally, the experimental measurements are presented and compared with simulation results for validation.

A. Dummy Stator Manufactured
A plastic stator with the same geometry of the SPM machine designed, with dimensions reported in Table I, is first 3-D printed and wound, as shown in Fig. 13(a), for an early feasibility trial study on the slot fill factor. Similar to other novel cooling concepts implemented in prototypes [41], the SCCs require the development of a careful manufacturing procedure. One possible way for the implementation is to use "coolingchannel-formers" that capture the dimensional and positional properties of the designed cooling channels. The formers are used during the winding process and removed post stator impregnation. For this, it is important to carefully choose the material properties of the resin and the former, in a way that the resin does not attach to the former, hence easing the procedure for the removal of the former post impregnation.

B. Experimental Test Rig
The high power density SPM machine is prototyped and tested with the materials and geometry summarized in Table I. Air release valves are installed on the machine housing to release any air trapped in the machine to ensure that the machine is fully flooded. Peak temperature reduction is the main focus of this work, which limits the maximum current (and corresponding power) of the machine. Multiple K-type thermocouples are placed within the stator slots in  the stator core and the end-windings to extract the hot spot temperatures and validate the thermal model. The K-type thermocouples are calibrated by placing them in the same thermally insulated segment at room temperature, and the variation is within 0.1 • C between different thermocouples. The physical locations "A" and "B" of thermocouple placement are described in Fig. 14(a). As mentioned in Section II-B and illustrated within Fig. 14(b), there are three end-winding sections for each end-region and 11 sections within the active length. Table III summarizes the thermocouple locations used within the experimental campaign. Both transient and steady-state temperatures are recorded and compared to the analysis.
In Fig. 15, the prototype is installed on a test bench to validate the design and proposed thermal modeling of Section II and thermal results of Section III. Fig. 15(a) shows the test rig flow loop, including the cooling unit (tool temp: 188), 10-μm oil filter, pressure transmitters at the inlet and outlet of the motor, and temperature measurement devices (Pico logger). The oil used in this research is Eastman Turbo Oil 2389, and the overall flow rate is 4 L/min, which is the same cooling medium condition as in the application where the drive motor is fully integrated with an oil pump. The oil is preheated to guarantee an inlet temperature of 90 • C, with the lab ambient temperature being 30 • C.

C. Experimental Results
The SPM machine runs at rated power and speed, 9.6 kW and 8700 rpm for 900 s during which the thermal steady state is reached after approximately 300 s, as can be seen from Fig. 16, where the simulation and experimental results are plotted and compared for TC1 and TC7. In Fig. 16, the solid lines (−) represent the experimental results, while the simulation results are represented with dashed lines (-). The system pressure drop extracted from the pressure transmitters installed at the inlet and outlet of the machine is 22.5 kPa.
The maximum temperature of 191 • C (TC7) is recorded in the axial middle section of the core, while it is 183.32 • C (TC1) in the end-winding region, as shown in Fig. 16 and Table IV. The experimental data matches well with the peak temperature in the simulation results, with the predicted temperatures to within 3% compared to the experimental results, as shown in Table IV. The axial temperature profile as predicted by the described LPTN also shows good agreement to the experimentally measured one, as shown in Fig. 17. Importantly, with slot channel cooling added to the flooding oil cooling system, the temperature is reduced significantly from 269 • C to 191 • C (about 28.9%), and a continuous current density of 26.84 A/mm 2 can be sustained. The continuous current density can be increased to be 30 A/mm 2 (+11.77%) before reaching the wire thermal limit of 210 • C.

V. CONCLUSION
Improved thermal management targeting the coil area and associated copper losses is critical for the next-generation emachines with a step change in power density. Slot channel cooling is considered to be a promising cooling technique in the thermal management of electrical machines. In this article, first, the thermal performance of the SCCS combined with FFOCS for an SPM machine is investigated with a bespoke designed thermal network, which captures the slot channel geometry variation. The slot channel geometry along with its location in the slot is optimized for the maximum thermal benefits. Finally, an SPM machine prototype is built and tested based on the optimized SCC. The experimental results are in overall good agreement with the simulation results. With the combined cooling system, the prototype can generate the required output torque with a high current density (26.84 A/mm 2 ) successfully sustained continuously, while the current density can be furtherly increased to beyond 30 A/mm 2 to work within the wire thermal limit (210 • C), compared to 22.1 A/mm 2 with only FFOCS. For a comprehensive view, the maximum current density that can be achieved for a more common wire type of 180 • C thermal limits is 25.38 A/mm 2 . From a manufacturing perspective, the proposed cooling technique that combines fully flooded oil cooling and slot channel oil cooling can be carefully embedded in high power density electrical machines, for both concentrated windings or distributed windings, with the help of spacers specially designed for the winding process and proper sealing mechanisms. In such a cooling system, where the air gap is oilflooded, mechanical losses, such as friction losses generated within the air gap, will reduce the machine efficiency. At the rated speed of 8700 rpm, the windage losses are 139 W or 8.6% of the total losses, with the corresponding efficiency of the machine being 85.7%. A suitable cooling medium for such a system is desired to have high thermal conductivity, high electrical resistivity, and low viscosity. Other properties, such as compatibility with insulation and oxidation stability, should also be considered for this kind of application.