In the following section, the results of the models based on the coupling procedure de-scribed in Section3.5are presented. At first, the results from the 2.5-D analysis in which 2-D EMAG model was coupled with Thermal Model II described in Section3.2are presented.
Then the second coupled model based on the coupling between 3-D EMAG model with Ther-mal Model II is discussed.
The main aim of the coupling procedure was to simulate the motor work based on the fundamental electric parameters like currents waveforms and voltages that were measured in the experimental campaigns and then compute the power losses of the machine. In the next step, the power losses were transferred from the EMAG to CFD solver in a form of volu-metric heat sources. This all means that the measured power losses could be used to validate the model, but not as the input data as it was defined in Thermal Model I and Thermal Model II.
4.3.1 2.5-D coupled model
In the current section, the results of the first coupled model based on 2.5-D analysis is presented. In the first part, only the EMAG model results are introduced. Then the results of the coupled solution of Thermal Model II and 2-D EMAG approach are discussed.
In Fig. 4.29, the current waveforms resulted from the presented 2-D EMAG model were compared with experimental recordings collected by the clamp meter connected with the oscilloscope. The presented results of the EMAG model are shown at Operating Point #1 pre-sented in Tab.2.2. The shape of the current waveforms is the result of the coupling between the 0-D auxiliary circuit model with the EMAG model. In the figure, the orange dots rep-resent the experimental records obtained for one out of three phases, while the blue cross marks show the current values predicted by the coupled model after each time step of the selected period. Additionally, the two next phases, being the results of the 2-D EMAG model, are presented in the background in a grey colour. One can see that experimental and numer-ical current waveforms are consistent with each other at a satisfactory level.
In Fig.4.30, the field of the magnetic flux at Operating Point #1 is presented for the stator and rotor core for six selected time steps with the rotor angle increment of 20◦. This field is presented in 2-D domain described in Section3.3. Additionally, vectors representing the
Figure 4.29: Current waveforms in A obtained from the experiment (orange) and 2-D EMAG model (blue) at Operating Point #1
current density are presented in the winding area. In this figure, red and blue colours of the magnets represent their different polarity.
At the first rotor position denoted as -20◦, the maximum current can be observed and its density achieved the value of approx. 8·106Am−2at the first phase. However, in this case, the maximum magnetic flux is observed at the neighbouring tooth where the second phase is wounded. In the second phase, the smaller current density values are observed.
The rotor position denoted as 0◦ shows smaller values when compared to the previous rotor position. In this position, the magnetic field is closed through all the stator teeth.
Therefore, it is also visible that the current density is directed oppositely comparing these two cases.
For the rotor position of 40◦, the maximum current values are also not reached and sim-ilarly it is visible for the rotor position of 60◦.
In the rotor position of 80◦, the maximum current density is visible again in the neigh-bouring phase. In the period of the presented rotor motion from -20◦ to 80◦, each phase achieved at least three times the maximum current value and at least once different current direction was observed.
Figure 4.30: Magnetic flux field in T and current density vectors in A ·m−2 for selected six time steps with the rotor angle increment of 20◦
As it was presented in Section3.3.3, the copper and core losses are current-dependent in the conductor and induction-dependent in the laminated core, respectively. Therefore, the previously presented results from the EMAG model were used to the power loss estimations.
In Fig. 4.31, the power losses varying in time are presented for both windings and core.
In the figure, the presented losses were calculated as a surface integral of the local losses multiplied by the core length. The blue curve represents the copper losses, the yellow curve shows hysteresis losses in the core, the eddy current losses in the core are depicted using orange colour and the excess losses are expressed by grey colour. All the losses resulted from the 2-D EMAG model do not cover the power losses that occur in the end-windings, bearings and copper paths located in the PCB plate because all these components were not included in this simplified model. Therefore, it had to be estimated separately from the experimental part and then implemented in the model in the same way as described in Section3.2.
Due to the transient character of the EMAG model, the final losses were calculated on the basis of the averaged loss paterns in the range marked by green frame presented in Fig.4.31.
The averaging of the local losses was conducted on the basis of each loss field calculated
Hysteresis Losses [W] Eddy Current Losses [W] Excess losses [W] Copper Losses [W]
Losses, W
time, ms
Figure 4.31: Power losses in W computed using 2-D EMAG model at Operating Point #1
for each time step being in the range of the green frame. The loss values in the core and in the windings part located along the core consisted of the losses estimated during the exper-imental tests listed in Tab. 2.3. As it was mentioned, the copper losses derived from the 2-D EMAG model were underestimated due to the lack of the end windings and copper paths in the numerical domain. After the field loss estimations, the averaged local losses were inter-polated and transferred to the CFD model according to the coupling procedure described in Section3.5.
The implemented losses from the EMAG model were used as non-uniform heat sources in the CFD model. The representative results, characterised by the highest temperature, are presented. The temperature field of the motor representing the external surfaces and wind-ings are presented in Fig. 4.32. The temperature field of the external motor parts, presented in Fig.4.32(a), can be compared with already presented thermography image shown in Fig.
4.28(a). In the presented field, the highest temperature occurred in the region where the core is placed, namely the front part of the motor housings. In Fig.4.32(b), the temperature field of the windings is introduced. The highest temperature occurred in the region of the external surfaces of the windings being in contact with internal air. The lowest temperature is noticeable in the corners where the contact with plastic bobbin is the largest. In addi-tion, the windings segmentation is marked and it is consistent with segmentation discussed during the description of Fig.3.6.
Figure 4.32: Temperature fields in◦C resulted from 2.5-D coupled model for Operating Point
#1 in Variant A: (a) external surfaces of the motor and (b) windings with marked segmenta-tion
4.3.2 3-D coupled model
The results presented in the following section are consistent with the data, such as cur-rent waves and magnetic fields varying in time, presented in the previous section describing 2.5-D analysis. Similarly, the 0-D circuit model was integrated with 3-D EMAG model.
In Fig. 4.33, the current waveforms resulted from the presented 3-D EMAG model were compared with the experimental recordings collected by the clamp meter connected with the oscilloscope. The presented results of the EMAG model are also shown at Operating Point #1 as it was introduced in the 2-D case. In this figure, the orange dots represent the experimental data obtained for one out of three phases, while the black cross marks show the current values predicted by the coupled model after each time step of the selected period.
Additionally, the two next phases, being the results of the 3-D EMAG model, are presented in the background as the grey lines. One can see that experimental and numerical current waveforms are consistent with each other.
Figure 4.33: Current waveforms in A obtained from the experiment (orange) and 3-D EMAG model (blue) at Operating Point #1
The fields of magnetic flux and current density was analogous with already presented in the previous Section Fig.4.30. Therefore, it is not presented in the current Section.
Similarly, as in the 2-D EMAG model, the computed power losses resulted from the tran-sient EMAG simulations. As in the 2-D case, the previously presented results of the current waveforms and calculated fields of induction allowed for the power loss estimation at the investigated operating points. As it was mentioned, in the 3-D EMAG analysis, the influence of the end-windings was taken into account. Therefore, in Fig. 4.34, the losses varying in time from the 3-D EMAG analysis are presented. In the figure, the presented losses were calculated as a volume integral of the local power losses. Moreover, in this figure, the blue curve represents the copper losses, the yellow curve shows hysteresis losses in the core, the eddy current losses in the core are depicted using orange colour and the excess losses are expressed by grey colour. As it was mentioned, the sum of the hysteresis, eddy current and excess losses constituted the core losses. In particular, the losses resulted from the 3-D EMAG model also covered the losses of the end-windings. Therefore, the estimated copper losses are higher than those presented in the 2-D analysis. The losses of the bearings and copper paths located in the PCB plate were not considered in the EMAG analysis. However, besides the coupling, these losses were applied in the same manner as in Thermal Model II. The av-eraging process of the local power losses was conducted in a similar way as in the 2-D EMAG case. The 3-D fields of the local losses for each time step were averaged within the time range marked by green frame presented in Fig. 4.34. The averaged local losses were then interpo-lated to the point representing the cells centres defined by the CFD mesh. Next, the lossfield was transferred to the CFD solver.
The implemented losses from the EMAG model were used as non-uniform heat sources in the CFD model. The representative results, characterised by the highest temperature, is presented for the Operating Point #1 for Variant A and this one is presented. The tempera-ture field of the motor representing the external surfaces are presented in Fig. 4.35(a). The temperature field of the external motor parts can be compared with already presented ther-mography image shown in Fig.4.28(a). The highest temperature occurs in the region where the core is placed. In Fig. 4.35(b), the temperature field of the windings is introduced. One can see that the temperature field consisted of the temperature field presented for the 2.5-D coupled analysis.
Hysteresis Losses [W] Eddy Current Losses [W] Excess losses [W] Copper Losses [W]
time, ms Losses, W
Figure 4.34: Power losses in W for windings (blue) and core losess computed using 3-D EMAG model at Operating Point #1
Figure 4.35: Temperature fields in◦C resulted from 3-D coupled model for Operating Point
#1 in Variant A: (a) external surfaces of the motor and (b) windings with marked segmenta-tion
Chapter 5
Summary, conclusions and future work
5.1 Summary
The dissertation begins with a literature review, explaining the motivations dragging to undertaking the specific research tasks. The introduction is followed by a description of ex-perimental activities carried out during the research. The measurement procedures are de-scribed as two independent experimental campaigns performed on the dedicated test rig.
The first one focused on the measurements of the air velocity within and around the anal-ysed motor and simultaneously on the temperature measurements. In the first experimen-tal campaign, constant temperature anemometers were positioned at twenty eight points to collect the values of the vertical velocity component of the hot air above the motor installed in the test rig. Moreover, two velocity components were recorded using the Laser Doppler Anemometry technique within the rear part of the investigated motor. During the first ex-perimental campaign, thermal measurements were also conducted using a set of twenty two calibrated thermocouples. The temperature and velocity measurements allowed for the in-vestigation of the natural convection phenomena that occur during the motor power loss dissipation. In these experimental tests, the operating points of the motor work including a combination of four resistor loads and three rotational speeds were investigated. The sec-ond experimental campaign was focused only on the thermal measurements using a set of the calibrated thermocouples and infrared thermography. In this campaign, different pas-sive heat dissipation intensification concepts were also investigated. During the second ex-perimental campaign, 6 operating points were recorded for each variant of heat dissipation enhancement.
The experimental part of the research was used to validate the formulated CFD models.
These models were built on the basis of the motor complex geometry. During the numeri-cal research, two thermal models were introduced to investigate thermal motor behaviour.
The first one was formulated and validated in the conditions occurring in the first experi-mental campaign. The second thermal model was created and next validated on the basis of the second experimental campaign. Moreover, it covered the proposed heat dissipation intensifications. The models were based on the standard governing equations used in CFD, while many of the model properties were implemented as dependent on temperature us-ing the user-defined functions. The motor losses were implemented in thermal models as the volumetric heat sources allocated to windings, core and bearings. Heat sources deriving from the Joule heating were calculated as a function of temperature. Moreover, the electro-magnetic model of the motor work is also shown and its coupling procedure with the second thermal model is described. In the developed thermal model, one of the important aspects of the model properties was the anisotropic character of the winding thermal conductivity.
The results of the numerical models showed a satisfactory consistency with the conducted experimental tests.
The proposed numerical model was used to express computationally all the operating points investigated experimentally. The results agreement between the model and experi-ment was reasonably good and location of the hotspot was correctly identified in the mo-tor winding region. The hotspot location was consistent with all tested loads. The average winding temperatures that were measured during experiments at the reference state with the motor rated parameters and no heat removal enhancement reached the value of 68 K above the room temperature. The numerical model results indicated the same tendency. The first variant of heat dissipation improvement was based on covering the external surfaces of the machine by high emissivity material that allowed for the decrease of the windings tempera-ture by approx. 4 K to the level of 64 K above the room temperatempera-ture. This first variant of heat enhancement was accompanied by the other heat removal improvements in further vari-ants. The second variant of heat dissipation improvement was based on the application of two types of radiators on the external motor surfaces. The usage of the first radiator charac-terised by a smaller outer area allows for reducing the winding temperature by 9 K compared to the case without heat removal enhancement. The application of the second radiator hav-ing bigger outer area resulted in the windhav-ing temperature drop by 16 K when compared to the case before modifications and referred to the rise above the ambient temperature. In the next variant, an application of the thermal filler in the form of potting material within the motor housing in the free space of the stator was used. That configuration allowed for the average temperature reduction of the windings by 18 K when compared to the unmodified
machine and referred to the ambient temperature. The last thermal motor test was also con-ducted with the thermal filler and with the bigger radiator. This combined method allowed for reducing the temperature by approx. 30 K comparing to the original motor construction and referring it to the ambient temperature. All numerical results of the investigated heat im-provements showed satisfactory consistency with the experimental results. Lower accuracy was only observed with the thermal filler application concepts.