Assessment of Edgewise Insulated Wire Bend Radius Impact on Dielectric Properties of Turn-to-Turn Insulation Through Thermal Aging

This study aims to evaluate the impact of the bending radius of edgewise insulated wires on dielectric properties such as partial discharge inception voltage (PDIV), partial discharge extinction voltage (PDEV), dielectric dissipation factor (DDF), and insulation capacitance (IC) through thermal aging. The tests are performed at room temperature (20 °C) and atmospheric pressure (1013 mbar) on unaged and thermally aged polytetrafluoroethylene (PTFE)-wrapped pairs of edgewise insulated wires, models of the turn-to-turn insulation. The accelerated thermal aging is carried out at 250 °C (i.e., 50 °C higher than thermal class) for two different aging periods: 156 and 312 h. To manufacture a coil, it is generally demanded to shape 90° bends out of edgewise enameled winding wires. Therefore, the obtained experimental results are helpful for the coil manufacturers, providing a clue how the bending radius can impact the dielectric properties of turn-to-turn insulation and introducing an optimum radius, which can present better insulation performance.


Assessment of Edgewise Insulated Wire Bend
Radius Impact on Dielectric Properties of Turn-to-Turn Insulation Through Thermal Aging H. Naderiallaf , M. Degano , Senior Member, IEEE, and C. Gerada , Senior Member, IEEE R ECTANGULAR wires are used to manufacture trac- tion motor coils in electric vehicles (EVs) and hybrid EVs (HEVs) to increase the motor performance [1].There are some advantages to employing rectangular insulated wires (e.g., edgewise ones) to manufacture the winding of electrical machines over random wound winding.The primary motivation is obtaining a high filling factor for reduced copper losses and higher power density [2].Moreover, rectangular insulated wires usually have a larger insulation thickness than custom round wires, offering a higher partial discharge inception voltage (PDIV).In addition, controlling the turn-to-turn voltage is more manageable compared to random wound winding, as the location of the wires is more defined.However, there are some disadvantages to using rectangular insulated wires, such as skin effect losses, restricting the drive frequency [3], and more complex coil manufacturing (e.g., to the need for a forming machine with good accuracy, laser soldering, and reinsulating the soldering points).
The turn-to-turn winding insulation is known as the most vulnerable insulation system in electrical machines playing a definitive role in the reliable machine's operation.Once the electric field corresponding to the turn-to-turn insulation exceeds the partial discharge (PD) inception field, PD activity is initiated, promoting insulation degradation, especially if the winding wires are insulated only with organic materials.In this case, PD can result in premature failure, reducing the life to a few days, if not hours [4], [5], [6].Therefore, work [7] advises that inverter-fed machines with Type I insulation (i.e., organic-only insulating materials) must be designed based on the PD-free criterion, i.e., the maximum peak voltage between two adjacent turns should remain lower than the minimum peak of PDIV [6].Unfortunately, PDIV can reduce due to the inevitable insulation aging (e.g., thermal aging) and degenerating PD activity, thus affecting the whole insulation system reliability [8].
The considered edgewise wire for this experimental study has Type I insulation (i.e., organic insulating materials), where PD is known as the end-of-life criterion [7].It is very important to investigate whether any damage or fraction for the insulation occurs during coil manufacturing through the bending of edgewise wires.As it is usually required to make 90 • bends out of edgewise enameled winding wires (Fig. 1), it will be necessary to know whether the bending radius can affect the dielectric properties or if there is any optimum radius that can give better insulation performance.
Thus, this article discusses the relationship between PD activity and the bending radius of turn-to-turn winding insulation made from edgewise insulated wires subjected to thermal aging.The electric field in the turn-to-turn insulation systems follows a capacitive distribution.Therefore, the dielectric dissipation factor (DDF) and insulation capacitance (IC) measurements are also performed on turn-to-turn samples (both straight and bent specimens), investigating their correlations with PDIV values.
The structure of this article is organized as follows.Section II introduces the test samples, measurement conditions, the approach used to measure PDIV and PD extinction voltage (PDEV), and the method to measure the DDF and IC.In Section III-A, the measured PDIV and PDEV as a function of bend radius for different levels of thermally aged specimens are presented.Section III-B explores the correlation between PDIV and the insulation thickness as a function of thermal aging for different bend radii.Section III-C examines the influence of bend radius on the DDF and IC at different aging levels.Section III-D proposes an equivalent circuit for the bent turn-to-turn samples.Section III-E shows the dispersion level of measured PDIV and PD charge amplitude versus thermal aging time for different bend radii.Section III-F discusses the measured PD charge amplitude, PD repetition rate, and the severity index under PDIV as a function of bend radius at different aging levels.Section III-G presents the measured DDF tip-up values for straight and bent samples for fresh and thermally aged samples.In Section III-H, the derived correlation between the measured PDIV and the dissipation parameters for turn-to-turn samples with different bend radii is shown.Section VI summarizes the main results of the work.

A. Test Samples Characteristics and Manufacture
Fig. 2 shows the test sample preparation procedure.PD, DDF, and capacitance measurements are carried out on unaged and thermally aged polytetrafluoroethylene (PTFE)wrapped pairs of edgewise insulated wires, mirroring the turn-to-turn insulation system [Fig.3(b)] [8].The test samples are categorized into two main groups: 1) straight (STR.) and 2) bent turn-to-turn specimens.Three different bend radii, 8.5, 6.5, and 3.5 mm, are considered to make 90 • bends out of edgewise enameled winding wires (Fig. 3).The bend radius of 3.5 mm is often the smallest bend radius for edgewise insulated wires used by the coil manufacturers.The two additional radii, both larger than 3.5 mm, are included with the sole purpose of establishing a trend for various parameters derived from at least three different bend radii.It is noteworthy to mention that the samples are bent using the same procedure employed in manufacturing an actual coil with edgewise wire.Furthermore, PTFE is employed to hold the two wires securely and near each other.Based on this, it is presumed that real coils also possess similar vulnerable points at the extreme sections, where PD activity may occur due to the bending and separation of the wires.This assumption is made because the bending technique and coil manufacturing process are identical to those used for the sample preparation.
Fig. 4 shows the wire cross section, indicating bare wire dimensions and the insulation thickness.The dimensions are the mean value obtained from 32 wires by differencing the dimension values before and after stripping wires using a micrometer with an accuracy of 1 µm [9].The wire is stripped using a laser wire stripping machine, ensuring that only the insulation is removed and not any portion of copper.The thermal class of the wires enamel is 200 • C, composed of a polyester-imide/polyester basecoat and a polyamide-imide overcoat.The length of each wire to assemble the test samples is 22 cm.
The accelerated thermal aging process is performed at 250 • C (i.e., 50 • C higher than the thermal class) for two different aging periods: 156 and 312 h.In this study, the "10 • " approach is employed to provide a rough estimation of insulation deterioration.According to this rule, the lifespan of insulation is divided by two for every 10 • C increase in temperature above the thermal index (TI) [10].For example, if the expected longevity of an insulation system at TI (e.g., TI = 200 • C for the insulation under investigation) is 20 000 h, exceeding the temperature by 50 • C above TI (i.e., 250 • C) dramatically reduces the lifetime to approximately 625 h.To investigate the impact of higher temperatures on insulation performance over an extended period, the accelerated thermal aging is carried out at 250 • C for 156 and 312 h, corresponding to approximately 25% and 50% of the overall insulation lifetime, respectively.Consequently, the results obtained from this experimental study pertain to the turn-to-turn winding insulation and its behavior up to 50% of the insulation's total lifespan.The decision to limit the aging duration to the timeframe of 312 h is also driven by the objective of preventing excessive brittleness of the insulation, Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.which could lead to the formation of cracks during the sample preparation process when wrapping the thermally aged pairs with PTFE.In other words, the insulation's brittleness at 250 • C and after 312 h reached a manageable level, making it possible to wrap the wires with PTFE by hand without causing cracks in the insulation.Thus, the chosen approach allows for a focused investigation of the impact of thermal aging without the interference of additional factors arising from excessive brittleness or cracks.Fig. 5 shows the insulated wires in the oven before starting the accelerated thermal aging process.After finishing each thermal aging period, the two portions of insulated wire (i.e., straight or bent wires) are wrapped in PTFE tapes.Therefore, the test samples are categorized into three main groups from the aging condition point of view: pristine samples and thermally aged specimens for 156 and 312 h.For each aging condition, four sets

B. Common Measurement Conditions: PD and (DDF, IC) Tests
All the measurements are performed at room temperature (20 • C), atmospheric pressure (1013 mbar), and relative humidity (30% ± 5%).The environmental conditions are controlled by a fully controlled Memmert pass-through oven.The specimens are located inside the oven, with one wire of the turn-to-turn samples connected to the ac power supply and one to the ground (Fig. 6).All the tests for each combination of the turn-to-turn samples construction (i.e., straight and bent: three different radii) and aging conditions (i.e., pristine and thermally aged: 156 and 312 h) are performed using five specimens to collect the set of data for each case study.It is important to emphasize that PD is a destructive phenomenon, especially when it occurs in organic insulation, such as polyamide-imide.In this context, PD is regarded as a crucial insulation endof-life criterion [7].Upon conducting repeated tests on the same specimen, it is observed that the PDIV values decreased.Therefore, each sample is tested only once to prevent the possible impact of previous measurements (e.g., PDIV drop caused by previous discharge activities) [11].

C. AC PDIV and PDEV Testing
PD tests are performed under a sinusoidal waveform excitation at a power supply frequency of 50 Hz.It is noteworthy to highlight that the IEC [7] allows PDIV testing using either sinusoidal or impulsive voltages to be used for the turnto-turn insulation qualification of inverter-fed motors.The PDIV values tend to be lower when subjected to sinusoidal excitations compared to that of PWM with rise times shorter than 1 µs [12].Even when PDIV is measured using a 50-Hz sinusoidal power source, the results remain consistent, offering a slightly more cautious evaluation of PDIV for turn-toturn insulation under the influence of a two-level inverter or surge generator, as discussed in [13], [14], and [15].In addition, comparative PDIV tests conducted on twisted pairs as described in [16] revealed that ac excitations result in the lowest PDIV values compared to PWM.Based on these findings, a conservative approach is adopted in this investigation, where PDIV measurements are carried out under ac 50-Hz excitations.Apart from that, it should be noted that the primary objective of this article, however, is to analyze and evaluate how the bending radius affects PDIV.This evaluation can be conducted using any waveform, but it is crucial to maintain a consistent waveform throughout the study.
The ac source is GPT-9802, manufactured by GW Instek.The peak voltage is raised in steps of 10 V every 30 s [8], [17] and monitored through a Teledyne LeCroy WaveSurfer 510 oscilloscope (1)-GHz bandwidth and 10-GS/s sampling rate) using a CT4079-NA differential probe (50-MHz bandwidth, 2000:1 voltage ratio, and 50-impendence).The peak value of the voltage is recorded as PDIV as soon as PD activity is initiated.After measuring PDIV, the peak voltage is reduced in steps of 10-V peak each minute [8], [17].When the PD activity is extinct, the peak voltage is recorded as PDEV.
PD detection is carried out with a conventional indirect circuit schematized in Fig. 7.A PD-free 1-nF coupling capacitor is used in parallel with the test specimen to amplify the PD signal, improving the detection sensitivity.The PD sensor to acquire the PD pulses is a ferrite-core high-frequency current transformer (HFCT) with 1-60-MHz bandwidth manufactured by Techimp HQ.The discharge current signal is recorded and processed through IEC 60270-compliant instrumentation (i.e., Techimp PDBaseII detector).The PDBaseII features a range of frequency acquisition from 16 kHz to 48 MHz, a sampling rate of 200 MSa/s, and delivers the PD pulse waveform [18].The PD calibration is accomplished for each test specimen separately using PDCAL PLUS manufactured by Techimp HQ, which can produce up to 800-pC pulses.It is worthwhile to note that the PD BaseII is equipped with specialized IEC software that allows for the conversion of PD charge amplitude values from millivolts (mV) to picocoulombs (pC) in the pattern after calibration.This functionality ensures that the PD magnitude is measured within a specific frequency bandwidth, consistent with the range defined by IEC 60270, which spans from 115 to 440 kHz.

D. DDF and IC Measurements
The DDF and IC tip-up testing (i.e., tanδ and C) of each turn-to-turn sample are accomplished at 50 Hz using a sinusoidal voltage waveform by adopting a Megger Delta4000 [19].The peak of applied sinusoidal voltage at 50 Hz is increased from 0 to 1.27 kV with a step of 39 V.The choice of selecting 1.27 kV as the voltage level is driven by the intention to use a value well above the PDIV.This higher voltage level allows for the clear visualization of two distinct regions in the dissipation factor curve compared to the applied voltage.The first region in the curve exhibits an almost flat dissipation factor versus the applied voltage, indicating the presence of conduction and polarization losses.The second region, on the other hand, displays a rising trend of the dissipation factor as the applied voltage increases, signifying the occurrence of ionization losses.As for the choice of 39 V, it is made to ensure that there are sufficient data points after ionization has taken place.This decision ensures that the trend in the ionization region can be observed and analyzed with clarity.Five specimens are tested for each case study.Also, each sample is measured once to avoid the influence of the previous measurement.Finally, the average value is reported.

III. EXPERIMENTAL RESULTS AND DISCUSSION
The PD measurement results are depicted and plotted as a function of thermal aging time using boxplots to provide a clear visual insight.Each box summarizes a dataset by showing the median value q 2 (central line in the box), the 25th, q 1 , and 75th, q 3 , percentiles (the edge of the box), and the data range [using whiskers at q 2 -1.5 × (q 3 -q 1 ) and q 2 + 1.5 × (q 3 -q 1 )].The small square symbol inside each box demonstrates the mean value [17], [20].The connection line of the boxes relates to the mean value, illustrating that the PD quantities (i.e., PDIV, PDEV, PD charge amplitude, and repetition rate) trend as a function of bend radius for each thermal aging period.

A. PDIV and PDEV Analysis
Fig. 8 shows the measured peak value of PDIV as a function of thermal aging time for straight and bent samples.Table I shows the mean values of PDIV corresponding to Fig. 8.
Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.According to Fig. 8 and Table I, PDIV diminishes as a function of thermal aging time for all the samples, regardless of the bend radius.Considering the fresh specimens and each thermal aging period (i.e., 156 and 312 h), interestingly, bent turn-toturn samples deliver a higher PDIV than the straight ones.It means that the bending process with a bend radius down to 3.5 mm does not lead to fraction/cracking in the insulation.Although the bent samples with a larger bend radius provide a higher PDIV before starting the aging process (i.e., for the pristine specimens), the results after thermal aging show that the bent samples with the bend radius of 6.5 mm provide the highest PDIV values, suggesting an optimum bend radius for the edgewise insulated wire.
Fig. 9 shows the measured peak value of PDEV as a function of thermal aging time for straight and bent samples.Table II reports the mean values of PDEV corresponding to Fig. 9. Fig. 9 and Table II show that similar to PDIV, the bent turn-to-turn samples deliver higher PDEV values than the straight samples.The bent specimens with a bend radius of 6.5 mm provide the highest PDEV both before and after the thermal aging process.

B. Correlation Between PDIV and Insulation Thickness Shrinking
The reason for the PDIV mitigation consequent to thermal aging is most likely due to the insulation thickness reduction caused by the polyamide-imide vaporization resulting from the thermal aging at 250 • C. Fig. 10 shows the insulation thickness shrinking of the edgewise insulated wire as a function of thermal aging time for both flatwise and edgewise directions.Each bar chart reports the mean value of 32 measured samples.The insulation thickness is reduced by 16.9 and 15.47 µm for flatwise and edgewise sides, respectively, after 312 h of thermal aging at 250 • C. Fig. 11 shows the correlations between PDIV drop and insulation thickness shrinking (i.e., flatwise side) due to thermal aging.The values are normalized to unaged straight mean.

C. DDF and IC Analysis
The reason for a slightly higher PDIV for the bent specimens than the straight ones can be ascribed to the bent samples' higher conducting losses and lower capacitance [21].Fig. 12(a) and (b) shows the variation of starting value or absolute DDF value, tanδ 0 , indexing the conducting and  polarization losses [22], and capacitance, C 0 , at a low voltage (39 V rms ) as a function of bend radius and thermal aging time, respectively, normalized to unaged straight mean.
Fig. 12 shows that the bent turn-to-turn samples have higher conducting and polarization losses (i.e., tanδ 0 ) and lower capacitance (C 0 ) regardless of the aging condition of the specimens.In addition, Fig. 12 shows that the bent samples with a bend radius of 6.5 mm, which deliver the largest PDIV and PDEV after thermal aging (Figs. 8 and 9), feature also the highest and lowest tanδ 0 and C 0 , respectively, after starting the thermal aging process.

D. Equivalent Charging Circuit for the Bent Turn-to-Turn Samples
Fig. 13 shows the equivalent circuit of the bent samples to clarify the contribution of the bending point resistance and capacitance to delivering a higher PDIV.It is worthwhile to recall that dielectric materials can be represented using either parallel or series equivalent circuits, and both approaches are considered acceptable.However, for this specific case, the series configuration is chosen, linking the straight part to the bending point.The reason behind this choice is that the series configuration allows for easier management and provides a clear and distinctive view of both the straight part and the bending point of the bent turn-to-turn wire insulation.In Fig. 13, R BP and C BP are bending point resistance and capacitance, respectively, V BP is the voltage drop across the bending point, V SP is the voltage across the straight part, R c and C c are resistance and capacitance of defect where discharges occur, respectively, R b and C b are resistance and capacitance of the dielectric in series with the defect, respectively, and R a and C a are the remaining resistance and capacitance of the test sample, respectively.R BP and C BP have a reverse and direct relationship with the insulation cross section at the bending point, A, respectively According to the equivalent circuit introduced in Fig. 13, the bending point impedance, Z BP , is equal to Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.When the edgewise insulated wire is bent, and if the insulation is still healthy at the bending point (i.e., free from any fraction/cracking), the insulation cross section, A, reduces due to squeezing at the bending site, leading to an increase and decrease of R BP and C BP , respectively.Therefore, the bending point impedance, Z BP , increases, resulting in a higher voltage drop across the bending point, V BP .Eventually, a higher test voltage, V test , thus PDIV, will be needed to incept PD.Considering (3), it can be speculated that the PDIV difference between straight and bent samples can be negligible if the supply frequency increases due to the reduction of Z BP .

E. Dispersion of PDIV and PD Charge Amplitude Under PDIV
Figs. 14 and 15 quantify the dispersion level of the peak value of PDIV and PD charge amplitude, q PD , under PDIV consequent to thermal aging time for straight and bent samples, relying on the shape/slope parameter of the two-parameter Weibull distribution (i.e., β).Fig. 14 shows that β PDIV steadily reduces or PDIV dispersion increases for the straight samples as a function of the thermal aging period.For the bent turn-to-turn specimens, while it promotes after 156 h of aging, the opposite occurs for a longer thermal aging period.β PDIV diminishes/PDIV dispersion increases for all the samples after 312 h of thermal aging time, manifesting a diagnostic marker for the thermal aging of polyamide-imide.
Regarding the PD charge amplitude, at least 2000 PD pulses are acquired to perform the Weibull analysis automatically using the PD BaseII software.Fig. 15 shows that β of q PD increases or PD charge amplitude dispersion decreases as a function of the thermal aging period, manifesting another aging indicator for polyamide-imide.In addition, Fig. 15 shows that before 312 h of thermal aging, the contribution of surface PD is more than the internal PD activity (i.e., β < 2).However, the opposite occurs after 312 h (i.e., 2 < β < 4.8) when more contribution of internal PD in the voids caused by the thermal aging process would be plausible [23].After 312 h, β of q PD for r Bend = 6.5 mm is still lower than two, verifying the less possibility of void creation in the insulation consequent to the thermal aging compared with other bend radii.repetition rate, N w , in pulse per period (ppp), respectively, under PDIV over the thermal aging period, comparing the straight samples versus the bent ones with different bend radii.
In general, both Q max,95% and PD repetition rate are decreased as a function of thermal aging time, which can be attributed to lower PDIV for the thermally aged samples (Fig. 8).After 312 h, Q max,95% is comparable for different bend radii [Fig.16(a)], while the PD repetition rate is higher for smaller bend radii [Fig.16(b)].
To better describe and compare the straight samples versus the bent ones based on harmfulness or damage associated with PD activity under PDIV through thermal aging time, a diagnostic dimensionless quantity known as the PD severity index (I d x) can be introduced such as one devised in [24], [25], [26], and [27], by simply multiplying Q max,95% and N w values as follows: Fig. 17 reports the I d x trend during the thermal aging time, comparing the straight edgewise insulated wires against the bent ones with different radii, using (4) based on the mean values reported in Fig. 16.I d x shows the lowest and highest values for r Bend = 6.5 and 8.5 mm, respectively, for unaged turn-to-turn samples.However, after thermal aging, I d x is minimum for the straight turn-to-turn specimens while manifesting a higher value for a smaller bend radius.As a result, the harmfulness or damage associated with PD activity is higher for the bent turn-to-turn samples than the straight ones consequent to thermal aging.In addition, there is a reduction trend for I d x with respect to the thermal aging time, which can be ascribed to lower PDIV for the thermally aged specimens (Fig. 8).

G. DDF TIP-UP Testing (Fresh Versus
Thermally Aged Four 312 h) Fig. 18 summarizes the measurement results of tanδ tip-up using a sinusoidal waveform at 50 Hz, comparing the straight with bent turn-to-turn samples and the pristine specimens versus the thermally aged ones for 312 h.Fig. 18 shows  that the thermally aged straight turn-to-turn samples give the highest DDF after ionization (i.e., when tanδ steeply starts to rise).Considering only the unaged specimens or the thermally aged ones, the straight samples deliver the highest Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
ionization losses since the more electric field is concentrated across the defect once ionization starts due to the absence of R BP (Fig. 13).For the bent turn-to-turn samples, tanδ starts to increase due to ionization at a higher voltage level than the straight specimens, verifying the obtained results via PD tests (i.e., PDIV STR.< PDIV Bent ).

H. Correlation Between PDIV and Dissipation Parameters
Fig. 19(a) and (b) shows the correlation plots between the PDIV and the dissipation parameters of tanδ H−L and C H−L , respectively, consequent to the thermal The value of tanδ H−L is the difference between the DDF corresponding to the highest applied voltage (i.e., 900 V rms ) and the DDF measured at the lowest applied voltage (i.e., 39 V rms ).A similar definition stands for C H−L , introducing the capacitance difference between the lowest and highest applied voltages.The values in Fig. 19 are normalized to unaged straight mean.
A higher value of tanδ H−L and C H−L can be a marker for enhancing the number and size of voids or the formation of delamination in the insulating system [21], [22].Overall, Fig. 19 shows that the bent turn-to-turn specimens with a radius of 6.5 mm deliver a lower tanδ H−L and C H−L , implying less possibility for voids or delamination in the insulation system.In addition, Fig. 19 shows that higher dissipation parameters (i.e., tanδ H−L and C H−L ) reflect lower PDIV values, indicating a good correlation.

IV. CONCLUSION
This contribution shows that bending edgewise insulated wires carried out during coil manufacturing can result in higher PDIV and PDEV than the straight turn-to-turn insulation.It is demonstrated that among the considered bend radii, r Bend = 6.5 mm gives the highest PDIV and PDEV after thermal aging of the turn-to-turn samples.Moreover, higher conducting and polarization losses (i.e., tanδ 0 ) and a lower turn-to-turn capacitance belong to the bent specimens, especially r Bend = 6.5 mm, after thermal aging.It means that bending wires result in PDIV rise in the cost of increasing losses, which should be considered during coil manufacturing.Regarding the capacitance reduction due to bending, it should be mentioned that it can be favored for inverter-fed motors using steep-fronted square waveform excitations.It is due to this fact that a lower capacitance of the coil helps the rise time to be shorter, closer to the one designed for the inverter's waveform, while the opposite limits the rise time of the excitation.
It is substantiated that PDIV drop after thermal aging is due to the insulation thickness shrinking.The severity index calculations show that the harmfulness or damage associated with PD activity can be higher for the bent turn-to-turn samples with a smaller bend radius after thermal aging.In addition, two aging indicators are introduced for the edgewise insulated wire (i.e., polyamide-imide): β PDIV and β of q PD .The former and latter decrease and increase, respectively, consequent to thermal aging after 312 h.Finally, good correlations between PDIV and dissipation parameters (i.e., tanδ H−L and C H−L ) are introduced for edgewise insulated wires.To create a coil, edgewise enameled winding wires must typically be bent at 90 • .The acquired experimental results are useful for coil makers because they show how the bending radius can affect the dielectric characteristics of turn-to-turn insulation and introduce an ideal radius that can show superior insulation performance.

Fig. 5 .
Fig. 5. Edgewise insulated wires inside the oven before the accelerated thermal aging process.
(i.e., straight, 8.5, 6.5, and 3.5 mm) of ten (i.e., five samples for PD tests and five specimens for DDF and capacitance tests) PTFE-wrapped pairs each are manufactured out of the edgewise winding wires.

Fig. 7 .
Fig. 7. Circuit and connections layout for PD test setup.

Fig. 12 .
Fig. 12. Variation of (a) starting value or absolute DDF, tanδ 0 , and (b) capacitance, C 0 , at a low voltage (39 V rms ) as a function of bend radius and thermal aging time, normalized to unaged straight mean.

Fig. 14 .
Fig. 14.PDIV dispersion level as a function of thermal aging time.

Fig. 15 .
Fig. 15.PD charge amplitude dispersion level as a function of thermal aging time under PDIV.

Fig. 16 .
Fig. 16.Q max,95% and N w trend of PD detected under PDIV during the thermal aging time for straight (STR.)specimens versus bent ones with three different radii.

F
Fig.16(a) and (b) report the trend of Q max,95% (the 95th percentile of PD charge amplitude distribution) in pC and PD

Fig. 18 .
Fig. 18.Variation of DDF as a function of test voltage measured at 50 Hz for the straight turn-to-turn samples with the bent ones and the unaged specimens versus the thermally aged ones for 312 h.

Fig. 19 .
Fig. 19.Correlation plots between the PDIV and the dissipation parameters of (a) ∆tan δ H−L and (b) ∆C H−L , normalized to unaged straight mean.

TABLE I MEAN
VALUES OF PDIV RELEVANT TO FIG.8

TABLE II MEAN
VALUES OF PDEV RELEVANT TO FIG.9