Non-Destructive Electromagnetic Evaluation of Material Degradation Due to Steel Cutting in a Fully Stacked Electrical Machine

Abstract

Although it is well known that the magnetic properties of electrical steel can be deeply affected by the cutting process, it is still not clear how to accurately evaluate these effects on a prototyped machine on its final shape, especially at a high frequency or a high power rating. This research provides a more practical method for accurate measurement of magnetic losses in electrical steel with consideration of material degradation effects due to the cutting process. Unlike other similar studies, these investigations are conducted not only on a few laminations but also on a complete electrical machine core. For a fair comparison between both cases, backlack bonding is used for lamination stacking since it is the most non-damaging joining technique. Two different test setups are used to measure the steel performance at a wide range of frequency and input power. Furthermore, a full axial length stator of a switched reluctance machine (SRM) is used as a case study to identify the magnetic properties of NO20 electrical steel. Additionally, by comparing the results obtained from the individual laminations and the assembled stator, the extra losses due to the cutting process and material degradation are extracted accurately.

1. Introduction

With the significant growth in the different industrial applications, the interest in high power density electrical machines is remarkedly increasing. One of the main components that define the magnetic loading capacity in the design of electrical machines is the electrical steel, which is usually iron alloy in form of coated laminations with specific magnetic properties such as small hysteresis area or low core loss. For instance, a bad selection of core material could cause an excessive amount of losses as well as poor efficiency. Therefore, choosing electrical steel with high permeability as well as economic efficiency is necessary [1,2,3].

Recently, many legal regulations regarding the manufacturing of industrial motors are calling for increased efficiency. Accordingly, many manufacturers are deeply interested in the design of energy-efficient electrical machines [4,5]. A precondition for such an ideal motor engineering is the use of low-loss electrical steel combined with high magnetic permeability. Most importantly, material processing by careful selection of the joining technique can effectively help in maintaining of magnetic properties of the material used.

The family of high-performance non-oriented electrical steel (NO) is not just thin silicon-steel sheets with minimum power loss. It extends to be a new collaborative approach to the development of pioneer applications for low-hysteresis laminated steel with maximum magnetic permeability. Starting from NO10 with a lamination thickness of 0.1 mm up to 0.3 mm in NO30, NO steel is considered one of the widest range of grades in the market. Moreover, according to standards (EN 10303:2015 and IEC 60404-8-8:2017), the NO20 not only fulfills the standard requirements but also has better magnetic and mechanical properties, i.e., (the specific iron loss, manufacturability, etc.), compared to many other steel grades with the same nominal thickness.

Basically, when the rotation of the magnetic field is associated with alternative currents, core losses in the electrical steel occur due to two main factors which are eddy current and hysteresis components. Many mathematical models and analytical methods have been used to estimate these losses in the steel laminations such as the Bertotti approach and Steinmetz equation [6,7,8,9,10,11,12,13]. However, most of these studies did not account for the manufacturing effects, such as laser cutting or lamination stacking. Only a few studies focused on the manufacturing effects [14,15,16,17,18]. However, even those studies were conducted on only a few laminations with certain limitations on the frequency and the input power. To the best of the authors’ knowledge, there is no published literature related to the core loss measurements of a full axial length core of an electrical machine, especially those which require higher input power and a wider frequency range.

In this paper, a more practical investigation is carried out to study the material degradation due to electrical steel cutting. The main target is not only to assess the manufacturing effects associated with steel cutting on the magnetic characteristics of the individual laminations but also to accurately evaluate and measure the overall degradation effect on the performance of the magnetic material of a full stack core. Different test setups are used for the characterization of the steel samples at a wide range of frequency and input power. Moreover, this investigation includes a case study for a stator core of an SRM made from NO20 steel. Furthermore, a non-damaging joining technique is used to stack the laminations. The standard mechanical, physical, and magnetic properties of NO20 are shown in Appendix A. Additionally, a comparison between the different joined and un-joined samples is presented for accurate measurements of the extra core losses due to the cutting process.

The paper is arranged as follows. In Section 2, an overview of material degradation due to electrical steel cutting is presented. Section 3 will be devoted to magnetic measurement and core losses computations. In Section 4, a case study of SRM will be explored and the laser-cutting process will be evaluated. Finally, conclusions will be provided in Section 5.

2. Degradation Effects Due to Cutting Process and Lamination Stacking

It is widely known that electrical steel cutting will not only introduce mechanical deformations near the cutting edge, but also it will impair the energy conversion process. Different topologies can be used to cut the laminations of an electrical machine such as wire electrical discharge machining [19], water jetting [20], punching [21,22], and laser cutting [23,24,25]. Electrical discharge machining and water jet cutting are the ones with the least effect on the magnetic properties of electrical steel, but both techniques have relatively low speeds so they are only used for small quantities. On the other hand, punishing and laser cutting are known to have the highest impact on the magnetic material quality. In the manufacturing of large quantities of electrical steel laminations, the punching technique is considered to be reasonable in terms of costs. However, in small quantities, the laser cutting method is more economically feasible.

Depending on the cutting technique, the insulation layer on the steel lamination is subjected to damage and stresses on the cutting edges [26]. This may also lead to short circuits during the core stacking. Additionally, the electrical steel itself may be subjected to thermal stress and other changes in microstructure level leading to the impairment of the magnetic properties. All these reasons will result in extra core losses in the electrical steel after the cutting phase.

Many researchers have studied the degradation effect on static and dynamic magnetic properties [27,28]. In [29], the deterioration of the electrical steel is examined from a micromagnetic point of view using the different aforementioned cutting techniques. In [30], the local material degradation near cutting edges is analyzed and modeled. Furthermore, the effect of the cutting speed is studied. Moreover, the local magnetic hysteresis properties are modeled and compared with the macroscopic measurements. In [31,32,33,34,35], the effect of different cutting techniques is studied on different grads of industrial electrical steel laminations, such as NO27, M100-23P, M270-35A, and M400-50A.

After the laminations are cut, it goes through another assembling process which indeed affects the steel characteristics. During this phase of the manufacturing process, a lot of physical modifications happen to the steel lamination. Some of them are made on purpose to account for other factors such as fixing purposes on the outer circumference of the stator. Other random deformations may occur incidentally during the compression of the laminations.

Different stacking methods can be used for lamination joining such as interlocking, welding, clamping, or bonding [36,37]. The lamination joining is important for creating a stable core with minimum eddy current losses. The selection of the technique depends mainly on the electrical machine application, and design with the consideration of other economic aspects. For instance, conventional laser welding is fast and cheap but it causes extremely high losses for certain applications due to the local material degradation and the interlaminar short circuit path between laminations along the entire stack length. On the other hand, face-to-face gluing is relatively expensive, but it causes no damage to the lamination, and accordingly, the core losses are much lower.

In this paper, in order to study the effect of lamination laser cutting accurately on the full stacked core, backlack bonding is selected for the case study SRM because the full-face bonding of steel lamination is considered the most non-damaging and flexible process among other methods. This technique combines optimized performance and high efficiency for different reasons:

• It allows complete freedom of design without accounting for any locations for welding or interlocking.
• No impact on magnetic properties.
• Improved heat dissipation and higher thermal conductivity. Furthermore, a better heat transfer between the laminations and the housing, thus a lighter and lower cost cooling unit can be utilized.
• Dimensional accuracy of the stacked laminations with the narrowest manufacturing tolerances.
• Mechanical stability without any tensions or heat expansion is similar to those caused by traditional welding.
• Damping acoustic noise and vibration of the laminations to minimize sound levels in the electrical machines, especially at high frequencies [38,39].

3. Magnetic Material Computations

In order to perform the magnetic measurements for a steel sample, a double coil setup is frequently used. Hence different tools can be used such as Epstein frame [40,41], single sheet tester [42], or ring core [43]. However, the main function stays the same, which is creating a common magnetic path for both primary and secondary coils. This study only uses the ring core technique in order to avoid the air gap that can be found on the other tools.

Figure 1 shows the magnetic measurements schematic for a double coil setup. As noticed, an amplifier is usually used as a variable AC input source to supply the primary coil and produce flux through the closed magnetic circuit. The secondary induced voltage is then measured and integrated to obtain the magnetic flux density. Then, the data acquisition system instantaneously records the measured data and is also used to compare the measured signal with the reference one in order to generate the control signal to obtain the desired magnetic flux. Additionally, the model should also provide an offset control in order to detect and compensate for any small dc voltage measured in the secondary voltage. That undesired dc small component will create a remarkably increasing ramp in the flux density waveform which increases continuously with time after the integration of the secondary voltage.

Measuring the magnetic properties of a material requires measuring the field strength $H\left(t\right)$ and the magnetic flux density $B\left(t\right)$ simultaneously. The field strength can be estimated from the measured primary winding current $i_p\left(t\right)$ as follows:

$$H\left(t\right)=\frac{N_p}{l_{Fe}}·i_p\left(t\right)$$

(1)

where, $N_p$ is the number of turns in the primary winding and $l_{Fe}$ is the average magnetic path length. The corresponding time varying magnetic flux density can be obtained by the integration of the back EMF induced into the secondary coil which has a number of turns of $N_{sec}$:

$$B\left(t\right)=\frac{1}{N_{sec}{A}_{Fe}} {\displaystyle \int }{V}_{Sec}\left(t\right) dt$$

(2)

where, $A_{Fe}$ is the core cross-section area. Additionally, the average value of the specific core loss in W/kg is obtained from the measured dynamic $B\left(H\right)$ loop, during one period of the fundamental frequency $f$ as shown in the equation:

$$P_{core}=\frac{f}{\rho } {{\displaystyle \int }}_0^{1/f}H\left(t\right) \frac{dB\left(t\right)}{dt} dt$$

(3)

or another alternative from

$$P_{core}=\frac{N_p f}{N_{sec} A_{Fe} l_{Fe} \rho } {{\displaystyle \int }}_0^{1/f}{i}_p\left(t\right) V_{sec}\left(t\right) dt$$

(4)

where $\rho$ is the mass density of the electrical steel in kg/m³.

Typically, the mean magnetic path length of the ring core ($l_{Fe}$) is determined from the yoke’s outer and inner diameters $(D_o, D_i)$ using this formula $\pi \left(D_o+D_i\right)/2$. However, this is only applicable if the magnetic field is uniform over the cross-section area. In a nonuniform magnetic field, the magnetic path length is calculated as follows [44]:

$$l_{Fe}=\frac{\pi \left(D_o+D_i\right)}{ln (\frac{D_o}{D_i})}$$

(5)

Using finite element modeling (FEM), the stator of the case study SRM, with the dimensions listed in

Table 1. SRM machine stator dimensions.

Parameter	Value
Stack length	78.5 mm
Stator outer diameter	120 mm
Stator inner diameter	62.5 mm
Yoke height	11 mm
Pole arc	30°
Pole width	17.5 mm

, is simulated using the physical and magnetic properties of NO20 listed in Appendix A. The results of the magnetic flux density in NO20 are shown in Figure 2. As can be seen, the toroidal primary coil around the stator yoke allows the flux to travel through the closed magnetic path of the yoke with a maximum magnetic flux density of 1.6 T. This value has been verified more practically through the experimental work. It is also clear from the FE model that the magnetic field is not uniform over the cross-section area due to the fringing of the flux lines near tooth parts. Therefore, Equation (5) is used for the estimation of the magnetic path length.

4. Sample Preparation and Measurement Results

Two different setups are used to test the different samples of the SRM at wide ranges of frequency and input power. The specifications of both experimental setups are shown in

Table 2. Test setup specifications.

Experimental Setup	Setup 1	Setup 2
Input power requirements	Up to 1 kVA	1–25 kVA
Effective frequency range	5–400 Hz	50 Hz–5 kHz
Data acquisition card model	National Instrument 6366	dSP MicroLabBox 1202
Secondary voltage limits	±10 V	±1000 V
Sampling time	1 × 10−3	1 × 10−4
Tested sample example	Few laminations	Complete core
Selected input source	KEPCO	Spitzenberger
Model type	BOP 50–8M	PAS 10,000
Peak output power	800 VA	10,000 VA
Source max voltage	50 V	270 V
Source max current	8 A	50 A

. As can be noticed, Setup 1 is preferable at low power (up to 1 kVA). That is for higher sensitivity and lower distortion in the waveforms, especially with low impedance loads. The effective maximum frequency range is practically limited to 400 Hz. In some limited cases, the frequency can be increased to 1 kHz. However, the number of data samples in each cycle will remarkably reduce due to the relatively lower sampling time, which in turn affects the accuracy of the results.

As for the 2nd test setup, it is used specifically for high power ratings and high-frequency operations. So, a powerful amplifier will be required, 1–25 kVA based on the size of the tested core. Furthermore, a DAQ with a higher sampling time will be used so that the system can perform the measurements accurately at the targeted high frequency (1 kHz), or even higher if necessary. Moreover, a higher rating voltage and current measurement devices will be required.

4.1. Individual Lamination Characterization

First, a 5-lamination sample is tested under the first setup at low frequencies starting from 5 Hz up to 400 Hz. The specifications of the primary and secondary coils are listed in

Table 3. Specifications of double coil test samples.

Test Sample	Sample 1	Sample 2
Material	NO20	NO20
Description	5 laminations	SRM full stator
Manufacturing process	Laser cutting	Laser cutting and backlack joining
Primary wire	0.25 mm2–8 Amp	10 mm2–80 Amp
Primary coil	150 turns	24 turns
Secondary oil	10 turns	10 turns

. The test platform is shown in Figure 3. As can be seen, an 800 VA bipolar operational power amplifier (Model: KEPCO) is selected as an input source for this small test sample. Furthermore, the primary current waveform is measured using a current sense resistor. Additionally, an analog integrator circuit is used to obtain the magnetic flux density waveform from the secondary coil voltage. Finally, a NI 6366 DAQ is used for interface and data acquisition. The core losses are measured at the different pre-mentioned frequencies and flux density levels as shown in Figure 4. In Figure 4a, the gap between hysteresis losses and dynamic magnetic losses is demonstrated by an example of 5 Hz and 50 Hz, respectively. Moreover, The losses are shown in Figure 4b for two different magnetic induction levels (1 T and 1.5 T) with respect to all the frequencies. These losses will be compared later on with the standard losses of NO20 in order to determine the extra losses caused by the laser cutting of the laminations.

For additional verification, the same 5-lamination sample is tested at the other setup, which will be described in detail in the next part. The comparison between the results at 100 Hz is shown in Figure 5a. As can be noticed in the waveform of B(t) and H(t), there is a very good agreement between both setups under the same operating condition of magnetic flux density. There is also a slight mismatch between the BH loops of both setups which can be explained by the lower sensitivity of the second setup, which is designed for higher impedance loads.

4.2. Characterization of Full Axial Length Stator

After the backlack-coated laminations are cut, the core is assembled and put through a two-step thermal treatment process. In the first phase, the temperature is only high enough to decrease the viscosity of the coating until complete melting. Then, the temperature is further increased in the second phase so that the backlack loses its thermoplastic properties and the adhesive strength increases remarkably until it reaches its hard form.

The double coil system is then prepared in the full axial length stator for magnetic material characterization as shown in Figure 6. As can be noticed, a very thick wire with rubber coating is used specifically for high power requirements. The primary turns are uniformly distributed around the stator yoke to insure uniform magnetic flux density along the magnetic path length. Furthermore, the secondary search coil is placed as close as possible to the stator yoke, so that it does not detect any stray or leakage flux. The specifications of the primary and secondary coils are listed in Table 3.

The complete stator is then tested on the 2nd setup that is shown in Figure 7. As can be seen, a 10 kVA amplifier (Model: Spitzenberger [45,46]) is selected to perform the test. Furthermore, voltage and current probes are used to measure the required V and I signals. Additionally, a dSpace MicroLabBox 1202 is used for data acquisition with a sampling time of 0.1 ms.

The test is performed for a wide range of frequencies starting from 100 Hz up to 1 kHz. The results for one of the test frequencies—400 Hz—are presented in Figure 8, Figure 9, Figure 10 and Figure 11. Figure 8a shows the flux density variation with time at different excitation levels which has a sinusoidal behavior similar to the waveforms of the induced voltage of the secondary search coil shown in Figure 8b. The 90° phase shift between the two waveforms is mainly because of the integration. As for the waveforms of the primary current and magnetic field, both waveforms have the same profile and phase angle as shown in Figure 8c,d, respectively. The distortion in this waveform is more visible at high current values near saturation level. Moreover, the simultaneous waveforms of B and H are plotted together to form the BH loop as shown in Figure 9. Furthermore, the maximum flux density point is spotted on each loop in order to draw the normalized BH curve which is shown in Figure 10a. Moreover, the relative magnetic flux permeability (µr) of NO20 is calculated at each level of magnetic flux density as demonstrated in Figure 10b. As can be noticed, µr reaches its maximum point at low induction and as B increases, µr decreases gradually until it reaches its lowest value at the saturation magnetic flux density. Additionally, the multiplication of B and H of each hysteresis loop is measured instantaneously with time as shown in Figure 11a. This expression represents the area of the hysteresis loop, which will be used for the calculation of the core losses as explained back in Equations (3) and (4). Accordingly, the specific core losses for each loop are measured in W/kg as shown in Figure 11b.

It is also worth mentioning that the magnetic field distortion tends to be very high at low frequencies, especially at high magnetic flux densities, due to the low impedance and the extremely high currents. An example at 100 Hz is demonstrated in Figure 12. As can be noticed, as the magnetic induction increases from 0.5 T to 1 T, the total harmonic distortion increases by 8%. However, at 1.5 T, the THD increases remarkably by over 25% compared to the case of 1 T.

All the aforementioned results are measured for all frequencies up to 1 kHz. The top magnetic flux density—practically achieved—decreases steadily as the frequency increases. Furthermore, the corresponding input source power required is measured as well as shown in Figure 13a. The source power reaches its maximum point at 500 Hz with 400 Hz in the second place. Additionally, the input current and voltage components are shown in Figure 13b. It is clear that as the frequency increases, the impedance increase and the input current decreases, and the input voltage increases.

Moreover, the variation of the magnetic flux density (B) with its own time derivative (dB/dt) at each frequency is shown in Figure 14a. As can be noticed, for the same B value, dB/dt remarkably increases as the frequency rises, which will cause more core losses as was mentioned in Equation (3) despite having the same magnetic induction. By a way of example, under the same magnetic induction of 1T, dB/dt increase from near 400 T/sec at 100 Hz to 4300 T/sec. Correspondingly, the specific core losses increase from only 6 W/kg at 100 Hz, remarkably, to 70 W/kg at 1 kHz under the same 1 T flux density as shown in Figure 14b.

The same example is investigated once again but using the hysteresis loop profile as shown in Figure 15. It is obvious that there is a big difference between the area of the BH loop at a relatively low frequency of 100 Hz compared to its area at 1 kHz at the same magnetic flux density. At the low frequency (i.e., low dB/dt), the loop is slim and its area represents the hysterics losses only. However, at the higher frequency (i.e., high dB/dt), the loop is much wider and its area represents the dynamic magnetic losses. Subsequently, the core losses rise due to the higher dB/dt. This is also more clear in the thermal images of the core in Figure 16. As can be noticed, the core temperature is around 35 °C at 100 Hz, but it can exceed 90 °C at certain points at 1 kHz.

4.3. Core Losses Comparison

In this part, the core losses of NO20 are compared after the different machining and assembly stages. Furthermore, the core losses due to manufacturing effects are calculated separately to evaluate the final core performance. In Figure 17, the specific core losses (W/kg) of the full stator are compared with those of the standard losses and the case of the individual lamination at different frequencies. There is a remarkable increase in the case of the individual unbonded lamination compared to the standard losses. This difference presents the effect of laser cutting on lamination. Moreover, there is an additional increase in the case of the full stator with respect to the laser-cut lamination with at least an equal value. This is because of the overall degradation effect after the assembly and bonding processes as well as the associated heat treatment.

By subtracting the standard core losses (obtained from the supplier datasheet [47]) from their values in the case of the laser-cut lamination sample, the extra core losses due to laser cutting of the enjoined steel laminations can be accurately measured at different flux densities and frequency levels as shown in Figure 18. Additionally, in order to extract iron losses due to the overall degradation effect in the backlack-joined stator, the difference between the core losses in the complete stator and the standard NO20 sheets is calculated.

In Figure 18, it can be noticed that the percentage of the extra core losses due to cutting in both samples is decreasing as the frequency increases despite the remarkable increase of the overall core losses with frequency. Additionally, when the magnetic induction is increased from 1 T to 1.5 T, the percentage of the extra core losses increases remarkably in the case of individual laminations, especially at low frequencies. On the other hand, the corresponding increase of the extra core losses in the case of full stack core is more obvious at higher frequencies.

Finally, if this case study SRM is going to work at 1.5 T in a 50 Hz application, it is expected that the cutting process will cause almost 200% extra core losses. However, if it will be used for a 400 Hz application, an extra 50% core losses should be expected.

5. Conclusions

This paper highlights the importance of incorporating cutting effects and material degradation into the design process in order to engineer a more efficient electric motor. Furthermore, this study introduces a more practical approach for accurate measurement of the core losses in a complete core of an electrical machine at a wide range of input power and frequency. Two different setups are built and used to measure the precise level of the core losses of two different samples of the case study SRM. The first sample consists of a few laser-cut NO20 un-joined laminations. This sample is tested at the low frequency and low power setup. The second sample is a full stack stator made from the same laminations but with an additional non-destructive stacking process using backlack bonding. A higher power rating test setup is then used for the full axial length core so that it can be tested under higher excitation currents and higher frequencies. By comparing the results, the extra core losses due to the laser cutting process are extracted separately for both samples. Moreover, the overall extra core losses due to the entire material degradation are accurately measured for the assembled machine.