Nanocrystalline foil sensors for local detection of in-plane magnetic flux in laminated machine cores

Abstract

Recently, we presented a novel tangential induction sensor for measurements of local induction values in soft magnetic lamination of machine cores. In contrast to traditional tangential field-coil sensors, it exhibits a soft magnetic ribbon as a nucleus. Sensors with SiFe ribbon were effectively tested for transformer core surfaces, while showing poor reproducibility in the interior. Also tests with thin Fe-based amorphous ribbons proved to lack reproducibility, mechanical stress through core clamping causing magneto-elastic effects. An alternative would be Co-based amorphous ribbons of minimum magnetostriction, however, linked with low saturation magnetization. As an optimum solution, we present here nanocrystalline ribbons of 20 $\mu$ m thickness, combining high permeability, high saturation magnetization and almost zero-magnetostriction. The tendency of saturation due to very high permeability is met by optimized shortening of the nucleus, the demagnetizing effect offering good performance in a wide range of induction. Assembling by a combined 3D/2D printing procedure promises a total sensor thickness of the order of 100 $\mu$ m. Prototypes were effectively tested in the interior of a 3-package transformer core. While earlier sensors were restricted to peak induction measurements, the nanocrystalline ones offer detection of the waveform of local inductions. For the first time, measurements are possible in completely non-destructive way.

Keywords

Nanocrystalline ribbons magnetic sensors induction distribution laminated machine cores transformer cores

1. Introduction

Laminated soft magnetic cores of transformers, electric motors, generators, rotating machines etc. exhibit complex 3D flux distributions. Variation of magnetic induction, normal to the laminations can be measured non-destructively by means of thin search frame sensors [1, 2]. However, for measurements of in-plane induction, not any methodology was existing so far that works in non-destructive ways.

For optimization of laminated cores, the information about flux distributions in the plane of magnetization is highly essential. The magnetic flux varies not only in the individual packages of a core, but also in the different regions of a given package. As it is well known, these local variations of induction cause increases of regional core losses [3], and audible noise due to regional increases of magnetostriction [4].

Nowadays, many researcher worldwide use numerical methods for induction calculations, mainly based on FEM [5, 6]. However, laminated machine cores show very complex flux distributions due to anisotropy, non-linearity, flux in normal direction, effects of overlaps, impairments etc. Problems to consider all these effects by numerical models tend to restrict the practical relevance of their results.

As an alternative, experimental methods provide more reliable results, however with high expenditure of manual work. The conventional methods for measuring of in-plane fluxes are to drill holes through individual laminations and to arrange search coils [7, 8], or to wind wires around individual packages [9, 10, 11]. However, apart from high expenditure, the holes tend to influence flux distributions in significant ways, due to thicknesses of wires and due to burrs from the drilled holes.

As more recent alternatives, so-called needle methods were developed [12, 13, 14], mainly for applications in Rotational Single Sheet Testers. However, particularly for transformer cores, these methods are restricted to surface regions, since removing of coating and electrical contacts to the conducting metal are needed. As an improvement, the concept of inter-laminar film sensors was presented in [15]. By means of the latter, measurements are possible also in core interior, however, with high expenditure of manual work.

Recently, we launched a novel concept of magnetic “dummy” sensors [16]. We developed the so-called pin sensor [17, 18]. The latter was used for evaluation of an induction profile in a 3-limb transformer core, stacked from three packages of different width. The methodology tends to be highly effective, exhibiting high sensitivity for induction peak values above 1.5 T, but measuring channels through the investigated core are necessary. The establishment of such channels proves to be a very difficult task.

In [19], the concept of tangential induction sensors was presented. Contrary to traditional tangential field sensors, a soft magnetic ribbon is arranged as a nucleus of the coil. Around the ribbon, a thin wire is wound for signal detection. While tangential field sensors need high turn numbers of the order of 1000, induction sensors show the advantage that 10 turns are sufficient. First sensors were prepared with a nucleus of a silicon iron sheet of 270 $\mu$ m thickness. It was tested at a 3-package, 3-phase model transformer core, measurements on the periphery being highly effective. On the other hand, for interior measurements, thinner materials would be preferable, which was the aim of the here reported study. For the first time in the current work, a method for non-destructive, interior induction measurements of laminated machine cores is presented.

2. Optimization of nucleus material

In order to illustrate the significance of the nucleus material, Fig. 1 shows the basic concept of a tangential induction sensor. The latter consists of a high permeability ribbon as the nucleus, and of a thin wire wound around it. Let us now assume that a local region of a laminated soft magnetic core exhibits a given induction value $B$ . Then, the closely contacted tangential sensor is magnetized with an induction $B_{S}$ which however tends to deviate from $B$ in significant ways. One reason is given by differences of the permeability functions of the investigated core material and the nucleus, respectively. A second one results from the demagnetizing field that depends strongly on the geometry of the nucleus. In practice, the evaluation of the unknown induction $B$ is achieved by means of an a priori determined transfer function $B(B_{S})$ , as described in Section 4.

Figure 1.

Basic concept of a tangential induction sensor (after [16]). In the depicted case, a sensor is placed in the corner of a 3-limb, 3-package transformer core.

As already mentioned, tangential induction sensors with a Si-Fe sheet as the nucleus are most attractive for measurements on the surface of a machine core. For interior tests, a lower thickness would be preferable. For a significant reduction, we applied Fe-based amorphous alloys of 20 $\mu$ m thickness. However, arranged in the interior of a core, the results proved to lack reproducibility, repeated measurements at the same location providing significantly different results. Closer studies revealed strong influences of mechanical stress as resulting from differences of the cores consolidation through clamping. As an explanation, the high saturation magnetostriction ( $\lambda_{S}\approx$ 50 ppm) of Fe-based materials yields strong variations of permeability in the course of magneto-elastic effects. An alternative would be given by a Co-based amorphous ribbon that exhibits almost zero magnetostriction ( $\lambda_{S}<$ 0.2 ppm), however, with the drawback of very low saturation magnetization (typically 0.3 T, up to 0.7 T).

Finally, we attained a break-through by Fe-based nanocrystalline ribbons (VITROPERM 800R, produced by VACUUMSCHMELZE) of about 20 $\mu$ m thickness. Their unique combination of high permeability, low thickness, high saturation magnetization and almost zero-magnetostriction tends to overcome all challenges that are put on the sensors.

3. Sensor design and performance

The manufacturing of sensors with nanocrystalline nucleus was performed in a project that is aimed on different types of so-called foil sensors of low effective thickness, of the order of 100 $\mu$ m. By means of a 2D/3D assembler, conductive and plastic components are printed on a 25 $\mu$ m thick polyamide foil as a carrier that can easily be arranged in different regions of a laminated core. The magnetic ribbon as the nucleus of e.g. 15 mm width and 50 mm length is manually placed on lower winding elements prior to print-on of upper elements. This procedure should offer simple and rapid assembling.

However, the case of a nanocrystalline nucleus yielded specific problems of its preparation. As a severe one, due to the highly brittle material, it proved to be impossible to prepare a ribbon of exact shape by means of cutting procedures. Instead, the material was sandwiched, and the sensor ribbon was broken off, shape anomalies reducing the homogeneity of inner flux distribution. As a consequence, individual calibration is needed for each individual sensor. As a further problem, very sharp edges complicate the arrangement of very thin windings. Thus first sensors were prepared with 10 turns of about 50 $\mu$ m thick wire, yielding a sum thickness close to 150 $\mu$ m.

The already mentioned transfer function $B(B_{S})$ was determined in an Epstein Frame Tester. The sensor was placed on one of the four SiFe sheet stacks assembled of the same grade as given in the core of planned sensor application. For calibration, $B_{S}$ of the ribbon was determined as a function of $B$ of the SiFe stack.

The results of calibration indicate that the sensor performance can be optimized in very flexible ways by the variation of the ratio of length $L_{s}$ and width $w_{s}$ of the nucleus. In order to attain a good average over the large grain sizes of modern SiFe materials, the width was selected rather large with $w_{s}=$ 15 mm. The upper curve of Fig. 2 represents a result of $B(B_{S})$ for a high ribbon length $L_{S}=$ 75 mm. The curve indicates high sensitivity – and thus also resolution – for low induction values $B$ . However, saturation effects ( $B_{S}=>$ 1.3 T) start at about $B=$ 0.6 T. With this performance, the sensor is highly attractive in cases of low intensities of induction.

On the other hand, transformers are operated with nominal inductions up to about 1.8 T. Here, very simple optimization was reached by taking advantage of effects of demagnetization. Very short ribbons with $L_{S}$ between 10 mm and 20 mm (not shown in Fig. 2) yielded very high resolution in instants of very high values of $B>$ 1.5 T, but failing for low values. Finally, optimum solutions were found with a length of about 30 mm, according to the lower curve of Fig. 2. Due to the extremely high initial permeability of the nano-crystalline material (order of 100.000), the transfer function $B_{S}(B)$ shows a very high slope for $B$ up to 0.6 T, corresponding to a high sensitivity for the given induction range. Above that, the slope is lower, but still sufficient for effective detections, supported by the fact that high signal intensities are given here.

Figure 2.

Examples for transfer functions for nanocrystalline nuclei of different length $L_{s}$ .

The above described characteristics mean that the sensor can be applied in the full range of induction as arising during the period of magnetization. In fact, applications of all earlier types of dummy sensors were restricted to analyses of local variations of induction peak values. On the other hand, the nanocrystalline sensor allows for the first time to determine also dynamic variations $B(t)$ of induction as arising during the period of magnetization. Signal evaluation is based on a look-up table for the transfer function $B(B_{s})$ of the individually applied sensor. The function $B_{S}(t)$ is automatically translated by a software script prepared in Matlab into $B(t)$ . Figure 3 shows an example of transfer from $B_{s}$ to $B$ for a period of magnetization. The addressed sensor position exhibits a sensor induction that is close to trapezoid shape with a slight peak. From transfer to $B$ , this peak becomes very pronounced. As a reason of difference, the nanocrystalline nucleus gets close to saturation while the core steel is not really close to it.

Figure 3.

An example of a measurement for a nominal magnetization $B_{\text{NOM}}=$ 1.7 T.

Figure 4.

Schematic of a nanocrystalline induction sensor foil, placed within a laminated core.

Figure 5.

A novel nanocrystalline tangential foil sensor, placed in the interior of a 3-limb, 3-package transformer core.

4. Practical sensor application and its results

According to the above, due to very low thickness, the sensors are highly fragile and not self-supporting. Free mounting of individual sensors in exactly defined interior regions of a machine core would be most difficult and laborious. Thus mounting is performed on a kapton (polyamide) foil of typically 25 $\mu$ m thickness. As illustrated by Figs 4 and 5, the foil length exceeds the width of the investigated core component. For example for a transformer of 150 mm sheet width, a foil length of about 250 mm is applied, so that the overhanging ends can be used as handles for positioning. In the depicted case, the left end is also used for contacting, considering that interior contacts would cause increased inter-laminar air gaps. With this concept, a given sensor can be used for test series at different locations, for the assessment of flux distributions in the whole of a given laminated core. It can be rearranged between different core layers, and in addition can be shifted there to different sub-regions. The sensors can also remain in a finished core, to be re-used later, e.g. for diagnoses of machine defaults.

For a 3-package, 3-phase model transformer core according to Fig. 5, two results of local measurement are shown in Fig. 6 for a nominal magnetization with 1.7 T. Figure 6a shows a result for the peripheral location “OL” of the left outer limb and the corresponding amplitude spectrum. It indicates that the limbs outer periphery shows a distinctly distorted waveform with low amplitude of not more than 1.5 T, among others due to increased effective path length along the periphery. The amplitude of the 3 ${}^{\text{rd}}$ and 5 ${}^{\text{th}}$ harmonics, measured at position “OL” show values of about 0.15 T, representing about 10% of the fundamental component.

Figure 6.

Local induction waveform $B(t)$ and the corresponding amplitude spectrum, detected by a nanocrystalline sensor within the main package of the investigated transformer core for a sinusoidal global induction with $B_{\text{NOM}}=$ 1.7 T. The third, fifth and the seventh harmonics are also given related to the fundamental component in percentage. (a) Measurement in the peripheral part of the outer limb (Fig. 5, position OL). (b) Measurement in the outer part of the T-joint (Fig. 5, position T).

Figure 6b shows a result for the T-joint region “T” below the V-shaped ends of middle limb laminations. It reveals a very high amplitude, of about 1.85 T, i.e. significantly beyond the nominal magnetization. As an interpretation, flux through the total of yoke can hardly pass through the V-triangle due to the high anisotropy of the given highly grain oriented material which yields a peripheral overload. The waveform shows a distinct widening with three peaks. They result from peaks of fluxes that arise at different instants of the period of magnetization, originating from different limb excitations. This distinct distortion of the waveform is reflected by harmonics up to 20% of the fundamental component. As a matter of fact, the nanocrystalline sensor offers such temporary-spatial information in completely non-destructive ways.

5. Conclusions

The current work reports significant recent improvements of tangential induction sensors for measurement in the interior of electric machine cores. Intensive research proves that nanocrystalline ribbons are the best type of material for the preparation of a sensor nucleus, due to its unique properties: low thickness, extremely high initial permeability, relatively high saturation magnetization, and almost zero magnetostriction, hence, non-sensitivity to mechanical clamping stress. The described sensor prototypes show a total thickness of about 150 $\mu$ m as resulting mainly from wires as used for windings. It is expected that the thickness will be further decreased by a novel combined 3D/2D printing procedure, as already tested. High robustness and multiple usage of the sensors is provided by mounting on a specific polyamide foil of low thickness as a carrier.

Effective testing was attained in the interior of a laminated model transformer core. As a novelty, the sensor detects not only the local peak values of induction, but also the entire waveform, due to the high sensitivity in the low induction range. The latter represents a significant benefit regarding the evaluation of the higher harmonics. It is claimed that the novel sensors allow non-destructive tests as not attainable with any other existing methodology.

Footnotes

Acknowledgments

We thank for support from Austrian Science Fund (FWF): Project number P28481-N30 MagFoilSensors and also from VACUUMSCHMELZE Gmbh & Co. KG for providing the nanocrystalline material.

References

Fauzi

and Ahmad

D.M.M.

, Normal direction of flux distribution in the mix 60

{}^{\circ}

–23

{}^{\circ}

T-joint of three phase transformer cores, in: Int. Pow. Eng. Conf. Malaysia, 2010, pp. 570–572.

Shilyashki

Pfützner

Hamberger

Aigner

Hofbauer

Matkovic

and Kenov

, The impact of off-plane flux on losses and magnetostriction of transformer core steel, IEEE Trans. Magn.50(11) (2014), 8401604, 1–4.

Pfützner

Shilyashki

Hamberger

Aigner

Hofbauer

Palkovits

Gerstbauer

Matkovic

and Galabov

, Automatic 3-dimensional building factor analysis of a grain oriented model transformer core, IEEE Trans. Magn.50(4) (2014), 8400604, 1–4.

Pfützner

Mulsalihovic

Yamaguchi

Sabic

Shilyashki

and Hofbauer

, Rotatitional magnetization in transformer cores – a review, IEEE Trans. Magn.47(11) (2011), 4523–4533.

Witczak

and Majer

, The influence of magnetic anisotropy on magnetostriction forces and vibration in three-phase three limb transformers, Int. J. Appl. El. Mech.51 (2016), 419–429.

Tang

Guo

and Wang

, Magnetic flux distribution in power transformer core with mitred joints, J. Appl. Phys.117 (2015), 17D522, 1–4.

Shilyashki

Pfützner

Anger

Gramm

Hofbauer

Galabov

and Mulasalihovic

, Magnetostriction of transformer core steel considering rotational magnetization, IEEE Trans. Magn.50(1) (2014), 8400115, 1–15.

Ahmad

and Fauzi

, Influence of the mix 60

{}^{\circ}

–23

{}^{\circ}

in different layers of T-joint of 3phase transformer core on longitudinal direction of flux distribution, in: 4th Int. P. Eng. and Opt. Conf. Malaysia, 2010, pp. 566–569.

Loizos

Kefalas

Kladas

and Souflaris

, Flux distribution analysis in three-phase Si-Fe Wound Transformer Cores, IEEE Trans. Magn.46(2) (2010), 594–597.

10.

Marketos

Marnay

and Ngnegueu

, Experimental and nummerical investigation of flux density distribution in the individual packets of a 100 kVA transformer cores, IEEE Trans. Magn.48(4) (2010), 1677–1680.

11.

Belehosur

Marketos

Moses

and Vincent

, Packet-to-packet variation of flux density in a three-phase, three-limb power transformer core, IEEE Trans. Magn.46(2) (2010), 642–645.

12.

Brix

Hempel

K.A.

and Schulte

F.J.

, Improved method for the investigation of the rotational magnetization process in electrical steel sheets, IEEE Trans. Magn.20(5) (1984), 1708–1710.

13.

Hasenzagl

Weiser

and Pfützner

, Novel 3-phase excited single sheet tester for rotational magnetization, J. Magn. Magn Mater.160 (1996), 180–182.

14.

Yamaguchi

Senda

Ishida

Sato

Honda

and Yamamoto

, Theoretical analysis of localized magnetic flux measurements by needle probe, J Phys iV France8 (1998), 717–720.

15.

Pfützner

and Mulasalihovic

, Thin film technique for interior magnetic analysis of laminated machine cores, Prz. Elektrot./El. Rev.85 (2009), 39–42.

16.

Pfützner

and Shilyashki

, Magnetic dummy sensors – A novel concept for interior flux distribution tests, Int. J. Appl. El. Mech.48 (2015), 213–218.

17.

Shilyashki

Pfützner

Hamberger

Aigner

Gerstbauer

Palkovits

and Trenner

, Pin Sensor for Interior Induction Measurements in Transformer Cores, IEEE Trans. Magn.51(1) (2015), 4001904, 1–4.

18.

Shilyashki

Pfützner

Hamberger

Aigner

Palkovits

Trenner

and Gerstbauer

, Automatic 3-dimensional flux analyses of a 3-phase model transformer core, Int. J. Appl. El. Mech.48 (2015), 277–281.

19.

Shilyashki

Pfützner

Palkovits

Hamberger

and Aigner

, A tangential induction sensor for 3D-analyses of peripheral flux distributions in transformer cores, IEEE Trans. Magn.51(6) (2014), 4003306, 1–6.