Magnetic detector bands for interior 3D-analyses of laminated machine cores

Abstract

The physical state of laminated soft magnetic machine cores has high technical and economical relevance, in particular if advanced materials are involved, as in the case of transformer cores. Experimental analyses tend to be restricted to core surface regions due to the high expenditure of interior arrangements of sensors. Further, the latter cause inter-laminar spaces, as a reason of artefacts. The paper presents a completely novel sensor concept that avoids the above problems. A magnetic detection band of e.g. 1 m length consists of different kinds of sensors. They are arranged on a substrate foil by combined 2D/3D-printing, the total band thickness being below 100 $\mu$ m. Using the ends as handles, the band is located through the core. Electric connection sockets to electronics remain outside, thus avoiding increased inter-laminar spacing. They also allow for later failure diagnostics, by permanently arranged bands. Sensors concern local distributions of magnetic flux, losses, temperature, strains and vibrations. Manufacturing is aimed on versatile low-cost products. This paper describes procedures that are focused on flux sensors, in particular for in-plane flux that represents the most complex sensor type. For the first time, it enables interior measurements in non-destructive ways.

Keywords

Induction sensors magnetic sensors nano-crystalline ribbons 3D printing

1. Introduction

Cores of electric machines like transformers or motors are built up from laminated soft magnetic materials. Core masses may be up to hundreds of tons, and the corresponding magnetic performance shows high energetic and economic relevance. Thus, both industry and the public (see e.g. the EC initiative [1]) put strong effort on the optimization of core performances. This concerns optimum magnetic use of material in connection with 3D flux distributions. It also concerns a minimization of power losses. In more recent time, increasing interest is also given on reductions of audible noise, as emitted through magnetostriction and magneto-static forces. As a specific problem, most core designs represent complex 3D-systems [2], a pre-condition that is essential for effective optimizations.

1.1 Methodologies of modelling

A major tool for core optimizations is given by numerical modelling like MACC [3] or FEM. The latter is strongly complicated by pronounced multi-directional non-linearity of magnetic characteristics [4, 5, 6]. On the other hand, it proves to be highly effective for temperature distributions [7]. As an alternative, optimizations are supported in experimental ways. One option is the study of model cores which however involves problems of effective scaling-down. The second option is given by full-sized cores. however being complicated by restricted access to the core.

Experimental studies are focused on core surface regions that favor the attachment of sensor elements. However, as already stressed, a core tends to represent a 3D system which means that the interior physical performance may differ in significant ways. In the case of bulk cores, the only option for inner sensor application is to prepare thin channels into the core – or even through the entire of core. In [8], we reported a methodology where interior distributions of the magnetic induction $B$ are analyzed by so-called pin-sensors. They are moved through channels of about 2.5 mm width in automatic ways. Since most cores are assembled from a high number of individual laminations, an inter-laminar arrangement of sensors offers itself as a more simple methodology.

1.2 Inter-laminar sensors

In effective ways, the inter-laminar arrangement of sensors is possible for 3D analyses of inner distributions of temperature. Thermistors or thermo-couples can be arranged on selected lamination surface regions, e.g. to identify interior hot spots of the given core. The sensors also yield local losses, applying the rise of temperature method in the course of short-time magnetization.

Since decades, inter-laminar sensors are also used for the assessment of magnetic off-plane fluxes, as e.g. arising between individual packages of transformer core. So called frame coil sensors are established by thin wires that are fixed on lamination surfaces. For example, Fig. 1 (from [9]) depicts 29 sensors on selected surface regions of an interior transformer core layer, using wires of 150 $\mu$ m thickness. However, the corresponding inter-laminar air gaps can be expected to yield substantial reductions of off-plane flux density. This means that demands exist for much thinner sensors, such as described further down.

Figure 1.

Example for traditional arrangement of 29 induction sensors manufactured from 150 $\mu$ m thick wire (from [9]).

While being of highest interest, the inter-laminar measurement of in-plane components of the induction vector was not possible so far. The traditional way is to prepare search coils of thin wire through small holes drilled into individual laminations. Non-destructive methods were not reported so far – the present technique being a first approach.

The above mentioned sensors represent effective tools that yielded deeper understandings for the interior performance of machine cores. However, their application is characterized by major problems:

The thickness of sensor elements tends to be of an order that exceeds 100 $\mu$ m. This is true even for smallest sensor components, due to their inter-laminar connection to wires that are needed for the pick-up of sensor signals. The corresponding inter-laminar gaps may cause artefacts. They can be explained by the fact that the normally arising effective distance of soft magnetic laminations does not exceed the order of 10 $\mu$ m. This minimum value corresponds to insulating coatings. However, increases may result from lamination thickness tolerances, as well as from restricted core packaging and clamping.

The arrangement of sensors on defined surface locations of selected laminations involves high expenditure of manual fabrication work. The latter is increased by the arrangement of lamination within the core of machine. In practice, it proves to be difficult to avoid the damage of sensors and of the thin wires that are needed for contacts to electronics.

The described procedures lack flexibility. In particular, the arrangement of treated laminations at other core regions proves to be complicated – or even impossible, in the case of regional variations of lamination geometry.

As a specific problem, the traditional in-plane induction sensors are of destructive nature, as a source of artefacts.

The following describes a novel family of detection bands that avoids all four above mentioned areas of problems.

2. Basic concept for magnetic detector bands

2.1 Global aspects of design

The basic idea of the novel sensor concept is illustrated by Fig. 2 for the case of a laminated transformer model core. As it is well known, industrial cores tend to exhibit a dozen of packages of different width (up to more than 1 m) in order to attain circular limb cross sections, in approximation. Here, a model core is depicted with just two packages P1 and P2, of different lamination width.

Figure 2.

Schematic outline of two detector bands that are arranged in a model transformer core (e.g. 1 m $\times$ 1 m) which exhibits two packages.

The figure indicates two types of detector bands. The left one is a permanent one that is arranged in defined ways on a selected lamination surface, during assembling of the core. It can be slightly fixed by thinnest agglutination, before it is permanently located through pressure of core clamping. The right band illustrates a temporarily located band. After short-time release of clamping, its sensors are positioned by the help of handles as given by the band ends. Measurement is performed after re-clamping. Finally, according to so far experience, many courses of re-positioning are possible. It should be mentioned that a given band may comprise different types of sensors, for multi-parametric analyses.

Further above, four problems of conventional sensors were listed. They are considered by the novel methodology in the following ways:

The sum thickness of novel magnetic detection band is well below 100 $\mu$ m. This is attained by preparing sensor elements with a novel 3D-assembler. Well known methods of 2D printing and 3D printing (e.g. [10, 11]) are combined here in an automatic system.

On a substrate foil of ca. 20 $\mu$ m thickness, several sensors (optionally of different types) are arranged, together with printed contact bands to the band end.

The band can be placed at different regions, and it can be moved. It also can be left within the machine core, without affecting its performance, e.g. for diagnoses, years after core production.

Non-destructive detection of in-plane flux is possible.

2.2 Manufacturing procedure

A bit more in detail, magnetic detector bands are manufactured by the already mentioned 3D-assembler (Fig. 3). It exhibits a large printing platform of about 1200 mm length and 800 mm width. This allows for high band lengths that ensure that the band ends remain outside the investigated core for free contacting and handling. Conductive sensor elements are printed e.g. with carbon material by means of an extruder with a nozzle of e.g. 100 $\mu$ m width. Plastic materials for sensor elements – and their fixing – are printed by polyactide (PLA) operated with 220 ${}^{\circ}$ C. This rather low temperature level is significant with respect to compatibility with conductive layers.

Figure 3.

Concept of 3D assembler for automatic manufacturing of detector bands.

The type of substrate foil differs according to the given sensor type. Specific demands concern magnetostriction sensors were foils of sufficient elasticity are needed. On the other hand, minimum over-all thickness is needed for off-plane induction sensors. Attempts to print frame coils on plastic foils of just 10 $\mu$ m thickness (or even less) prove to be a challenge. All other types of sensors allow for higher sum-thickness.

As general demands, the foil should be thin, even and robust. In optimum ways, this is fulfilled by ca. 20 $\mu$ m thick and 50 mm wide plastic foil that yields well defined, precise results of printing. However, the mechanical contact between substrate and printed sensor elements tends to be insufficient for compounds that are repeatedly mounted at different core positions. An effective alternative is offered by 20 $\mu$ m thick paper, mechanical contacts being enhanced by soak processes. As a drawback, the latter yields waviness in connection with low precision of printed elements. In cases where the affected elements determine the sensor signal (as e.g. for off-plane induction sensors), individual sensor calibration becomes necessary.

Further down, Section 3.2 describes novel in-plane induction sensors in detail. All other so far prepared sensor types are based on well known physical principles. Off-plane induction sensors follow the frame-coil principle [9] through a conductive meander layer, the main challenge being minimum sum thickness (see Section 3.1). Thermal thermo-couple sensors [12] exhibit a contact region of two conducting layers of conductive inks, the challenge being to rise sensitivity. Strain sensors show conductive meander-bands on a rubber substrate [13], the main problem resulting from non-linear performance.

Two very specific sensor types need the incorporation of thinnest soft magnetic ribbons with print-on windings. This concerns vibration sensors that are based on a curvature-sensitive bilayer of Fe-based amorphous ribbon and non-magnetic steel ribbon, as described in [14]. Further it concerns in-plane induction sensors. A completely novel sensor type with an incorporated nano-crystalline ribbon is described further down (Section 3.2) for the first time. So far, these ribbons are positioned in manual ways, with interruption of printing processes. However, an automatic incorporation is planned, as indicated in Fig. 3.

3. Induction detector bands and their performance

Induction bands are aimed on analyses of 3D magnetic flux distributions in machine cores. As a general demand for the mounted sensors, they should have considerably large effective areas, in order to average over a sufficient amount of grains. In particular, this is relevant for highly grain oriented SiFe sheets that may show grain sizes round 10 mm. 2D flux distributions within individual layers of soft magnetic laminations are studied by means of in-plane sensors. Inter-laminar $z$ -flux [9, 15, 16] in the third so-called $z$ -direction is analyzed by off-plane sensors. They are more simple, thus being discussed at first.

3.1 Off-plane induction sensors

Thinnest sensors for off-plane induction $B_{z}$ – as an indication of $z$ -flux – are offered by frame coils. They detect the flux $\Phi_{z}=AB_{z}$ that passes through the frame. $A$ is the effective area that can be increased through rising number of spiral turns. Figure 4 shows a foil that carries four sensors of four turns each.

Figure 4.

Design of a detector band with sensors for off-plane flux density $B_{z}$ multi-turn frame coils yielding demands of 3D-printing.

In contrast to the simplicity of the sensor principle, main problems arise from the request of very low sum thickness. As already mentioned, the effective spacing of magnetic laminations may be as small as 10 $\mu$ m. An increase of this spacing $s$ through a sensor yields a decrease of the local induction $B_{z}$ that represents the primary target of measurement.

A correct measurement would need a substrate thickness below 1 $\mu$ m which, however, it is not realistic. Thinnest substrates are offered by vegetable films with thicknesses down to 5 $\mu$ m. However, their handling proves to be a most specific problem. For flattening the extremely thin substrate, specifically designed foil tension frames have to be constructed that must remain outside the core. As well, stable printing proves to be most difficult.

As a compromise concerning thickness, first sensors were printed on 20 $\mu$ m thick paper foils. Waviness and the need of isolating bridges between conductive print bands yield further local increases of sum thickness $s$ . However, within a clamped core, this can be assumed to be insignificant due to elasticity of the involved materials. As reported elsewhere [17], the bands prove to be highly effective for analyses of time changes $B_{z}(t)$ . A given sensor can be arranged at a high number of different core locations, and different core layers, respectively. However, for the applied sensors, double-side preparation was used, while the here described are single-side printed.

3.2 In-plane induction sensors

According to the above, increased inter-laminar spacing $s$ through the sensor arrangement tends to be critical for off-plane detection. On the other hand, it has no significant effects on in-plane flux distributions. This means that the total sensor thickness is not critical here. On the other hand, the sensor principle is much more complex. In fact, a non-destructive measurement of interior in-plane induction $B$ was not possible up to our development of tangential induction sensors.

As closer described in [18], the novel sensor type is based on the arrangement of a very thin soft magnetic ribbon between two core layers. It acts as a nucleus that is magnetized by the local inter-laminar field, analogous to a tangential field sensor. However, as is well known, in order to produce an effective signal, the latter would need about 1000 windings on a plastic carrier of millimeter thickness. For interior measurements, this is not acceptable since producing a large air gap. Here, we apply a nucleus the thickness of which does not exceed the order of 20 $\mu$ m. Instead of a wired coil, some few print-on flat turns of conductive material prove to be sufficient to produce an effective sensor signal.

The principle of measurement is illustrated by Fig. 5a. It depicts local sections of three laminations of an inner region of a core, a sensor being arranged between the 2 ${}^{\text{nd}}$ and 3 ${}^{\text{rd}}$ lamination. The task of sensor is to measure the course of time $B(t)$ of the local induction of SiFe material. It should be stressed that $B(t)$ differs completely from the induction in air that can be assumed to be almost zero.

Figure 5.

In-plane induction sensors. (a) Principle of measurement, three laminations with an inter-laminar sensor seen from the side (notice: strong scaling differences). (b) Photo of a practical design of sensor on paper substrate, seen from above.

An appropriate material for nucleus manufacturing proves to be given by 18 $\mu$ m thick nano-crystalline ribbons of type VAC Vitroperm 800-R. A critical task is given by the optimization of the nucleus length $l_{N}$ . For a high length $l_{N}$ , the field $H_{N}(t)$ of nucleus would approach the field $H(t)$ of SiFe. Due to extremely high permeability values $\mu_{N}$ ( $>$ 100000), the nucleus would tend to get saturated (up to ca. 1.5 T). For a representative local averaging over the grain structure of SiFe, we set $l_{N}$ to about 40 mm. This yields a pronounced – advantageous – effect of demagnetization according to $H_{N}<H$ .

In practice, core induction values should be detectable up to values $B=$ 1.85 T (10% below saturation of SiFe). To take full advantage of the nucleus, we set $w_{N}$ to about 10 mm so that saturation (close to 1.5 T) is reached for the corresponding field conditions. This nucleus optimization yields very high sensor resolution for instantaneous values $B$ up to about 1.2 T. For higher values up to $B=$ 1.85 T, the resolution is weaker, but still sufficient, the noise level of signal being low here.

The above nucleus optimization yields an effective transfer of the to-be-measured induction $B(t)$ to the nucleus induction $B_{N}(t)$ . The latter is detected by some windings round the nucleus. For calibration, the sensor can be placed in the center of a limb of Epstein Frame Tester filled with the given type of core material. The induction peak value $B_{\textit{PEAK}}$ is varied in steps of e.g. 0.2 T, and the corresponding values $B_{\textit{N,PEAK}}$ are measured in order to attain a transfer function for the given type of SiFe.

In our so far work, the function was assumed to be valid also for instantaneous values, according to $B_{\textit{PEAK}}(B_{\textit{N,PEAK}})=B(B_{N})$ . However, it is evident that this assumption exhibits systematic errors. They result from the fact that the two types of involved material show different performance of hysteresis. According to so far tests, the corresponding errors of course of time $B(t)$ are acceptable. However, a clear assessment is impeded by the lack of a golden-standard comparison method.

Figure 5b shows an example of a 20 $\mu$ m thick paper foil with a single sensor. Its fabrication comprises the following four steps: (i) Printing of conductive bands for electric sensor connection and for lower winding elements. (ii) Positioning of nano-crystalline ribbon with support of slight agglutination. (iii) Mechanical fixing of ribbon by plastic print. (iv) Printing of upper winding elements to complete an effective pick-up coil.

Figure 6 shows a result of measurement in an outer limb region of a model transformer core magnetized with 1.7 T (according to Fig. 1, left from location 1). The course of time indicates strongly distorted induction $B(t)$ which can be attributed to phenomena of both path length differences and circular magnetization.

Figure 6.

Result of measurement of in-plane induction $B(t)$ as detected in a transformer model core.

Here it should be stressed, that our earlier versions of induction sensors applied a nucleus either from crystalline SiFe material or from Fe or NiFe based amorphous material. All three cases are characterized by rather low permeability in the range of low induction values. This restricted the application to analyses of peak induction values, as a function of location. On the other hand, nano-crystalline ribbons, as here introduced for the first time, allow for the resolution of courses of time, due to very high values of initial permeability. Further, the ribbons are characterized by approximate zero-magnetostriction. This means that the performance in a clamped core is not influenced by mechanical stress, as a significant advantage towards the above mentioned crystalline or amorphous types of nucleus.

4. Discussion and conclusions

The above section was focused on the design of detector bands for local distributions of magnetic induction that tends to represent the most significant characteristic of a soft magnetic machine core. Industrial demand exists for a combination with sensors for local losses and core temperatures. Attempts are made to detect them by thermo-couple sensors, mounted on the same substrate. A spot of conductive print material is directly prepared on the substrate foil. A second spot of different composition is printed with slight overlapping, in order to establish a thermo-couple. Basically, this is a simple task. But in practice, the attained sensitivity proves to be influenced by many impact factors.

A further band type concerns distributions of mechanical core characteristics. Flexural vibrations can be detected analogous to in-plane induction, replacing the nano-crystalline ribbon by a magnetic bilayer ribbon that is incorporated according to Fig. 5. As closer described in [14], the latter compound consists of a highly magnetostrictive amorphous ribbon and a non-magnetic steel ribbon. Its task is to transfer curvature of the compound in tension of the amorphous ribbon. A second quantity concerns local magnetostriction of the core material, successful studies made by printing meander-like strain structures on already mentioned elastic substrates [13].

Provided that all above sensor types can be realized in reliable ways, the novel detection bands allow assessments of the most significant core characteristics for interior 3D analyses of laminated machine cores. Planning improved core designs, bands can easily be inserted in a given model core, and measurements can be taken after clamping. After repeated processes of de-clamping, the band positions can be modified for more detailed studies. As well, possibilities exist to modify core characteristics, like the design and pressure of clamps. After completion of study, the bands can be applied in other cores, if still being intact.

Further applications concern industrial full size cores. Bands can be located at selected regions during the assembling of a core. Clamping of the core favors the stable positioning of sensors in flat state. Bare band ends that act as mere handles for arrangement can be removed. On the other hand, measures have to be set to avoid that active contact ends remain without disturbance of core behavior. During normal, long-term core operation, the novel detection bands enable for the first time to study the “real” physical behavior of the finished product, without causing artefacts.

As a further industrially relevant aspect, further measurements are possible in the course of diagnostics for defects that may appear after years of core manufacturing. For example, this may concern anomalies of temperature distributions in the course of insulation defects, or inhomogeneous core vibrations in the course of mechanical defects.

As a main conclusion, magnetic detection bands can be placed into soft magnetic machine cores without affecting the physical state of core. Connections to electronics remain outside, through band lengths that can exceed one meter. Several sensors are arranged on the band by means of combined 3D/2D printing. This allows for rapid manufacturing of low costs. Sensors concern all significant physical core characteristics like magnetic induction and losses, temperature, as well as mechanical strain and vibrations. The total effective band thickness remains well below 100 $\mu$ m, even for specific sensor types that comprise a soft magnetic nano-crystalline or amorphous nucleus.

Footnotes

Acknowledgments

The authors thank for support from the Austrian Science Funds FWF (project MagFoils No. P 28481-N30). They also acknowledge very valuable experimental work of M. Palkovits and G. Trenner.

References

European Commission (2014), Commission Regulation (EU), No. 548.

Pfützner

Shilyashki

and Mulasalihovic

, Modern transformer cores – 3-dimensional magnetic systems of under-estimated complexity, Int. J. Appl. Elmagn. Mech 48 (2015), 143–151.

Shilyashki

Pfützner

Gerstbauer

Trenner

Hamberger

and Aigner

, Numerical prediction of rhombic rotational magnetization patterns in a transformer core package, IEEE Trans. Magn 52 (2016), 7200110, 1–10.

Preis

Biro

and Ticar

, FEM analysis of eddy current losses in nonlinear laminated iron cores, IEEE Trans. Magn 41 (2005), 1412–1415.

Komeza

Juszczak

E.N.

Barba

Napieralski

Lecointe

J.P.

and Hihat

, Using the FEM meshes adaption and genetic algoriths for identification of permeability in normal direction of anisotropic sheets, IEEE Trans. Magn 48 (2012), 211–214.

Hollaus

and Schöberl

, Multi-scale FEM for the eddy current problem in laminated media, in: Proc. Trans. Magn. Conf. CEFC, 2014, pp. 1–4.

Cao

Hongsheng

and Suxiang

, Research on the 3D temperature field of transformer winding based on finite element analysis, Adv. Mater. Res (2010), 129–131.

Shilyashki

Pfützner

Hamberger

Aigner

Palkovits

Trenner

and Gerstbauer

, Automatic 3-dimensional flux analyses of a 3-phase model transformer core, Int. J. Appl. Elmagn. Mech 48 (2015), 277–282.

Yao

X.G.

Moses

and Anayi

, Normal flux distribution in a three-phase transformer core under sinusoidal and PMW excitation, IEEE Trans. Magn 43(6) (2007), 2660–2662.

10.

Secor

E.B.

Prabhumirashi

P.L.

Puntambekar

Geier

M.L.

and Hersam

M.C.

, Inkjet printing of high conductivity flexible grapheme patterns, Phys. Chem. Lett 4 (2013), 1347–1351.

11.

Leigh

S.J.

Bradley

R.J.

Purssell

C.P.

Billson

D.R.

and Hutchins

D.A.

, A simple low cost conductive composite material for 3D printing electronic sensors, PLoS ONE 7(11) (2012), e49365.

12.

Chu

Bilir

D.T.

Pease

R.F.W.

and Goodson

K.E.

, Thin film nano thermocouple sensors for applications in laser and electron beam irradiation, in: IEEE Proc. 12.Int. Conf. Solid State Sens, 3B3.2, 2003, pp. 1112–1115.

13.

Pfützner

Shilyashki

Windischhofer

Trenner

and Giefing

, Printed detector bands for measurements of magnetostriction in core interior of laminated machines, in: Abstr. In. Conf. Magn. Meas, M2.3, 2017, p. 22.

14.

Pfützner

Kaniusas

Kosel

Mehnen

Meydan

Vazquez

Rohn

Merlo

A.M.

and Marquardt

, Magnetostrictive bilayers for multi-functional sensor families, Sens. & Act. A 129 (2006), 154–158.

15.

Hihat

Lecointe

J.P.

Duchesne

Napieralska

and Belgrand

, Experimental method for characterizing electrical steel sheets in the normal direction, Sensors 10 (2010), 9053–9064.

16.

Kefalas

T.D.

Loizos

and Kladas

A.G.

, Normal flux distribution at step-lap joints of Si-Fe wound cores, Mat. Sci. Forum 670 (2011), 284–290.

17.

Shilyashki

Pfützner

and Huber

, Interlaminar magnetic flux assessment of a transformer core measured by an extra-thin printed foil detector, IEEE Trans. Magn 53 (2017), 8400706, 1–6.

18.

Shilyashki

Pfützner

Palkovits

Hamberger

and Aigner

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