Eddy-Current-Based Contactless Speed Sensing of Conductive Surfaces

Transcription

1 1 IEEE Proceedings of the Southern Power Electronics Conference (SPEC 1), Auckland, New Zealand, Deceber 5-, 1 Eddy-Current-Based Contactless Speed Sensing of Conductive Surfaces A. Tüysüz, M. Flankl, J. W. Kolar, A. Mütze This aterial is published in order to provide access to research results of the Power Electronic Systes Laboratory / D-ITET / ETH Zurich. Internal or personal use of this aterial is peritted. However, perission to reprint/republish this aterial for advertising or prootional purposes or for creating new collective works for resale or redistribution ust be obtained fro the copyright holder. By choosing to view this docuent, you agree to all provisions of the copyright laws protecting it.

2 Eddy-Current-Based Contactless Speed Sensing of Conductive Surfaces Arda Tüysüz, Michael Flankl and Johann W. Kolar Power Electronic Systes Laboratory ETH Zurich Zurich, Switzerland Eail: {tuysuz, flankl, Annette Mütze Electric Drives and Machines Institute Graz University of Technology Graz, Austria Eail: Abstract Contactless speed sensors are used in a broad area of applications in various industries such as achining, assebly lines and transportation. Coonly used technologies are based on optics (e.g. caeras, encoders), or electroagnetic effects (e.g. variable reluctance sensors, Hall sensors). However, these sensors require a non-unifor property of the oving target that can be detected. For exaple, variable reluctance sensors rely on the variation of the air gap, and Hall sensors require a agnetic field whose spatial distribution is dependent on the position of the over. A clear disadvantage of all these systes is the fact that they require odifications on the target s geoetry and/or agnetic properties, since they cannot easure the speed of a sooth body/surface. Moreover, soe of the are sensitive to environental conditions; e.g. dirt in case of optical encoders and high teperature in case of peranent agnets can render these systes ineffective. Therefore, an eddy-current-based contactless speed sensor is developed in this work for easuring the speed of sooth, electrically conductive surfaces in harsh operating conditions. An injection coil is used to induce eddy currents in the over whose speed is to be detected, and two differentially wound pick-up coils are used to detect the speed-dependent deforation of the eddy-current field. Two-diensional finite-eleent ethod (-D FEM) is used for odeling the syste and optiizing the sensor geoetry as well as the injection frequency. Measureents taken fro a prototype verify the validity of the design procedure and the analyzed speed-sensing concept. I. INTRODUCTION Measuring the speed of a oving (i.e. rotating or translating) body is essential for realizing a closed-loop control in a broad range of industries such as achining, assebly lines and transportation. Direct contact of the speed sensor with the oving body s surface is either prevented by the operating context, or undesirable due to reliability concerns in nuerous applications. Therefore, several types of contactless speed sensors are used such as variable reluctance sensors, Hall sensors or encoders. However, these sensors require a nonunifor property of the oving target that can be detected, e.g., variable reluctance sensors rely on the spatial variance of the agnetic pereance, Hall sensors of the agnetic field distribution and encoders of the optical properties. Hence, these sensors cannot be used for easuring the speed of sooth surfaces. More advanced speed sensing ethods such as using iage processing (as done in optical coputer ice) can easure the speed of rather sooth surfaces, but they are Injection coil Yoke Pick-up coils v Mover Fig. 1: Construction of the analyzed speed sensor coprising a agnetic yoke, an injection coil and two pick-up coils. A cut-away view is shown in the iage to visualize the coils. The yoke of the actual syste covers all three coils. not considered here since they cannot operate reliably in harsh environents. An interesting contactless speed sensing ethod is presented in [1], where a U-shaped inductor is used to place a agnetic ark on a oving steel band, whose speed is estiated based on the tie delay between the detections of the agnetic ark by two agnetic field sensors placed at different downstrea locations. However, this ethod cannot be used for estiating the speed of non-agnetic etals such as aluinu. The agnetic arker is eliinated in [], where two eddy-current sensors are used for detecting the inhoogenities in the rail above which they are oving. The speed is estiated via the cross-correlation of the two sensor output signals. In this work, a single, differentially wound eddy-current sensor is eployed for easuring the speed of an electrically conductive surface, which does not need to be ade of a agnetic aterial or feature any inhoogenities. A sketch of a possible sensor arrangeent is shown in Fig. 1. When an AC current is flowing in the injection coil and the over is in otion (v = ), there is a slight difference in the induced voltages in the pick-up coils, which is easured and processed for estiating the speed. Siilar speed sensor arrangeents have been presented in literature (e.g. []), but these works /1/$1. 1 IEEE 1

3 Magnetic yoke Pick-up coil 1 iinj Injection coil Pick-up coil u diff v > v v1 = udiff Band-pass filter Aplitude deodulation u lp u lp v > v i inj v 1 = J v 1 = /s v > /s Mover (a) 1/f inj (b) Fig. : Operating principle of the speed sensor used in this work. (a) An injection coil is used for generating the eddy currents (J) inthe over. The current distribution is depicted for two different speeds: v 1 = /s, and v > /s. Since the current distribution is skewed due to the speed of the over, the reaction field induces different voltages in the two differentially wound pick-up coils. (b) The difference of the induced voltages is easured, filtered and aplitude deodulated. The resulting output voltage u lp is proportional to speed. have not considered a agnetic yoke. In this work, the sensor topology is further iproved by the addition of a agnetic yoke. The principle of operation is detailed further in Sec. II. Sec. III describes the design aspects (i.e. goals, constraints and degrees of freedo of optiization) as well as the electroagnetic odeling ethod. Effects of different degrees of freedo on the sensor perforance are also analyzed. The hardware prototype and the test setup are introduced in Sec. IV, and the results of the earlier analysis are copared to the easureents. Finally, Sec. V recaps the iportant findings and concludes the paper. II. PRINCIPLE OF OPERATION An AC current injected in the injection coil excites eddy currents in the electrically conductive over. The eddy-current distribution in the over depends on several factors such as the aplitude î inj and the frequency f inj of the injected current i inj, the air gap length g, the aterial properties of the over as well as its speed v. The eddy-current distribution is syetric at standstill, but a non-zero speed results in a skewing of the eddy-current distribution, as shown in Fig. (a). This skewing results in a slight difference of the agnetic flux density on the two sides of the injection coil. This speed-dependent effect is detected via the two pick-up coils arranged syetrically on both sides of the injection coil, which are wound in opposite directions, such that the difference of induced voltages (u diff ) can be obtained with a single-ended voltage easureent. The easured voltage is band-pass filtered with a filter whose center frequency is at the injection frequency f inj, in order to reove any possible easureent noise, as well as any coponents resulting fro haronics of the injected current. The speed inforation can then be extracted fro the easured voltage by aplitude deodulation as shown in Fig. (b). A phase-sensitive rectification, as discussed in detail in [], can be used for differentiating positive and negative velocities. On the other hand, in applications where the direction of oveent is not required, the speed can also be estiated by a sipler, phase-insensitive rectification. III. MODELING, DESIGN ASPECTS AND OPTIMIZATION Models describing the electroagnetic behavior of the considered speed sensor are needed for optiizing the sensor geoetry as well as the injection frequency for a given air gap and speed range. Iportant work on the analytical odeling of a siilarly arranged eddy-current-based speed sensor for a oving solid body has been presented in literature. In [] and [5], the authors discuss the case of circular injection and pick-up coils, whereas a rectangular injection coil of arbitrary orientation towards the conducting plane is odeled in [] and [7]. However, these works do not consider the use of a agnetic yoke, which shields the sensor fro its environent and increases the air gap flux density. Even though it ay be coputationally ore deanding copared to the aboveentioned odeling approaches, finite-eleent ethod (FEM) siulations offer an easy-to-set-up and very flexible odeling environent. Therefore, two-diensional (-D) tie-transient FEM siulations are used for odeling the speed sensor featuring a agnetic yoke. Key degrees of freedo in the speed sensor design are the geoetries of the coils and the yoke, as well as the aplitude and the frequency of the injected current. For a given speed range and air gap, the design goal is to obtain a high sensitivity that results in a voltage range that can be easured using siple signal electronics. A linear sensor characteristic is also desired since it akes the processing easier.

4 t y g w y w inj Mover t p w c t p w c Yoke Coil forer Fig. : Cross-sectional view of the speed sensor geoetry (left), and the geoetry of a -D printed coil forer (right), with key geoetric paraeters. TABLE I: DESIGN PARAMETERS Sybol Paraeter nae Value w y Yoke width t p Plastic wall thickness. t y Yoke thickness 5.1 w c Coil side width. l c Coil length 5 g Air gap (noinal) v Mover speed.. /s a Mover acceleration 5 /s Yoke aterial Ferrite N7 Mover aterial Aluinu Ac-1 Neglecting agnetic saturation, the easured voltage u diff (and hence the sensitivity) for a given air gap g and speed v can be increased by increasing the total Apere-turns of the injected current. Here, possible liits are the energy consuption of the sensor and the heating of the injection coil due to power dissipation. The relationship between the sensitivity and the injection frequency is ore coplicated due to the skin effect in the over, as it will be analyzed in detail later in this paper. Further design considerations are the overall sensor volue as well as the partitioning of the sensor area between the injection and pick-up coils. The iniization of the sensor size is not a priary goal in this work; therefore, at a first step the sensor area is set by choosing a coercially available, [ x 5 x 5.1] ferrite block as the yoke. The partitioning of the injection and the pick-up coil areas, on the other hand, plays a ajor role in the perforance of a sensor with given volue. Hence, the injection coil width is regarded as a design degree-of-freedo in the optiization, together with the injection frequency. Fig. shows the key design paraeters together with Tab. I. Both the injection and pickup coils are wound using -D printed coil forers. For ease of anufacturing, the design is liited to only non-overlapping (i.e. concentrated) coils. Fig. shows the siulation results for the aplitude of the differential voltage û diff for î inj = 1 A and one-turn injection and pick-up coils. The over speed is v = /s. Air gaps of g =[,, ], as well is injection frequencies of f inj = [1,, ] Hz are analyzed. It can be seen that, an optiu w inj exists with the axiu sensitivity for each analyzed air gap and injection frequency. For all the analyzed l c û diff / î inj û diff / î inj û diff / î inj g = Hz Hz Hz g = g = w inj ( Fig. : Siulation results showing the ratio û diff/î inj for air gaps of g = (top), (iddle) and (botto. Three different injection frequencies (f inj = [1 ] Hz) are considered. The speed of the over is v = /s. It can be seen that the injection coil width of w inj = 15, and the injection frequency of f inj = Hz lead to the highest û diff/î inj (hence, highest sensitivity) for ost of the siulated cases. TABLE II: PARAMETERS OF THE SENSOR PROTOTYPE Sybol Paraeter nae Value w inj Injection coil width 15 N inj Injection coil winding turns 7 R inj Injection coil DC resistance. Ω L inj Injection coil inductance ( Hz) 5 μh N pick-up Pick-up coil turns nuber air gaps, w inj = 15, which results in identical injection and pick-up coil geoetries, and f inj = Hz leads to the highest sensitivity. IV. EXPERIMENTAL ANALYSIS Following the initial analysis described above, a speed sensor prototype is built in order to verify the siulations. Fig. 5 (top) depicts the sensor prototype and Tab. II lists its key paraeters. The test setup depicted in Fig. 5 (botto is used for the experiental analysis. A drive achine is used to rotate an aluinu wheel, whose surface speed is easured with the speed sensor prototype that is ounted on an adjustable positioning table for an easy variation of the air gap. The encoder at the drive achine s shaft also provides a reference rotational speed easureent. A digital signal generator and a linear aplifier are used for injecting an AC current with a fixed aplitude into the

5 Pickup coil Injection coil Pickup coil u diff,ap (V).5 /s /s /s /s /s g = Ferrite yoke.5 Rotating wheel Fixture Ø 5 Coil leads Drive achine Encoder u diff,ap (V) g = Tie (s) Adjustable positioning table v Fixture Coil leads Fig. 5: (Top) Sensor prototype with equally sized injection and pickup coils (w inj = 15, cf. Fig. ). The -D printed coil forers are glued onto the ferrite yoke, which is glued on the fixture. (Botto Test bench coprising a drive achine, an aluinu wheel and the speed sensor prototype ounted on an adjustable positioning table. The encoder ounted on the shaft of the drive achine provides reference speed easureents. injection coil. For all the easureents presented in this paper, the aplitude of the current is set to î inj = 5 A (I inj = 5 A RMS ), resulting in a power dissipation of P inj = Iinj R inj = W in the injection coil. Since the goal is the verification of the electroagnetic design, the construction of dedicated filtering and aplitude deodulation hardware is oitted at this stage, and a digital oscilloscope is used for easuring the voltage u diff,ap, which is obtained by aplifying the voltage u diff ties using a siple operational aplifier circuit in a non-inverting configuration. For having full flexibility, the filtering and deodulation steps are realized as post-processing in a coputer, once the voltage wavefor is recorded with the oscilloscope. All the easureent results presented in the further sections are band-pass filtered using a nd -order Butterworth filter whose lower and upper cut-off frequencies are set to.f inj and 1.f inj, respectively. Fig. : Measureent results for air gaps of g = (top), and g = (botto. The injection frequency is f inj = Hz, and the injection current aplitude is î inj = 5 A. It can be seen that a non-zero voltage exists for v = /s, as a result of asyetry due to anufacturing and positioning tolerances. A. Offset and Verification of FEM Siulations Fig. depicts the easured voltage u diff,ap for f inj = Hz and g = [, ]. It can be seen that a non-zero voltage exists for v = /s, which will henceforth be called the offset voltage. It results fro an unequal flux linkage of the pick-up coils at standstill, and is caused by anufacturing tolerances (i.e. the sensor not being perfectly syetrical). For a direct coparison of the results with the FEM siulations, where a syetrical structure is assued, the offset is reoved by an offline, tie-doain subtraction fro the easured u diff,ap at non-zero speeds. However, ore practical ways of offset reoval are also discussed in the following sections. The coparison of u diff,ap wavefors calculated by FEM siulations to the easureent results is done in Fig. 7, for f inj = Hz and g =. It can be seen that, the siulations are able to predict both the aplitude s stronger, and the phase s weaker dependencies on the speed accurately. B. Sensitivity and Linearity Since the aplitude has a higher sensitivity to the speed, the phase change is disregarded at the first step for evaluating the effect of the injection frequency on the sensitivity and linearity of the sensor. Fig. shows the siulated and easured aplitudes of the easured voltage u diff,ap for different injection frequencies and air gaps. It can be seen that around v = /s, f inj = 1 Hz is resulting in a strongly nonlinear response, showing that the iniu injection frequency should be chosen considering the axiu speed to be easured. It is also seen that an optiu f inj exists, i.e., increasing f inj above Hz leads to a saller sensitivity,

6 u diff,ap (V).. -. Siulation Tie (s).. Measureent -. /s /s /s /s Tie (s) Fig. 7: Siulation (left) and easureent (right) results for an air gap of g =, injection frequency of f inj = Hz, injection current aplitude of î inj = 5 A. It can be seen that the siulation and easureent results agree very well after the offset is reoved fro the easureents. which is attributed to the skin effect in the over. Finally, it is also noted that the siulations and easureents agree over this wide design range with a total average isatch below 1%, which is expected to be originating fro -D (end) effects, as well as the discrepancies between the assued and actual electrical conductivities of the over, and between the assued linear otion and the actual curvature of the aluinu wheel. C. Effect of the Air Gap The effect of the air gap g on the easured voltage u diff,ap can be seen in Fig., where easured voltage is plotted, with and without offset correction (by a tie-doain subtraction), for different values of f inj, g and v. It can be seen that the air gap plays a significant role in the resulting sensitivity, and the effect of the offset becoes less pronounced at saller air gaps. D. Aplitude Deodulation and Dynaic Behavior Finally, both the reoval of the offset through a phasesensitive deodulation, and the dynaics of the sensor are deonstrated together in Fig. 1, where the speed of the over is changed with a = ±5 /s. In this case, there is no tiedoain subtraction; instead, the band-pass filtered signal is deodulated by a ultiplication with a clock signal in the for of u clk = sign(i inj ). Even though this is done offline in software, a real-tie hardware realization would be possible with a zero-crossing detector and siple signal electronics []. The deodulated signal is low-pass filtered with a nd -order Butterworth filter whose cut-off frequency is set to Hz, and a gain of 57 /Vs is applied to obtain the speed as plotted in Fig. 1. V. CONCLUSION Measuring the angular or translational velocity of a solid body in otion is crucial for the onitoring and/or control of û diff,ap (V) û diff,ap (V) v ( Siulation g = Siulation g = 15 v ( v ( Measureent g = 1 Hz Hz Hz Hz Measureent g = 15 v ( Fig. : Siulation (left) and easureent (right) results for air gaps of g = (top) and g = (botto. Injected current aplitude is î inj = 5 A. It can be seen that the siulation and easureent results agree very well, with an average isatch below 1% for g = and below % for g =. nuerous industrial processes. A direct echanical contact of the speed sensor to the over is either undesired or strictly prohibited, e.g. due to wear and reliability concerns. Several contactless speed sensors are widely used today, such as agnetic or optical devices. However, these devices are either not well-suited for operation in harsh environents (e.g. optical sensors are susceptible to dirt), or they require a spatial variation of the physical quantity that they easure, and cannot be used for easuring the speed of sooth surfaces (e.g. variable reluctance sensors rely on the spatial variation of a agnetic air gap). Therefore, a contactless speed sensor that is well-suited for estiating the speed of an electrically conductive, sooth surface is developed in this work. The analyzed speed sensor coprises an injection coil, which induces eddy currents in the over. At standstill, the eddy-current distribution is syetrical, but it gets skewed with increasing speeds. The reaction field is detected via two differentially wound pick-up coils placed on both sides of the injection coil. Signal conditioning steps of filtering and aplitude deodulation are used for extracting the speed inforation fro the difference of the voltages induced in the pick-up coils. 5

7 (b) 1 (a) (c) 1 (d) Fig. : Measureent results showing udiff,ap for different over speeds and air gaps, for finj = Hz ((a) and (b)), finj = Hz ((c) and (d)). The aplitudes are plotted as easured in (a) and (c); whereas, the offset is corrected by a tie-doain subtraction in (b) and (d). of dedicated hardware for the ipleentation of filtering and deodulation steps, as well as signal injection. Also the trade-off between the sensor volue and sensitivity will be investigated for various air gaps, and overs ade of different aterials. 15 Speed ( Encoder Sensor Prototype 1 ACKNOWLEDGMENT 5 1 Tie (s) 5 7 Fig. 1: Mover speed, as recorded by the encoder and the speed sensor prototype. The encoder reading is converted to surface speed as v = ωr where ω is the echanical angular velocity and r = 1. The low-pass filtered output voltage of the eddy-current-based speed sensor is converted to speed by ultiplication with the constant 57 /Vs. The air gap is g =, and the injection frequency is finj = Hz. The odels presented in literature are enhanced by including a agnetic yoke, which shields the sensor and aplifies the sensitivity for a given power consuption by intensifying the air gap flux. The outer volue of the sensor is fixed in this work, but the optial partitioning of the injection and pickup coil areas as well as the optiu injection frequency are studied. A prototype is built and the design procedure is verified experientally. It is found that the voltages induced in the two pick-up coils are not identical at standstill, as a result of sall asyetries in the construction due to anufacturing tolerances. However, a siple phase-sensitive deodulation approach is shown to alleviate the effect of this offset voltage on the sensor perforance. The future work will focus on the design and construction The authors would like to express their sincere appreciation to Nabtesco Corp., Japan, for the financial and technical support of research on energy harvesting technologies at the Power Electronic Systes Laboratory, ETH Zurich, which provided the basis for achieving the results presented in this paper. In particular, inspiring technical discussions with K. Nakaura, Y. Tsukada and Y. Ono are acknowledged. Moreover, the authors acknowledge the support of CADFEM (Suisse) AG concerning the ANSYS software. R EFERENCES [1] J. M. Lopera, M. J. Prieto, F. F. Linera, G. Vecino, and J. A. Gonzlez, A new speed easureent syste, IEEE Industry Applications Magazine, pp. 51, Noveber/Deceber 5. [] S. Hensel, T. Strau, and M. Marinov, Eddy current sensor based velocity and distance estiation in rail vehicles, IET Science, Measureent Technology, vol., no. 7, pp. 75 1, 15. [] N. Takehira, A. Tanaka, and K. Toda, Analysis of a speed eter utilizing eddy current effect, Electrical Engineering in Japan, vol. 7, no. 5, pp. 1, Septeber 177, translated fro Denki Gakkai Ronbunshi, vol. 7A, no., pp. 57-, Septeber 177. [] A. Tu ysu z and J. W. Kolar, New position-sensing concept for iniature lateral-stator achines, IEEE Transactions on Industrial Electronics, vol., no., pp., June 1. [5] N. Takehira and A. Tanaka, Analysis of transission type speedoeter utilizing eddy current effect, Electrical Engineering in Japan, vol. 1, no. 5, pp. 1 7, Septeber 1, translated fro Denki Gakkai Ronbunshi, vol. 1A, no., pp. -, Septeber 1. [] A. Tanaka, N. Takehira, and K. Toda, Analysis of a rectangular coil facing a oving sheet conductor, Electrical Engineering in Japan, vol. 11, no., pp. 15, August 11, translated fro Denki Gakkai Ronbunshi, vol. 11A, no., pp. 5-, August 11. [7] A. Tanaka and N. Takehira, Eddy current speed eter using galvanoagnetic devices, Electrical Engineering in Japan, vol. 1, no., pp., June 1, translated fro Denki Gakkai Ronbunshi, vol. 1A, no., pp. 7-7, June 1.