Eddy-Current Sensor for Measuring the Profile of Collectors of Electrical Machines

A. I. Proshin, L. Ya. Zinner, A. I. Skorospeshkin

(Recommended by the seminar of the departments of electrical machines and general electrical engineering)

The main difficulty that arises in the development of instruments for measuring the profile of collectors under dynamic conditions is associated with the creation of a sensor sufficiently sensitive to small changes in the profile. Of the known transducers of mechanical displacements into electrical signals, capacitive and eddy-current sensors are most often used; these have significant advantages over others. These are, first of all, high sensitivity and resolving power and, which is especially important, the possibility of measurements by a contactless method.

Concerning the question of developing sensors for measuring the profiles of collectors of operating electrical machines, when the collector is at a high potential and operation of the sliding contact is accompanied by considerable sparking, we note that the eddy-current sensor is the most acceptable in this case. This is explained by the fact that an eddy-current sensor, unlike a capacitive one, does not require electrical contact with the collector. In addition, an eddy-current sensor has high sensitivity, better noise immunity, and is less sensitive to instability of the capacitance of the connecting leads.

The principal drawback of an eddy-current sensor is the dependence of its readings on temperature changes in the electrical conductivity of the collector material and insufficient resolving power. Below we discuss measures for eliminating them.

Figure 1 shows the design, and Fig. 2a the external appearance, of an eddy-current sensor for collectors with segments not less than 5 mm wide. This sensor belongs to the transformer type and is used in a circuit with amplitude modulation of the high-frequency voltage. It has two windings placed on a U-shaped core made of 600NN ferrite. The excitation winding \(W_{\mathrm{v}}\) is supplied from a current generator of stable frequency, having a large output resistance. The measuring winding \(W_{\mathrm{i}}\) is included in a measuring resonant circuit, tuned at infinite clearance to a frequency somewhat higher than the excitation frequency, so that the operating point lies in the upper part of the most linear section of the resonance curve.

The core with the wound windings is embedded in a massive copper screen, which serves as the sensor housing. Only the end faces of the core and the winding leads emerge at the surface. To prevent a short-circuited turn, a gap of \(0.1\)—\(0.2\) m is left in the middle part of the screen (between the branches of the core).

Fig. 1. Design of an eddy-current sensor

Fig. 2. a) eddy-current sensor with a linear characteristic;
b) eddy-current sensor with a narrowly localized field

The shield concentrates the electromagnetic field of the sensor in the half-space adjoining the end faces of the core, directing the lateral leakage fluxes into this region. Experiment showed that, under identical supply conditions for the sensors, the field intensity at the end surfaces of the shielded sensor is 2.5–3 times greater than for an unshielded one. This is explained by the fact that the magnetic flux of the eddy currents in the shield, directed opposite to the leakage fluxes that induce these currents, is added to the main field.

When a conducting nonmagnetic plate is brought close to the end surfaces of the sensor, the electromagnetic field of the eddy currents excited in it, acting on the electromagnetic field of the sensor, changes the electrical parameters of the measuring winding $W_i$.

The introduced inductance, having a negative sign, increases the resonant frequency of the measuring circuit, while the introduced active loss resistance reduces its quality factor. In addition, owing to the separate arrangement of the windings (the excitation winding is in the middle part, and the measuring winding is at the ends of the core), when the plate is brought close, the coupling coefficient between them decreases.

Each of these factors separately causes a decrease in the voltage at the sensor output as the distance between it and the conducting surface is reduced (Fig. 3). An increase in the resonant frequency of the measuring circuit causes the operating point to shift downward along the slope of the resonance curve. A decrease in the quality factor leads to a reduction of the resonance peak and a widening of the passband. A decrease in the coupling coefficient is manifested in a drop in the ratio of the voltages on the measuring winding and the excitation winding.

Figure 4 shows the initial section of the output characteristic of the sensor, recorded at different frequencies with the same stabilized excitation current. Sensor parameters: $W_v = 10$ turns, $W_i = 30$ turns, capacitance of the measuring circuit $C_i = 30$ pF. The resonant frequency of the sensor at an infinite gap is 2 MHz. The excitation current

Fig. 3. Voltage in the measuring circuit of the sensor as a function of the gap magnitude and frequency

Fig. 4. Initial portion of the sensor characteristic

The plate material forming the variable gap with the sensor is electrical copper. Its thickness is tens of times greater than the depth of penetration of the electromagnetic field into copper at the given frequencies.

When measuring the profile of a commutator, the gap between its surface and the sensor, in order to increase the resolving power of the latter, should not exceed 0.5 mm.

It can be seen from Fig. 4 that, by selecting the operating frequency, it is possible to obtain satisfactory linearity of the initial portion of the characteristic. At a frequency of 2 MHz (Fig. 3), the ratio of the output voltages corresponding to infinite and zero gaps is maximal and is equal to 50 (2 V and 0.04 V), but the most linear section with the greatest sensitivity lies in the interval 1.2–1.7 mm.

With a decrease in frequency to 1.8 MHz, the sensitivity falls somewhat, but the characteristic in the distance interval 0–1.0 mm is practically linear. The sensitivity at a current of 3 mA at a frequency of 1.8 MHz is 0.21 mV/µ. For comparison, we note that the sensitivity of an unshielded sensor consisting of two coaxial coils 35 mm in diameter is 2 mV/µ at a current of 50 mA and a frequency of 2.0 MHz [1].

The excitation frequency can be raised to 5 MHz and higher by reducing the circuit capacitance \(C_i\) and, if necessary, the number of turns \(W_i\) and \(W_v\). In this case the resolving power of the sensor improves, the influence of temperature variation in the electrical conductivity of the collector material decreases, and the sensitivity increases in proportion to the square root of the frequency [2, 3].

The high-frequency current generator must have a large output resistance, so that the sensor excitation current does not depend on its input resistance, which increases with increasing gap. This circumstance increases the sensitivity and linearity of the output characteristic.

The electronic measuring circuit following the sensor must have a high input resistance, on the order of tens of kilohms, and a minimum input capacitance, so that fluctuations of these parameters have little effect on the properties of the measuring circuit.

Despite the fact that the width of the working surface of the sensor (0.7 mm) is much less than the width of the lamella, the magnitude of the latter, if it is less than 5 mm, strongly affects the measurement result. This occurs because the region over which the eddy currents spread along the conductor surface extends far beyond the working surface of the sensor, and the sensor begins to respond to the approach of the edge of the plate already at 2–2.5 mm. As a result, the voltage pulse during passage over a lamella has strongly rounded edges, and the sensor allows the profile of collectors with lamellae at least 5 mm wide to be measured with permissible error. Further narrowing of the lamella causes changes in the pulse height just as an increase in the gap does.

For collectors with narrower plates, a sensor with a narrowly localized electromagnetic field was developed. Its external appearance is shown in Fig. 2b, and its design and overall dimensions in Fig. 5.

The ferrite Π-shaped core is composite. Its two poles 1, 2, with grooves preliminarily turned for the measuring winding \(W_i\), are glued through a shielding copper wedge-shaped plate so that the minimum distance between them is 0.15–0.2 mm. The poles are connected by a carefully ground yoke 3 with the excitation winding placed on it.

The opposite end of the pole block is ground to a four-sided cone.

The block of poles and the yoke with the windings applied to them are soldered into the sensor housing. Only the ends of the poles of the housing core emerge at the surface, forming the working surface of the sensor, measuring \(0.7 \times 0.2\) mm.

The dependence of the resolving power of the sensor on frequency was investigated. Fig. 7 shows the shape of the voltage pulse on the measuring winding, corresponding to the passage of a lamella, recorded at various frequencies. The voltage was measured with a vacuum-tube millivolt—

Fig. 5. Design of an eddy-current sensor with a narrowly localized field

Fig. 6. Dependence of the resolving power of the sensor on frequency.
\(U_{\delta=\infty} - U_{\delta=0} = \text{const},\ \delta = 25\ \mu\text{m}\)

meter. In order to increase the accuracy of the experiment, the voltage at the zero gap was compensated by means of an additional sensor connected in opposition to the main one. This made it possible to use the entire scale of the instrument.

The measuring circuit was tuned to the required frequency by changing its capacitance \(C_{\mathrm{u}}\). Since, as the frequency increases, the sensitivity of the sensor increases, the voltage difference at infinite and zero gaps was kept constant by reducing the excitation current.

The voltage distribution was recorded over one and the same plate 2.9 mm wide. In all cases the distance of the sensor from it was the same and equal to 25 \(\mu\)m.

At a frequency of 5 MHz the peak of the voltage pulse has a clearly pronounced flat portion. This is not observed at a frequency of 500 kHz. A further increase in frequency should lead to a widening of the flat portion. The dotted curve is the limiting case, when the resolving power is determined only by the ratio of the dimensions of the sensor and the lamella and by the magnitude of the gap between them, but does not depend on any further increase in frequency, since the region over which the eddy currents spread along the surface already almost does not extend beyond the region over which the electromagnetic field spreads.

The resolving power of the sensor makes it possible to measure the profile of commutators with lamellae at least 2 mm wide.

When measuring the absolute value of a gap up to 150 µm, the error due to the edge effect for plates more than 3 mm wide is practically zero; for plates 3–2.5 mm wide it is not more than \(+2\) µm, and for plates 2.5–2.0 mm wide it is not more than \(+5\) µm.

When measuring the value of the gap deviation and a 10 percent spread in the lamella width of a given commutator, the error due to the edge effect is:

for lamellae wider than 2.5 mm — not more than \(\pm 0.5\) µm,
for lamellae 2.0–2.5 mm — not more than \(\pm 1\) µm.

Fig. 7 gives the dependence of the sensor output voltage on the magnitude of its distance from a copper plate. The operating frequency is 5 MHz. Increasing the resolving power by reducing the dimensions of the working surface of the sensor led to a sharp deterioration in the linearity of its characteristic and to a reduction in the range of measured gaps to 0.3 mm. In developing the profilometer circuit, it is necessary to provide for correction of the sensor characteristic by automatic regulation of the excitation current as a function of the gap value, or for parametric correction during signal detection. The second method is less complicated and provides better stability of the characteristic.

Fig. 7. Characteristic of a sensor with a narrow-local field

The resolving power of the sensor depends strongly on its distance from the surface of the commutator; therefore measurements should be made at the minimum permissible gaps of 100–150 µm, depending on the expected value of the commutator runout.

The dependence of the sensor readings on temperature-induced changes in the electrical conductivity of the commutator material was also checked experimentally.

The electrical conductivity of copper when heated from 20 to 120°C decreases—

is reduced by 40 percent. According to the graphical relationships presented in [2, 3], with a gap of 0.5 mm and a frequency of 5 kHz, a 40 percent decrease in the electrical conductivity of copper causes an increase in the inductance of the sensor windings by approximately 2 percent, and at a frequency of 75 kHz—by 1 percent.

With a further increase in frequency, the influence of the conductivity value continues to decrease [2].

Since the operating point of the sensor lies on the steepest part of the resonance curve, the relative change in output voltage when the collector temperature rises is considerably greater than the relative change in the inductance of the measuring winding.

At a frequency of 50 kHz, heating of the collector to \(120^\circ\text{C}\) causes a decrease in the voltage difference at infinite and zero gaps by approximately 11–12 percent. At a frequency of 5 MHz this decrease does not exceed 2–3 percent.

A change in the temperature of the sensor itself by \(\pm 20\)–\(25^\circ\text{C}\) causes only a shift of the zero of the characteristic. The voltage difference corresponding to infinite and zero gaps, and consequently the sensitivity, remains practically unchanged.

Thus, an increase in the operating frequency is advisable in order to improve the resolving power of the sensor, increase its sensitivity to the gap, and reduce sensitivity to fluctuations in the electrical conductivity of the collector material. It is also possible to improve the linearity of the characteristic, if the resolving power of the sensor permits a compromise increase in the size of its working surface.

In addition, at high frequencies the influence of the speed of motion of the conductor relative to the sensor on the readings of the instrument is smaller. At frequencies on the order of hundreds of kilohertz and higher it is negligibly small [3].

A limitation on increasing the operating frequency may be caused by the frequency properties of the ferrite and by the use of semiconductors in the electronic circuit.

Thus, the eddy-current sensors developed by the authors, possessing high sensitivity and good resolving power, make it possible to investigate the behavior of collectors of electrical machines in dynamic conditions. At the same time, the sensors ensure high measurement accuracy, which makes it possible to estimate the relative displacement of segments with an accuracy of \(+0.5\ \mu\).

REFERENCES

1. N. N. Shumilovsky, G. G. Yarmolchuk, V. P. Grabovetsky, M. A. Prusov. A Method of Eddy Currents for Monitoring Production Parameters. ILiM, 1964.

2. D. L. Dorofeev. Nondestructive Testing by the Eddy-Current Method. Oborongiz, 1961.

3. V. V. Nikolaenko, N. N. Rudnyi. Investigation of the Influence of the Speed of Motion of the Tested Specimen on Measurements. ONTIPribor, 1964.