TECHNICAL FIELD

The present invention relates to a device for speed measurement in a rail-mounted vehicle.

BACKGROUND ART

It has long been desired to be able to measure the speed of a rail-mounted vehicle accurately and with a high reliability over the whole speed range of the vehicle and under all operating conditions. An accurate speed value is desirable to be able to obtain, by integration, the distance covered by the vehicle and hence the position of the vehicle along the track, which information is required by superordinate traffic control systems. Further, it is desirable to obtain an accurate speed value for information to systems for control of the vehicle slip during acceleration or deceleration, and it is then important to obtain a good speed measure also at a very low speed.

Further, inter alia for reasons of reliability and cost, it is desirable that means for speed measurement and position determination be arranged in their entirety on the vehicle and that they be completely, or to the greatest possible degree, independent of external means, such as stationary signalling or measurement systems arranged at the track or at some other location.

It is previously known to use tachometer generators connected to the wheels of the vehicle. However, slipping of the wheels when driving or braking entails unavoidable measurement errors in such equipment. Further, the measures of speed and distance, which are obtained from a tachometer generator, are dependent on the current wheel diameter. This is changed with time, both by wear and by the wheels being turned down, which

is done at regular intervals. The influence of the diameter • change may to a certain extent be compensated by recurring calibrations and adjustments of the measurement system, but the need thereof entails an essential drawback, and under all circumstances a tachometer generator system can probably not provide a higher accuracy in, for example, the distance measurement than some 10 to 30 percent.

In, inter alia, the article entitled "Hastighetsmatning med korrelationsmetod" , Andermo, Mόrk, Sjόlund, Teknisk Tidskrift 1976, No. 3, pages 18-21, it has been proposed (Figure 3 with description) that the speed of a rail-mounted vehicle may be measured optically in a contactless manner with the aid of a correlation method. A sensor mounted in a bogie has two light-emitting diodes which illuminate the rail at two different locations at a known distance from each other. The reflected radiation is sensed at both locations with the aid of p otodiodes. One of the sensed signals is displaced in time until a maximum correlation is obtained between the two time-variable signals. The time displacement together with the known distance between the measuring locations then provides the speed of the vehicle and, by integration, the distance covered. However, in practice, it has been found that optical systems are sensitive to the heavy fouling of detectors, etc., which is unavoidable during vehicle opera- tier.. Further, particles present between the rail and the senεcr, such as, for example, raindrops, snow, and brake dust, result in disturbances of the measurement, among other thi gs by heavy damping of the optical signals. Therefore, it has proved to be difficult, or impossible, to obtain a high reliability and a high measurement accuracy during operation in vehicle environment using equipment of the above-mentioned type.

In Topping, Wennrich: "Radargestϋtzte Weg- und Geschwindig- keitsmessung auf Schienenfahrzeugen, Signal+Draht, 85 (1993), pages 360-364, a system for speed and road measurement during

vehicle operation with the aid of a doppler radar is descri¬ bed. Such a system has proved to be less sensitive to fouling than an optical system. During vehicle operation in the win¬ ter in a Nordic or arctic climate, however, it has proved that the unavoidable presence of snow and ice coatings obstructs the radar radiation to such a high extent that the system cannot be used under these conditions. Further, in measurement equipment of this kind, it has proved to be difficult to obtain the required accuracy of measurement at a low vehicle speed.

US patent specification 4 179 744 describes a device for checking the function of electric rail-mounted vehicles. The device has one or more stationary sensors placed along the rail, which are connected to stationary measurement ampli¬ fiers and signal processing equipment. When the vehicle passes the sensors, these detect the electromagnetic fields from the traction equipment of the vehicle. This makes possible control and analysis of the function of the traction equipment. By arranging two such sensors at a known distance from each other along the rail, and by allowing the signal processing equipment to determine the time displacement between the signals from the two sensors, the speed of the vehicle may be calculated. The device requires stationary installations and it may only give information about the vehicle speed at that moment when the vehicle passes the sensors and cannot give the continuous speed information which is required for, for example, position determination or slip control.

US patent specification 4 283 031 describes a device for use in connection with railway crossings. It has stationary sen¬ sors arranged along the rail for determining, for example, the length of the train, the number of cars, the train speed and direction. Each sensor is arranged near the rail and sen¬ ses those changes in an electromagnetic field, generated by the sensor, which are caused by wheel passages. By deter i-

ning the time between the passages by a vehicle wheel past two sensors arranged at a known distance from each other, the train speed may be determined. The device involves the same disadvantages as the device described in the preceding paragraph.

US patent specification 5 141 183 describes a device in a handling system (e.g. an overhead travelling crane or an industrial robot arranged on a trolley) comprising a carriage movable along a rail. The carriage has current collectors running along stationary contact rails. On the rails, magne¬ tized strips are arranged which have regions with alternately opposite magnetization directions. The current collector is provided with a sensor which senses the field from the strip and which counts the regions which are passed. If the regions have known dimensions, the speed of the car may be deter¬ mined. Also this device requires stationary members (the magnetic strips) and may, therefore, only give speed infor¬ mation upon the very passage of the stationary members.

SUMMARY OF THE INVENTION

The invention aims to provide a device of the kind described in the introductory part of the description, which within the whole speed range of the vehicle exhibits a high reliability and a high accuracy of measurement also under very severe operating conditions, and which is able to work completely independently of means or systems arranged outside the vehicle.

Further, the invention aims to provide a device which makes possible a reliable detection of non-movement of the vehicle.

The invention also aims to provide a device which makes possible detection of rail defects, such as cracks and rail failures.

What characterizes a measurement system according to the invention will become clear from the appended claims.

BRIEF DESCRIPTION OF THE DRAWING

The invention will be described in greater detail in the ^■ following with reference to the accompanying Figures 1-9. Figure 1 shows a sensor means according to the invention, wherein Figure la shows the means viewed from the side, Figure lb the means viewed from above, Figure lc the means viewed in the direction of movement of the vehicle, Figure Id a sensor means in an alternative embodiment, viewed from above, and Figure le an example of the mounting of the sensor means on the vehicle. Figure 2a shows a block diagram of a measurement device according to the invention. Figure 2b shows the configuration of the electronic system, arranged in the sensor itself, in the means according to Figure 2a. Figure 2c shows the configuration of the circuits for signal processing of each one of the two sensor signals in the means according to Figure 2a. Figure 2d shows how the device accor¬ ding to Figure 2a may be supplemented in order to work alter¬ nately at two different frequencies. Figure 3 shows in the form of a vector diagram the components of the output signal of the sensor coil. Figure 4 shows an alternative embodiment of the sensor means according to the invention, with sensor coils in two directions orthogonal to each other. Figure 5 shows a sensor means with three sensor units and with a possibility of choosing between two different measurement distances. Figure 6 shows how the switching between the measurement distances may be made in the device according to Figure 5. Figure 7 shows an example of how the device accor¬ ding to the invention may be supplemented for detection of non-movement of the vehicle. Figure 8 shows an example of how the device according to the invention may be supplemented with means for detection of defects in the rail. Figure 9 shows an alternative embodiment of the magnetization and sensor coils of the sensor means.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

The examples of electric and mechanical dimensioning infor¬ mation occurring in the following description are only approximate.

Figure la shows a sensor means according to the invention. It is arranged on a vehicle bogie above a rail 2, the longitudi¬ nal direction of which coincides with the direction of move- ment of the vehicle and lies in the plane of the paper. The sensor means comprises a housing 1 made of electrically con¬ ducting material, for example aluminium. The electrically conducting house walls provide screening between the sensors and against disturbances from external fields. The housing has three spaces 11, 12 and 13. In each of the spaces 11 and 12 a sensor, Gl and G2, respectively, is arranged. Each sensor has a coil frame 110 and 120, on which a magnetization coil 111 and 121, respectively, is arranged. The magnetiza¬ tion coils have substantially vertical longitudinal axes. Each coil has a length of 80 mm, a diameter of 22 mm, con¬ sists of 150 turns and is fed with an alternating voltage with a frequency of 100 kHz. In the lower parts of the coil frames 110 and 120, grooves are milled perpendicular to the direction of movement, and in these grooves sensor coils 112 and 122 are arranged. Each sensor coil has a height of 7 mm and a width (perpendicular to the direction of movement of the vehicle) of 25 mm and consists of 250 turns. The sensor coils are arranged so as to be rotatable to a certain extent around axes perpendicular to the plane of the paper for adjustment of the coils such that their sensing directions becor.e perpendicular to the direction of the magnetizing field.

The distance d between the lower part of the sensor means and the upper surface of the rail 2 is, for example, 50-100 mm.

It is adapted with respect to the deflection of the bogie and to the reduction of the diameter of the vehicle wheels which

arises when the wheels are turned down which is normally done at certain intervals.

The distance L in the direction of movement between the axes of the two sensors is about 100 mm in the example shown. Each sensor system (with at least two sensors) is measured indivi¬ dually to obtain an equivalent "electrical distance", L'EL- which is then stored in a non-volatile memory (e.g. an E^ memory). This distance, 'E - is then utilized as calibration value by the signal processing unit.

In the common space 13 between the two sensors, certain electronic equipment is arranged on a printed circuit-board 131. In the example described, this electronic equipment consists of pre-amplifiers for the signals from the sensor coils and of bandpass filters for the sensor signals.

The housing 1 is provided with connection devices (not shown) for supply voltages to the magnetization coils and for the output signals from the printed circuit-board 131 and with the necessary connections between the sensor coils and the printed circuit-board.

Further, Figure la shows the coordinate system used herein- after in the description. The X axis of the system is verti¬ cal and parallel to the longitudinal axes of the magnetiza¬ tion coils. The Y axis is parallel to the longitudinal direc¬ tion of the rail, and hence to the direction of movement of the vehicle. The Z axis is horizontal and perpendicular to the longitudinal direction of the rail.

Figure lb shows a section through the sensor means viewed from above. Figure lc shows a section through the sensor means viewed from the front.

The magnetization coil of a sensor generates a magnetic alternating field - the magnetization field - with a substan-

tially vertical main direction. Since the coil lies relati¬ vely close to the rail, the rail will influence the field. Factors which influence the field are the magnitude of the air gap between the sensor and the rail, the geometry of the rail (e.g. dimensional changes, damage, interruptions), and the permeability and conductivity of the rail. During move¬ ment of the vehicle, variations in these factors will gene¬ rate correlatable variations of the magnetic field configu¬ ration. However, the variations in the field are small compa- red with the magnitude of the magnetization field. Since the sensor coil of each sensor is separate, however, it may be oriented in space so as to select and sense that magnetic component which best represents the changes in the field which are caused by these variations in the properties of the rail. By orienting the sensor coils in the manner shown in Figure 1, that is, orthogonally to the magnetization field, the sensitivity of the coils to the strong magnetization field is reduced to a very great extent. The field which is orthogonal to the magnetization field will in this way con- stitute a greatly increased percentage of the output signal from a sensor coil, and in this way it has been possible to increase the sensitivity and the accuracy of the detection of the magnetic-field variations to a very great extent.

Figure Id shows, viewed from above, an alternative embodiment of the housing 1 of the sensor means. The housing consists of an extruded alumium profile with two circular parts, which form the spaces 11 and 12 for the two sensors and which are separated by one part with plane-parallel walls which form the space 13 for the common electronic unit (the printed circuit-board 131) .

Figure le shows an example of the mounting on the vehicle of the sensor means 1 shown in Figures la - lc. The means is mounted on the underside of one of the bogies of the vehicle, namely, the bogie 3 with its two wheel sets 31 and 32.

Figure 2a shows the sensor means according to Figure 1 with associated equipment for feeding the field coils and for signal processing of the output signals of the sensor coils. A supply unit SU feeds the magnetization coils 111 and 121 with an alternating voltage with the frequency 100 kHz with the aid of a sine-wave oscillator OSC and a power amplifier PA. The output voltages un and ui2 from the sensor coils 112 and 122 are supplied to the electronic circuits 131ι and 1312 arranged on the printed circuit-board 131. The output signals udl and ud2 from these circuits are supplied to signal pro¬ cessing circuits SBl and SB2, which generate the digital signals SI and S2. Each such signal constitutes a measure of the instantaneous value of the phase position of the field sensed by the respective sensor coil. The signals SI and S2 are supplied to a calculating unit CE, which by means of, inter alia, correlation of the two signals, calculates measured values of the speed v of the vehicle and the dis¬ tance s covered.

Figure 2b shows the configuration of the electronic circuit 131j _. . The sensor coil 112 is connected to a load resistance Rl cf 50 kohms. The voltage uϋ is supplied to an amplifier and impedance converter Fll with an amplification of 5-10 times and an output impedance of 50 ohms. The output signal of the amplifier is filtered in a bandpass filter BPl for filtering away other signals than those which are derived frorr. the magnetization field, which has the frequency 100 kHz. In the example now described, the filter has a pasεband with upper and lower limit frequencies 150 kHz and 50 kHz, respectively. The output signal from the bandpass filter is designated udl-

The electronic 1312 is built up in the same way as the circuit 131ι.

Figure 2c shows the configuration of the signal-processing circuit SBl shown in Figure 2a. The signal u l is supplied to

a phase-locked loop PLLl with a large time constant, one or a few seconds. This circuit will have a phase position which corresponds to the mean value of the phase position of the input signal udl• The circuit generates two output signals with the same frequency as the input signal, that is, 100 kHz. An output signal U 21 has the same phase position as the input signal and is supplied to a circuit AGCl for control of the working point of the sensor means. A second output signal consists of a square-pulse train rll which is phase-shifted 90° from the former output signal and is supplied to an input of an exclusive OR circuit XORl. The signal u _r n serves as a phase-position reference when determining the phase position of the voltage generated in the sensor coil.

The circuit AGCl is a circuit with a controllable gain. The output signal udml of the circuit has the same phase position and curve shape as the input signal U 21 b t ^a variable amp¬ litude. The input of the circuit for control of the ampli¬ fication is supplied with the signal udl- The peak value of this signal is detected, for example in an envelope detector, and controls the amplitude of the output signal of the circuit in such a way that the amplitude of the output signal almost - but not quite - corresponds to the amplitude of the measured signal urll. The circuit AGCl has a large time constant, for example one or a few seconds, and the output signal udml will therefore have the same frequency as the measured signal udl and an amplitude and a phase position which nearly correspond to the mean values of the amplitude and the phase position of the measured signal di.

In a differential amplifier F21 the signal udml is subtracted fror. the measured signal udi and the difference constitutes the output signal u'dl of the amplifier. The circuit PLLl - AGCl - F21 now described will control the working point of the r.eans such that the component in the output signal of the sensor coil which is caused by the magnetization field is eliminated to the desired extent. In this way, the sensiti-

vity and the accuracy in the detection of the field varia¬ tions caused by the vehicle movement are increased, which variations are small compared with the magnetization field.

It has proved to be suitable not completely to eliminate the voltage component caused by the magnetization field, but to allow the output signal of the sensor coil to contain, for example, 100 mV of this component as phase reference. The XOR gate requires an input signal u'ddl (Figure 2c), which on average should be 90° phase-shifted relative to u _r ll which is obtained from the circuit PLLl. The latter circuit has found its phase position substantially from the component of the magnetization field. Therefore, a sufficiently large component from the magnetization field should also be present in the signal u'ddl and thus also in u'dl-

The output signal u'dl of the amplifier F21 is supplied to a comparator CMPl which emits a logic one if the input signal is larger than zero and a logic zero in the opposite case. The output signal u'ddl of the comparator, which signal is a square pulse train with the same phase position and frequency as the input signal u'dl» is supplied to a second input of the XOR circuit XORl.

If the two input signals to the XOR circuit are in phase, the output signal of the circuit becomes zero. If the input signals are in anti-phase, the output signal becomes 1. On average, the input signal u'ddl will have the same phase position as the measured signal u l, that is, the phase difference between the two input signals to the XOR circuit will, on average, be 90°. On average, therefore, the output signal of the circuit will have the value 1/2, that is, the working point will, on average, lie in the centre of the dynamic range of the circuit, which entails an optimum utilization of the dynamic range.

The output signal (pp of the XOR circuit consists of a pulse train with the frequency 100 kHz and with a mean value which, on average, has the value 1/2 and which may vary between the above-mentioned limits 0 and 1. In a low-pass filter LPl, the 100 kHz component and harmonics of this component are suppressed, and the output signal φ _a of the filter is an analog signal which varies concurrently with the phase posi¬ tion of the output signal of the sensor coil. The output signal of the filter is amplified and converted into digital form in an A/D converter ADl with the output signal Si.

The signal-processing circuit SB2 in Figure 2a is built up in the same way as the circuit SBl described above.

The signals SI and S2 from the signal-processing circuits SBl and SB2 in Figure 2a are supplied to a correlation unit CE. This suitably consists of a microprocessor programmed to perform speed determination with the aid of, inter alia, correlation of the two signals SI and S2 and to calculate, by integration/summation of the speed values, the distance covered by the vehicle.

Each one of the signals SI and S2 is stored continuously as a sequence of a predetermined number of digital values, which thus always reproduce the variation of the signal during a certain time prior to the moment in question.

A continuous calculation of the correlation between the signals Ξl and S2 is made when these are displaced by a varying time interval τ relative to each other. The time displacement τ _m which provides the highest value of the correlation integral is used as one subset for the speed determination. Further, the result of previous measurements (the previous history) , modelling of the dynamic properties of the vehicle (the train) , possibly other (less accurate) speed sensors, as, for example, a tachometer generator, are

used as input data. The evaluation program, which is a statistical probability calculation with adaptive weights of the various input data, then provides an MLE (Maximum Likelihood Estimation) of the instantaneous speed of the vehicle. The speed of the vehicle is obtained as

L EL v = tMLE

where

L'EL i the equivalent "electrical distance" between the two sensors (see Figure 1 with associated description) t LE s the value of t which gives the best possible correlation according to MLE.

Further, the device may be simply adapted to determine the direction of movement of the vehicle by shifting between Si and S2 during the correlation and investigating in which order between the two signal patterns the correlation is obtained.

The microprocessor is adapted to carry out correlation analysis with a predetermined frequency, for example 10 measurements per second.

Figure 2d shows how increased reliability in the speed determination may be obtained by allowing the sensor means alternately to operate at two different frequencies, for example 70 kHz and 100 kHz. A control signal f _c from the calculating member CE switches with a suitable periodicity, for example between each measurement, the oscillator frequency between these two values. Where necessary, filter circuits etc. in the signal-processing units SBl and SB2 are alεc switched synchronously therewith. Since the depth of penetration of the field into the rail is different for the two frequencies, the sensed signal patterns will vary in

different ways during the movement of the vehicle. However, the speed values calculated at one frequency shall, in principle, correspond to the values which are determined at the other frequency. If the values do not correspond, it is possible (if the difference is small) to form the mean value thereof, or (if the difference is great) to take this as an indication of a fault in the sensor means.

As an alternative to allowing the sensor means to alternately operate at different frequencies, two or more sensor systems and measurement channels, operating at different frequencies, may be arranged.

If desired, of course, two or more identical sensor means may be used on a vehicle to obtain a higher availability and increased reliability.

Figure 3 shows in the form of alternating-voltage vectors the output voltage ui of a sensor coil, which voltage is composed of the two components ujx and iy. That component in the output signal of a sensor coil which is directly caused by the magnetization field may, in practice, never be eliminated by adjusting the orientation of the coil. However, by the circuit for control of the working point of the means, described above with reference to Figure 2c, as mentioned, this component may be further reduced to the desired degree. However, it has proved to be suitable not to eliminate the component completely, and therefore, in the output signal of the sensor coil, there is a component ui _x which is caused by the magnetization field. The field variations in the direc¬ tion of sensing of the sensor coil (orthogonally to the mag¬ netization field) , which are caused by the movement of the vehicle, will substantially provide a component iy of the output signal of the sensor, which component has a 90° phase shift relative to the component ui _x . The variations of the component uiy cause variations of the phase position φ of the output signal of the coil relative to the phase position of

the component i _x generated by the magnetization field. As mentioned, the orientation of the sensor coils orthogonally to the magnetization field entails a great reduction of the influence of the magnetization field on the output signal of a coil. The variations in the phase position of the sensor signal which are caused by the variations in the voltage component uiy therefore become greatly increased, which entails a good sensitivity and accuracy in the detection.

Figure 4 shows, viewed from above, a sensor means of the same type as that shown in Figure 1. A sensor, Gl and G2, respec¬ tively, is arranged in each of the spaces 11 and 12 and each sensor has, in the same way as in Figure 1, a magnetization coil (not shown) which generates an alternating field with a vertical main direction. Also, in the same way as in Figure 1, each sensor has a sensor coil 112y and 122y, respectively, with their sensing directions in parallel with the Y-axis. In the means shown in Figure 4, each sensor has an additional sensor coil, 112z and 122z, respectively, with its sensing direction in parallel with the z-axis.

From the sensor coils 112y and 122y, two sensor signals, here designated udly and ud2y- are obtained, in the same way as described with reference to Figures 1 and 2, via electronic circuits 131 arranged in the sensor, which sensor signals are signal-processed and correlated with each other for forming a measure of the speed v of the vehicle in the manner described with reference to Figure 2. The signals uc!i _z and ud2z are processed in the same way, either by separate signal- processing circuits and calculating means, or by using the same circuits alternately for determining the vehicle speed from the signals from one of the pairs of coils and alterna¬ tely from the signals from the other pair of coils. Possibly, one of the pairs of coils with its signal-processing circuits and calculating means may be used in the normal case and the other pair of coils with its signal-processing circuits and calculating means serve as a pair of stand-by coils to be

activated in the event of a fault in the normally used system.

It has been found that sensor coils with their sensing direction in the y-direction are insensitive to fields which are caused by traction and signal currents flowing in the rail, and that this coil orientation may be preferable. This is not the case with coils which have their sensing direction in the z-direction, but the disturbing influence of the above-mentioned currents may to a great extent be reduced with the aid of some disturbance elimination method known per se, for example according to the Swedish patent with publi¬ cation number 441 720.

Figure 5 schematically shows a sensor unit according to an alternative embodiment of the invention. It has three sen¬ sors, each one designed as, for example, the sensors in Figure 1. It has a sensor Gl and a sensor G2 in the same way as the sensor means of Figure 1. The distance between the centre lines of the sensors constitutes the measuring dis¬ tance L. The sensor signals from the sensors Gl and G2 are correlated with each other in the manner described with reference to Figure 2, and the speed of the vehicle is cal¬ culated by means of the measuring distance L.

Between the sensor Gl and the common electronic space 13, a third sensor G2 ' of the same kind is arranged adjacent the sensor Gl and forms together therewith a shorter measuring distance L' with a length of, for example, 40 mm. At low vehicle speed, this shorter measuring distance provides considerably faster speed determination than the longer measuring distance L. As shown in Figure 6, the choice of measuring distance may preferably be made automatically in dependence on the vehicle speed. The signals ud2 and u'd2 from the sensors G2 and G2 ', respectively, are supplied to the signal-processing unit SB2 via electronic switching members SWl and SW2. A level-sensing circuit NVl is supplied

with the calculated speed value v. if the speed is greater than a certain predetermined value vo, the longer measuring distance is used and the signal ud2 is switched via the switching member SWl into the signal-processing unit SB2. If the speed does not exceed the value vo , the shorter measuring distance is activated by instead switching the signal u'd2 via the switching member SW2 into the unit SB2.

Although the system described above with a suitable dimen- sioning may provide a good measurement result down to a very low speed, the measurement system unavoidably ceases to function when the speed approaches zero. In many applica¬ tions, therefore, it is desirable to complete the system with an indication as to whether the speed of the vehicle is zero.

When the vehicle is stationary, Ξl and Ξ2 will be uncorrela¬ ted time sequences for all time displacements τ between the sequences. One method is to test for total independence when the estimated value of the speed is below a predetermined limit. When this hypothesis is verified at a given signifi¬ cance level, uncorrelated sequences are indicated. In addi¬ tion, SI and S2 will be approximately static sequences when the speed is zero. A low variance (RMS value) is, therefore, also an indication of the speed being zero.

Figure 7 shows how these tests may be combined. The calcu¬ lating unit CEa is supplied with the signals SI and S2 and delivers - if the above-mentioned test of the absence of correlation is fulfilled - a signal NC which indicates that the two signals are uncorrelated. The circuits CEb and CEc deliver signals LVl and LV2 if the variance of the signals SI and Ξ2, respectively, lies below predetermined levels. The signals NC, LVl and LV2 are supplied to an AND circuit AC which delivers an indicating signal V=0 if all three tests are fulfilled.

As an alternative to the tests described in the preceding paragraph, other known statistical standard tests for stationary state may be used.

The circuits shown in Figure 7 suitably consist of parts of the program for a microprocessor which constitutes a control and calculating unit for the speed sensor.

Figure 8 shows how the device described above may be supple- mented with means for detection of defects in the rail, such as cracks or rupture. The sensor signal SI is supplied to a circuit CD which calculates one or more predetermined charac¬ teristics Csl of the signal, for example maximum amplitude or rate of change. That value or those values CM of the corre- sponding characteristics, which occur at the defect or defects which are to be detected, are stored in advance in a memcry M. A comparison circuit COMP2, a pattern recognition circuit, continuously compares the characteristic quantities Csl and CM and delivers a detection or alarm signal SL at a predetermined degree of correspondence.

Alternatively, the defect detection may be made by comparing that signal pattern, which the values of Si for a certain period of time constitute, with the corresponding signal pattern stored in advance in the memory. The comparison between the signal patterns may possibly be made by time- shifting one of the patterns, in the same way as with the speed measurement described above, in relation to the other pattern until maximum correlation is obtained, whereby a fault is considered to have been detected if at least a pre¬ determined degree of correlation is obtained.

Figure 9 schematically shows an alternative and advantageous embodiment of the magnetization and sensor coil of a sensor (e.g. Gl or G2 in Figure 1). The magnetization coil 111 is designed as a flat sheet-wound coil with a vertical axis. The coil has a considerably smaller length (extent in the verti-

cal direction in the figure) than diameter. By making the coil short, all its winding turns will be as close to the rail as possible, which provides a high magnetizing field intensity at the rail surface. The sheet winding is suitably performed in the manner shown in the figure, with a large number of turns of a thin sheet, which provides a large effective area and hence a lower effective resistance and a higher current-handling capacity than a corresponding wire winding (at the frequencies used, the depth of penetration will be small because of the skin effect) . Alternatively, however, the magnetization coil may, of course, be designed as a wire-wound coil.

The sensor coil 112 is arranged adjacent to the magnetization coil and at the same height d above the rail 2 as this. As in the sensors described above, the sensor coil has a horizontal longitudinal axis and sensing direction. With the location shown - as well as with the sensors described above - the sensing direction of the coil will be perpendicular to the direction of the magnetizing field at the sensor coil. Since the sensor coil is arranged adjacent to the magnetization coil, the latter may be arranged nearer the rail, which provides a higher field intensity at the rail surface. It has proved to be particularly advantageous to arrange the sensor coil 112 displaced in the longitudinal direction of the rail by a distance from the magnetization coil which is approxi¬ mately half (d/2) of the distance d between the magnetization coil 111 and the rail. At this distance, the magnetizing field intensity at the rail is highest, and a maximum induc- tion is obtained in the sensor coil. The embodiment of the sensor shown in Figure 9 has proved to provide good detection properties.

In the sensor means described above, the signals which are correlated are formed by sensing the phase position of the sensor signal. Alternatively, the amplitude component of the sensor signal may also be used.

Further, in the sensor means described above, each one of the signals which are correlated is obtained from one single sen¬ sor coil. Alternatively, several sensor coils may be connec¬ ted together to form such a signal, in which case the coils are oriented with different sensing directions, chosen in a suitable manner, for optimization of the sensitivity of the total output signal to the desired field variations and/or insensitivity to direct influence by the magnetization field.

For monitoring the function of the device, a separate monito¬ ring winding may be adapted to sense the amplitude and/or phase of the magnetization field. In that case, a monitoring unit is adapted to trigger an alarm in the event of loss of the magnetization field, or if the characteristics of the field deviate from the desired ones.