FR2766274A1 - Method for locating faults in a multi-pair communication cable - Google Patents

Abstract

Models the cable condition by studying and measuring the propagation of a wave on the cables electrostatic screen and not on the cable pairs. Sends a short duration electrical pulse, of set amplitude and duration into one end of the cable (1), between the cable screen (11) and the ground cover. Detects echoes and reflections from the cable fault (14). Determines position of fault from propagation time of signal to fault and back. Pulse duration is short with respect to the time necessary for its propagation along the cable. INDUSTRIAL STANDARDS - INSTRUMENTATION AND TESTING - The device is designed to detect cable faults.

Description

METHOD FOR LOCATING DEFECTS ON A METAL CABLE
AND DEVICE FOR CARRYING OUT THIS METHOD
The present invention relates to a method for locating faults on a metallic cable of a communication network, more particularly to a method of locating faults on a metallic multipair telephone cable.

 The invention also relates to a device for implementing this method.

 In what follows, and without this limiting in any way the scope of the invention, we will consider the case of a local distribution telephone network by multipair metal cable.

 In fact, for local distribution telephone networks, there are two main types of cables: twisted metal pair cables and fiber optic cables. The latter can, a priori, carry a greater amount of information simultaneously. However, fiber optic systems are much more expensive than traditional metal pair systems and are more complex to implement.

In addition, if we do not consider the case of high speed arteries, but the aforementioned local distribution networks, efficient communication techniques make it possible to greatly increase the speed of twisted pairs. By way of non-exhaustive example, we can cite the "HDSL" and "ADSL" techniques. These techniques ultimately make the use of fiber optic networks unattractive, all the more so since the existing conventional fleet of metallic cables is very large. . It has been estimated at around 140 billion francs for France.

 There is therefore a need to maintain even this existing park in good condition. More precisely, it is advantageous to carry out predictive maintenance, that is to say which makes it possible in particular to detect and, above all, locate in advance the appearance of a fault at a specific location of a transmission cable.

 As illustrated in FIG. 1 placed in the appendix to this description, a transmission cable 1 usually comprises the following elements: an outer sheath 10, waterproof and electrically insulating, for example made of polyethylene, and a tubular electrostatic screen 11, inside which are arranged tens, even hundreds of twisted pairs 12, conveying the signals to be transmitted. For reasons of simplification of the drawings, only the two extreme pairs have been shown: pl and pn, n being the total number of pairs. The electrostatic screen 11 is produced, for example, from a thin aluminum strip (typically 80 μm thick), wound in contiguous turns. This screen 11 is commonly called "Alupe" in terms of profession.

 There are three main cable configurations: cables in the ground, cables in buried pipe, the pipe being made for example of polyvinyl chloride ("PVC"), and overhead cables.

 Whatever the configuration of the cables, there are two main sources of faults.

 The first type of fault, the most common (around 80% of cases), is of mechanical origin and concerns splices in connection boxes, overhead or buried. The origin of these faults is diverse: vibrations (proximity to a road or railway), rodents, permanent attack on water or mechanical stresses introduced when laying cables. These faults are generally easily located, especially for buried cables or pipes, because water infiltration in the sleeve leads to anomalies. The invention is not affected by this type of defect.

 The second type of fault concerns alterations to the cable itself. These are faults that appear along it.

 It is easily understood that, even for overhead cables, these defects are generally not detectable by a simple visual inspection (otherwise impossible for buried cables).

 A cable has a set of physical parameters which allows it to be characterized: permittivity, conductance, characteristic impedance, etc. These parameters can be "vustw of the ends of a cable. In other words, when there is a defect or an alteration along the cable, there is a local variation of some of these parameters which are reflected or, at the very least, can be measured or detected at the ends of the cable.

 We could think of modeling the behavior of a cable as a prerequisite for various assault simulations on this cable. The data collected, associated with each type of attack, would make it possible to characterize the resulting faults and, possibly, to give the possibility of locating the place of appearance of the defect.

 However, until recently, the mathematical approach was reduced due to the limited automatic calculation means available, the phenomena involved being very complex. Only methods by experimentation and / or observation could be implemented.

 The appearance of very powerful, miniaturized and relatively inexpensive IT tools has modified this approach. However, if the modeling of a simple cable, for example coaxial, or of a simple pair is now quite feasible by these means, the modeling of much more complex cables, comprising tens or even hundreds of pairs, remains still out of reach of these IT tools, in time and for realistic costs. Indeed, each conductor being coupled to its neighbors, the number of interdependent parameters and the resulting level of complexity make the computational load prohibitive.

 It is also necessary to create models for each of the recalled cable configurations (buried, under pipe or overhead) and, for each of these models, sub-models taking into account, if not all possible conditions, but at least one significant set of conditions typical of the ambient environment in which the cable is immersed. It would suffice to simulate and study certain borderline cases, difficult, expensive, long or even impossible to recreate in reality and to measure the corresponding behavior of the cable.

 There is however a possible simplifying hypothesis. The vast majority of cable alterations, statistically observed, are due to degradations of its surface layer, that is to say of the protective sheath: porosity, etc.

 Under conditions that can be described as "normal" operating (intact cable), the electrostatic screen is isolated from the ambient environment. When degradations of the protective sheath appear, suddenly or gradually, that is to say an "abnormal" regime, there is a correlative appearance of electromagnetic phenomena between the electrostatic screen and the ambient medium, since this screen is no longer insulated by the sheath, or at least the degree of insulation drops below a given threshold.

 However, by studying these phenomena, the Applicant has been able to establish, surprisingly, that it was possible to model the behavior of a multipair cable simply by studying and measuring the propagation of waves on the electrostatic screen, and not on one and / or several pairs.

 The invention takes advantage of these observations and proposes a method of locating a fault along a multi-pair cable.

 To do this, use is made of a reflectometry method, known per se, consisting of injecting a signal at one end of a cable and studying the characteristics of the return signal, after reflections, possibly multiple, or only at the end of the line with distributed constants that the cable forms (if there are no intermediate faults), that is to say precisely on the singularities that the faults form.

 According to a main characteristic of the invention, this signal is injected between the electrostatic screen and the ambient medium (earth, etc.).

 In a preferred embodiment, a temporal reflectometry technique is implemented by injecting a signal of very short duration compared to the time necessary for its propagation along the cable.

 Still in a preferred embodiment, a pulse of Gaussian form is used.

 According to another characteristic of the invention, a mathematical processing of the reflected signal is carried out. Indeed, in a large number of situations, the reflected signal will be noisy. By way of example, the ambient medium can be intrinsically subjected to intense electromagnetic disturbances or, if the defect in the cladding is slight, the resulting echo is slight and, a priori, noisy.

To extract the useful information from the noise, the reflected signal is subjected to a processing step by auto correlation, or preferably to a processing step by intercorrelation.

 The method according to the invention makes it possible to locate a fault on a cable with great precision. However, in certain configurations, the precision obtained may prove to be insufficient. It is then necessary to carry out important work to effectively go back to the fault. This concerns more particularly buried cables. It is particularly important that the cable area with a defect is determined with the greatest possible accuracy, to avoid having to dig up a long section of cable.

 Also, according to a complementary aspect of the method of the invention, once the fault has been located, with a determined precision, it is possible to carry out an additional step consisting in measuring the electromagnetic field radiated by the cable, when a signal is injected electric at its entrance. We can thus make the location of the fault along the cable more precise and reduce the area of uncertainty to its strict minimum.

The subject of the invention is therefore a method of locating at least one defect along a multi-pair metallic cable, said cable comprising a plurality of pairs of conductive wires intended for carrying information signals, an electrostatic conductive screen of the electricity surrounding these pairs and an external sheath of electrically insulating material, said cable being immersed in a determined medium, characterized in that it comprises at least the following steps
- generation of an electrical pulse, of shape,
of specified amplitude and duration
- injection of this pulse, for the first time
end of said cable, between said electrostatic screen
and the material constituting said determined medium,
so as to obtain the propagation of a wave
electromagnetic along said electrostatic screen
- time analysis of electrical signals
appearing at said first end of the cable, on a
specified period, to detect the presence of echoes
corresponding to reflections of said wave
electromagnetic on said fault, this fault is
translating by interactions between said screen
electrostatic and said determined medium
- and determination of the distance between said fault
of said first cable end from time
necessary for said electromagnetic wave to
propagate until said fault and return to said fault
first end and the propagation speed of
the electromagnetic wave in the cable.

 The invention also relates to a fault location device along a multi-pair metal cable, for the implementation of this method.

The invention will be better understood and other characteristics and advantages will appear on reading the description which follows with reference to the appended figures, among which
- Figure 1 schematically illustrates an example of multi-pair wire rope
- Figure 2 schematically illustrates a simplified model of this cable
- Figure 3 schematically illustrates a cable model according to the method according to the invention
- Figures 4a to 4d illustrate this same model, completed to take account of different particular configurations of cables
- Figures 5a to 5c illustrate these same models affected by defects and taking into account the characteristics of the environment in which they are immersed
- Figures 6a to 10b are function diagrams explaining the method of the invention
- Figure 11 illustrates an example of a fault location device along a multi-pair metal cable, for the implementation of the method of the invention
- And Figures 12a and 12b illustrate an additional variant of the method according to the invention for locating a fault more precisely.

The fault location method in a multipair wire rope essentially comprises the following two phases
The first phase, or preliminary phase, consists in defining models for the main configurations of multipair metallic cables which have been recalled, and correlatively of the faults appearing on these cables, and in carrying out simulations making it possible to determine the characteristics of the cables subjected to alterations, under different measurement conditions. This phase makes it possible to acquire characteristic data for each of the models, and therefore for the behavior of the multipair metallic cables associated with these models.

The second phase or phase of measuring the signals on the cable uses a reflectometry technique poorly and proceeds as follows
- injection of a signal at the end of the cable, between the electrostatic screen and the environment
- measurement of signals on the cable.

 The second phase can be completed by a mathematical processing of the signals which will be specified below.

 We will now detail these different phases.

 Concerning the first phase, the modeling of a simple, monopair or coaxial cable became possible with the powerful computing resources currently available. It consists in applying the theory of the propagation of the waves, ie to solve the equation known as "of the telegraphists", to reproduce the real behavior of the cable.

 When the cable becomes more complex (tens or hundreds of pairs), which is the case of the multipair metallic cables considered in the context of the invention, the fine modeling becomes unusable.

Each driver is in fact coupled to its neighbors and the number of interdependent parameters (several thousand) becomes prohibitive.

 It is then illusory to try to finely model a cable comprising a large number of pairs.

 Experience has shown that it is sufficient to consider only one of the pairs, and to simulate the others by a single solid conductor. The diameter of this solid conductor is the "copper" equivalent of the other pairs.

 The simplified model Ms of the cable becomes that illustrated in FIG. 2. It breaks down, like a real cable, into a sheath 10 and an electrostatic screen 11, or shielding. However, the set of pairs 12 can be simulated by a single pair px (x being an arbitrary index between 1 and n) and the aforementioned single full conductor, referenced 120.

 The model shown in Figure 2 is only interesting if the internal conductors (pairs) can be used for detecting and locating a cable fault.

 However, experimentation has also shown, surprisingly, that it was necessary to take into account, not the internal pairs 12, but the electrostatic screen 11, and only this electrostatic screen, to be able to detect a fault at level of the insulating sheath 10, and above all to locate it.

 Experience has indeed shown that the propagation of waves on this electrostatic screen 11 was not affected by the presence of the internal pairs 12.

 The Mo model in accordance with the method according to the invention, and illustrated by FIG. 3, can therefore be summarized as an electrostatic screen 11 and an insulating sheath 10, these two elements being surrounded by the material constituting the external ambient environment Ma. L the electrostatic screen 11 itself surrounds a dielectric material 13 simulating the interior of the cable.

 This model is therefore particularly simple. To fix the ideas, we will consider in what follows that the diameter of the electrostatic screen 11 is 5.56 mm (the latter forming an aluminum tube, the wall of which is 80 mm thick ), that the outside diameter of the cable (sheath 10) is 7.1 mm and that the internal and external dielectric materials on the electrostatic screen 11 have a dielectric constant equal to 2.3.

 In reality, this basic Mo model can be broken down into three main "sub-models", depending on whether one of the three configurations of multi-pair wire rope is considered: buried, under pipe or overhead. It is indeed necessary to also model the environment, that is to say the ambient environment, which differs greatly from one configuration to another.

 FIG. 4a illustrates the standard model Mot of a buried multipair metallic cable, the ambient environment Mat being constituted by earth. It is of course necessary to retain some simplifying assumptions.

 The cable in the ground is supposed to be buried at a depth of about 1 meter (which corresponds to the "FT Standard"). This value is important in modeling, because it defines the quantity of matter which is "seen" by the electromagnetic wave injected into the electrostatic screen 11 (as will be specified in the following), during its journey along cable 1.

 The earth forming the ambient medium Mat is assumed to be homogeneous. Its CT conductance and ErT permittivity are variable parameters simulating the different geological layers and humidity.

 Figure 4b illustrates the standard Moc model of a multi-pair wire rope in buried pipe. The cable in buried pipe is not in direct contact with the ground. It is arranged in a pipe 2, made of synthetic material (for example polyvinyl chloride), with a typical diameter of 80 mm. This pipe 2 is supposed to be filled with water (ambient environment Mae). The presence of water is frequent and is very important for an electromagnetic wave as a conductive element with high permittivity. The conductance and permittivity of water have been referenced in FIG. 4b, aT and erT.

 FIG. 4c schematically illustrates a real aerial multipair metallic cable 1 ′. It comprises, as before, a plurality of pairs 12, an electrostatic screen 11 and a sheath 10. The latter comprises a pocket or a conduit 100 in which is threaded a rope or carrier cable 3, usually made of steel. The ends of this cable 3 are fixed to telephone poles (not shown).

 Experience has shown that the carrying cable 3 does not influence the propagation of the waves, as is the case for the internal pairs, and can be ignored. The model equivalent to an overhead multipair metallic cable therefore comes down to the model of the cable in the ground, that is to say that shown in FIG. 4a, but with air as the ambient medium.

 However, the critical point in modeling such a cable configuration lies in the fact that the air, depending on the time, day and season, has very different dielectric characteristics.

 As an example, it is preferable to look for faults in the early morning when the cable is covered with a thin layer of dew: e = 0.5 mm for example. The permittivity of the Moar ambient medium is therefore that of oatmeal water and its conductivity sEau.

 FIG. 4d illustrates the standard Moar model of an overhead multipair metallic cable under the conditions set out above, that is to say surrounded by dew (ambient Maar environment).

 The modeling in accordance with the invention of the various cable configurations having been specified, it remains to model the faults encountered.

 First of all, as has been indicated, only the faults encountered along the cable, and more precisely those due to alterations in the insulating sheath 11, are concerned by the method for locating faults of the invention.

The faults noted along a cable, whether in the ground, in buried or overhead pipes, are of various origins as indicated below
A stone, a pickaxe, a public works machine or a rodent can damage the outer polyethylene sheath of the cable. Ultimately, this lesion in the presence of water will result in more or less frank contact of the electrostatic screen with the external environment. This results in a disturbance of the network.

 A branch during a thunderstorm, or, quite simply, due to its growth, will wear out the polyethylene sheath of the overhead cable. During the next rain, an insulation fault will be detectable.

 Shotguns are also the cause of injuries to polyethylene sheaths.

 It is useful to remember that within the framework of the invention, a defect is defined as an alteration of the insulating characteristics of the outer sheath 10 of the cable 1, because statistically, it has been observed that these alterations constitute the main cause of this type fault (fault along cable 1). The outer sheath 10 therefore no longer constitutes a perfect insulator and the injured part can then be impregnated with moisture.

 A fault can therefore be modeled as a local variation in the dielectric characteristics of the outer sheath 10 of the cable 1.

 The length ld of a fault is configurable, and the permittivity and resistivity of the fault are also configurable.

 If we call ad the "resistivity of the fault", typically this resistivity is included in the range going from ad = 10 1 to ad = 10. Since, a priori, 18 alteration of the sheath 10 is accompanied by a more or less significant penetration of water, the "permittivity of the defect" is chosen equal to that of the water, ie: erd - 10. We then obtain three main models or complete standards, models accounting for the presence of a fault referenced 14. These models are illustrated, respectively, by FIGS. 5a to 5c.

 The elements common to the previous figures have the same references and will only be described again as necessary.

 FIG. 5a schematically illustrates a complete standard model Most of a buried cable 1 and having a localized defect 14, that is to say an alteration of the material of the insulating sheath 10. The conductivity of the electrostatic screen 11 is referenced ab and the permittivity of the sheath 10 is referenced srg. The other parameters have been defined previously. Finally, it is assumed, as before, that the cable 1 is buried at a distance l (for example 1 = 1 m) from the surface of the ground Ss. The length of the defect 14 is ld.

 FIG. 5b schematically illustrates a complete standard Mosc model of a cable 1, under pipe 2 and comprising a defect 14 of length ld. The various conductivity and permittivity parameters have already been defined.

 FIG. 5c schematically illustrates a complete standard Mosa model of an overhead cable 1 ′ comprising a defect 14 of length ld. The various conductivity and permittivity parameters have already been defined. The cable 1 ′ is assumed to be disposed at a height h (for example h = 3 m) relative to the surface of the ground Ss.

 These models being defined, we will simulate their behavior by studying the propagation of an electromagnetic wave along the cable 1, or more precisely along the electrostatic screen 11. This study amounts to solving the equation known as "telegraphists" , from the Maxwell equation, for sets of variable conditions and corresponding to the aforementioned models. To do this, computer tools with programmed logic can be used. Specialized software allows, in particular, to perform such calculations.

 The data collected will then make it possible to test, in real size, physical cables responding, a priori, to one or the other of the models, under pre-established conditions. The location of faults may be carried out as will be specified below.

 However, it can now be indicated that a method based on reflectometry is used, a method known per se, which consists in injecting a signal at a first end of a transmission line with distributed constants, and in analyzing and measure the characteristics of the reflected signal, by the other end of the line and / or by discontinuities encountered by the wave along the line.

 In the known art, the wave is propagated between active wires carrying the signals to be transmitted, for example the two wires of a pair, or between the core and the ground braid of a coaxial cable.

 On the contrary, according to one of the main characteristics of the invention, the signal is injected between the ambient medium (earth, air or water depending on the cable configurations) and the electrostatic screen 11.

 In reality, there are two main methods of reflectometry: ref lectometry in the frequency domain and ref lectometry in the time domain.

 Depending on whether one or the other of these techniques is used, the injected wave consists either of a sinusoidal wave variable in frequency, or of a brief electric pulse.

 In the frequency domain, we are interested in the amplitude and the phase of the wave along the cable and the disturbances brought on them by the presence of fault 14.

 However, simulations and measurements show that the interpretation of the results in this frequency domain quickly proves to be delicate and complicated.

 In the time domain, the echo reflected by the defect 14 carries clear information on the position thereof.

 Also, in a preferred embodiment of the method according to the invention, this temporal technique is retained and only this will be described below.

 The electromagnetic wave propagating on the electrostatic screen 11 "communicates" with the external medium due to dielectric and resistive losses. At defect 14, there is discontinuity in the propagation of the wave. This discontinuity results in a return of energy towards the source or echo. This echo carries information on the severity of the fault and on its position along the cable.

 Pulses of various shapes, in particular rectangular, trapezoidal, or sinusoidal, are entirely usable in the context of the invention. However, in a preferred embodiment of the invention, a pulse of Gaussian form is chosen which has a certain number of advantages, as will be shown below.

 Indeed a Gaussian-shaped pulse during its propagation along the cable 1, and more particularly of the electrostatic screen 11, and during its reflection on the fault 14, has the advantageous characteristic of retaining, at all times, its characteristics of Gaussian. This is not the case for other pulse forms, such as rectangular, trapezoidal or sinusoidal pulses. Indeed, the integration, in the cable 1 of a pulse of this type, considerably modifies the shape and the phase of this pulse and it then becomes very difficult, even impossible to keep references linked to the shape, to locate the time to go and return to fault 14, therefore to locate fault 14 precisely.

The mathematical equation of a Gaussian centered on the abscissa axis, is as follows

 By way of example, its graphic representation for a = 10 (X axis) is illustrated in Figure 6a. The vertical axis Y gives the amplitude of the normalized curve to the unit.

 In reality, the waveform starts at t = 0 and has a duration at the base of around 3 Hs, as shown in Figure 6b. To fix the ideas, we choose an impulse whose width at mid-height is typically 2.4 ds.

The horizontal axis t is the time axis (in ys) and the vertical axis A is the signal amplitude axis, normalized to unity.

 It is this signal which is injected at the entry of the cable 1, between the ambient medium and the electrostatic screen 10.

 In a faultless transmission line, the only discontinuity encountered by the injected wave is the impedance of termina return of the wave), until the attenuation accompanying these paths is sufficient for the measuring devices used can no longer detect echoes.

 This phenomenon also exists when discontinuities are encountered along the line. As indicated, the electrostatic screen 11 behaves like a line with distributed constants. If the sheath 10 is altered at any point on the cable 1, a defect 14 appears, which will have the consequence of disturbing the propagation of the wave. A premature reflection of the wave then occurs. The study and measurement of the reflected signal will therefore make it possible to locate the defect 14 (the time to go and return being shortened), and its gravity (by measuring the amplitude of the echoes).

 However, when the defect 14 of the cable 1 is slight, the echo will be slight and probably noisy. The ambient environment can also be intrinsically very noisy. Also, in a preferred embodiment of the invention, to determine with the required precision the position of the fault along the cable, the signal is subjected to a mathematical processing to extract the useful data from the noise.

 There are many methods of extracting a polluted and even drowned in noise signal, methods used for example in radar techniques, in radio astronomy or in seismology.

 One of these methods is particularly suitable in the present case: it is autocorrelation and intercorrelation.

The autocorrelation function of a function is written for an actual signal x (t) as follows

 Autocorrelation measures the "resemblance" between the injected signal x (t) and its strict copy offset after a time offset T.

As is well known, the properties of the autocorrelation function are as follows:
- it has a parity
- the autocorrelation function of a signal
centered random (signal to which we have subtracted its
mean value) tends to 0 when X increases
indefinitely: r = () = 0
- The autocorrelation function is maximum for

 We conclude that it is a practical and efficient way to calculate the phase difference between the two waves. In the context of the application targeted by the invention, the phase shift is directly linked to the position of the fault.

ou, ce qui est équivalent, que la transformation de Fourier inverse de la densité de la densité spectrale énergétique f(v) est égale à la fonction d'autocorrélation.It is also known from signal processing theory that the Fourier transform of the autocorrelation function of a signal x (t) is strictly equivalent to its energy spectral density, ie

or, which is equivalent, that the inverse Fourier transformation of the density of the energy spectral density f (v) is equal to the autocorrelation function.

We also know that the Fourier transform of a Gaussian is a Gaussian, that is

 We deduce immediately that the result of the autocorrelation is also a Gaussian function whose standard deviation is multiplied by a factor a compared to the initial function. However, the pulse selected for a preferred embodiment of the method according to the invention has a Gaussian form.

In practice, the wave reflected by the echo is not strictly identical to its generator: it is of course a
Gaussian, of smaller amplitude, but whose standard deviation a is slightly larger.

 It is therefore more accurate to speak in this case intercorrelation.

 According to the method of the invention, an intercorrelation is carried out between the injected signal x (t) and its echo e (t), measured at the input of the injured cable. In other words, we seek, using a mathematical tool, the "resemblance" between the injected signal x (t) and its measured and analyzed echo e (tJ.

 The advantage of this approach is that all uncorrelated noise present in the signals x (t) and e (t) is minimized, as will be shown below in practical examples.

Mathematically, the intercorrelation rxe (t) for the two real signals x (t) and e (t) is thus defined

This function has a maximum for a particular value of X
In the application, X represents a time difference between the two functions, this time difference being directly linked to the time taken by the wave to reflect on the fault. The value of T, expressed in seconds or Fs is very exactly the time taken by the electromagnetic wave to propagate until fault 14, to reflect on it, and to return to the entry of the cable 1. This value of T, associated with the speed of propagation of the electromagnetic wave, makes it possible to calculate directly and with precision the position of the defect 14 relative to the entry of the cable.

avec a et b des constantes de proportionnalité.We will more particularly consider the case of two Gaussian pulses, for example the two signals x and y such that

with a and b constants of proportionality.

The intercorrelation function between these two functions is equivalent, in this case, to the convolution product of these two functions x and y. We then show that the result is such

If there is no time offset between the two signals x and y, the function r (T) is maximum at
xy the abscissa 0. If y (t) and maximum in t + t0, we show, by a simple change of variable, that r (T) is maximum
xy for T = tO.

 We will show the possibility of processing a noisy signal by an autocorrelation method on a particular example.

 We consider an emitter of a signal x (t) located at a distance L from a reflecting surface, distance L that we want to measure. The signal propagates at speed v. If we call y (t) the reflected signal measured at the location of the transmitter, y (t) is the image of the signal x (t), but weakened by a coefficient c and offset with respect to x (t ) of a delay t0 = 2L / v.

 The signal y (t) is assumed to be embedded in a "white" background noise B (t).

The reflected signal therefore obeys the relation
y (t) = cx (t-to) + B (t) (7).

Intercorrelation is by definition given by the following relation

soit en changeant de variable et en posant t' = t-to :

ce qui représente exactement la fonction d'autocorrélation du signal x(t) dont il est connu qu'elle passe par un maximum pour une valeur du temps de corrélation (T-to) égale à 0.Since the intercorrelation of a signal with white noise is zero, we simply obtain

either by changing the variable and setting t '= t-to:

which exactly represents the autocorrelation function of the signal x (t) which is known to go through a maximum for a value of the correlation time (T-to) equal to 0.

 This simplification of the intercorrelation function into an autocorrelation function is only possible if the signal returns to a form identical to the transmitted signal. On the other hand, this result is valid whatever weakening, the form or the nature of the signal used sinusoid, slots, Gaussian, electromagnetic, sound waves, etc.

 In this simple case, the autocorrelation is directly interpretable, and its maximum gives access to the distance L = v / 2to sought.

de telle sorte que le maximum de la fonction r (T) est
xy caractéristique de la position L du défaut, source de l'écho mesuré et analysé, tel que L=v/2to, t0 étant la valeur pour laquelle la fonction r (T) est maximale.In the case of cable 1, the shape pulse
Gaussian injected keeps its Gaussian character after thinking about the defect. However, it is possible that its width at half-height or, what is equivalent, its standard deviation a is modified, without losing its Gaussian character. In this case, the cross-correlation is no longer simplified in autocorrelation but, as it has been shown, the cross-correlation of two Gaussian signals remains Gaussian and the cross-correlation function takes the following simple form

so that the maximum of the function r (T) is
xy characteristic of the position L of the fault, source of the echo measured and analyzed, such that L = v / 2to, t0 being the value for which the function r (T) is maximum.

 The Applicant has conducted simulations and tests on different cable configurations, different types of faults, whether in a noisy environment or not. Some examples of the results obtained will now be detailed to better highlight the important characteristics of the process according to the invention.

 We will first consider that the environment is ideal, that is to say not noisy. As an example, we will consider the case of a buried multipair metal cable.

To do this, we refer again to the model in FIG. 5a.

 The following hypotheses are carried out: the defect 14 is a wound in the outer sheath 10 of the cable, located 100 m from the entrance, over a length of 10 cm. As a result, the sheath became porous and became impregnated with humidity and od = 1 s / m is obtained.

 The wave measured and analyzed at the entry of the cable 1 in the time domain is illustrated by the curve of the time diagram in FIG. 7a. The vertical axis V represents the amplitude of the input pulse Ve in Volts and the horizontal axis is the time axis t, graduated in ps.

 In reality, the curve includes two very distinct parts: a single pulse C1 of high amplitude (positive in the example), and a series of pulses (negative in the example), globally referenced C2, of very lower amplitudes and decreasing. The curve C1 represents the injected pulse and the various extremes of the curve C2 the multiple echoes due to the reflection of the wave on the fault 14, attenuated due to the losses suffered during the propagation.

 It can be seen that the multiple echo C2 (several round trips between fault 14 and the input) is clearly visible in this ideal case.

 In practice, an oscilloscope or the screen of a microcomputer is used to display the curve. I1 is then possible to locate, "by eye", the position of the defect 14 along the cable 1 with fairly good accuracy.

Indeed, the abscissa can be graduated in meters instead of microseconds, since the speed of propagation of the signal along the cable 1 is known.

 Since the ambient environment was considered to be ideal, that is to say not noisy or noiselessly noisy, a mathematical processing of the signal is not necessary. However, the echoes are greatly attenuated. In other words, it is possible to "extract" through an autocorrelation treatment the echo "drowned" in the thickness of the abscissa line. A simple "rule of 3" then gives, in percentage, relative to the total length of the cable 1, the position of the defect 14.

 From a digital file of the wave illustrated in FIG. 7a, obtained after measurements, and that of the injected signal (known a priori), the calculation of the intercorrelation curve is carried out using a computer tool suitable, for example using specialized software such as MATHLAB (registered trademark) mathematical software.

 The cross-correlation function, as shown in the YX diagram in Figure 7b, is very pure. As before, it comprises two distinct zones: a peak C'î, centered (in the example described) on the value 2048, and a series of peaks, referenced globally C'2, substantially centered on the values 1950, 1840 and 1720. The peak C 'is shown clipped due to its very high amplitude.

The abscissa scale X is linked to the parameter X of the intercorrelation formulas, therefore directly linked to the location of the defect such that
L = V (T / 2) (12)
L being the distance of the fault with respect to the entry of cable 1 and V the speed of propagation of the pulse in this cable.

 The large peak, C'i, centered on the value 2048 of the abscissa axis, is the maximum given by the autocorrelation of the input signal on itself (FIG. 7a: C1). This peak does not contain any information about the fault.

 On the other hand, the other peaks, C'2, centered substantially on the abscissa X = 1950, 1840 and 1720, appear only because of the presence of the defect 14.

This is a clear indication that multiple echoes "resembling" the injected signal are present.

 In Figure 7b, the length of cable 1, assumed to be 1000 m, has been cut into 2048 elements. Therefore T actually varies from -2048 to +2048 or more exactly from 0 to 4096.

 In other words, a unit on the abscissa scale represents 50 cm. Typically, the location of the fault can be performed with an accuracy of 1 meter.

It can be obtained by searching for the mathematical maximum of the curves on the calculation files.

 Similar results have been obtained for other cable configurations and there is no need to detail them in full. However, the determination and localization of a defect can be further improved using a differential method.

 We will describe such a method by taking the case of a cable buried under pipe and by referring again to the model in Figure Sb. It is assumed, as before, that the ambient environment is not noisy (ideal environment). It is assumed that fault 14 is a clear fault, that is to say causing a significant disturbance.

 In order to maximize the disturbance due to fault 14, a differential method of the signature of an injured cable is applied, the signature of this same cable without injury is subtracted. Thus, it is the variations of the electromagnetic characteristics linked to fault 14, which are highlighted by intercorrelation.

In the example described with reference to FIG. 8, it is considered that fault 14 has a conductance
EC = 0.1 s / m, a length of 1 cm, because line 2 is assumed to be filled with water.

 It can be seen that, in FIG. 8, the fault appears alone, the intercorrelation of the signal injected with itself no longer appearing.

 Finally, we will detail the results obtained under very unfavorable conditions, namely a slight defect in the sheath 10 of an overhead cable 1 ′, corresponding to the model illustrated in FIG. 5c. In addition, it is assumed that the soil is dry, that the air is not saturated with humidity and that the defect, which is assumed to be slight, has a low conductance. More precisely, the hypotheses are as follows: defect from 1 cm to 100 m, with the conductivity of the defect defect = 0.0001, air oair = 0.1, and soil osol = 0.01. Despite these completely unfavorable conditions, there remains a weak but exploitable signal.

 FIG. 9 illustrates the differential intercorrelation of the signal measured at the input of the cable 1. It should be noted that the speed of propagation of the wave is greater in the case of an overhead cable than in the case of a buried cable .

 Although useful in the case of an ambient environment called "ideal" (that is to say not noisy), mathematical processing proves to be practically essential in the case of a noisy environment, in any event if it is a highly noisy environment.

 Again, these unfavorable conditions can be simulated on the models which have been previously defined in the context of the method according to the invention. It is indeed possible, using the simulation software used, for example the mathematical signal processing software "MATHLAB" mentioned above, to add noise, for example so-called "white" noise, this in greater or lesser proportion.

 Different noise levels have been simulated. The experiments carried out by the Applicant have shown that even when the echo is drowned in noise a hundred times, or even in extreme cases a thousand times greater than itself, the extraction of useful information (echoes) is still possible. By way of example, we will describe with reference to FIGS. 10a and 10b, the extreme case of a very highly noisy signal.

 It is assumed that the defect 14 is 10 cm long and is 100 meters from the entry of cable 1 and that the parameter ad is equal to unity. We take the case of a buried cable, the model of which is described in FIG. Sa. The injected signal and the echo should be theoretically similar to what is illustrated in FIG. 7a.

 Figure 10a illustrates the signal injected (Gaussian of 1 V maximum) and the corresponding echo. These signals are completely invisible, because they are completely embedded in the noise: the signal to noise ratio is around 1/1000.

Advantageously, to extract the useful signal, the processing is carried out in two stages - stage 1: Average over several noisy measurements, by
example 50 measurements - step 2: Intercorrelation on the signal thus "filtered".

 The intercorrelation operation makes it possible, despite the noise, to locate the fault 14, as shown in FIG. 10b. Parasitic "lines" appear in this figure, but there are exploitable peaks corresponding to the injected signal, C "1, and the echo signals, C" 2, useful for locating the fault 14. In addition, it is possible to improve this result by increasing the number of successive measurements on which an average is carried out.

 If we move away from this extreme case, for average noise levels (for example for a signal / noise ratio of the order of 1/100) or low, the parasitic lines disappear, leaving only the corresponding peaks echoes and the signal injected (or only echoes, if a differential method is used).

 We will now describe, with reference to FIG. 11, an example of a fault location device 14, on a multi-pair metal cable 1, for implementing the method of the invention.

 In the example described, the architecture of the fault location device 14 is built around a microcomputer 40. The microcomputer 40 controls a pulse generator 41, for example for the generation of a pulse of Gaussian shape , in a preferred embodiment of the invention. The pulse generated is applied between the electrostatic screen 11 and the ambient medium symbolized by the earth.

 In a practical embodiment, the generator 41 may consist of a specialized card inserted in the microcomputer 40, such as the "Compugen 1100" card, sold by the company "Le Croy".

It therefore receives instructions developed by the central unit of the microcomputer 40 under the guidance of a recorded program. This card 41 generates any waveform from the content of a memory (for example of capacity 16 * 10 6 words of 12 bits) connected to the input of a digital-analog converter (not shown).

 The temporal resolution is 12.5 ns (80 MHz) and the amplitude resolution is 4096 points (12 bits).

 The shape, duration and amplitude of the pulse generated are a function, in particular, of the known characteristics of the cable 1 and of its configuration (buried, buried in pipeline or overhead).

 In practice, this involves generating an electrical pulse starting at 0 Volt for t = 0, the characteristics of which have been described with reference to FIG. 6.

 The electrical pulses corresponding, on the one hand, to the pulse injected at the input of cable 1 and, on the other hand, to the echoes produced by fault (s) along cable 1 are digitized by two additional specialized cards, under the unique reference 42, for example of the "Compuscope 8012 / PCI" type, also sold by the company "Le Croy". These cards 42 carry out the acquisition in real time of the signals present at the cable input 1, by sampling, so as to carry out the reverse analog-to-digital conversion. The sampling period is 12.5 ns. The cards 42 are also inserted in the microcomputer 40.

 The microcomputer 40 moreover performs, under the guidance of recorded programs, all the necessary mathematical processing, in particular the autocorrelation or intercorrelation operations (for example under the guidance of the abovementioned "MATELAS" software or similar software) and develops the instructions necessary for the proper functioning of the pulse generation card 41 and of the signal acquisition cards 42.

 The signals acquired and analyzed, after possible mathematical processing, are stored in conventional memory organs associated with the central unit of the microcomputer 40 (random access memory and / or mass memory) and are displayed on the screen 400 of the microphone. -computer 40, in real time or delayed.

 The method according to the invention which has just been described in its different variants, and in particular those using mathematical post-processing, makes it possible to locate a fault with great precision, even if the fault is slight and if the cable is immersed in a noisy environment.

However, in certain configurations, there remains an uncertainty which can be considered too great.

In particular, for buried cable configurations, such as the configuration described with reference to FIG. 5a (which is modeled by FIG. 5b), it is extremely important that the location of the fault along the cable is as precise as possible. In this case, it is indeed necessary to dig up the cable to make repairs. It is easy to see that this operation requires significant resources and proves to be expensive.

 Also, according to an additional characteristic of the invention, once the location has been carried out according to the main steps of the method (measurement of the reflection of the injected signal and possible mathematical processing of the measured signal), the fault can be located more precisely by measuring the electromagnetic radiation of the cable when a periodic signal or a series of pulses which are propagated along the cable is injected.

 Figures 12a and 12b schematically illustrate this complementary variant of the method according to the invention.

 It has been assumed, as shown more particularly in FIG. 12a, that the cable 1 was of the buried type (see FIGS. 5a and 5b) and that it had a defect 14 of determined length ld. In practice, the position of the defect 14 is determined with finite precision. If L is the exact position of the center of the defect 14 with respect to the input (arbitrarily X = 0 on the abscissa axis shown in Figure 12b), the measured distance is equal to L + [i'd / 2] . The value of ld is generally low (for example of the order of 10 cm), i'd therefore represents the area of uncertainty on the exact location of the fault 14, in a direction parallel to the cable 1.

 To reduce the length d of the uncertainty zone, electrical signals, for example pulse signals, are injected at the input of the cable 1, and the electromagnetic radiation H emitted by the cable 1 is measured.

 More precisely, once the first steps of the process have been completed and the location of fault 14 acquired with the precision recalled above (L + [1 'd / 2]), a symbolized intervention team is sent to the site, in FIG. 12a, by a single operator Op. This is provided with a mobile device 5 for measuring the electromagnetic radiation H and moves in the area where the cable 1 has a fault 14.

 The measuring device 5 essentially comprises an antenna 50 located in the lower part thereof and which can therefore move parallel to the surface of the ground Ss.

The apparatus comprises an electronic unit 51 provided with conventional electronic circuits receiving the signals coming from the electromagnetic radiation H received by the antenna 50.

The housing 51 includes various other members and more particularly a display screen 54. A tubular mechanical structure 52 subjects the antenna 50 to the electronic housing 51. The antenna 50 can be fixed (advantageously inclined relative to the tubular structure 52, to ergonomic reasons) or have an adjustable tilt (using a ball joint or similar). The tubular structure 52 is advantageously provided with handles 53, for example of the "handlebar" type, or any other gripping member, which allows the operator to firmly hold the device 5 by hand, to move it, and to maintain the antenna 50 parallel to the surface of the ground Ss.

 When the operator Op is near cable 1, just vertically of it, the signal received is maximum. When it moves away from it, the signal received decreases. The operator Op can therefore locate the passage of cable 1 and then follow a path parallel to it.

 The diagram on the lower part of FIG. 12b illustrates the variation of the amplitude of the signal received as a function of the distance from the cable entry 1 curve C3 representing the amplitude of the electromagnetic radiation H as a function of the abscissa X The regular decrease in the amplitude of H is due to the natural weakening of the electrical signals injected at the input during their propagation along the cable 1.

 On the other hand, in the immediate vicinity of fault 14, the signal received, as illustrated on curve C3, there is a significant discontinuity which results in a sudden increase in the amplitude of the signal received, followed by an equally rapid decrease. This Hp peak, centered on defect 14, is very localized, which allows great selectivity. It is this discontinuity which is detected during the movement of the device 5. The operator Op can thus very precisely locate the location of the fault. Then just dig the ground there, and only there, to dig up the cable.

In summary, the more precise localization phase of fault 14 comprises the following steps
a / injection of an electrical signal at cable entry 1
b / location of the cable passage by "hanging" on the electromagnetic radiation emitted by it
c / precise location of the position of fault 14 (at ground level Ss) by detecting the aforementioned discontinuity.

 The signals received are processed by electronic reception circuits (adaptation, amplification and possibly digitization circuits) coupled to the antenna 50 and included in the housing 51. Once processed, the signals are transmitted to the member. display 54, to be displayed in any suitable form: curve, digital display, etc. The display member 54 can optionally be coupled to a sound device, the intensity of the sound emitted is, for example, proportional to the amplitude of the electromagnetic field measured.

 Although the signal input device at the cable entry has been referenced differently (reference 4 ′), it is possible, a priori, to use a device very similar if not identical to device 4 of FIG. 11. it can however be necessary to provide additional electronic circuits to accommodate the specifics of the variant of the method according to the invention which has just been described. The choice of such circuits is within the reach of the skilled person.

 Likewise, it is generally necessary to coordinate the operations carried out at the entry of cable 1 and those carried out in the field by the operator Op. This can be carried out manually or automatically. In both cases, it is necessary to provide data transmission means (not shown), which may or may not be integrated in the electronic box of the device 5. The latter is naturally provided with an autonomous power supply device (batteries, batteries, etc.). Finally, it can include, in addition to the aforementioned electronic reception circuits, an auton digital data processing unit.

 On reading the above, it is easy to see that the invention achieves the goals it has set for itself.

 it should be clear, however, that the invention is not limited only to the examples of embodiments explicitly described, in particular in relation to FIGS. 3 to 12b.

In particular, the numerical examples have only been given to better specify the main characteristics of the method according to the invention. Although particularly suitable, by these specific properties, the Gaussian type pulses are not the only ones usable in the context of the invention, as has been indicated. In addition, the duration and the amplitude of the pulse injected on the electrostatic screen are a function of the characteristics of the multi-pair wire rope, of its configuration and, possibly, of the characteristics of the ambient environment (humidity, etc.).
The test device enabling the location of the fault along the cable is also not limited to the only device illustrated with reference to FIG. 11, the architecture of which is based on a universal microcomputer, supplemented by cards of specialized electronic circuits. One can in particular make a specific measuring device, advantageously portable. Similarly, the signal processing operations, in particular the autocorrelations or intercorrelations can be carried out by specialized electronic circuits, and not by software as it has been described, without departing from the scope of the invention.

 It is also possible to add to the microcomputer 40, in a conventional manner, a printer or a plotter for printing the graphics obtained, for example on graph paper.

 Finally, the cable can present, not just one defect, but several. We will then obtain several series of echoes staged in time, which will make it possible to discriminate the different fault positions.

Claims (14)

 1. Method for locating at least one fault (14) along a multi-pair metallic cable (1), said cable comprising a plurality of pairs of conductive wires (12) intended to convey information signals, a screen electrostatic (11) conductor of electricity surrounding these pairs (12) and an external sheath (10) of electrically insulating material, said cable (1) being immersed in a determined medium (Mat), characterized in that it comprises at minus the following steps  - generation (41) of an electrical pulse, of given shape, amplitude and duration  injection of this pulse, at a first end of said cable (1), between said electrostatic screen (11) and the material constituting said determined medium (Mat), so as to obtain the propagation of an electromagnetic wave along said electrostatic screen (11)  - temporal analysis of the electrical signals appearing at said first end of the cable (1), over a determined period, to detect the presence of echoes corresponding to reflections of said electromagnetic wave on said fault (14), this fault being reflected by interactions between said electrostatic screen (11) and said determined medium (Mat)  - And determination of the distance separating said fault (14) from said first end of cable (1) from the time necessary for said electromagnetic wave to propagate to said fault and return to said first end and the speed of propagation electromagnetic wave in the cable (1).  2. Method according to claim 1, characterized in that it comprises a preliminary phase consisting in creating models (Most, Mosc, Mosa) describing the electromagnetic characteristics of said multi-pair metallic cable (1) and said defect (14), and to acquire data describing its behavior during the injection of said electrical pulse in the presence of simulated faults, of variable characteristics, and in that each of said models (Most, Mosc, Mosa) is attached to a so-called standard configuration of the cable (1 ) and under conditions characteristic of said given medium (Mat, Mae, Maar) -  3. Method according to claim 2, characterized in that said cable (1) described by said models comprises said electrostatic screen (11), the latter surrounding a dielectric material (13), said external sheath (10) and said determined medium (Mat, Mae, Maar), and in that said defect (14) is modeled by modifications of the conductivity and permittivity characteristics of the insulating material of the outer sheath (10) resulting in an interaction between said electromagnetic wave propagating along of said electrostatic screen (11) and said determined medium (Mat, Mae, Maar) in which the cable (1) is immersed.  4. Method according to claim 3, characterized in that a first of said standard models (Most) is associated with a configuration of buried multipair metallic cable (1), at a determined depth (l), relative to the surface of the soil (Ss), and in that said determined medium (Mat) is earth.  5. Method according to claim 3, characterized in that a first of said standard models (Mosc) is associated with a configuration of multipair metallic cable (1) under pipe (2) made of plastic material, and in that said determined medium ( Mae) is water filling the pipe (2).  6. Method according to claim 3, characterized in that a first of said standard models (Mosa) is associated with an aerial multipair metallic cable configuration (1 ′), said cable being retained, by a carrying rope (3), a determined height (h), relative to the surface of the ground), and in that said determined medium (Maar) is air (Maar), at a determined humidity rate.  7. Method according to any one of claims 2 to 6, characterized in that said electrical pulse is a Gaussian-shaped pulse, and in that the electrical characteristics of this pulse and said propagation speed are determined from the one of said standard models (Mat, Mae, Maar) and said acquired data.  8. Method according to claim 7, characterized in that said electrical signals appearing at said first end of the cable (1) being tainted with noise, it comprises an additional step consisting in subjecting these signals to a mathematical processing and in that this processing mathematical includes an intercorrelation operation, so as to extract them from said noise.  9. Method according to any one of the preceding claims, characterized in that it comprises an additional phase making it possible to make said fault location (14) more precise and comprising at least the following steps  a / injection of an electrical signal at a first end (4 ′) of said cable (1), this signal causing electromagnetic radiation (H) throughout the cable (1), said radiation (H) having a local discontinuity ( Hp) of its amplitude in a zone centered on the defect (14)  b / and more precise localization of said fault (14) by detection of the position, along the cable (1), of said discontinuity (Hp) of electromagnetic radiation (H).  10. Method according to claim 9, characterized in that, said cable (1) being of the buried type, said additional phase comprises an intermediate step consisting in locating the passage of the cable (1) by the detection of said electromagnetic radiation (H) .  11. Device for implementing the fault location method according to any one of the preceding claims, characterized in that it comprises means (41) for generating a pulse of shape, amplitude and duration determined, and injecting this pulse, at a first end of said cable (1), between said electrostatic screen (11) and the material constituting said determined medium (Mat, Mae, Maar), acquisition means (42) in real time electrical signals appearing at said first end of the cable (1), over a determined period, and means (40) of control and calculation, with programmed logic, for carrying out a temporal analysis of these signals, so detecting the presence of echoes corresponding to reflections of said electromagnetic wave due to at least one fault (14) and locating this fault (14) along said cable (1), and in that said means (40) of order e t of programmed logic calculation include a display screen (400) for displaying said echoes.  12. Device according to claim 10, characterized in that said control and calculation means with programmed logic consist of a microcomputer (40) and said means of pulse generation (41) and acquisition (42) by specialized electronic circuit boards included in said microcomputer (40).  13. Device according to claims 11 or 12, characterized in that it is also associated with a mobile measuring device (5) of electromagnetic field (H) radiated by said cable (1) and caused by the injection of an electrical signal at a first end of the cable (1), said measuring device (5) being intended to be moved along the cable (1) and to detect discontinuities (Hp) in the amplitude of said electromagnetic field (H) due to said faults (14), so as to locate these faults (1) with improved precision.  14. Device according to claim 13, characterized in that said measuring device (5) comprises at least one antenna (50) sensitive to said electromagnetic radiation (H), a display member (54), electronic circuits, included in a housing (51), transforming the electromagnetic radiation received (H) into electrical signals displayable on said display member (54) and a mechanical structure (52) supporting said housing (51), the display member (54) and said antenna , said mechanical structure being provided with gripping members (53) allowing an operator (Op) to move said measuring device (5) by hand.