Method and system for the structural diagnosis of concrete railway sleepers

Abstract

TITLE: Method and System for Structural Diagnosis of Concrete Railway Sleepers This structural diagnostic method comprises four steps: an application step, in which a mechanical stress is applied by impact to a concrete railway sleeper (2, 3) to be diagnosed; a detection step, in which a vibration generated by the mechanical stress and propagating through the sleeper being diagnosed is detected on the surface of the sleeper and transformed into a vibrational time-domain signal; a processing step, in which the vibrational time-domain signal is electronically processed to obtain a frequency spectrum; and an analysis step, in which the frequency spectrum is compared to a reference to deduce a structural damage characterization of the sleeper being diagnosed. Figure for the abstract: Figure 1

Description

Title of the invention: Method and system for the structural diagnosis of concrete railway sleepers

[0001] The present invention relates to a method and system for the structural diagnosis of concrete railway sleepers.

[0002] Railway sleepers are pieces laid across a railway track, to which the rails are fixed in such a way as to maintain the gauge and inclination of the latter, and which transmit the loads moving on the rails to a ballast on which the railway sleepers rest. We are interested here more specifically in so-called "concrete" sleepers, these sleepers being either monoblock, in which case being made of a single piece of concrete extending over the entire longitudinal dimension of the sleeper, or biblock, in which case being made of two end blocks of concrete, connected to each other by a central metal crossbar.

[0003] Concrete railway sleepers have a long service life, potentially reaching several decades. However, under certain circumstances, these railway sleepers exhibit structural damage that is more pronounced than their expected normal wear. The causes or reasons for this situation are varied and lead to equally varied structural consequences: without being exhaustive, examples include contact erosion between the underside of the sleeper and the ballast, spalling and similar damage related to transport and / or maintenance operations, overstress cracking in service, chemical cracking resulting from alkali-aggregate reactions, thermal shrinkage cracking during sleeper manufacturing, etc.In all cases, it is difficult to assess structural damage to sleepers non-destructively, typically through simple visual observation in situ, i.e., directly on the railway track. Indeed, observation conditions are often poor, and a large portion of the sleepers is obscured by ballast or under rails, unless the sleeper is removed from the track and, if necessary, examined later in a testing laboratory, which is particularly tedious. Furthermore, even when cracks are visible on the surface of the sleepers, it is virtually impossible to reliably determine whether the corresponding structural damage is critical enough to require urgent replacement. More generally, there is a real need to be able to structurally diagnose concrete railway sleepers in situ in order to effectively manage their maintenance in a broad sense.

[0004] The object of the present invention is to propose a method and a system which allow for the structural diagnosis of concrete railway sleepers in a reliable, fast and simple manner, where appropriate directly on the railway track.

[0005] To this end, the invention relates to a method for the structural diagnosis of concrete railway sleepers, comprising:

[0006] - an application step, in which a mechanical stress is applied by impact on a concrete railway sleeper to be diagnosed.

[0007] - a detection step, in which a vibration generated by the stimulus mechanical and propagating within the diagnosed crossbeam is detected on the surface of the diagnosed crossbeam and transformed into a vibratory time-domain signal.

[0008] - a processing step, in which the vibratory time signal is processed electronically in order to obtain a frequency spectrum, and

[0009] - an analysis step, in which the frequency spectrum is compared to a reference to deduce a characterization of structural damage to the diagnosed crossmember.

[0010] The invention also relates to a structural diagnostic system for concrete railway sleepers, comprising:

[0011] - at least one impactor suitable for applying a stress per impact mechanical work on a concrete railway sleeper to be diagnosed,

[0012] - at least one sensor adapted to detect, on the surface of the cross member once diagnosed, a vibration generated by mechanical stress and propagating through the diagnosed cross member, and to transform the vibration into a vibratory time signal,

[0013] - an electronic processing unit, adapted to process the time signal vibrational in order to obtain a frequency spectrum, and

[0014] - analytical means adapted to compare the frequency spectrum to a reference to deduce a characterization of structural damage to the diagnosed crossmember.

[0015] One of the ideas underlying the invention is to seek to characterize the structural damage of a concrete railway sleeper based on a simple, non-destructive impact applied to the sleeper to be diagnosed, particularly in situ, i.e., directly on the railway track where the sleeper is in use, which is particularly simple and practical to implement. The structural damage characterization is dimensionless, in the sense that it does not quantify a geometric dimension or a dimensional parameter of the sleeper being diagnosed, but rather this structural damage characterization associates the diagnosed sleeper with one structural state among several states. Thus, the structural damage characterization advantageously provides information on, with a margin of controlled error, (i) the presence of at least one structural defect belonging to one or more given defect types and / or (ii) structural wear that differs from, and is therefore potentially greater than, nominal structural wear. To achieve this, the invention provides that the vibration generated in the sleeper by the aforementioned impact is detected on the surface of the sleeper and transformed into a time-domain vibration signal, in practice by means of one or more ad hoc sensors. This signal is then processed, in particular by an ad hoc electronic processing unit, to convert from the time domain to a frequency domain, in the form of a frequency spectrum, advantageously provided in situ.Frequency spectrum analysis then allows for the determination of the structural damage characterization of the diagnosed sleeper, by comparison to a predetermined reference. This reference is obtained, for example, from a reference sleeper, typically new or considered sound, or from the prior statistical processing of a homogeneous population of sleepers. This frequency spectrum analysis can be performed either partially by a human operator, who is advantageously shown the frequency spectrum and the aforementioned reference on a screen in situ, or entirely by ad hoc computer means, possibly partially remote, as explained in detail later. In all cases, the corresponding analysis tools are particularly practical and allow for control over the reliability of the diagnosis provided, that is, the structural damage characterization that is deduced by or thanks to these analysis tools.

[0016] According to advantageous optional features of the structural diagnostic method and / or system according to the invention, taken individually or in all technically possible combinations:

[0017] - the application step, the detection step and the processing step, as well as at least Part of the analysis stage is implemented on railway tracks.

[0018] - in the analysis step, the frequency spectrum and a reference spectrum constituting said reference is displayed to a human operator who deduces, by visual observation, the characterization of structural damage of the diagnosed crossbeam.

[0019] - in the analysis step, (i) the frequency spectrum is compared computationally to a reference spectrum constituting said reference, by calculating a distance score that is representative of a mathematical distance between the frequency spectrum and the reference spectrum, and (ii) the structural damage characterization is determined from the calculated distance score.

[0020] - in the analysis step, (i) the frequency spectrum is compared computationally to several reference spectra constituting said reference, by calculating distance scores that are respectively representative of mathematical distances between the frequency spectrum and the reference spectra, and (ii) the structural damage characterization is determined from the calculated distance scores.

[0021] - in the application step, mechanical stress is applied to the cross member to be diagnosed by an impact having an energy of less than 3 joules, preferably less than 1 joule.

[0022] - in the processing step, the vibratory time signal is processed so as to extract a significant part, before applying to this significant part a bandpass filter and then a mathematical transformation on a frequency domain to obtain the frequency spectrum.

[0023] - the bandpass filter has a low cutoff frequency substantially equal at 3,000 Hz, preferably 3,500 Hz, and a high cutoff frequency substantially equal to 20,000 Hz, preferably 15,500 Hz.

[0024] - the impactor(s), the sensor(s) and the electronic processing unit, as well that at least part of the analytical equipment forms a portable hardware set, usable by a human operator on a railway track.

[0025] The invention will be better understood upon reading the following description, given solely by way of example and made with reference to the drawings in which: - [Fig.1] [Fig.1] is a schematic perspective view of two concrete railway sleepers and a structural diagnostic system for such sleepers; - [Fig.2] [Fig.2] is a flowchart of a structural diagnostic process, implemented by the structural diagnostic system of [Fig.1]; - [Fig.3] [Fig.3] is a schematic cross-section of part of the structural diagnostic system of [Fig.1], used on a cross-section, this cross-section being schematically associated with signals produced by the structural diagnostic system; - [Fig. 4] [Fig. 4] includes inserts a) and b) corresponding respectively to frequency spectrum patterns implemented in the structural diagnostic process, as well as an insert c) corresponding to a distance score pattern, also implemented in the structural diagnostic process; and - [Fig.5] [Fig.5] includes two inserts a) and b) which correspond respectively to diagrams including a representation space, implemented in the structural diagnostic process.

[0026] Figure [1] shows a structural diagnostic system 1 allowing non-destructive control of the concrete structure of concrete railway sleepers, such as sleepers 2 and 3 shown in Figure [1] as examples.

[0027] The sleeper 2 is a so-called monoblock sleeper and comprises a concrete body 2.1 extending continuously from one end of the sleeper 2 to the other, opposite each other in a longitudinal direction of the sleeper 2. The concrete body 2.1 includes two end portions 2.2 and 2.3, which are opposite each other in the longitudinal direction of the sleeper 2, as well as an intermediate portion 2.4, which connects the end portions 2.2 and 2.3 and which is possibly thinner than the end portions 2.2 and 2.3. The sleeper 2 is provided, on each of its end portions 2.2 and 2.3, with features 2.5 for the attachment of a railway rail, not shown in the figures. Crossbeam 2 optionally includes a metal wire reinforcement, embedded in the concrete body 2.1.

[0028] The sleeper 3 is a so-called bi-block sleeper and comprises two concrete blocks 3.1 and 3.2, which form two end sections of the sleeper 3, opposite each other along a longitudinal direction of the sleeper 3, and which are connected to each other by a metal spacer 3.3. Similar to the sleeper 2, the sleeper 3 is provided, on each of its concrete blocks 3.1 and 3.2, with features 3.4 for securing a railway rail.

[0029] The specific features of the crossbeam 2 and the crossbeam 3 are not limiting as long as their structure is mainly, but not necessarily exclusively, made of concrete, like the concrete body 2.1 for the crossbeam 2 and the concrete blocks 3.1 and 3.2 for the crossbeam 3. Thus, the concrete of these crossbeams 2 and 3 is for example reinforced, prestressed or post-stressed.

[0030] In all cases, as is customary in the field and without this being shown in [Fig.1], sleeper 2 and sleeper 3 are used interchangeably in a railway which comprises both a support, typically a ballast, on which rests a group of sleepers, such as sleepers 2 and 3, distributed regularly along a longitudinal direction of the railway, and rails, which extend lengthwise along the longitudinal direction of the railway and which are fixed to the sleepers, at dedicated fittings of the latter, such as fittings 2.5 for sleeper 2 and fittings 3.4 for sleeper 3.

[0031] Regardless of the specific characteristics of sleepers 2 and 3, their concrete structure necessarily evolves over time, undergoing structural damage. This structural damage results from normal wear of sleepers 2 and 3, linked in particular to the stresses that the sleepers transmit between the ballast and railway equipment running on the track. As mentioned in the introductory section of this document, under certain conditions of use and / or due to certain manufacturing conditions of sleepers 2 and 3, the structural damage to these sleepers is likely to be greater than normal wear. In particular, cracks or other structural defects are likely to appear and substantially and rapidly worsen the structural damage. As soon as such structural defects appear in the concrete structure of sleepers 2 and 3, the service life of these sleepers is likely to be shorter than that of the same sleepers free from such structural defects.

[0032] In order to characterize the actual structural damage to concrete railway sleepers, such as sleepers 2 and 3, while these sleepers are in use in a railway track, it is proposed to implement a structural diagnostic procedure, which will be described in detail below. In what follows, this structural diagnostic procedure is applied, for example, to sleeper 2, but it is understood that this structural diagnostic procedure is equally applicable to sleeper 3. An embodiment of this structural diagnostic procedure is schematically illustrated in [Fig. 2] and is advantageously implemented by the structural diagnostic system 1, appropriately controlled by a human operator.

[0033] As illustrated in [Fig.2], the structural diagnostic process comprises four main successive steps, namely an application step 101, a detection step 102, a treatment step 103 and an analysis step 104. Before describing in detail each of these four steps successively, it should be noted that the application step 101, the detection step 102 and the treatment step 103, as well as at least part of the analysis step 104, are advantageously implemented in situ, i.e. directly on the railway track, in particular while the concrete sleepers of the latter are left as is, in other words maintained in an operational situation within the railway track.

[0034] In application step 101, a mechanical stress is applied by impact to a concrete railway sleeper to be diagnosed, here sleeper 2, so as to, as schematically illustrated in [Fig.3], generate a vibration V in the sleeper 2, more precisely in the concrete structure of the latter, here the concrete body 2.1. Once generated by the mechanical impact stress, the vibration V propagates in the sleeper 2, in particular through the concrete structure of the latter, here the concrete body 2.1, while being reflected by the external surfaces of the sleeper 2, as well as by internal defects or, more generally, internal structural heterogeneities, such as cracks, such as the one illustrated in [Fig.3] under reference F, but also such as voids, segregation interfaces, etc.

[0035] In practice, the impact that mechanically stresses the sleeper 2 to generate the vibration V during application step 101 is of sufficiently low intensity so as not to damage the sleeper 2, or even displace it relative to the support on which the sleeper rests, typically the ballast when application step 101 is implemented on a railway track. Furthermore, the intensity of this impact is sized so that the resulting vibration V in the concrete structure of the Traverse 2 exhibits a wavelength suitable for the structural damage that the structural diagnostic process aims to characterize. For this purpose, this impact advantageously has an energy less than 3 joules, preferably less than 1 joule.

[0036] For the purposes of implementing application step 101, the structural diagnostic system 1 includes an impactor 10, which is schematically shown in Figures 1 and 3. The impactor 10 is adapted to apply the aforementioned mechanical stress by impact to a cross member to be diagnosed, here cross member 2. By way of non-limiting example, the impactor 10 includes an electromagnet that drives the displacement of an impact weight. Of course, other embodiments are conceivable for the impactor 10, provided that it allows the cross member 2 to be mechanically stressed by impact in order to generate vibration V in the latter.

[0037] In the detection step 102, the vibration V is detected on the surface of the diagnosed crossbeam, here crossbeam 2, and is transformed into a vibrational time signal, as schematically illustrated in [Fig. 3], on which this vibrational time signal is referenced as SV. The vibrational time signal SV is thus representative of the evolution, over time, of surface movements of the concrete body 2.1, induced by the vibration V.

[0038] For the purposes of implementing the detection step 102, the structural diagnostic system 1 includes a sensor 20 shown schematically in Figures 1 and 3. The sensor 20 is adapted to detect the vibration V and to transform it into the vibration time-domain signal SV. By way of non-limiting example, the sensor 20 includes a contact microphone with a piezoelectric transducer. Of course, other embodiments are conceivable for the sensor 20 as long as it emits a vibration time-domain signal, such as the signal SV, obtained by detecting and transforming the vibration V by this sensor.

[0039] In the embodiment shown in the figures, the impactor 10 and the sensor 20 are advantageously integrated into a handheld tool 30 of the structural diagnostic system 1. This handheld tool 30 can be operated manually by the operator using the structural diagnostic system 1, particularly in situ. This handheld tool 30 comprises, for example, a housing 31, in which the impactor 10 and the sensor 20 are arranged, and which can be applied against the crossbeam 2 so as to bring the impactor 10 and the sensor 20 into contact with the crossbeam, as schematically illustrated in [Fig. 3]. The handheld tool 30 also includes a handle 32, which extends from the housing 31 and allows the operator to conveniently control and manipulate the handheld tool 30. In particular, the handle 32 is equipped with a manual trigger 33, the actuation of which by the operator controls the impactor 10.In practice, the shape and dimensions of handle 32 are not limiting.

[0040] In processing step 103, the vibratory time-domain signal SV is electronically processed to obtain a frequency spectrum, schematically illustrated in [Fig. 3] under the reference SF. The frequency spectrum SF thus represents a spectral amplitude as a function of frequency, typically quantized in hertz, with the symbol Hz. According to an efficient and effective embodiment, in processing step 103, the vibratory time-domain signal SV is, at least in part, mathematically transformed over a frequency domain to obtain the frequency spectrum SF: the corresponding mathematical transformation is, for example, a Fourier transform, in particular a fast Fourier transform.Furthermore, what is subjected to the aforementioned mathematical transformation is advantageously only a portion of the vibrational time-domain signal SV, namely a filtered portion, which is obtained by applying a bandpass filter to the vibrational time-domain signal, in particular to only a significant part of this signal, extracted from the vibrational time-domain signal using, for example, a threshold value applied to the amplitude of this signal. Following a preferred design that results in remarkable performance for the structural diagnostic process, the aforementioned bandpass filter has a lower cutoff frequency of 3,000 Hz, preferably 3,500 Hz, and an upper cutoff frequency of 20,000 Hz, preferably 15,500 Hz.

[0041] In practice, it should be noted that the absolute values ​​of the amplitude of the SF frequency spectrum, obtained at the end of processing step 103, are not relevant to the structural diagnostic procedure, in the sense that these absolute values ​​are not intended to be used in this procedure. On the other hand, the variation of this amplitude of the SF frequency spectrum as a function of frequency is of interest, particularly for identifying one or more peaks, as discussed later.

[0042] For the purposes of implementing the processing step 103, the structural diagnostic system 1 includes an electronic processing unit 40, which is shown schematically in Figures 1 and 3. The electronic processing unit 40 is adapted to process the vibrational time-domain signal SV in order to obtain its frequency spectrum SF. The electronic processing unit 40 is advantageously adapted to successively extract a significant portion of the vibrational time-domain signal SV and apply the aforementioned bandpass filter, followed by the aforementioned mathematical transformation. In all cases, the embodiment of the electronic processing unit 40 is not limiting.In practice, the electronic processing unit 40 can either include an assembly of analog electronic components, dedicated to the processing operations implemented by this electronic processing unit 40, or include a set of digital electronics, typically including . a microprocessor or similar, programmed to carry out the aforementioned processing operations.

[0043] In the embodiment shown in the figures, the electronic processing unit 40 is advantageously integrated into a portable computing device 50, such as a tablet, a laptop, or a specific terminal. The advantages of this portable computing device 50 will become apparent later.

[0044] In analysis step 104, the frequency spectrum SF is compared to a reference to deduce a structural damage characterization of the diagnosed crossbeam, here crossbeam 2. Various embodiments are conceivable for this analysis step 104, depending in particular on the nature of the aforementioned reference, as well as on how the frequency spectrum SF is compared to this reference. Before explaining this aspect of the structural diagnostic process in more detail, it should be noted that, for the purposes of implementing analysis step 104, the structural diagnostic system 1 includes analysis means 60, schematically illustrated in [Fig. 1]. These analysis means 60 are adapted to determine the structural damage characterization of the diagnosed crossbeam by comparing the frequency spectrum SF provided to the analysis means 60 by the processing unit 40 to the aforementioned reference.Various embodiments of the analysis means 60 will be described in more detail later, in association with the different embodiments of the analysis step 104, explained further on. That being said, in all cases, at least a part of the analysis means 60 is advantageously computer-based, being able to process the frequency spectrum SF provided by the electronic processing unit 40. In particular, this computer-based part of the analysis means 60 includes: .

[0045] - a computer processing unit 61, which is integrated into the equipment portable computer 50 and which, where applicable, shares all or part of the digital electronics of the electronic processing unit 40, and / or

[0046] - a remote computing device 62, such as a computer or a server, which is capable of reading a data carrier from the portable computing equipment 50 or which is in wireless communication with this portable computing equipment 50.

[0047] In the embodiment envisaged in [Fig.1], the impactor 10, the sensor 20 and the electronic processing unit 40, together with the computer processing unit 61 of the analysis means 60, form a portable hardware assembly, which is usable by an operator on a railway and which, here, combines the portable tool 30 and the portable computer equipment 50.

[0048] A first embodiment of analysis step 104 is illustrated by each of the insets a) and b) of [Fig. 4]. In this first embodiment, the reference used in analysis step 104 consists of a predetermined reference spectrum SR, which is represented in dotted lines on inserts a) and b) of [Fig.4] and which is for example stored in an ad hoc memory of the analysis means 60. In practice, this reference spectrum SR is for example acquired either by prior implementation of the application steps 101, detection 102 and processing 103 on a reference sleeper, typically a sleeper whose structural damage is known to correspond to normal wear, or by prior statistical processing of results obtained by implementation of the application steps 101, detection 102 and processing 103 on a population considered homogeneous of sleepers.

[0049] Furthermore, in the first embodiment of the analysis step 104, the reference spectrum SR and the frequency spectrum SF, as derived from the application 101, detection 102, and processing 103 steps applied to a sleeper to be diagnosed, such as sleeper 2, are displayed, typically on a screen of the analysis means 60, here a screen 51 integrated into the portable computer equipment 50, so that, by direct visual observation, the operator can compare the reference spectrum SR and the frequency spectrum SF and deduce the structural damage characteristic for the sleeper being diagnosed. This deduction is thus performed by the operator in situ, i.e., directly on the railway track.In practice, the respective displays of the reference spectrum SR and the frequency spectrum SF on screen 51 are advantageously simultaneous and superimposed: as corresponding schematic illustrations, the frequency spectrum associated with a first crossbeam to be diagnosed is represented in solid lines on inset a) of [Fig.4], labeled El, and the frequency spectrum associated with a second crossbeam to be diagnosed is represented in solid lines on inset b) of [Fig.4], labeled E2, while on these two insets, the reference spectrum SR is represented in dashed lines.

[0050] In practice, structural damage characterization involves, for example, the operator determining whether the structural damage of the sleeper being assessed is either normal, i.e., similar to that of the reference sleeper, as in the case illustrated in inset a) of [Fig. 4], or abnormal, i.e., substantially different from that of the reference sleeper, as in the case illustrated in inset b) of [Fig. 4]. On a railway, the operator can thus quickly determine, among all the sleepers on a railway track, those with abnormal structural damage.

[0051] A second embodiment of analysis step 104 is illustrated by insets a), b) and c) of [Fig.4].

[0052] Here again, the reference used in analysis step 104 consists of a reference spectrum, here the SR reference spectrum described above for the first embodiment.

[0053] Furthermore, in this second embodiment, the frequency spectrum SF, as obtained from the application steps 101, detection 102 and processing 103, is compared computationally to the reference spectrum SR, by calculating a distance score that represents a mathematical distance between the reference spectrum SR and the frequency spectrum SF. The specifics of the calculation of this distance score are not limiting, it being noted that various calculation formulas are known in the technique, such as the so-called DTW formulas, which is the acronym for the English expression "Dynamic Time Warping", the so-called DTW formulas being based on the difference between respective peaks of the reference spectrum and the analyzed frequency spectrum, as schematically illustrated by the double-headed arrows, schematically drawn on inset b) of [Fig.4].

[0054] In all cases, for each diagnosed sleeper, an associated distance score is calculated, and then the structural damage characteristic is determined from this calculated distance score. As a schematic illustration, distance scores calculated respectively for the frequency spectrum labeled El on inset a) of [Fig. 4] and for the frequency spectrum labeled E2 on inset b) of [Fig. 4] are plotted on inset c) of [Fig. 4] on a distance score axis D whose origin zero corresponds to the distance score, necessarily zero, for the reference spectrum SR. It is thus possible to assess the extent to which the structural damage of a diagnosed sleeper differs from the structural damage of the reference sleeper and therefore, for example, to classify the diagnosed sleepers along a railway into two or more categories.

[0055] In practice, the calculation of the distance score of each diagnosed sleeper is advantageously carried out by an ad hoc computer program of the computer part of the analysis means 60. This amounts to saying that the analysis means 60 are configured to computerize the frequency spectrum of the diagnosed sleeper to a reference spectrum, calculating a corresponding distance score, and to determine the structural damage characterization for the diagnosed sleeper from the calculated distance score, it being noted that this determination can be carried out both computerically and / or by a human operator.

[0056] A third embodiment of analysis step 104 is illustrated by insets a) and b) of [Fig.5].

[0057] In this third embodiment, the reference used in analysis step 104 consists of several reference spectra, which are individually similar to the SR reference spectrum described above in the first and second embodiments, but which are distinguished from each other by corresponding to spectra These are frequency-domain spectra, respectively representative of predetermined structural damage states, which are distinct from one another. As a schematic illustration, in inset a) of [Fig. 5], four distinct structural damage states are considered, each associated with a reference spectra labeled SRI, SR2, SR3, and SR4. Of course, the number of reference spectra, here four, is merely a non-limiting example. These four reference spectra are illustrated in a two-dimensional representation space, defined by two orthonormal axes: a first distance score axis DI and a second distance score axis D2, whose respective distance score calculations differ from one another. Thus, it is understood that the representation space is reduced along two principal axes of variation.More generally, and as mentioned again later, we understand, through this example, that it is possible to use other dimension reduction techniques than the one illustrated in [Fig.5], and this, if necessary, by providing for a number of dimensions other than two as in the example illustrated in [Fig.5].

[0058] Furthermore, in the third embodiment of analysis step 104, for each diagnosed sleeper, the distance scores associated with the diagnosed sleeper are calculated respectively along the distance score axes DI and D2, which makes it possible to assess the "position" of the diagnosed sleeper in the aforementioned representation space, as schematically illustrated in inset b) of [Fig. 5]. From this "position", a structural damage characterization for the diagnosed sleeper can be determined, for example by classifying the diagnosed sleepers along a railway track into various categories according to the respective "positions" of these sleepers in the representation space.

[0059] Here again, in practice, the calculation of the distance score of each diagnosed sleeper is advantageously carried out by an ad hoc computer program of the computer part of the analysis means 60. This amounts to saying that the analysis means 60 are configured to computerize the frequency spectrum of each diagnosed sleeper to several reference spectra, calculating a corresponding distance score, and to determine, in particular computerically, the structural damage characterization for the diagnosed sleeper from the calculated distance score.

[0060] Other embodiments of analysis step 104. By way of non-exhaustive examples:

[0061] - dividing the measurements taken into groups in order to define groups of similarity with respect to the distance between frequency spectra, using machine learning techniques such as k-means partitioning; and / or

[0062] - representation of the position of the measurements along reduced axes using non-linear dimensionality reduction techniques, in order to evaluate the positioning of the measures relative to each other in a reduced dimensionality representation that is more easily interpretable, such as a one-dimensional, two-dimensional or three-dimensional representation.

[0063] Regardless of how analysis step 104 is carried out, it is understood that the structural diagnostic process results in determining a structural damage characterization for each of the diagnosed crossbeams. This structural damage characterization associates the diagnosed crossbeam with one of several predetermined and / or a posteriori classified structural states. Thus, the structural damage characterization provides information, with a controlled margin of error, particularly on:

[0064] - the presence of at least one structural defect falling under one or more types of given defects, such as the F crack mentioned above, and / or

[0065] - more pronounced structural wear than the nominal structural wear.

[0066] Also, regardless of the embodiment of analysis step 104, its implementation potentially involves the operator in deducing this structural damage characterization, in which case the analysis means 60 provide the operator, through computer processing, with the necessary assessment information, for example in the form of at least one comparative display, as in the first embodiment detailed above. Alternatively, the implementation of analysis step 104 is entirely computer-based, which may require significant computing power, particularly that provided by the remote computing equipment 52.

[0067] Finally, various modifications and variations to the structural diagnostic procedure, as well as to the structural diagnostic system 1 described so far, are conceivable. By way of example: - vibration V is detectable and measurable by several sensors of the structural diagnostic system, such as sensor 20, these different sensors being able to be distributed on the crossmember to be diagnosed; - Vibration V can be generated by several impactors of the structural diagnostic system, such as impactor 10, and may then be detected by a single sensor, such as sensor 20, these different impactors being distributed across the crossmember to be diagnosed; and / or - rather than the entire analysis step 104 being implemented in situ, i.e. on the railway, part of this analysis step 104 can be implemented remotely from the railway, for example in an office.

Claims

Demands

1. A method for the structural diagnosis of concrete railway sleepers (2, 3), comprising: - an application step (101), in which a mechanical stress is applied by impact to a concrete railway sleeper (2, 3) to be diagnosed, - a detection step (102), in which a vibration (V) generated by the mechanical stress and propagating in the sleeper being diagnosed is detected on the surface of the sleeper being diagnosed and is transformed into a vibration time signal (SV), - a processing step (103), in which the vibration time signal (SV) is electronically processed so as to obtain a frequency spectrum (SF), and - an analysis step (104), in which the frequency spectrum (SF) is compared to a reference (SR;SRI, SR2, SR3, SR4) to deduce a structural damage characterization of the diagnosed crossbeam, characterized in that, in the analysis step (104): - the frequency spectrum (SF) is compared computationally to a reference spectrum (SR) constituting said reference, by calculating a distance score which is representative of a mathematical distance between the frequency spectrum and the reference spectrum, and - the structural damage characterization is determined from the calculated distance score.

2. A method according to claim 1, wherein the application step (101), the detection step (102) and the processing step (103), as well as at least part of the analysis step (104) are carried out on a railway.

3. A method according to any one of the preceding claims, wherein, in the analysis step (104): - the frequency spectrum (SF) is compared computationally to several reference spectra (SRI, SR2, SR3, SR4) constituting said reference, by calculating distance scores which are respectively representative of mathematical distances between the frequency spectrum and the reference spectra, and - the structural damage characterization is determined from the calculated distance scores.

4. A method according to any one of the preceding claims, wherein, in the application step (101), the mechanical stress is applied to the cross member to be diagnosed (2, 3) by an impact having an energy of less than 3 joules, preferably less than 1 joule.

5. A method according to any one of the preceding claims, wherein, in the processing step (103), the vibratory time signal (SV) is processed so as to extract a significant part of it, before applying to this significant part a bandpass filter and then a mathematical transformation on a frequency domain to obtain the frequency spectrum (SF).

6. A method according to claim 5, wherein the bandpass filter has a low cutoff frequency substantially equal to 3,000 Hz, preferably 3,500 Hz, and a high cutoff frequency substantially equal to 20,000 Hz, preferably 15,500 Hz.

7. Structural diagnostic system (1) for concrete railway sleepers (2, 3), comprising: - at least one impactor (10) adapted to apply a mechanical stress by impact to a concrete railway sleeper (2, 3) to be diagnosed, - at least one sensor (20) adapted both to detect, on the surface of the sleeper being diagnosed, a vibration (V) generated by the mechanical stress and propagating in the sleeper being diagnosed, and to transform the vibration (V) into a vibration time-domain signal (VDS), - an electronic processing unit (40), adapted to process the vibration time-domain signal (VDS) so as to obtain a frequency spectrum (FS), and - analysis means (60), adapted to compare the frequency spectrum (FSDS) to a reference (SR; SRI,

8. SR2, SR3, SR4) to deduce a characterization of the structural damage of the diagnosed crossmember, characterized in that the analysis means (60) are configured to implement an analysis step (104) in which - The frequency spectrum (FS) is compared computationally to a reference spectrum (RS) that serves as a reference, calculating a distance score that represents a mathematical distance between the frequency spectrum and the reference spectrum, and - The structural damage characterization is determined from the calculated distance score. System according to claim 7, wherein the impactor(s) (10), the sensor(s) (20) and the electronic processing unit (40), together with at least one part (61) of the analysis means (60) form a portable hardware assembly, usable by a human operator on a railway.

Images