Computer-implemented method, data processing system, and computer program product for data curation

Abstract

This invention provides a computer-implemented method. The method includes: receiving (S1) vibration data (VD), wherein the vibration data (VD) is inferred from backscattered laser light from an optical fiber cable (21) and includes information on the vibration energy of a plurality of measurement segments (22, 23) of the optical fiber cable (21); determining (S2) whether the vibration energy change during a recent time window (Δtr) in at least one measurement segment exceeds a predefined amplitude; if it exceeds the predefined amplitude, determining (S3) whether the vibration energy change in the at least one measurement segment (22, 23) exceeds a predefined duration (D), and / or whether the vibration energy change in measurement segments adjacent to the at least one measurement segment (22, 23) exceeds a predefined spatial extent (SE), thereby determining a data set representing the vibration energy change. The invention provides an event signal (S4) if a cluster (DC) exists in the vibration data (VD); and if such a data cluster (DC) is identified in the vibration data (VD). Additionally, the invention provides a data processing system and a computer program product.

Description

[Technical Field]

[0001] This application relates to computer implementation methods, data processing systems, and computer program products. [Previous Technology]

[0002] To monitor linear assets (e.g., railway tracks), fiber optic cables may be laid along the linear asset. Laser light propagating through the fiber optic cable may be backscattered by scattering points within the fiber optic cable. This backscattered laser light infers vibration data containing information about the vibration energy of multiple measurement segments of the fiber optic cable. The fiber optic cable may be several kilometers long and may include thousands of measurement segments. Vibration data can be collected in real time for each measurement segment. Possibly, the measurement resolution may be less than one millisecond, for example, about 500 microseconds. Therefore, the vibration data may contain a large amount of data. However, only a small portion of the vibration data contains relevant information. Such relevant information may refer to events occurring in one or more measurement segments that cause vibrations propagating along the fiber optic cable as seismic waves and / or sound waves. Such events may be rare, and discovering them in such a large amount of vibration data may be challenging. Moreover, if all vibration data were uploaded to the cloud or a computing network to discover and classify such events in the available data, bandwidth limitations would be quickly reached. [Summary of the Invention]

[0003] One object of the present invention is to provide a computer implementation method for efficiently implementing data curation and pre-selection. Another object of the present invention is to provide a data processing system and computer program product for efficiently implementing data curation and pre-selection.

[0004] These objectives are achieved through independent claims. Other embodiments are the subject of subsidiary claims.

[0005] According to at least one embodiment, a computer-implemented method is provided. This computer-implemented method describes an algorithm that can be stored on a computer or storage medium. This computer-implemented method can be configured to run on a computer. This computer-implemented method may be simply referred to as the method below. The method may be a method for data management and / or data pre-selection and / or data acquisition.

[0006] According to at least one embodiment, the method includes receiving vibration data, wherein the vibration data is inferred by backscattered laser light from an optical fiber cable and includes information on the vibration energy of a plurality of measurement segments of the optical fiber cable.

[0007] The received vibration data can be inferred from an optical signal (which represents the vibration detected by the fiber optic sensor). In particular, the optical signal can be a backscattered signal provided as an input signal to the fiber optic cable. The fiber optic cable can be simply referred to as an optical fiber below. The optical fiber can extend along a linear asset. The optical fiber can be laid underground near the linear asset. It is also possible to lay the optical fiber above ground near the linear asset. The optical fiber extends generally parallel to the linear asset. The input signal can be an optical signal, such as a laser pulse. The input signal is provided to the optical fiber at its input end. A portion of the laser light is reflected back to the input end because the laser light is scattered at a scattering point (e.g., impurities in the optical fiber, which can be natural or artificial). The variation in the backscattered signal is related to physical changes in the optical fiber, which can be caused by noise, structural noise, vibration, or acoustic waves along the optical fiber. Therefore, the backscattered signal can be detected to determine the noise occurring around the linear asset. By evaluating the backscattered signal, the location of the noise can be determined. Therefore, the backscattered signal can be evaluated to infer vibrations at different locations and / or different time points along the linear asset corresponding to the measurement segment of the fiber optic cable. Each measurement segment may correspond to one meter or several meters, for example, at least one meter and at most 20 meters of the fiber optic cable.

[0008] According to at least one embodiment, the method includes determining whether the change in vibration energy during a recent time window in at least one measurement segment exceeds a predefined amplitude.

[0009] The measurement segment of this fiber optic cable is susceptible to vibration, fluctuations, and noise at various times. In other words, the vibration data contains noise, and the amplitude of the measured vibration energy is very small. However, if the vibration energy variation exceeds a predefined amplitude, this may indicate that an event has occurred in the measurement segment within the recent time window. The recent time window may refer to the recent measurement period. The recent time window may cover several seconds, for example, the last 1 to 10 seconds of the measurement.

[0010] According to at least one embodiment, the method includes determining whether the vibration energy change in the at least one measurement segment exceeds a predefined duration if the predefined amplitude is exceeded, and / or whether the vibration energy change in a measurement segment adjacent to the at least one measurement segment exceeds a predefined spatial range, thereby determining whether a data set representing the vibration energy change exists in the vibration data.

[0011] The vibration energy change exceeding the predefined duration may mean that an increase in vibration energy can also be observed in the at least one measurement segment, respectively, for previous or subsequent measurement periods or time windows. For example, the predefined duration includes two, three, five, or ten subsequent time windows corresponding to at least 2 seconds and at most 100 seconds. The vibration energy change exceeding a predetermined spatial range may mean that an increase in vibration energy can also be observed in other measurement segments adjacent to or near the at least one measurement segment. For example, if an increase in vibration energy is observed in the next, the second next, the third next, and / or the fourth next measurement segment, then the vibration energy change can be considered to have a spatial range along the optical fiber cable. For example, the predefined spatial range corresponds to two, three, five, ten, or fifty adjacent measurement segments.

[0012] Determining that the vibration energy change exceeds a predefined duration and / or a predefined spatial range means that the vibration data includes a data cluster corresponding to the measurement segment and / or time window that presents an increase in vibration energy. Therefore, the proposed method can be used to discover and identify data clusters with increased vibration energy in the vibration data.

[0013] According to at least one embodiment, the method includes providing an event signal when such a data cluster is identified in the vibration data.

[0014] This could mean implementing the event signal as an alarm, indicating an event occurring along the fiber optic cable corresponding to the data cluster. However, it is also possible to implement the event signal as a data structure based on the data cluster. For example, the event signal may include a portion of the vibration data corresponding to the data cluster, or data derived from the data cluster. The event signal may contain information about the impact location, and / or the amplitude of the vibration energy change, and / or the propagation speed of the seismic wave along the linear asset, and / or the frequency band of the vibration energy change, and / or the duration of the vibration energy change, and / or its spatial extent.

[0015] According to at least one embodiment, a computer-implemented method includes: receiving vibration data, wherein the vibration data is inferred by backscattered laser light from an optical fiber cable and includes information on vibration energy of a plurality of measurement segments of the optical fiber cable; determining whether a change in vibration energy during a recent time window in at least one measurement segment exceeds a predefined amplitude; if it exceeds the predefined amplitude, determining whether the change in vibration energy in the at least one measurement segment exceeds a predefined duration, and / or whether the change in vibration energy in measurement segments adjacent to the at least one measurement segment exceeds a predefined spatial range, thereby determining whether a data cluster representing the change in vibration energy exists in the vibration data; and, if such a data cluster is identified in the vibration data, providing an event signal.

[0016] Among other things, the computer implementation method described herein is based on the following considerations.

[0017] Vibration data collected via fiber optic sensing during the monitoring of linear assets can represent a large amount of data, making it difficult to identify events of interest. Such events may represent hazardous situations, and it is important to record and process them. The proposed method uses an efficient algorithm to detect such events by identifying sudden changes in vibration energy at the impact location and up to hundreds of meters away. Clusters of data associated with these vibration changes are identified within the vibration data, thus enabling a general energy change detector independent of specific signal shapes or sources. The proposed algorithm searches for energy changes and impacts, thereby identifying clusters of energy changes within the vibration data.

[0018] Additionally, it may be desirable to identify events corresponding to the data clusters in the vibration data. For example, if the linear asset (along which the fiber optic cable is laid) is a railway track, it may be important to understand whether the vibration energy change is caused by a moving rail vehicle or by falling rocks. However, due to the sheer volume of data, uploading all the vibration data to a computing network / cloud / artificial intelligence to classify event types is computationally inefficient, or even impossible. The proposed method can be advantageously used to manage, pre-select, and / or filter the vibration data, thereby further processing only the potentially interesting portions of the vibration data. This means a reduction in the amount of data to be transmitted or uploaded, thus avoiding bandwidth issues.

[0019] According to at least one embodiment of the computer-implemented method, the method is performed during the monitoring of a linear asset, the fiber optic cable being arranged along the linear asset. This may mean performing the method to monitor the linear asset and / or to analyze and / or filter and / or manage vibration data collected during the monitoring of the linear asset. This may further mean performing the method immediately, i.e., as soon as the vibration data becomes available. Advantageously, the linear asset can be monitored for potentially hazardous events and / or events that are not part of the normal operating mode of the linear asset.

[0020] According to at least one embodiment of the computer implementation method, the linear asset is one of railway tracks, roads, pipelines, power lines, cables, and fences. Therefore, different linear assets can be monitored by fiber optic sensing. Accordingly, vibration data can be collected and analyzed from a plurality of different linear assets or from fiber optic cables arranged along such linear assets.

[0021] According to at least one embodiment of the computer-implemented method, the event signal represents at least one event type. The event may be at least one of the following types: movement of a person, animal, or object; movement of a rail vehicle; movement of a rail vehicle on a broken or damaged track; partial or complete derailment of a rail vehicle; rockfall; landslide; mudslide; flashover; maintenance work; thermal expansion; pipe leak; excavation. In order to be detectable by fiber optic sensing, the event must occur within a specific area or volume surrounding the fiber optic cable. Therefore, different event types can be identified by the computer-implemented method. The computer-implemented method can filter the vibration data for different events of interest.

[0022] According to at least one embodiment of the computer implementation method, the fiber optic cable is arranged along a linear asset, and the method further includes determining whether the data cluster in the vibration data represents the normal operating mode of the linear asset, and suppressing the event signal in this case.

[0023] In the example of the railway track described above, the normal operating method of the linear asset can be the movement of railway vehicles. It is possible to detect the movement of railway vehicles using fiber optic sensing, and in this case, it may be necessary to filter the vibration data. However, this method can also be used to detect and identify unusual and potentially dangerous events. To distinguish between normal and abnormal events, the method can implement a pattern recognition algorithm. For example, the method can filter data clusters in the vibration data that present typical vibration energy variation patterns for the normal operating mode. In particular, such vibration energy variations may occur in a specific frequency band and / or exhibit specific propagation along the linear asset. In this case, the transmitted data is further reduced and false alarms are avoided by suppressing the event signal.

[0024] According to at least one embodiment of the computer implementation method, the method further includes transmitting the event signal to a user and / or a computing network and / or artificial intelligence to classify the event type represented by the event signal.

[0025] As described above, based on the identified data set, the event signal can represent a reduced amount of data. The user and / or the computing network and / or the artificial intelligence can classify the event type based on the event signal. This may mean that the event signal contains information that enables the user and / or the computing network and / or the artificial intelligence to identify the event. For example, the artificial intelligence can be trained using training data to identify the event type based on the event signal. As described above, the event signal may include the portion of the vibration data corresponding to the data set, and / or information derived from the data set. Identifying the type of event occurring during the measurement may be crucial. This method can advantageously pre-select and filter the vibration data, thereby outputting a selection (i.e., the event signal) for further processing and classification.

[0026] According to at least one embodiment of the computer implementation method, determining whether the vibration energy change during a recent time window in at least one measurement segment exceeds a predefined amplitude includes Z-score normalization of the vibration data for each measurement segment.

[0027] Z-score normalization (also known as standardization) is a statistical method used to transform and recalibrate data to conform to a standard normal distribution. In this process, data points in a dataset are transformed so that the mean is 0 and the standard deviation is 1. Z-score normalization makes different datasets or variables comparable by placing them on the same scale. This is particularly useful when dealing with features measured in different units or ranges. Additionally, Z-scores can help identify outliers in a dataset. Data points with Z-scores significantly different from the mean can be considered outliers, aiding in the detection and handling of anomalies. In the current case, Z-score normalization can be performed on each measurement segment. Advantageously, this allows for efficient determination of vibrational energy variations.

[0028] According to at least one embodiment of the computer implementation method, the vibration data sample used to calculate the Z score is updated in real time.

[0029] This means that at each time point, the vibration data sample includes the most recent measurements. The vibration data sample used to calculate the Z-score in each measurement segment may include a fixed number of measurements. For each new measurement in the sample, the oldest measurement can be removed from the sample. Therefore, this method can implement a computationally efficient "rolling" algorithm to form the vibration data sample.

[0030] According to at least one embodiment of the computer implementation method, the vibration data sample used to calculate the Z score includes vibration data of the recent time window and vibration data of the past time window. According to at least one embodiment, the ratio of the duration of the recent time window to the duration of the past time window is constant and less than 1.

[0031] For example, the recent time window covers at least 1 second and at most 10 seconds of the measurement period. A first mean of the vibration energy is calculated using the measurements from the recent time window. For example, the past time window covers at least the last 30 seconds of the measurement period. A second mean of the vibration energy is calculated using the measurements from the past time window. Additionally, the standard deviation of the vibration energy is calculated using the measurements from the past time window. The Z-score is calculated by subtracting the second mean from the first mean and dividing the result by the standard deviation. Therefore, this method can implement a computationally efficient "rolling" algorithm to determine the vibration energy variation.

[0032] According to at least one embodiment of the computer implementation method, determining the spatial range of the vibration energy change includes applying a tolerance function to ignore measurement segments adjacent to the at least one measurement segment that do not exhibit the vibration energy change.

[0033] Although an event has occurred, some measurement segments may not show changes in vibration energy. For example, these measurement segments may be weakly coupled to the environment, so the impact may not be transmitted to them. The measurement segment is unaffected by the vibration energy changes caused by the impact. If this is the case, the data clusters in the vibration data may have gaps that prevent the method from identifying them as data clusters. Such gaps can be ignored by applying a tolerance function. For example, a threshold number of consecutive unaffected measurement segments can be defined. A data cluster is identified as long as the number of measurement segments corresponding to the gap is less than the threshold number.

[0034] According to at least one embodiment of the computer implementation method, the method further includes selecting at least one predetermined frequency band of the vibration data to determine the vibration energy change.

[0035] The vibration data may contain data corresponding to multiple frequency bands. However, it may be sufficient to look for changes in vibration energy in only one or two frequency bands. For example, monitoring changes in vibration energy in the 1-200 Hz band and / or in the 750-1500 Hz band. Other frequency bands are also possible. By limiting the vibration data to a specific frequency band, the amount of data to be transmitted can be further reduced. In addition, irrelevant frequency bands can be ignored.

[0036] According to at least one embodiment of the computer implementation method, determining whether a data set representing the change in vibration energy exists in the vibration data is performed simultaneously on multiple subsets of the vibration data. According to at least one embodiment, the separately identified data sets are combined according to predefined rules to provide the event signal.

[0037] For example, a subset of vibration data refers to different frequency bands. However, these subsets may also refer to characteristics of the vibration data other than frequency. For example, a subset may refer to data derived from the vibration data as a spatial or temporal derivative. In another example, only values ​​above a specific amplitude level of the vibration energy belong to a subset. A subset may refer to a sample of the vibration data. Determining whether a data cluster exists in the vibration data can be performed in parallel for multiple subsets. For example, the method may search for vibration energy variations in the 10-60 Hz band and the 900-1000 Hz band in parallel. For example, combining the separately identified data clusters according to predefined rules to provide the event signal may mean that an event signal is provided only if a data cluster is identified for each subset. This may also mean that an event signal is provided if X of the Y subsets present the corresponding data cluster (where X and Y are natural numbers). Other rules are also possible. For example, if a data cluster is identified in the 10-60 Hz frequency band and in at least one other subset of the vibration data, an event signal is provided. Advantageously, by considering different subsets of the vibration data, the confidence level of event detection can be increased.

[0038] Furthermore, a data processing system is provided. This data processing system can be configured to execute the computer-implemented method described above. Therefore, features related to the computer-implemented method are also disclosed with respect to this data processing system, and vice versa. This processing system can be implemented as a data acquisition system.

[0039] According to at least one embodiment, the data processing system includes a receiving unit configured to receive vibration data, wherein the vibration data is inferred from backscattered laser light from an optical fiber cable and includes information on the vibration energy of a plurality of measurement segments of the optical fiber cable. The receiving unit can receive the vibration data from a detection unit of an optical fiber sensor. Therefore, the input of the receiving unit can be coupled to the output of the detection unit. The receiving unit can receive the vibration data in real time. The vibration data can represent the measurement value of the optical fiber sensor.

[0040] According to at least one embodiment, the data processing system includes an evaluation unit configured to determine whether a vibration energy change during a recent time window in at least one measurement segment exceeds a predefined amplitude, and if it exceeds the predefined amplitude, to determine whether the vibration energy change in the at least one measurement segment exceeds a predefined duration, and / or whether the vibration energy change in measurement segments adjacent to the at least one measurement segment exceeds a predefined spatial range, thereby determining whether a data cluster representing the vibration energy change exists in the vibration data. The evaluation unit may be coupled to the output of the receiving unit on its input side.

[0041] According to at least one embodiment, the data processing unit includes a transmitting unit configured to provide an event signal when such a data cluster is identified in the vibration data. The transmitting unit may be coupled to the output of the evaluation unit on its input side. As described above, based on the identified data cluster, the event signal may be implemented as an alarm or a data structure.

[0042] The receiving unit, the evaluation unit, and / or the transmitting unit can be implemented using hardware or software. The data processing system may be part of the fiber optic sensor or may be located outside the fiber optic sensor. The data processing unit may be a computer or may be composed of computers. The data processing system may be connected to a computing network. In particular, the transmitting unit may be configured to transmit the event signal to a computing network for further analysis of the data set.

[0043] According to at least one embodiment, a data processing system includes: a receiving unit configured to receive vibration data, wherein the vibration data is inferred by backscattered laser light from an optical fiber cable and includes information on the vibration energy of a plurality of measurement segments of the optical fiber cable; an evaluation unit configured to determine whether a change in vibration energy during a recent time window in at least one measurement segment exceeds a predefined amplitude, and if it exceeds the predefined amplitude, to determine whether the change in vibration energy in the at least one measurement segment exceeds a predefined duration, and / or whether the change in vibration energy in measurement segments adjacent to the at least one measurement segment exceeds a predefined spatial range, thereby determining whether a data cluster representing the change in vibration energy exists in the vibration data; and a transmitting unit configured to provide an event signal if such a data cluster is identified in the vibration data.

[0044] This data processing system has the same advantages as described above in conjunction with the computer implementation method. In particular, this data processing system is able to discover events of interest in a large amount of vibration data. Furthermore, before uploading the relevant data to a computing network or cloud for event classification, this data processing system performs data management and / or data pre-selection of the vibration data. Therefore, bandwidth issues can be avoided.

[0045] According to at least one embodiment, the data processing system further includes a processing unit configured to process the vibration data. For example, the processing unit is configured to filter the vibration data in a frequency space. The processing unit can be implemented as hardware or software. The processing unit can be arranged between the receiving unit and the evaluation unit. The processing unit can prepare a plurality of subsets of the vibration data, each subset corresponding to different characteristics of the vibration data. For example, each subset corresponds to a different frequency band of the vibration data. The evaluation unit can determine in parallel for each subset whether a data cluster exists in the vibration data. By filtering specific frequencies, the amount of data to be transmitted can be further reduced. In addition, by considering different frequency bands, the level of confidence in event identification can be increased.

[0046] Furthermore, a computer program product is provided. According to at least one embodiment, the computer program product includes instructions that, when executed by a computer, cause the computer to perform the computer-implemented method described above. Therefore, features related to the computer-implemented method are also disclosed in this computer program product, and vice versa. This computer program product has the same advantages as those described above in connection with the computer-implemented method. [Simplified Explanation of the Diagram]

[0047] The following description of the accompanying drawings further illustrates and explains the computer implementation method, the data processing system, and the computer program product. Components and parts of the data processing system that have the same function or effect, and steps of the computer implementation method, are represented by the same element symbols. Identical or substantially identical steps, components, and parts may be described only with respect to the accompanying drawings in which they first appear. Their descriptions are not necessarily repeated in subsequent drawings.

[0048] Figures 1 and 2 show example embodiments of fiber optic cables laid out along linear assets.

[0049] Figure 3 shows an example representation of vibration data visualized in a space-time relationship diagram.

[0050] Figure 4 shows an example embodiment of the computer implementation method.

[0051] Figures 5 and 6 show other example embodiments of the computer implementation method based on vibration data representation.

[0052] Figure 7 shows an example embodiment of the data processing system.

Implementation Method

[0053] Figure 1 shows a fiber optic sensor cable 21 arranged along a linear asset 10 (e.g., a railway track 10). The fiber optic cable 21 may be part of a fiber optic sensor 20. The fiber optic sensor 20 includes the fiber optic cable 21 (which may be simply referred to as an optical fiber below). The optical fiber 21 includes a plurality of measurement segments, such as measurement segment 22 and another measurement segment 23. Moreover, the fiber optic sensor 20 includes a detection unit 25. The detection unit 25 collects vibration data VB from the fiber optic cable 21. This vibration data VB is inferred by laser light backscattered from the fiber optic cable 21 and includes information on the vibration energy of the plurality of measurement segments 22, 23 of the fiber optic cable 21. For example, the laser light is emitted by a laser source integrated in the detection unit 25. The detection unit 25 is connected to a data processing system 30. The detection unit 25 collects the vibration data VB from the fiber optic cable 21 and transmits it to the data processing system 30. Therefore, linear assets 10 can be monitored using fiber optic sensing.

[0054] Figure 2 illustrates the principle of fiber optic sensing. As shown in Figure 1, fiber optic cable 21 extends along linear asset 10, which in this example is rail 10. Detection unit 25 is connected to fiber optic cable 21. The diagram illustrates that a rail vehicle 99 can move on rail 10. The top view of Figure 2 shows vibration data inferred from the backscattered signal detected by detection unit 25. The distance along fiber optic cable 21 is plotted on the x-axis, and the amplitude of the vibration is plotted on the y-axis. At the first position x1 along fiber optic cable 21, a person 98 is walking. Therefore, the amplitude of the vibration at the first position x1 is higher than at other positions where no mechanical vibration occurs. At the second position x2 along fiber optic cable 21, the rail vehicle 99 is moving. Therefore, the amplitude of the vibration at the second position x2 is higher than at the first position x1 where less mechanical vibration occurs.

[0055] Figure 3 shows the representation of the vibration data VD as a space-time relationship diagram. The distance along the fiber optic cable 21 is plotted in arbitrary units on the x-axis. The distance is related to different measurement segments 22, 23 along the fiber optic cable 21, as shown in Figure 1. For example, the x-axis corresponds to a 50-kilometer fiber optic cable with thousands of measurement segments. Time is plotted in arbitrary units along the arrows on the y-axis, meaning from bottom to top. For example, the y-axis corresponds to a measurement time of 1 hour. The third axis is not shown in this two-dimensional representation; however, the amplitude of the vibration energy can be plotted on this axis. As mentioned above, the fiber optic cable 21 can be arranged along a linear asset 10, which in this example is a rail 10. However, other linear assets 10 are also possible. For example, the linear asset 10 can be one of rails, roads, pipelines, power lines, cables, and fences.

[0056] Given the dimensions of the x-axis and y-axis, the vibration data VD contains a large amount of information / data. Under normal circumstances, almost no obvious features are identifiable in the vibration data VD. This means that the amplitude of the vibration energy does not increase for the white areas in the figure. At a fixed location (i.e., at a specific measurement section along fiber optic cable 21), a continuous signal CS is displayed throughout the measurement period. For example, this continuous signal CS corresponds to a bridge crossed by the rails, whereby the bridge is exposed to vibration. For these reasons, the vibration energy increases at this location. In addition, the figure shows a signal E0 that varies with space and time. This signal E0 may correspond to an event, which may be the movement of a rail vehicle. During the measurement period, the amplitude of the vibration data increases at the location where the rail vehicle is moving. Therefore, in this figure, the movement of the rail vehicle is represented by the area where the amplitude of the vibration data increases.

[0057] This figure shows other events E1 to E4, which are sparsely distributed in the vibration data (i.e., along the measurement segment and the measurement time). These events E1 to E4 may correspond to event types of interest. For example, event E1 corresponds to a person walking along or across linear asset 10. For example, event E2 corresponds to a rockfall on linear asset 10. For example, event E3 corresponds to maintenance work on linear asset 10. Finally, event E4 may correspond to thermal expansion of linear asset 10. Other events of interest are possible, such as the movement of animals or objects, the movement of rail vehicles on broken or damaged tracks, partial or complete derailment of rail vehicles, landslides or debris flows around linear asset 10, flashovers, pipe leaks, and excavation. All of these events generate vibrations. Therefore, they lead to an increase in vibration energy at or near the fiber optic cable. In addition, the vibration propagates as seismic waves and / or sound waves. If the event occurs in the first measurement segment, the vibration may also be observed in the adjacent measurement segment. The events described above may represent dangerous situations, and it is important to record and address them. However, such events of interest may be difficult to identify within the large volume of vibration data (VD).

[0058] Figure 4 schematically illustrates a computer-implemented method according to one embodiment. This method can be used to discover events of interest in vibration data. Generally, this method can be used for data management and / or data pre-selection, especially in situations where a large amount of vibration data exists, but only a small number of events of interest are "hidden" in this data. The method includes the following steps, which may not necessarily be performed in this order, but can be performed in this order.

[0059] In step S1, vibration data VD is received. This vibration data VD is inferred from backscattered laser light from the fiber optic cable 21 and contains information on the vibration energy of multiple measurement segments 22, 23 of the fiber optic cable 21. As mentioned above, this vibration data VD may represent a large amount of data, and typically, no obvious features are identifiable in this vibration data VD because events leading to an increase in vibration energy are rare.

[0060] In step S2, it is determined whether the change in vibration energy during a recent time window Δtr in at least one measurement segment exceeds a predefined amplitude. This vibration data can be updated in real time. Therefore, this vibration data can refer to live data. Therefore, the recent time window Δtr can refer to the current measurement period. This vibration energy is susceptible to noise. Therefore, the vibration data VD contains noise. However, certain types of events significantly increase this vibration energy, thus distinguishing the signal from the noise floor. Therefore, in this step of the method, a sudden change in this vibration energy during the measurement segment (e.g., the impact location) is detected.

[0061] In step S3, if the predefined amplitude is exceeded, it is determined whether the vibration energy change in the at least one measurement segment (i.e., the impact location) exceeds a predefined duration D. Additionally or alternatively, it is determined whether the vibration energy change in a measurement segment adjacent to the at least one measurement segment exceeds a predefined spatial range SE. For example, the vibration energy change may be observed in a measurement segment hundreds of meters away from the impact location. Therefore, it is determined whether a data cluster DC representing the vibration energy change exists in the vibration data VD. By determining the data cluster DC in the vibration data VD, energy changes are detected in a general manner, i.e., independent of a specific signal shape or signal source. Referring to Figure 3, the data cluster DC is a cluster in the vibration data that extends along the x-axis and / or y-axis. In other words, events E0 to E4 may form a data cluster. The continuous signal CS (e.g., a bridge) may not form a data cluster because the vibration energy does not change with time and space.

[0062] In step S4, if such a data cluster DC is identified in the vibration data VD, an event signal ES is provided. In one example, the event signal ES is implemented as an alarm: a potentially dangerous event has occurred along the linear asset 10 monitored by the fiber optic sensor 20.

[0063] In another example, the event signal ES is implemented as a data structure based on the data cluster DC. This could mean that in optional step S41, the event signal ES is output data provided to the user and / or computing network and / or artificial intelligence to classify the event type represented by the event signal ES. The event signal ES may be a data segment corresponding to the data cluster DC, or it may be data derived from the data cluster DC. For example, the event signal ES contains information about the impact location, and / or the amplitude of the vibration energy change, and / or the propagation speed of the seismic wave / sound wave along the linear asset 10, and / or the frequency band of the vibration data VD, and / or the duration D of the vibration energy change, and / or its spatial range SE (in terms of the measurement segment in which the vibration energy change was detected). In other words, the vibration data VD is filtered and / or pre-selected so that only data belonging to the event of potential interest is transmitted to the user, the computing network and / or the artificial intelligence. Therefore, reducing the amount of data to be classified can avoid bandwidth issues.

[0064] In the optional step S31 shown in FIG4, it is determined whether the data cluster DC in the vibration data VD represents the normal operating mode of the linear asset 10. In this case, the event signal ES is suppressed. In this example, if the linear asset 10 is a railway track, the normal operating mode may be the movement of a rail vehicle. Such events (shown as event E0 in FIG3) may not constitute events of interest. The method can determine the normal operating mode by means of a suitable algorithm. For example, the method can identify a specific pattern in the vibration data VD caused by the movement of a rail vehicle. For example, if the linear asset 10 is a road, the normal operating mode may refer to the movement of road vehicles (e.g., cars).

[0065] In the optional step S32 shown in Figure 4, multiple detections of data clusters DC that are close in time and / or space are combined into a single detection. For example, if a rockfall occurs, multiple measurement segments at different locations may be affected, and the impacts of the same rockfall may be temporally separate. However, since the detected data clusters DC belong to the same event (rockfall in this example), they can be considered as a combined data cluster DC. Step S32 can be implemented such that if the distance between data clusters in space and / or time is less than a predefined limit, these clusters are considered as a single data cluster DC.

[0066] The method shown in Figure 3 can be implemented as a computer program product. This means that the program includes instructions that, when executed by a computer, cause the computer to perform the method according to Figure 3.

[0067] Furthermore, as shown in FIG4, the method may include additional optional steps S11 such as selecting at least one predetermined frequency band of the vibration data VD to determine the vibration energy variation. The vibration data VD may contain multiple frequency bands of vibration energy in the measurement segment. However, only a few frequency bands may be of interest and / or may be sufficient to monitor the linear asset 10. Therefore, the vibration data VD may be filtered by frequency. The vibration data VD may also be filtered and / or preprocessed for categories other than frequency. In particular, the vibration data VD (on which the data cluster DC is determined) may be the original data or data derived from the original data. The derived data may form a subset of the vibration data VD. In other words, a first subset of the vibration data VD may refer to a first frequency band of the vibration data VD, and a second subset of the vibration data VD may refer to a second frequency band of the vibration data VD. It is also possible that other subsets of the vibration data VD refer to the derivatives of the vibration data VD.

[0068] If the method includes step S11 to select a predetermined frequency band or another subset of the vibration data VD, then the method may further include steps S2', S3', S31' and / or S32' to analyze and process the vibration data VD for different frequency bands and / or subsets. For detailed descriptions of steps S2', S3', S31' and / or S32', please refer to steps S2, S3, S31 and S32 above respectively. Therefore, determining whether a data cluster DC representing the vibration energy change exists in the vibration data VD can be performed simultaneously for multiple subsets of the vibration data VD. The data clusters DC identified for each of the multiple subsets can be combined according to predefined rules to provide the event signal ES. For example, if X of the Y analysis subsets present corresponding data clusters (X and Y are natural numbers), then the event signal ES is provided, indicating that an event has occurred. In particular, if the vibration data VD is analyzed in three different frequency bands, and the data cluster DC exists in only two frequency bands, then the event signal ES may or may not be provided based on the predefined rule.

[0069] Step S2, which determines whether the change in vibration energy during a recent time window Δtr in at least one measurement segment exceeds a predefined amplitude, may include Z-score normalization of the vibration data VD for each measurement segment 22, 23. In particular, the vibration data sample used to calculate the Z-score is updated in real time. The vibration data sample used to calculate the Z-score includes the vibration data VD for the recent time window Δtr and the vibration data VD for the past time window Δtp, wherein the ratio of the duration of the recent time window to the duration of the past time window is constant and less than 1. This is illustrated in Figure 5 based on the representation of the vibration data VD including the data cluster DC. Figure 5 shows the recent time window Δtr and the past time window Δtp for a specific measurement segment. However, such time windows may be formed for each of the plurality of measurement segments 22, 23. The time window may be calculated for all measurement segments 22, 23 using a computationally efficient "rolling" method. This means that with each new element (new measurement) in the corresponding time window, the oldest element (oldest measurement) is removed. For example, the recent time window △tr covers the most recent one to ten seconds of the measurement time. The past time window △tp may cover at least the last 30 seconds of the measurement time. Thus, the Z score for each measurement segment can be calculated as follows:

[0069]

[0070] Where μr(t+tr) is the mean of the recent time window Δtr, μp(t+tp) is the mean of the past time window Δtp, and σp(t+tp) is the standard deviation of the past time window Δtp. By calculating the z-scores of each measurement segment 22 and 23, the vibration energy change can be determined, which correspondingly supports the determination of the data cluster DC in the vibration data VD, as shown in Figure 5.

[0071] Figure 6 shows the data cluster DC in the vibration data VD, which corresponds to the vibration energy change over a specific spatial range SE and a specific duration D. As shown in Figure 6, the data cluster DC includes gaps G at various locations or measurement segments. These measurement segments do not exhibit vibration energy changes, for example, due to the weak coupling of the fiber optic cable 21 to the environment at these locations. However, determining the spatial range SE of the vibration energy change may include applying a tolerance function to ignore measurement segments that do not exhibit the vibration energy change, thereby identifying the data cluster DC even if some measurement segments are not affected by the impact. For example, the tolerance function is implemented such that if a set of affected measurement segments contains no more than three, five, or ten consecutive unaffected measurement segments, then this analytical portion of the vibration data VD is identified as the data cluster DC.

[0072] Figure 7 shows an example embodiment of the data processing system 30. The data processing system 30 includes a receiving unit 31 configured to receive vibration data VD, wherein the vibration data VD is inferred by backscattered laser light from the optical fiber cable 21 and includes information on the vibration energy of a plurality of measurement segments 22, 23 of the optical fiber cable. For further explanation, please refer to step S1 in Figure 3.

[0073] The data processing system 30 further includes an evaluation unit 34 configured to determine whether the vibration energy change during a recent time window Δtr in at least one measurement segment exceeds a predefined amplitude, and if it exceeds the predefined amplitude, to determine whether the vibration energy change during the at least one measurement segment exceeds a predefined duration D, and / or whether the vibration energy change in a measurement segment adjacent to the at least one measurement segment exceeds a predefined spatial range SE, thereby determining whether a data set DC representing the vibration energy change exists in the vibration data VD. The evaluation unit 34 may be divided into two or more sub-units. For example, the first sub-unit 35 may be implemented as a Z-score normalizer to determine whether the vibration energy change during a recent time window Δtr in at least one measurement segment exceeds a predefined amplitude. For further explanation, please refer to step S2 of the method. For example, another subunit 36 ​​can be implemented as a transient energy detector to determine whether the vibration energy change in the at least one measurement segment exceeds a predefined duration D, and / or whether the vibration energy change in a measurement segment adjacent to the at least one measurement segment exceeds a predefined spatial range SE. Therefore, the transient energy detector can determine whether a data set DC representing the vibration energy change exists in the vibration data VD. For further explanation, please refer to step S3 of the method described above.

[0074] The data processing system 30 further includes a transmitting unit 39 configured to provide an event signal ES when such a data cluster DC is identified in the vibration data VD. For further explanation, please refer to step S4 of the method. For example, the event signal ES may be implemented as an alarm 50. However, the event signal ES may also be output data based on the data cluster DC. In this case, the event signal ES may be transmitted to a user and / or a computing network and / or artificial intelligence to classify the event type represented by the event signal ES. In Figure 7, the computing network and / or artificial intelligence is shown as a computing cloud 40.

[0075] The data processing system 30 may include other components corresponding to steps S11, S31, and / or S32 of the method described above. In particular, the system 30 may include a processing unit 33 configured to process the vibration data VD, particularly filtering the vibration data VD in the frequency space. It may also include a suppression unit 37 configured to suppress the event signal ES when it is determined that the data cluster DC in the vibration data VD represents the normal operating mode of the linear asset 10. The system 30 may further include a detection associator or merger unit 38. The detection associator or merger unit 38 is configured to evaluate event detections from several different filters and / or subsets of vibration data, and combine these detections into a single detection according to predefined combination rules, as described above. Additionally or alternatively, the detection associator or merger unit 38 is configured to combine multiple detections that are very close in time and space into a single detection, as described in conjunction with step S32 of the method.

[0076] The embodiments of the computer implementation method, data processing system 30, and computer program product disclosed herein have been discussed to familiarize the reader with the novelty of the ideas. Although preferred embodiments have been shown and described, those skilled in the art can make many changes, modifications, equivalents, and substitutions to the disclosed concepts without unnecessarily departing from the scope of the claims.

[0077] It should be understood that this disclosure is not limited to the disclosed embodiments and the content specifically shown and described above. Rather, features described in individual appendices or the specification may be advantageously combined. Moreover, the scope of this disclosure includes such variations and modifications that will be apparent to those skilled in the art and fall within the scope of the appended claims.

[0078] The term "comprising" as used in the claims or specification does not exclude other elements or steps of the corresponding feature or process. If the terms "a" or "an" are used with a feature, it does not exclude a plurality of such features. Furthermore, no reference numeral in the claims should be construed as limiting the scope.

[0079] This patent application claims priority to European Patent Application 24185681.4, the disclosure of which is incorporated herein by reference.

Claims

1. A computer-implemented method, the method comprising: Receive (S1) vibration data (VD), wherein the vibration data (VD) is inferred from backscattered laser light from the optical fiber cable (21) and includes information on the vibration energy of a plurality of measurement segments (22, 23) of the optical fiber cable (21), determine (S2) whether the change in vibration energy during a recent time window (Δtr) in at least one measurement segment exceeds a predefined amplitude, including Z-fraction normalization of the vibration data (VD) for each measurement segment (22, 23), and if it exceeds the predefined amplitude, then confirm (S3) Determine whether the vibration energy change in the at least one measurement segment (22, 23) exceeds a predefined duration (D), and / or whether the vibration energy change in the measurement segment adjacent to the at least one measurement segment (22, 23) exceeds a predefined spatial range (SE), thereby determining whether a data cluster (DC) representing the vibration energy change exists in the vibration data (VD), and if such a data cluster (DC) is identified in the vibration data (VD), provide (S4) an event signal (ES).

2. The computer implementation method as described in claim 1, wherein, The method is performed during the monitoring of a linear asset (10), along which the fiber optic cable (21) is laid.

3. The computer implementation method as described in claim 2, wherein, The linear asset (10) is one of rails, roads, pipelines, power lines, cables, and fences.

4. The computer implementation method as described in any one of claims 1 to 3, wherein, The event signal (ES) indicates at least one event type (E), specifically: movement of a person, animal, or object; movement of a rail vehicle; movement of a rail vehicle on a broken or damaged track; partial or complete derailment of a rail vehicle; rockfall; landslide; mudslide; flashover; maintenance work; thermal expansion; pipeline leak; excavation.

5. The computer implementation method as described in claim 1, wherein, The fiber optic cable (21) is arranged along the linear asset (10), and the method further includes determining (S31) whether the data cluster (DC) in the vibration data (VD) represents the normal operating mode of the linear asset (10), and suppressing the event signal (ES) in this case.

6. The computer implementation method as described in claim 1 further includes transmitting the event signal (ES) to a user and / or a computing network (40) and / or artificial intelligence to classify the event type (E) represented by the event signal (ES).

7. The computer implementation method as described in claim 1, wherein, The vibration data sample used to calculate the Z-score is updated in real time.

8. The computer implementation method as described in claim 7, wherein, The vibration data sample used to calculate the Z score includes vibration data (VD) for the recent time window (Δtr) and vibration data (VD) for the past time window (Δtp), wherein the ratio of the duration of the recent time window to the duration of the past time window is constant and less than 1.

9. The computer implementation method as described in claim 1, wherein, Determining the spatial range (SE) of the vibration energy change includes applying a tolerance function to ignore measurement segments adjacent to the at least one measurement segment (22, 23) that do not exhibit the vibration energy change.

10. The computer implementation method as described in claim 1 further includes (S11) selecting (at least one predetermined frequency band of the vibration data (VD) to determine the vibration energy variation.

11. The computer implementation method as described in claim 1, wherein, Determining (S3) whether a data set representing the change in vibration energy exists in the vibration data (VD) is performed simultaneously on multiple subsets of the vibration data (VD), wherein the separately identified data sets (DC) are combined according to predefined rules to provide the event signal (ES).

12. A data processing system (30), comprising: A receiving unit (32) is configured to receive vibration data (VD), wherein the vibration data (VD) is inferred from backscattered laser light from the optical fiber cable (21) and includes information on the vibration energy of a plurality of measurement segments (22, 23) of the optical fiber cable. An evaluation unit (34) is configured to determine whether the change in vibration energy during a recent time window (Δtr) in at least one measurement segment (22, 23) exceeds a predefined amplitude. The evaluation unit (34) includes a Z-score normalizer (35) to perform the operation on the vibration data (VD) for each measurement segment (22, 23). Z-score normalization, and if the predefined amplitude is exceeded, it is determined whether the vibration energy change in the at least one measurement segment (22, 23) exceeds the predefined duration (D), and / or whether the vibration energy change in the measurement segment adjacent to the at least one measurement segment (22, 23) exceeds the predefined spatial range (SE), thereby determining whether a data cluster (DC) representing the vibration energy change exists in the vibration data (VD), and the transmitting unit (39) is configured to provide an event signal (ES) if such a data cluster (DC) is identified in the vibration data (VD).

13. The data processing system (30) as described in claim 12 further includes a processing unit (33) configured to process the vibration data (VD), particularly filtering the vibration data (VD) in the frequency space.

14. A computer program product, including instructions that, when executed by a computer, cause the computer to perform the method described in any one of claims 1 to 11.

Images