Method for counting axes with computer-aided evaluation and error detection
The method improves axle counting reliability by normalizing and comparing measurement signals to identify genuine wheel passages and detect manipulation, addressing the challenge of distinguishing between wheel signals and interference, thereby ensuring secure and accurate counting in rail systems.
Patent Information
- Application Number
- EP2023171526
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-04
- Publication Date
- 2024-11-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing axle counting systems face challenges in reliably distinguishing between genuine wheel passage signals and interference or manipulation, leading to potential misinterpretation and security concerns, especially in rail infrastructure where malicious attempts to manipulate the count can occur.
A method that evaluates measurement signals by searching for maxima exceeding a threshold, normalizing amplitudes and time, and comparing patterns to identify wheel passages, generating a warning for non-matching signals to prevent incorrect counting, utilizing dynamic time warping and machine learning for improved pattern recognition and security.
Enhances the reliability of axle counting by reducing incorrect detections due to interference or manipulation, allowing for timely intervention and maintaining rail operation safety, with the ability to adapt to changing manipulation strategies through machine learning.
Smart Images

Figure IMGAF001_ABST
Abstract
Description
[0001] The invention relates to a method for counting axles, in which a wheel passes through an axle counting sensor mounted on a track, the axle counting sensor generates a measurement signal, and the course of the measurement signal is evaluated by computer, thereby identifying the wheel. Furthermore, the invention relates to a computer program product and a device for providing this computer program product, wherein the computer program product is equipped with program instructions for implementing this method.
[0002] When counting axles using axle counters, interference of various kinds is known to occur, from simple noise or environmental influences to dangling cables on trains or so-called pinwheel effects on bogies on tight curves. Therefore, there is a particular desire to detect interference similar to signals from wheels or bogies in order to reliably identify the wheel signals and, in particular, to uncover intentional attempts to interfere. Furthermore, railway infrastructures are difficult to protect against malicious tampering. For example, an axle counter could be influenced by passing steel objects past the sensor in such a way that additional counting pulses would be triggered. There is a great interest in detecting such interference as early as possible.
[0003] A related problem must be solved, for example, in the computer-aided recognition of handwriting. Claus Bahlmann et al. describe a suitable method for handwriting recognition in the paper "The Writer Independent Online Handwriting Recognition System frog on hand and Cluster Generative Statistical Dynamic Time Warping" in IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 26, NO. 3, MARCH 2004. The goal here is to recognize letters despite the differences resulting from different handwriting. However, this cannot be easily transferred to axle counters, since axle counters must distinguish between useful signals indicating wheel passages and noise signals. This principle has already been applied to pattern recognition in axle counters, according to EP 4124539 A1.
[0004] In this case, sufficient safety must also be achieved. It must be considered that the interference signals may reach such a level that they are misinterpreted as wheel passage. The following example illustrates this without limiting the generality.
[0005] From a measurement point of view, a bogie consists, for example, of two consecutive wheels, i.e. two maxima of the signal amplitude of the measurement signal with a certain plateau in between, which result in a recognizable pattern.
[0006] The object of the invention is to provide a method for counting axles that offers a comparatively high level of security against false detection of wheel passages due to malicious manipulation attempts (hereinafter referred to as manipulations). Furthermore, the object of the invention is to provide a computer program product and a device for providing this computer program product, with which the aforementioned method can be carried out.
[0007] This task is carried out with the subject matter of the claim (method) specified at the beginning of the when evaluating the measurement signal (U1 ... U2), the curve (VL1 ... VL2) of the measurement signal (U1 ... U2) is searched for at least one exceedance of a predetermined threshold value or a maximum (M1 ... M4) of the amplitude of the measurement signal (U1 ... U2), the curve (NV1 ... NV2) of the measurement signal (U1 ... U2) containing the maximum is compared with patterns (M1 ... M2) of at least one curve (VL1 ... VL2) for the measurement signal (U1 ... U2) when passing a wheel (RD), in the event of a match being found between the curve containing the maximum and one of the patterns, an axle is counted, According to the invention, this is achieved in that in the event that there is no sufficient agreement between the curve containing the maximum and one of the patterns, a warning message is generated.
[0008] The measurement signal is a temporal progression of the measured variable, preferably the signal voltage, which exhibits respective peaks due to the passage of the wheel of an axle, but also due to interference and possible manipulation. This means that the computer-assisted evaluation of the measurement signal can detect the event to be recorded, namely that a wheel has passed the axle counting sensor. However, interference and manipulation can also be falsely identified as such a wheel passage.
[0009] The maximum value does not necessarily have to be the starting criterion for the search. It can also begin after a threshold value has been exceeded during an evaluation. However, this is only an alternative for heavily noisy signals.
[0010] The advantage of the method according to the invention is that the method for counting axles is, so to speak, hardened against the unnoticed occurrence of manipulation attempts. In other words, the axle counting method achieves artificial intelligence (i.e., improved pattern recognition, particularly through machine learning) by knowing the expected axle counting results, whereby the signal curve of such results can be compared with the measured measurement signals. If a given measurement result does not match a regular pattern, i.e., one representing permissible axle counting results, the warning message generated provides an indication of this.
[0011] The warning message can lead to various measures. A tampering attempt in itself does not necessarily immediately pose a safety risk to rail operations. For example, if there are no trains running near the axle counter, the warning message can be used to further monitor the affected axle counter and, for example, dispatch maintenance personnel to inspect it. It may also be a technical error that can be corrected.
[0012] If trains are in the vicinity of the axle counter or if the incidents occur repeatedly, further consequences can be derived from the warning message (e.g., in terms of security monitoring), up to and including suspension of operations on the affected line. A pattern comparison can also be performed with regard to the error that has occurred, comparing measurement results that deviate from the regular patterns mentioned above to determine whether they indicate a pattern of error. This can be due to either a malfunction of the axle counter or a specific type of attempted manipulation (more on this below).
[0013] In the context of the invention, "computer-aided" or "computer-implemented" can be understood as an implementation of the method in which at least one computer or processor carries out at least one method step of the method.
[0014] The term "computer" or "computer" covers all electronic devices with data processing capabilities. Computers can include, for example, personal computers, servers, handheld computers, mobile devices, and other communication devices that process data in a computerized manner, as well as processors and other electronic devices for data processing, which can preferably also be connected to form a network.
[0015] In the context of the invention, a "processor" can be understood as, for example, a converter, a sensor for generating measurement signals, or an electronic circuit. A processor can be, in particular, a central processing unit (CPU), a microprocessor, a microcontroller, or a digital signal processor, possibly in combination with a memory unit for storing program instructions, etc. A processor can also be understood as a virtualized processor or a soft CPU.
[0016] In the context of the invention, a "storage unit" can be understood as meaning, for example, a computer-readable memory in the form of a random-access memory (RAM) or data storage device (hard disk or data carrier).
[0017] "Interfaces" can be implemented in hardware, for example, via a wired or wireless connection, and / or in software, for example, as interaction between individual program modules or program parts of one or more computer programs.
[0018] "Program modules" are understood to mean individual functional units that enable a program sequence of method steps according to the invention. These functional units can be implemented in a single computer program or in several communicating computer programs. The interfaces implemented in this way can be implemented in software within a single processor or in hardware if multiple processors are used.
[0019] According to one embodiment of the invention, it is provided that when evaluating the measurement signal, at least one maximum of the signal amplitude is searched for in the course of the measurement signal, the amplitude of the measurement signal is normalized during amplitude normalization such that the maximum is identical to a predetermined target value, a dynamic time normalization is carried out for the course of the measurement signal before and after the maximum, wherein the course of the measurement signal normalized by amplitude normalization and time normalization is compared with patterns of both at least one course for the measurement signal when passing a wheel and at least one course for the measurement signal when an error occurs.
[0020] According to this design, both amplitude normalization and dynamic time normalization, also referred to as Dynamic Time Wrapping (DTW), are applied to the measurement result. This has the advantage that the measurement signal is normalized both in terms of its maximum amplitude and the length of its temporal profile. This subsequently facilitates the comparison of the measured signal profile to be evaluated, with a specified duration and maximum amplitude, with patterns of different profiles (more on this below). This improves the reliability of pattern recognition for wheel passages and minimizes the probability of incorrect evaluation results, particularly due to manipulation.
[0021] Amplitude normalization is performed in such a way that the maximum of the measured signal curve under consideration is identical to a specified target value after normalization. Preferably, normalization can be performed to 1, meaning that the target value is equal to 1. However, this is not absolutely necessary. It is important that the specified target value of the maximum matches the maxima contained in the patterns with which the measured signal curve in question is to be compared.
[0022] The DTW is performed to identify a time-limited section of the entire measurement signal curve, extending before and after the maximum, which is to be compared with patterns for the purpose of detecting wheel passes or occurring errors. Since a wheel pass generates a signal curve that rises to a maximum and then falls again, each of the curves identified by the DTW contains a maximum.
[0023] Time normalization is performed to allow a comparison of the respective measurement signal curve with the samples. It is particularly important to consider that the measurement signal curve depends particularly on the speed of the vehicle passing the axle counting sensor. A higher speed produces a steeper, shorter rise to the maximum (followed by a corresponding subsequent decline). A lower speed, in comparison, produces a flatter, longer rise to the maximum (followed by a corresponding subsequent decline).
[0024] The principle of DTW is known, for example, from speech recognition (the recognition of speech characteristics during dictation): Here, individual words are to be recognized from a spoken text by comparing them with stored speech patterns. One problem is that words are often pronounced differently. Vowels, in particular, are often spoken longer or shorter. For a successful pattern comparison, the word should therefore be stretched or compressed accordingly, but not evenly, but primarily at the vowels that were spoken longer or shorter. The dynamic time warping algorithm performs this adaptive time normalization. Another application is handwriting recognition. Here, pattern recognition of individual letters takes place, with the goal of recognizing the letters in different handwritings.
[0025] This design leverages the knowledge that, compared to character recognition or speech recognition, the measurement signals of an axle counter are comparatively low in complexity. On the other hand, however, there are errors that can be mistaken for wheel rotation and therefore lead to incorrect results during evaluation. These must be reliably detected despite the comparatively low complexity of the patterns. This is where the invention comes in, by defining patterns not only for the events to be detected during a wheel rotation of different vehicles, but also for typically occurring errors, which are then recognized as such and cannot be mistaken for a wheel rotation.
[0026] In other words, this design aims not only to detect those events whose occurrence is desired and should be counted, but also to deliberately detect those events that should not occur, and therefore should not be counted, but could be mistakenly identified as events to be counted. If these events are reliably identified as errors, they can be excluded as counted events, even if their evaluation as a countable event of a wheel run would be uncertain. This represents the inventive added value of increasing detection reliability.
[0027] According to one embodiment of the invention, it is provided that the course of the measurement signal standardized by amplitude standardization and time standardization is compared with at least one pattern of a course for the measurement signal when a skewer gear occurs when bogies are cornering.
[0028] As already mentioned, a pinwheel is a measurable signal increase in the measurement signal of an axle counting sensor, which generates a maximum between the two maximums of the wheel passages of a bogie. A pinwheel occurs primarily when the axle counting sensor is installed in a curve and the measurement is taken while the bogie is cornering.
[0029] If possible pinwheel effects are defined as a pattern of occurring errors, then when a pinwheel occurs during measurement, an axle counting sensor, particularly within the context of the DTW, can generate a curve that can be assigned to this error after comparison with the pattern existing for the pinwheel. If this assignment is unambiguous, the respective measurement signal curve can be excluded from assignment to the event of a wheel crossing. This is particularly advantageous when assignment to a wheel crossing would be borderline and, in case of doubt, a non-existent axle would be counted incorrectly.
[0030] In other words, there are cases in which the method according to the invention can be used with greater reliability when counting axles belonging to a bogie. The occurrence of incorrectly counted axles can therefore be ruled out, or the probability of such an occurrence can at least be reduced. However, this also increases the probability that the detection of deviations is due to an attempted manipulation, and consequently, a warning signal must be generated. This can advantageously increase the probability of detecting attempted manipulation, particularly in passenger transport, where vehicles with bogies and double axles are used almost exclusively. At the same time, the probability that the occurrence of other errors will lead to a warning signal, which would, for example, trigger a service interruption, is reduced.
[0031] According to one embodiment of the invention, it is provided that an axle counter is used which has a first axle counting sensor and a second axle counting sensor arranged one behind the other in the direction of travel, wherein the method is carried out successively for the first axle counting sensor and the second axle counting sensor.
[0032] These are so-called dual-axle counters, which are widely used. The two installed axle counting sensors, i.e., the first axle counting sensor and the second axle counting sensor, therefore each generate the same maximum values in quick succession over the time course of the measurement signal, at least when there are no disturbances. In this case, the maximum values correspond to the counted wheels. Otherwise, interference signals can also be detected, leading to maximum values.
[0033] The use of two axle counting sensors does not change the functional principle of the axle counter. The first axle counting sensor and the second axle counting sensor function in the same way as the axle counting sensor of an axle counter in which only a single axle counting sensor is installed. Therefore, the statements made in connection with this invention apply equally to the axle counting sensor or the first axle counting sensor as well as the second axle counting sensor, unless otherwise stated.
[0034] The advantage of using a first axle counting sensor and a second axle counting sensor is that the axle counter offers greater protection against failure. As already described, this reduces the likelihood that errors will be mistakenly interpreted as tampering, resulting in a shutdown of operations after a warning signal is issued.
[0035] According to one embodiment of the invention, it is provided that the maxima in the first curve detected by the first axle counting sensor and the second curve detected by the second axle counting sensor are compared and only those maxima in the first curve and second curve of the measurement signal normalized by amplitude normalization and time normalization are compared with patterns that are present in both the first curve and the second curve.
[0036] This embodiment of the invention leverages the knowledge that the event of a wheel passing the axle counting sensor is reliably detected as a maximum in the measurement signal curve. Therefore, these maximums must also occur in both measured signal curves. If a maximum occurs in only one of the two measurement signal curves, it can be concluded that this is an interference signal that should not be counted. In this case, there is also a probability that this is an attempted manipulation. This can advantageously be reliably signaled by generating a warning signal.
[0037] According to one embodiment of the invention, it is provided that the maxima in the first curve detected by the first axle counting sensor and the second curve detected by the second axle counting sensor are compared and the curve of the measurement signal before and after a maximum, which is to be taken into account in the dynamic time normalization, is determined taking into account a time offset between a comparable maximum of the first curve and the second curve.
[0038] If maxima are found in the first and second curves that correspond to each other, the time offset determined from these can be advantageously used to obtain a speed-dependent measure for the temporal limits of the curve to be considered in the dynamic time normalization. This advantageously ensures that the curve in the dynamic time normalization has a sufficient range to contain the characteristics to be assessed for later comparison with the patterns. As a result, the reliability of the detection process increases, which is why attempts at manipulation can also be detected with greater certainty. As already described, this reduces the probability that errors will be falsely interpreted as manipulation and that operation will come to a standstill after a warning signal is issued.
[0039] According to one embodiment of the invention, it is provided that the course of the measurement signal standardized by amplitude standardization and time standardization is compared with patterns of both at least one course for the measurement signal when passing a single wheel and at least one course for the measurement signal when passing two wheels of a bogie.
[0040] This embodiment of the invention takes advantage of the finding that the double axle of a bogie, i.e. the two wheels that pass the axle counting sensor in this case, produce a characteristic pattern with two maxima. If these two maxima are identified by the DTW as belonging to the bogie, normalization can be carried out with reference to this double event. This can then be compared with the corresponding pattern. This further increases reliability. At the same time, such a double signal is more difficult to produce through malicious manipulation, thus advantageously increasing the probability that a manipulation attempt will be detected. This is particularly important in passenger transport, where manipulation can endanger human lives.However, in passenger transport, vehicles with bogies and double axles are almost exclusively used, so that an increase in safety can be achieved here.
[0041] According to one embodiment of the invention, a comparison with patterns of a course for the measurement signal when passing two wheels of a bogie is only carried out if the time offset of the maxima in the course of the measurement signal does not exceed a limit value predetermined as a function of the speed of a vehicle passing the axle counting sensor.
[0042] This measure is based on the knowledge that when passing a bogie, the axle counting sensor records two maximum values in quick succession. In other words, it can be ruled out that it is a bogie if the maximum values are not measured within a speed-dependent time interval typical for bogies. This also represents an additional security gain with regard to the detection of tampering. If the tampering fails to meet this criterion, the signals from the attempted tampering are not even recognized as a bogie. The downstream process step, in which a warning message must be issued due to an incorrect detection, thus does not occur at all, which advantageously enables rail traffic operations with fewer disruptions.
[0043] To specify the threshold value, the speed of the vehicle crossing the axle counting sensor must be known. There are various ways to do this. For example, the speed can be determined using another sensor and fed into the process as an input variable. For example, the speed can be measured in the vehicle and transmitted wirelessly to a computer, which performs the calculations of the method according to the invention.
[0044] Another possibility is to estimate the speed from the relationship between a pattern of maximum values (corresponding to the axle counting pulses). Bogies are typically installed on vehicles of a certain length, so that the bogies each generate maxima close to each other, followed by a longer pause (passing the center of the vehicle) or a shorter pause (between two coupled vehicles). The ratio of the pauses allows the speed to be estimated and thus the speed-dependent limit value to be determined.
[0045] Another option is to use so-called dual-axle counters, in which two axle counting sensors are installed in close succession. Since the distance between the axle counting sensors is known, the speed can be determined by determining the time offset between the maximum values generated by the same wheel in both axle counting sensors (more on this below).
[0046] According to a particular embodiment of the invention, it is provided that Speed is calculated taking into account two axles of a vehicle crossing the axle counter (AZL) or of one axle of the crossing vehicle which crosses the first axle counting sensor (AZ1) and the second axle counting sensor (AZ2) of the axle counter (AZL), it is checked whether the calculated speed lies within a permissible speed interval, the warning message is also generated if the speed lies outside the permissible speed interval.
[0047] This advantageously allows for additional plausibility checks of the axle counting signals. It is very difficult to generate multiple signals consecutively on one axle counter or different axle counters in such a way that the manipulated signals also withstand a test regarding the train's speed. Even if the individual signals are deceptively similar, improbable calculated speed values can be used to conclude that an error or attempted manipulation has occurred.
[0048] According to a particular embodiment of the invention, it is provided that the acceleration of the vehicle is calculated from the calculated speed of several axles of the vehicle, it is checked whether the calculated acceleration lies within a permissible acceleration interval, the warning message is also generated if the acceleration lies outside the permissible acceleration interval.
[0049] The above statement regarding speed also applies in particular to acceleration. If multiple speeds of the vehicle crossing the axle counter are measured, the change in speed during the measurement process can be used to determine whether the train is accelerating, braking, or traveling at a constant speed. In this case, certain acceleration values are unlikely due to inertia. However, it is also unlikely if the acceleration lies within the interval, but rapid jumps are detected in the temporal sequence of individual measurements. In particular, repeated jumps between positive and negative acceleration values, which indicate a deceleration, indicate an error or an attempted manipulation.
[0050] According to a particular embodiment of the invention, it is provided that in the event that no correspondence is found between the curve containing the maximum and one of the patterns, an error pattern representing the respective measurement signal is generated and stored.
[0051] This measure advantageously allows the axle counter to later detect potential errors, and especially attempts at manipulation. This allows (in the future) to determine, by evaluating the unusual measurement signal and also using a pattern comparison, whether additional information regarding the nature of the deviation from the expected result is already available. This comparison can be performed as follows.
[0052] According to a particular embodiment of the invention, it is provided that in the event that no agreement is found between the course containing the maximum and one of the patterns, the history containing the maximum is compared with stored error patterns, in the event of a match being found between the history containing the maximum and one of the error patterns, a match message is generated.
[0053] The compliance report can be used later in the process in conjunction with the detected error pattern to advantageously initiate more targeted measures. For example, if a tampering attempt has been reliably detected, the police or security personnel can be called in. If an error has been reliably detected, maintenance can be carried out or maintenance personnel can be dispatched to the axle counter in question.
[0054] To improve the operation of the axle counter and, in particular, to detect attempts at manipulation more reliably, artificial intelligence can be advantageously applied, which enables machine learning. This gradually improves pattern recognition, while simultaneously further improving the recognition of regular patterns (i.e., patterns describing permissible axle counting results).
[0055] It should be noted that, unlike typical errors, such as the aforementioned "spit walk" in double racks, it is not possible to classify manipulation attempts in advance in the form of error patterns, since the means by which the malicious manipulation is intended to be achieved are not yet known. This is where the invention comes in, applying machine learning so that the method can, so to speak, update itself, thus enabling rapid adaptation to changing manipulation strategies.
[0056] Artificial intelligence (hereinafter also abbreviated to AI), also referred to as artificial intelligence, is understood in the context of this invention in the narrower sense to mean the capability of computer-aided machine learning, also called machine learning (hereinafter also abbreviated to ML). This involves the statistical learning of the parameterization of algorithms, preferably for very complex application cases. Using ML, the system recognizes and learns patterns and regularities in the recorded process data based on previously entered learning data. With the help of suitable algorithms, ML can independently find solutions to emerging problems. ML is divided into three fields: supervised learning, unsupervised learning, and reinforcement learning. It has more specific applications, such as regression and classification, structure recognition and prediction, data generation (sampling), or autonomous action.
[0057] In supervised learning, the system is trained by the relationship between input and corresponding output of known data, thus learning approximate functional relationships. The availability of suitable and sufficient data is crucial, because if the system is trained with unsuitable (e.g., non-representative) data, it will learn incorrect functional relationships. In unsupervised learning, the system is also trained with sample data, but only with input data and without any connection to a known output. It learns how to form and expand data groups, what is typical for the respective use case, and where deviations or anomalies occur. This allows use cases to be described and error conditions to be discovered.In reinforcement learning, the system learns through trial and error by suggesting solutions to given problems and receiving a positive or negative evaluation of this suggestion via a feedback function. Depending on the reward mechanism, the AI system learns to perform corresponding functions.
[0058] Machine learning can be performed, for example, using artificial neural networks (ANNs). Artificial neural networks are usually based on the interconnection of many neurons, such as McCulloch-Pitts neurons or slight variations thereof. In principle, other artificial neurons can also be used in ANNs, such as high-order neurons. The topology of a network (the assignment of connections to nodes) must be determined depending on its task. After the construction of a network, the training phase follows, in which the network "learns." A network can learn, for example, using the following methods: Developing new connections Deleting existing connections Changing the weight (the weights from neuron j to neuron i) Adjusting the thresholds of the neurons, if they have thresholds Adding or deleting neurons Modifying the activation, propagation or output function
[0059] Furthermore, the learning behavior changes when the activation function of the neurons or the learning rate of the network is changed. In practice, an ANN learns primarily by modifying the weights of the neurons. Adjusting the threshold can be performed by a neuron. This enables ANNs to learn complex nonlinear functions using a learning algorithm that attempts to determine all of the function's parameters from existing input and desired output values using an iterative or recursive approach. ANNs are a realization of the connectionist paradigm, according to which a function consists of many simple, similar parts. Only when these parts are combined does their behavior become complex.
[0060] Furthermore, a computer program product with program instructions for carrying out the said method according to the invention and / or its embodiments is claimed, wherein the method according to the invention and / or its embodiments can be carried out by means of the computer program product.
[0061] Furthermore, a provision device for storing and / or providing the computer program product is claimed. The provision device is, for example, a storage unit that stores and / or provides the computer program product. Alternatively and / or additionally, the provision device is, for example, a network service, a computer system, a server system, in particular a distributed, for example, cloud-based computer system and / or virtual computer system, which stores and / or provides the computer program product, preferably in the form of a data stream.
[0062] The provision takes place in the form of a program data block as a file, in particular as a download file, or as a data stream, in particular as a download data stream, of the computer program product. However, this provision can also take place, for example, as a partial download consisting of multiple parts. Such a computer program product is, for example, read into a system using the provision device, so that the method according to the invention is executed on a computer.
[0063] Further details of the invention are described below with reference to the drawings. Identical or corresponding elements of the drawings are provided with the same reference numerals and are explained several times only to the extent that differences arise between the individual figures.
[0064] The exemplary embodiments explained below are preferred embodiments of the invention. In the exemplary embodiments, the described components of the embodiments each represent individual, independently considered features of the invention, which also further develop the invention independently of one another and are thus also to be considered as components of the invention, either individually or in a combination other than that shown. Furthermore, the described components can also be combined with the features of the invention described above.
[0065] They show: Figure 1an embodiment of the device according to the invention with its functional relationships schematically with a computer infrastructure as a block diagram, wherein the individual functional units contain program modules, each of which can run in one or more processors and the interfaces can accordingly be implemented in software or hardware, Figure 2 an embodiment of the method according to the invention, wherein the individual method steps can be implemented individually or in groups by program modules and wherein the interfaces according to Figure 1 are indicated as examples.
[0066] In Figure 1 A vehicle FZ is shown, which is traveling in a direction of travel FR on a track GL. The vehicle FZ has bogies DG, each equipped with two axles. These are Figure 1 indicated by wheels RD.
[0067] As soon as the wheels RD pass over an axle counter AZL, comprising a first axle counting sensor AZ1 and a second axle counting sensor AZ2, a pulse is generated in the course of the measuring signal U1, U2 (cf. Figure 2 ) (more on this below).
[0068] The axle counter AZL is connected to an evaluation unit AE, which has a first computer CP1. This computer CP1 is connected to both the first axle counting sensor AZ1 and the second axle counting sensor AZ2 via a sixth interface S6. Instead of two axle counting sensors, a single axle counting sensor AZ can also be used; therefore, one of the two axle counting sensors is designated by both the reference symbol AZ and the reference symbol AZ1.
[0069] The evaluation unit AE also houses a first storage device SE1, which is connected to the first computer CP1 via a fifth interface S5. This contains, for example, a program for implementing the method according to the invention as well as a library with various patterns M1, M2 (see FIG. Figure 2 ), which are used for certain curves to be measured VL1, VL2, represented by standardized curves NV1, NV2, NV3 (cf. Figure 2 ).
[0070] Furthermore, the first computer CP1 is connected to a second computer CP2 in a control center LZ via a third interface S3. The second computer CP2 is also connected to a second storage device SE2 via a fourth interface S4. The control center represents a trackside facility, such as a signal box or an automatic train control system.
[0071] The vehicle FZ and the control center LZ are equipped with antennas AT, allowing them to communicate with each other via a second interface S2. Furthermore, the vehicle FZ can communicate with a satellite STL via a first interface S1. This makes it possible, for example, to locate the vehicle FZ, where the satellite STL is a navigation satellite.
[0072] The method according to the invention comprises program modules that can run either on the first computer CP1 or on the second computer CP2. This depends on how "intelligent" the axle counting arrangement formed by the axle counter AZL and the evaluation unit AE is.
[0073] In Figure 2 The process of the method according to the invention is illustrated by a flow chart. Schematic representations of the signal curves are chosen to explain the individual process steps. In the upper part of the Figure 2The curves VL1 of the first axle counter sensor AZ1 and VL2 of the second axle counter sensor AZ2 are shown. For this purpose, a diagram has been chosen in which the measurement signal U1, U2 is represented as an output voltage over time t. In the lower part of Figure 2 The subsequent processing steps of a normalization are shown, resulting in the standardized curves NV1, NV2, NV3, NV4, and a comparison with patterns M1 and M2. The method already masters the comparison with these patterns at the beginning of the process flow of this exemplary embodiment. The comparison can be found or optimized using machine learning, as described.
[0074] In the Figure 2 As already mentioned, the axle counter AZL is calculated according to Figure 1with a first axle counting sensor AZ1 and a second axle counting sensor AZ2. It is also possible to use an axle counter with only one axle counting sensor AZ, where the Figure 2 would look similar, ie the diagram of the VL1 course and the associated measures, indicated by arrows, would be omitted.
[0075] Based on the curves VL1 and VL2, it can be seen that the axle counting sensors AZ1 and AZ2 are installed with a lateral offset in the direction of travel in the GL track. This leads to a time offset of comparable maximum values ZVM. This is shown in Figure 2 indicated by selecting the first maximum M1 generated due to the passage of the first wheel RD of the bogie DG in the first curve VL1 and in the second curve VL2.
[0076] Furthermore, it can be seen in the curves VL1, VL2 that two wheels (axles) of a bogie are passing over each other. This can be seen because, in addition to the time-shifted first maximum M1, the curves VL1, VL2 also show a second maximum M2, which is also shifted by the time offset ZVM and is very similar to the first maximum M1. The first maximum M1 and the second maximum M2 are each separated by a time offset ZVR between the wheel passages. This time offset ZVR corresponds exactly to the time difference between the wheel passage of the first wheel of the bogie DG and the second wheel RD of the bogie DG.
[0077] From the time offset, the speed v of the vehicle crossing the axle counter can be calculated. By deriving the speed v, the acceleration a can also be determined. However, this is only possible if additional axle counting pulses from other axles of the vehicle (which are in Figure 2 are not shown).
[0078] The calculated values for speed v and acceleration a can be examined in a plausibility check step (PL) to determine whether these values are realistic for the crossing vehicle. For this purpose, for example, a speed interval and an acceleration interval can be defined, which can also be defined vehicle-specifically. In the latter case, however, data from the pattern recognition described below must also be used to select the relevant speed intervals and acceleration intervals, which allows for the identification of a specific vehicle type.
[0079] If the plausibility check step PL shows that the speed and / or acceleration values are not plausible (e.g., they do not lie within the relevant speed and acceleration intervals), a warning message is generated and output, for example, via an output device.
[0080] In order to generate the standardized curves NV1, NV2, NV3, NV4, a standardization N is carried out in a manner not shown in detail according to the method according to the invention. This standardization includes an amplitude standardization of the measurement signal U1, U2 to a target value ZW, which in the exemplary embodiment according to Figure 2is 1. In addition, a dynamic time normalization is performed, whereby the first curve VL1 and the second curve VL2 are examined before and after the identified maximum M1, M2, M3, respectively, to such an extent that the curve associated with the maximum M1, M2, M3 can be characterized (and compared with patterns MU1, MU2, MU3; more on this below). This results in the normalized curves NV1, NV2, NV3, NV4 being created in time windows ZF1, ZF2, ZF3, which correspond in their temporal extent to the patterns MU1, MU2, MU3.
[0081] As can be seen, the evaluation of the first maximum M1 leads to the generation of the first normalized curve NV1, and the evaluation of the second maximum M2 leads to the generation of the third normalized curve NV3. Furthermore, a third maximum M3 and a fourth maximum M4 can be seen in both the first curve VL1 and the second curve VL2, which leads to the generation of a second normalized curve NV2 and a fourth normalized curve NV4.
[0082] In the last step, a pattern comparison of the standardized curves NV1, NV2, NV3 is carried out. This shows that the first standardized curve NV1 and the third standardized curve NV3 each match the first pattern MU1, which represents a wheel pass. This leads to a count of 2. The second standardized curve NV2 is identified by the second pattern MU2, which represents a pinwheel. Therefore, the standardized curve NV2 is excluded from counting (indicated by an X). Otherwise, in the embodiment according to Figure 2 No warning message is generated because the error that occurred was reliably identified as a phenomenon that can occur in bogies under certain conditions. Therefore, no action is required.
[0083] Unlike the first pattern MU1 and the second pattern MU2, the error pattern MU3 does not yet exist before the occurrence of the fourth measurement signal M4. Since the standardized measurement signal M4 could not be detected after its first occurrence, a warning message WM is generated and output upon its first occurrence.
[0084] At the same time, the measurement signal is saved, and machine learning (ML) is applied in the further course of the process. This allows for the identification of similarities to comparable incidents, i.e., the regeneration of a measurement signal with the characteristics of the fourth measurement signal M4. These findings, as a result of machine learning, contribute to the generation of an error pattern MU3. After this error pattern has been generated, the standardized profile of the fourth measurement signal can also be recognized. In these cases, detection of the error pattern MU3 generates a warning message WM and a match message UM, the latter being linked to additional specific measures depending on the detected error pattern MU3.
[0085] For example, if the manipulation attempt is known through the error pattern MU3, different measures can be initiated than if the error occurs that is known through the pattern MU3 (as already described).
[0086] In Figure 2 It is indicated that the sample MU1, the second sample MU2, and the third sample MU3 have a hatched confidence interval that allows for certain fluctuations with respect to the standardized curves NV1, NV2, and NV3. This takes into account the fact that the measured curves VL1 and VL2 are subject to certain tolerance fluctuations. In addition to measurement tolerance, it must also be considered that different vehicles generate different measurement signals, which depend, for example, on conditions such as the vehicle's wheel wear. List of reference symbols
[0087] LZControl center FZVehicle DGBogie RDWheel FRDirection of travel GLTrack ATAntenna STLSatellite AZLAxle counter AZ, AZ1, AZ2Axle counting sensor AEEvaluation unit CP1 ... CP2Computer SE1 ... SE2Storage device S1 ... S5Interface VL1 ... VL2Course M1 ... M4Maximum NV1 ... NV4Normalized course ZF1 ... ZF4Time window U1 ... U2Measurement signal MU1 ... MU3Pattern NNormalization ZWTarget value ZVMTime offset between comparable maxima ZVRTime offset between wheel passes 2Counting result XExclusion vVehicle speed aVehicle acceleration PLPlasibilitation step WMWarning message MLmachine learning step
Claims
1. Method for counting axles, in which • an axle counting sensor (AZ, AZ1, AZ2) mounted on a track (GL) is passed by a wheel (RD), • the axle counting sensor (AZ, AZ1, AZ2) generates a measurement signal (U1 ... U2), • the course (VL1 ... VL2) of the measurement signal (U1 ... U2) is evaluated by computer, whereby the wheel (RD) is identified, • when evaluating the measurement signal (U1 ... U2), the course (VL1 ... VL2) of the measurement signal (U1 ... U2) is searched for at least one exceedance of a predetermined threshold value or a maximum (M1 ... M4) of the amplitude of the measurement signal (U1 ... U2), • the course (NV1 ... NV2) of the measurement signal (U1 ... U2) containing the maximum is compared with patterns (MU1 ... MU2) of at least one course (VL1 ... VL2) for the measurement signal (U1 ... U2) when passing a wheel (RD), • in the case of a match found between the curve containing the maximum and one of the patterns, an axle is counted, characterized by thatin the event that no match is found between the curve containing the maximum and one of the patterns (MU1 ... MU3), a warning message (WM) is generated.
2. Method according to claim 1 characterized by that when evaluating the measurement signal (U1 ... U2) • the amplitude of the measurement signal (U1 ... U2) is standardized during amplitude standardization such that the maximum (M1 ... M4) is identical to a predetermined target value (ZW), • a dynamic time standardization is carried out for the course (VL1 ... VL2) of the measurement signal (U1 ... U2) before and after the maximum (M1 ... M4), wherein the course (NV1 ... NV2) of the measurement signal (U1 ... U2) standardized by amplitude standardization and time standardization is compared with patterns (MU1 ... MU3) of at least one likewise standardized course (VL1 ... VL2) for the measurement signal (U1 ... U2) when passing a wheel (RD).
3. Method according to claim 2, characterized by thatthe curve (NV1 ... NV4) of the measuring signal (U1 ... U2) normalized by amplitude normalization and time normalization is also compared with patterns (MU1 ... MU3) of at least one curve (VL1 ... VL2) for the measuring signal (U1 ... U2) when an error occurs.
4. Method according to one of the preceding claims, characterized by that the curve (NV1 ... NV2) of the measuring signal (U1 ... U2) standardized by amplitude standardization and time standardization is compared with at least one pattern (MU1 ... MU2) of a curve (VL1 ... VL2) for the measuring signal (U1 ... U2) when a skewer gear occurs when bogies (DG) are cornering.
5. Method according to one of the preceding claims, characterized by thatan axle counter (AZL) is used which has a first axle counting sensor (AZ1) and a second axle counting sensor (AZ2) arranged one behind the other in the direction of travel (FR), wherein the method is carried out successively for the first axle counting sensor (AZ1) and the second axle counting sensor (AZ2).
6. Method according to claim 5, characterized by that the maxima in the first curve (VL1) recorded by the first axle counting sensor (AZ1) and the second curve (VL2) recorded by the second axle counting sensor (AZ2) are compared and only those maxima in the first curve (VL1) and second curve (VL2) of the measuring signal (U1 ... U2) normalized by amplitude normalization and time normalization are compared with patterns (MU1 ... MU2) that are present in both the first curve (VL1) and the second curve (VL2).
7. Method according to claim 5 or 6, characterized by thatthe maxima in the first curve (VL1) recorded by the first axle counting sensor (AZ1) and the second curve (VL2) recorded by the second axle counting sensor (AZ2) are compared and the curve (VL1 ... VL2) of the measuring signal (U1 ... U2) before and after a maximum (M1 ... M4), which is to be taken into account in the dynamic time normalization, is determined taking into account a time offset between a comparable maximum (M1 ... M4) of the first curve (VL1) and the second curve (VL2).
8. Method according to one of the preceding claims, characterized by that• speed is calculated taking into account two axles of a vehicle crossing the axle counter (AZL) or of one axle of the crossing vehicle which crosses the first axle counting sensor (AZ1) and the second axle counting sensor (AZ2) of the axle counter (AZL), • it is checked whether the calculated speed (v) is within a permissible speed interval, • the warning message (WM) is also generated if the speed is outside the permissible speed interval.
9. Method according to claim 8, characterized by that • the acceleration (a) of the vehicle is calculated from the calculated speed of several axles of the vehicle, • it is checked whether the calculated acceleration lies within a permissible acceleration interval, • the warning message (WM) is also generated if the acceleration lies outside the permissible acceleration interval.
10. Method according to one of the preceding claims, characterized by that in the event that no agreement is found between the curve containing the maximum and one of the patterns (MU1 ... MU2), an error pattern (MU3) representing the respective measurement signal is generated and stored.
11. Method according to one of the preceding claims, characterized by that in the event that no match is found between the curve containing the maximum and one of the patterns (MU1 ... MU2), • the curve containing the maximum is compared with stored error patterns (MU3), • in the event of a match being found between the curve containing the maximum and one of the error patterns, a match message (UM) is generated.
12. Computer program product with program instructions for carrying out the method according to one of claims 1 - 7.
13. Provision device for the computer program product according to the last preceding claim, wherein the provision device stores and / or provides the computer program product.
Citation Information
Patent Citations
Method for counting axles with computer-aided evaluation
EP4124539A1
Redundancy switching of detection points
US10144439B2
Railroad signal aspect compliance monitoring systems and methods
US7890223B1
