A polygon detection method and device for train wheels

Abstract

The present invention discloses a method for detecting polygons of train wheels, comprising: obtaining optical fiber phase information of the wheel using an optical fiber sensor; constructing a trend curve based on the optical fiber phase information; constructing a power spectrum based on the trend curve; inputting the power spectrum and wheel information into a trained neural network model to obtain a frequency spectrum of the polygon data; and outputting wheel polygon detection results based on the frequency spectrum and wheel speed information in the wheel information. The present invention reduces signal entropy and useless information by preprocessing the optical fiber phase information by smoothing and removing wheel weight trends. Furthermore, by combining the power spectrum with wheel diameter and vehicle speed information to form multidimensional feature information as input to the neural network, the accuracy of the detection results is further improved, thereby resolving the problem of poor detection accuracy in traditional train wheel polygon detection methods.

Description

Technical Field

[0001] The present invention relates to the technical field of train wheel detection, and in particular to a polygonal detection method and device for train wheels. Background Art

[0002] When a train is running at high speed, the wheels are prone to periodic wear along the circumferential direction due to eccentricity caused by their own manufacturing, wheel-rail friction, track excitation and other factors. This is called wheel polygon problem. The wheel polygon problem will cause abnormal vibration of the wheel and rail when the vehicle is running, leading to loosening and premature fatigue damage of train components, causing hidden dangers to operational safety and causing discomfort to passengers on the train.

[0003] Currently, wheel polygon detection is usually carried out by detecting wheel surface displacement, such as using a laser sensor to directly measure the change in wheel tread profile when static, or indirectly measuring the rail deformation caused by wheel-rail contact during operation; by detecting wheel-rail vibration, using an acceleration sensor to detect rail vibration at the trackside, or detecting wheel vibration on board, and then deducing the wheel polygon state; detecting wheel-rail force changes, using a force sensor to detect rail force changes based on the wheel-rail force relationship.

[0004] However, using vibration accelerometers for detection requires a large number of sensors to be placed near each wheel of each vehicle. While placing accelerometers on the rails can detect multiple objects in real time, they are susceptible to electromagnetic and environmental interference, and the noise of the collected signals is significantly affected, affecting subsequent processing and the accuracy of the final detection results. At the same time, existing polygon processing algorithms for trackside detection require waveform splicing for acceleration, fiber Bragg gratings, and distributed optical fibers to cover the signals of an entire wheel cycle. Quality issues such as inaccurate end-to-end splicing can affect the back-end recognition effect. Comparison with normal wheels or feature extraction parameter calculation methods such as spectrum and time-spectrum (calculating polygon feature frequencies and then comparing them with theoretical values, calculating amplitudes to determine polygon equivalents) result in low recognition accuracy.

[0005] In summary, the traditional polygon detection method for train wheels has the problem of poor detection accuracy. Summary of the Invention

[0006] In view of this, the present invention provides a polygon detection method and device for train wheels, which solves the problem of poor detection accuracy in traditional polygon detection methods for train wheels by improving data acquisition methods and data processing methods.

[0007] To solve the above problems, the technical solution of the present invention is to adopt a polygon detection method for train wheels, including: obtaining optical fiber phase information of the wheel based on an optical fiber sensor; constructing a trend curve based on the optical fiber phase information; constructing a power spectrum based on the trend curve; inputting the power spectrum and wheel information into a trained neural network model to obtain the frequency spectrum of the polygon data; and outputting the detection result of the wheel polygon based on the frequency spectrum and the wheel speed information in the wheel information.

[0008] Optionally, obtaining the optical fiber phase information of the wheel based on the optical fiber sensor includes: obtaining the relative position information of each acquisition channel in the optical fiber sensor compared to the rail segment; obtaining the wheel diameter information of multiple wheels and the vibration signals, starting time information and the wheel speed information when the multiple wheels pass through the rail segment; segmenting the vibration signal based on the relative position information, the starting time information and the wheel speed information to generate the vibration signal of each wheel; and performing sliding mean filtering on the vibration signal of each wheel to generate the optical fiber phase information of each wheel.

[0009] Optionally, constructing a trend curve based on the optical fiber phase information includes: removing low-frequency signals from the optical fiber phase information of each wheel and then constructing the trend curve of each wheel.

[0010] Optionally, the method for training the neural network model includes: constructing an initialized network model; obtaining a training data set and a test data set consisting of power spectrum samples containing wheel speed data, wheel diameter data and pre-classified polygons or non-polygons; and training and testing the neural network model based on the training data set and the test data set.

[0011] Optionally, outputting the detection result of the wheel polygon based on the frequency spectrum and the wheel speed information in the wheel information includes: extracting the maximum frequency in the frequency spectrum and generating a polygon order in combination with the wheel speed information; extracting the maximum amplitude corresponding to the maximum frequency in the frequency spectrum; and generating the detection result based on the polygon order and the maximum amplitude.

[0012] Accordingly, the present invention provides a polygon detection device for train wheels, comprising: a fiber optic sensor for obtaining fiber optic phase information of the wheel; a data processing unit capable of constructing a trend curve based on the fiber optic phase information, and after constructing a power spectrum based on the trend curve, inputting the power spectrum and wheel information into a trained neural network model to obtain the spectrum of the polygon data, and outputting the detection result of the wheel polygon based on the spectrum and the wheel speed information in the wheel information.

[0013] Optionally, the optical fiber sensor is arranged at the waist of the rail segment to be measured, wherein the optical fiber sensor is also used to obtain wheel diameter information of multiple wheels and starting time information and wheel speed information when multiple wheels pass through the rail segment.

[0014] Optionally, the optical fiber sensor is a distributed optical fiber or a grating optical fiber.

[0015] Optionally, the data processing unit generates a polygon order by extracting the maximum frequency in the spectrum and combining it with the wheel speed information, and extracts the maximum amplitude corresponding to the maximum frequency in the spectrum, and then generates the detection result based on the polygon order and the maximum amplitude.

[0016] The primary improvement of the present invention lies in the polygonal detection method for train wheels. This method uses a fiber optic sensor located at the rail web of the rail segment to be tested to collect optical fiber phase information. This passive, outdoor acquisition method overcomes the electromagnetic interference issues of traditional sensors, minimizes optical signal loss, and significantly improves the signal-to-noise ratio. Furthermore, preprocessing the optical fiber phase information by smoothing and removing wheel weight trends reduces signal entropy and useless information. Furthermore, by combining the power spectrum with wheel diameter and vehicle speed information to form multidimensional feature information, which serves as input to a neural network, the accuracy of the detection results is further improved, thereby addressing the poor detection accuracy of traditional polygonal detection methods for train wheels. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a simplified flow chart of the polygon detection method of train wheels of the present invention;

[0018] Figure 2 This is a simplified unit connection diagram of the polygon detection device for train wheels of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] like Figure 1 As shown, a polygon detection method for train wheels includes: obtaining optical fiber phase information of the wheel based on an optical fiber sensor; constructing a trend curve based on the optical fiber phase information; constructing a power spectrum based on the trend curve; inputting the power spectrum and wheel information into a trained neural network model to obtain the frequency spectrum of the polygon data; and outputting the detection result of the wheel polygon based on the frequency spectrum and the wheel speed information in the wheel information.

[0021] Furthermore, the optical fiber phase information of the wheel is obtained based on the optical fiber sensor, including: obtaining the relative position information of each acquisition channel in the optical fiber sensor compared to the rail segment; obtaining the wheel diameter information of multiple wheels and the vibration signals, starting time information and the wheel speed information when the multiple wheels pass through the rail segment; segmenting the vibration signal based on the relative position information, the starting time information and the wheel speed information to generate the vibration signal of each wheel; and performing sliding mean filtering on the vibration signal of each wheel to generate the optical fiber phase information of each wheel.

[0022] During experimental testing, the inventors found that the collected optical fiber phase information contained some low-frequency signals that may not belong to the vibration frequency range caused by the wheel polygon. After setting up multiple groups of wheels with different wheel diameters as a control group for experiments, it was found that the frequency of the low-frequency signal was negatively correlated with the wheel diameter when the wheel speed remained unchanged, that is, the frequency of the low-frequency signal was negatively correlated with the wheel weight. Therefore, it was found that the optical fiber phase information contained low-frequency signals caused by the deformation trend of the rail due to the downward pressure of the wheel gravity. Therefore, constructing a trend curve based on the optical fiber phase information includes: after eliminating the low-frequency signals in the optical fiber phase information of each wheel, constructing the trend curve of each wheel, thereby reducing signal entropy and reducing useless information, effectively improving the accuracy of the neural network prediction results.

[0023] Furthermore, the method for training the neural network model includes: constructing an initialized network model; obtaining a training dataset and a test dataset comprising power spectrum samples containing wheel speed data, wheel diameter data, and pre-classified polygons or non-polygons; and training and testing the neural network model based on the training dataset and the test dataset, thereby enabling the neural network to perform a binary classification of the power spectrum based on the input multidimensional feature information, i.e., determining whether the power spectrum contains a polygonal spectrum or not. The neural network model can be a commonly used neural network such as YOLO-V3 or CNN.

[0024] Furthermore, outputting a wheel polygon detection result based on the frequency spectrum and wheel speed information in the wheel information includes: extracting the maximum frequency in the frequency spectrum and generating a polygon order in combination with the wheel speed information; extracting the maximum amplitude corresponding to the maximum frequency in the frequency spectrum; and generating the detection result based on the polygon order and the maximum amplitude. After generating multiple detection results, a more accurate polygon roughness dB value can be obtained by fitting the corresponding relationship based on the roughness dB values ​​of the multiple maximum amplitudes and their corresponding polygon orders.

[0025] The present invention collects optical fiber phase information using an optical fiber sensor positioned at the waist of the rail segment to be tested. This passive, outdoor acquisition method overcomes the electromagnetic interference issues of traditional sensors, minimizes optical signal loss, and significantly improves the signal-to-noise ratio. Furthermore, by preprocessing the optical fiber phase information by smoothing and removing wheel weight trends, signal entropy and useless information are reduced. Furthermore, by combining the power spectrum with wheel diameter and speed information to create multidimensional feature information, which serves as input to a neural network, the accuracy of the detection results is further improved, thereby addressing the poor detection accuracy of traditional polygonal detection methods for train wheels.

[0026] Correspondingly, such as Figure 2 As shown, the present invention provides a polygon detection device for train wheels, comprising: a fiber optic sensor for obtaining fiber optic phase information of the wheel; a data processing unit, capable of constructing a trend curve based on the fiber optic phase information, and after constructing a power spectrum based on the trend curve, inputting the power spectrum and wheel information into a trained neural network model to obtain the spectrum of the polygon data, and outputting the detection result of the wheel polygon based on the spectrum and the wheel speed information in the wheel information.

[0027] Furthermore, the optical fiber sensor is arranged at the waist of the rail section to be measured, wherein the optical fiber sensor is also used to obtain wheel diameter information of multiple wheels and starting time information and wheel speed information when multiple wheels pass through the rail section.

[0028] Furthermore, the optical fiber sensor is a distributed optical fiber or a grating optical fiber.

[0029] Furthermore, the data processing unit generates a polygon order by extracting the maximum frequency in the spectrum and combining it with the wheel speed information, and extracts the maximum amplitude corresponding to the maximum frequency in the spectrum, and then generates the detection result based on the polygon order and the maximum amplitude.

[0030] The above is a detailed introduction to a polygon detection method for train wheels and a device thereof provided by an embodiment of the present invention. The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principle of the present invention, the present invention can also be improved and modified in several ways, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

[0031] Professionals can also further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly with hardware, a software module executed by a processor, or a combination of the two. The software module can be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

Claims

1. A polygon detection method for train wheels, characterized in that: include: Obtaining optical fiber phase information of the wheel based on optical fiber sensor; constructing a trend curve based on the optical fiber phase information; constructing a power spectrum based on the trend curve; Inputting the power spectrum and wheel information into the trained neural network model to obtain the frequency spectrum of the polygon data; outputting a detection result of a wheel polygon based on the frequency spectrum and wheel speed information in the wheel information; The constructing a trend curve based on the optical fiber phase information includes: After removing the low-frequency signal from the optical fiber phase information of each wheel, constructing the trend curve of each wheel; The method of training the neural network model includes: Build an initialization network model; Obtaining a training data set and a test data set comprising wheel speed data, wheel diameter data and pre-classified polygonal or non-polygonal power spectrum samples; The neural network model is trained and tested based on the training data set and the test data set, so that the neural network can perform binary classification on the power spectrum based on the input multi-dimensional feature information and determine whether the power spectrum has a polygonal spectrum or not.

2. The polygon detection method according to claim 1, wherein: Obtaining optical fiber phase information of the wheel based on the optical fiber sensor includes: Obtaining relative position information of each acquisition channel in the optical fiber sensor relative to the rail segment; Obtaining wheel diameter information of multiple wheels and vibration signals, starting time information and wheel speed information of multiple wheels when passing through the rail segment; Segmenting the vibration signal based on the relative position information, the start time information, and the wheel speed information to generate the vibration signal of each wheel; The vibration signal of each wheel is subjected to sliding mean filtering to generate the optical fiber phase information of each wheel.

3. The polygon detection method according to claim 1, wherein: Outputting a detection result of a wheel polygon based on the frequency spectrum and wheel speed information in the wheel information includes: extracting the maximum frequency in the spectrum and generating a polygon order in combination with the wheel speed information; Extracting the maximum amplitude corresponding to the maximum frequency in the spectrum; The detection result is generated based on the polygon order and the maximum amplitude.

4. A polygonal detection device for train wheels, characterized in that: include: Fiber optic sensor, used to obtain fiber optic phase information of the wheel; The data processing unit is capable of constructing a trend curve based on the optical fiber phase information, and after constructing a power spectrum based on the trend curve, inputting the power spectrum and wheel information into a trained neural network model to obtain the spectrum of polygon data, and outputting the detection result of the wheel polygon based on the spectrum and the wheel speed information in the wheel information, wherein the constructing of the trend curve based on the optical fiber phase information includes: constructing the trend curve of each wheel after eliminating the low-frequency signal in the optical fiber phase information of each wheel; the method for training the neural network model includes: constructing an initialized network model; obtaining a training data set and a test data set consisting of power spectrum samples containing wheel speed data, wheel diameter data and pre-classified as polygonal or non-polygonal; training and testing the neural network model based on the training data set and the test data set, so that the neural network can perform binary classification on the power spectrum based on the input multi-dimensional feature information, and determine whether the power spectrum has a polygonal spectrum or not.

5. The polygon detection device according to claim 4, characterized in that: The optical fiber sensor is arranged at the rail waist of the rail section to be measured, wherein, The optical fiber sensor is also used to obtain wheel diameter information of multiple wheels, starting time information when multiple wheels pass through the rail section, and the wheel speed information.

6. The polygon detection device according to claim 5, characterized in that: The optical fiber sensor is a distributed optical fiber or a grating optical fiber.

7. The polygon detection device according to claim 4, characterized in that: The data processing unit generates a polygon order by extracting the maximum frequency in the spectrum and combining it with the wheel speed information, and extracts a maximum amplitude corresponding to the maximum frequency in the spectrum, and then generates the detection result based on the polygon order and the maximum amplitude.

Images