A method, device and storage medium for classifying and selecting ultrasonic guided wave modes in a switch rail

Abstract

The present invention relates to a method for classifying and selecting ultrasonic guided wave modes in a switch rail. The switch rail has a cross-section with a varying area. The ultrasonic guided wave modes of different cross-sections of the switch rail are classified according to the vibration mode characteristics. And based on the classification results, a class of modes that can cover each cross-section is selected as the detection mode, and the detection mode can be used to detect cracks at any position of the switch rail. Further, the optimal excitation point of the detection mode is determined, and an excitation transducer is installed at this point, which can excite modes that can propagate at each cross-section position, reduce mode conversion and the number of modes, and reduce the complexity of subsequent signal processing. The above method realizes the automatic classification and selection of numerous complex modes in the switch rail, has good detection sensitivity for defects at any position of the cross-section of the switch rail, improves the detection efficiency, and can meet the requirements of switch rail defect monitoring. On the basis of the above method, the present invention also proposes a corresponding device and a storage medium.

Description

Technical Field

[0001] The present disclosure relates to the field of detection of defects in switch points, and particularly to a method, device, and storage medium for classifying and selecting ultrasonic guided wave modes in switch points. Background Art

[0002] The switch point of a turnout is a key component for railway line switching. The tip part of the switch point is relatively weak and is easily damaged by wheels to generate cracks. Timely detection of switch point cracks is of great significance for ensuring the safe operation of railways.

[0003] Ultrasonic guided waves are a special type of ultrasonic waves generated when ultrasonic waves propagate in a waveguide medium and continuously reflect, refract, and convert between longitudinal and transverse waves at the medium boundary. Because ultrasonic guided waves can propagate over a long distance and cover the entire cross-section, they have great advantages in the field of medium- and long-distance non-destructive testing, and are thus widely used in detecting structural damages of composite plates, pipelines, cylinders, and steel rails. Ultrasonic guided waves can be generated and received through various physical means, such as mechanical, piezoelectric, electrostatic, electromagnetic, laser, etc. Figure 1 As shown, ultrasonic guided waves are emitted into the steel rail through a piezoelectric transducer and propagate towards both ends after reflection and refraction on the steel rail surface.

[0004] When ultrasonic guided waves encounter a crack, reflected echoes will be generated, and complex mode conversion phenomena will occur during the process. By collecting and analyzing the propagated guided wave signals through a transducer, it is possible to determine whether there is a crack and specific information about the crack, such as its location and size.

[0005] An active ultrasonic guided wave detection system mainly consists of a transducer, an arbitrary waveform signal excitation source, a linear power amplifier, and a multi-channel oscilloscope. Figure 2 As shown is an ultrasonic guided wave detection system built with general equipment. The signal generator serves as the arbitrary waveform signal excitation source. The signal is amplified by the power amplifier to increase its complex driving ability, and the amplified signal is loaded onto the ultrasonic transducer. The transducer is pasted on the steel rail surface using a coupling agent, so that ultrasonic guided waves can be transmitted to the steel rail with less loss, realizing the excitation of ultrasonic guided waves. The receiving transducer is pasted on the steel rail surface using a coupling agent. When the ultrasonic guided waves propagate to the receiving transducer, they will cause the piezoelectric wafer to vibrate, which is converted into a voltage signal through circuit processing and sent to the oscilloscope, realizing the acquisition of guided wave signals. The signals collected by the oscilloscope are exported using a storage medium such as a USB flash drive, and the guided wave signals are processed and analyzed using computer science calculation software to evaluate the health status of the steel rail.

[0006] The existing technologies are all based on ultrasonic guided wave technology to detect defects in equal-section steel rails. By utilizing the dispersion characteristics and vibration mode characteristics of guided waves in the medium, the interaction mechanism between the guided wave modes and the detected defects is further analyzed, and then the modal control, modal extraction of guided waves, and internal defect detection of the waveguide medium to be measured are realized. For the variable-section switch rail, when the guided wave propagates inside it, the number of modes is many times more than that of the basic rail. The group velocity and phase velocity of each mode are different. When the modes are mixed together, it is difficult to distinguish and identify each wave packet, which will have an extremely adverse impact on the reception, processing, and analysis of the guided wave signal. Summary of the Invention

[0007] In view of this, the main purpose of the present invention is to classify each guided wave mode in the switch rail, determine a class of modes with similar vibration characteristics in each cross-section, which helps to find the optimal excitation point. Applying excitation at this position of the switch rail can minimize the number of excited modes, so as to solve the problem that it is difficult to detect defects in the variable-section switch rail because the guided wave modes inside the cross-sections with different sizes are different, there are many modes propagating inside it, and the extraction of fault signal characteristics is complex.

[0008] In the first aspect, the present invention proposes a method for classifying and selecting ultrasonic guided wave modes in a switch rail. The switch rail has a cross-section with a changing area. The method classifies the ultrasonic guided wave modes of different cross-sections of the switch rail according to the vibration shape characteristics; and based on the classification result, selects a class of modes that can cover each cross-section as the detection mode, and the detection mode can be used to detect cracks at any position of the switch rail.

[0009] Preferably, in the method, the different cross-sections of the switch rail are determined as follows:

[0010] S100. Set the cross-sectional area change range, divide the switch rail into several sections according to the range, and regard the cross-section of the switch rail in each section as an equal-section.

[0011] S200. In each section, obtain a cross-section.

[0012] Preferably, in the method, the cross-sectional area change range is 0 - 5%.

[0013] Preferably, in the method, the vibration shape characteristics include the displacement of the modal vibration nodes in space and the phase velocity of each mode; the vibration shape characteristics are obtained through the dispersion curves of each cross-section, and the dispersion curves are calculated by the semi-analytical finite element method after finite element discretization of each cross-section.

[0014] Optionally, in the method, the cross-section obtained in S200 is located at the center point position of each section.

[0015] Preferably, in the method, it is characterized in that the classification method adopted includes one or a combination of the following methods: clustering algorithm, neural network, random forest and other classification methods.

[0016] Preferably, in the method, the detection mode is selected through the following steps:

[0017] S401. Sort the classification result modes according to the energy mean value, and select N modes with large energy mean values as candidate mode classes, where N is a set value;

[0018] S402. Select the one with the smallest energy variance from the candidate mode classes as the detection mode.

[0019] Preferably, in the method, the excitation point of the detection mode is determined through the following steps:

[0020] S411. According to the detection mode, in all cross-sections where the transducer can be installed, select the cross-section where the cross-sectional vibration shape with the highest energy value is located;

[0021] S412. For the selected cross-section, select the maximum vibration position of the mode as the excitation point.

[0022] In a second aspect, the present invention proposes an ultrasonic guided wave mode classification and selection device for a switch rail, including a memory and a processor, and a computer program capable of being loaded and executed by the processor as any one of the methods in claims 1 to 8 is stored on the memory.

[0023] In a third aspect, the present invention proposes a computer-readable storage medium, which is characterized in that: a computer program capable of being loaded and executed by the processor as any one of the methods in claims 1 to 8 is stored.

[0024] Compared with the prior art, the method of the present invention can realize the automatic classification of numerous complex modes in the switch rail, can realize the rapid selection of the detection mode of the switch rail, improve the detection efficiency, has good detection sensitivity to defects at any position of the cross-section of the switch rail, and can meet the requirements of switch rail defect monitoring. For the selected mode, exciting this mode for crack detection of the switch rail will minimize the number of modes propagating in the switch rail, reduce the difficulty of subsequent mode separation and signal feature extraction, and improve the accuracy of switch rail crack detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0026] Figure 1 is a schematic diagram of the ultrasonic guided wave detection technology;

[0027] Figure 2 is a schematic diagram of the composition of the ultrasonic guided wave detection system;

[0028] Figure 3 is a schematic diagram of the switch rail model;

[0029] Figure 4 is a schematic diagram of the typical cross-section of the switch rail in an embodiment of the present invention;

[0030] Figure 5 is a schematic diagram of the finite element discretization model of cross-section 1;

[0031] Figure 6 is a schematic diagram of the comparison of the dispersion curves of cross-section 2 and cross-section 3;

[0032] Figure 7 is a schematic diagram of the division of the cross-sectional area of the switch rail;

[0033] Figure 8 is a schematic diagram of the set of the 7th type of modal vibration shapes;

[0034] Figure 9 is a schematic diagram of the energy distribution of the 9th type of mode;

[0035] Figure 10 is a schematic diagram of the modal vibration shape of cross-section 7 in the 9th group classification. Specific implementation manner

[0036] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present application.

[0037] In the description of the present application, it should be understood that the term "ultrasonic guided wave" is a special type of ultrasonic wave generated when ultrasonic waves propagate in a waveguide medium and continuously undergo reflection, refraction, and conversion between longitudinal and transverse waves at the medium boundary. Among them, some typical guided waves are named after their researchers, such as Lamb wave and Rayleigh wave. Since it can propagate a long distance in the waveguide medium, it is often used for the health condition monitoring of large structural media. The term "mode" refers to the inherent vibration characteristics of a structural system. The term "dispersion" for an isotropic medium here refers to geometric dispersion. When using ultrasonic guided waves for long-distance detection of a waveguide medium, due to the existence of the dispersion phenomenon, the waveform of the guided wave will be distorted during propagation. At the same frequency, there are multiple modes of ultrasonic guided waves in the waveguide medium, and the relationship between the phase velocity and frequency, and the group velocity and frequency is usually represented by a dispersion curve.

[0038] Due to the variable cross-section characteristics of the switch rail, when guided waves propagate inside it, the number of modes is many times more than that of the stock rail. The group velocity and phase velocity of each mode are different. When the modes are mixed together, it is difficult to distinguish and identify each wave packet, and it is very difficult to extract a specific mode alone, which will have an extremely adverse impact on the reception, processing, and analysis of guided wave signals. Therefore, the present invention proposes a method for classifying and selecting ultrasonic guided wave modes in a switch rail, classifying the ultrasonic guided wave modes of different cross-sections of the switch rail according to the vibration shape characteristics; and based on the classification results, selecting a class of modes that can cover each cross-section as the detection mode, and the detection mode can be used for crack detection at any position of the switch rail. Further, find its optimal excitation point, and apply excitation at this position of the switch rail, which can minimize the number of excited modes and reduce the complexity of mode classification and fault signal feature extraction.

[0039] In one embodiment, defect detection is performed on the straight switch rail of the CNR6307-104 turnout. The total length of the straight switch rail is 12,480 millimeters, and the shape is as Figure 3As shown. Due to the variable cross-section characteristics of the switch rail, it cannot be simply regarded as a constant cross-section rail for analysis. When modeling the switch rail, considering that the cross-section of the switch rail changes slowly and continuously longitudinally, the propagation characteristics of guided waves are almost the same as those of a constant cross-section rail locally. Therefore, the switch rail of the turnout can be equivalent to a combination of multiple constant cross-section straight rails. Set the range of cross-sectional area change, such as 0 - 5%, and divide the switch rail into several sections according to the range. The cross-section of the switch rail in each section is regarded as a constant cross-section; within each section, obtain a cross-section. For example, if the cross-section of the switch rail changes uniformly, the cross-section can be selected at the center point of each section. It is also possible to obtain the position of the cross-section for each section according to the change of the switch rail cross-section. For the straight switch rail of CNR6307 - 104 turnout, a total of 8 representative cross-sections are selected, which are the switch rail cross-sections at 0, 397, 1589, 2781, 3972, 5641, 6663, and 12480 millimeters from the tip, respectively, and are marked as cross-sections 1 - 8, as Figure 4 shown.

[0040] The vibration mode characteristics of ultrasonic guided waves in different cross-sections can be obtained through the dispersion curves of each cross-section. The dispersion curves describe the relationship between the frequencies of different modes of ultrasonic guided waves in the switch rail of the turnout and the phase velocity and group velocity. Due to the irregular characteristics of the switch rail cross-section shape, there are many modes propagating inside it, and the dispersion curves are very complex. Before selecting the common modes of each cross-section, the frequency of the mode must be determined first. It can be seen from the dispersion curves that as the frequency increases, the number of guided wave modes that can propagate also increases. Considering the requirements of defect detection of the switch rail of the turnout, too low a detection frequency will reduce the ability to detect small defects, resulting in missed detection. Too high a detection frequency will increase the number of modes and make the dispersion characteristics complex, affecting the detection results. Considering the above factors and combining with on-site tests, the frequency of 60KHz is selected for analysis in this embodiment. Solve the dispersion curves of guided waves at 8 cross-sections in sequence. Taking the tip cross-section, i.e., cross-section 1, as an example to illustrate the specific solution process. Use the PDE toolbox of MATLAB to perform finite element discretization on the cross-sectional diagram. The PDE toolbox will automatically discretize the rail cross-section according to the size and shape of the rail cross-section. It is preferably to use triangular elements, and other element types can also be used. The shape functions of triangular elements are relatively simple, which can reduce the subsequent calculation workload. After being discretized by the PDE tool, the tip cross-section has a total of 179 nodes and 278 elements, as Figure 5 shown. After discretization, the semi-analytical finite element method is used to calculate to obtain the wave number and frequency. The dispersion curves can be drawn according to the obtained values. For the dispersion curve diagrams of 8 cross-sections, a total of 262 modal vibrations appear at 60KHz.

[0041] Taking the dispersion curves of cross-section ② and cross-section ③ as an example for comparison, as Figure 6As shown. By comparing the dispersion curves of similar cross-sections, it can be seen that: when the cross-section sizes and shapes are similar, the dispersion curves of cross-section ② and cross-section ③ are highly similar. Even though new modal components and phase velocity deviations occur due to differences in cross-sectional area and shape, it can still be seen that there are modes with basically the same vibration characteristics at the positions of similar cross-sections ② and ③. Comparing cross-sections 1 - 8, modes with similar characteristics can be found in all of them. Therefore, for the switch rail, although its cross-section is gradually changing, a mode can be found that is included in the dispersion curves of each cross-section of the switch rail. By finding this mode and exciting it for crack detection of the switch rail, the number of modes propagating in the switch rail will be minimized, the difficulty of subsequent mode separation and signal feature extraction will be reduced, and the accuracy of switch rail crack detection will be improved.

[0042] When obtaining such modes that can cover each cross-section, the ultrasonic guided wave modes of different cross-sections of the switch rail can be classified according to the vibration mode characteristics. When performing modal classification, the methods that can be used include one or a combination of the following methods: clustering algorithms, neural networks, random forests, and other classification methods.

[0043] In this embodiment, the K-means clustering algorithm in the clustering algorithm is taken as an example. The K-means clustering algorithm is an iterative clustering analysis algorithm that divides the data into k groups, randomly selects k objects as the initial clustering centers, then calculates the distance between each object and each seed clustering center, assigns each object to the clustering center closest to it, and then divides all the data into k classifications nearby. Among the 8 typical cross-sections selected above, the minimum number of modes corresponding to 60 kHz is 24, and the maximum is 38. Therefore, let the number of classifications k = 25, which can ensure that the selected modes can cover all cross-sections of the switch rail and at the same time make the classification effect detailed enough.

[0044] When performing modal classification, the phase velocity and the vibration energy of the mode are mainly used as characteristic indicators. Define the energy mean index as:

[0045]

[0046] where E represents the energy mean, i is the node number, n is the number of nodes in the cross-section, Δx i , Δy i , Δz i are the displacements of the i-th node in the x, y, and z directions.

[0047] To make the classification result more refined, the cross-section of the switch rail is further divided into 3 parts, the rail head, the rail web, and the rail base, as Figure 7As shown, the energy values are solved for these three parts of the region respectively, and used as several of the classification indicators.

[0048] Define the characteristic indicators of each mode to be classified, as shown in Equation (2):

[0049]

[0050] where C P is the phase velocity of the mode, and E x , E y , E z represent the average energy values of the overall nodes in the x, y, and z directions, and their definitions are as shown in Equations (3)-(5).

[0051]

[0052]

[0053]

[0054] represents the average energy values of the nodes in the rail head part in the x, y, and z directions, represents the average energy values of the nodes in the rail web part in the x, y, and z directions, represents the average energy values of the nodes in the rail base part in the x, y, and z directions.

[0055] For the 25 classification results, check which cross-section the mode belongs to, and eliminate the classification results that do not cover all typical cross-sections. Finally, 10 groups of classification results are obtained. As shown in Table 1. The first column is the number of the classification result, with 10 classification results from 1 to 10. The second column is the number of modes containing cross-section 1 in each classification result. The first group of classification results has 2 modes of cross-section 1, the eighth classification result has 4 modes of cross-section 1, and for the remaining classification results, each only contains 1 mode of cross-section 1. The third to ninth columns are the number of modes of cross-sections 2-8 in each group of classification results respectively.

[0056] Table 1

[0057]

[0058] The classification results of the 10 guided wave modes in Table 1 are distributed in each cross-section. Next, the optimal classification result will be selected from these 10 classifications, and this mode will be used to detect internal crack defects in the switch rail.

[0059] First, calculate the average energy value and the variance of the energy value according to the following Equations (6) and (7) respectively.

[0060]

[0061]

[0062] In the formula, is the energy of the modal whole, and M j is the total number of modes of the j-th section in the classification result, and E jm is the average energy of the m modes of the j-th section, and S 2 is the variance of the energy values of the modes.

[0063] Define as the average energy of the head, web, and bottom of the m-th mode of the j-th section, respectively. Use to replace E in formula (6) jm , and successively obtain the average energy values of the head, web, and bottom parts of each mode, which are respectively denoted as Use to replace E in formula (7) jm , and successively obtain the energy value variances of the head, web, and bottom parts of each mode, which are respectively denoted as

[0064] The average value of the total energy represents the vibration amplitude of the modal whole. The average energies of the head, web, and bottom represent the vibration conditions of each mode in the three regions. If a mode not only has a high total energy but also has relatively high average energies of the head, web, and bottom, it indicates that the energy distribution of the head, web, and bottom is relatively uniform. For the classification combination with a larger average energy, when each mode propagates in the switch rail, the energy is large, and the echo at the defect is also large, which is conducive to the analysis and extraction of the later defect signals.

[0065] The variance is used to measure the deviation degree between the vibration energy of each node of the mode and the average energy. For the modal combination with a smaller variance, the fluctuation of the modal shape is smaller. When the guided wave propagates in the switch rail, the vibration energies of the head, web, and bottom are relatively balanced, and there can be better echo signals for the defect signals of each part, and the probability of missed detection is smaller.

[0066] Secondly, according to the calculation results of the average energy, sort them according to the average value of the total energy, and select N modes with large average energies as candidate modes, where N is a set value; for example, if it is set to 3, it can be found from the calculated average energy results that the 7th, 8th, and 9th categories in Table 1 can be used as candidate categories. For the 7th group classification, although the average value of the total energy is relatively high, the energies of the head and web are very small, and the vibration of the whole mode is concentrated in the bottom part. Draw the modal shape of the 7th group classification, as Figure 8As shown, in the mode of the 7th category classification, the vibration energy is concentrated at the rail bottom, which is only suitable for detecting defects located at the rail bottom. For defects in the rail head and rail waist, the sensitivity is poor. In order to select a more suitable mode from the 8th and 9th categories, according to the smaller energy variance, the 9th group of modes is finally selected as the detection mode for crack detection at any position of the switch rail. And for the 7th category, in terms of energy variance, it is also significantly larger than the 8th and 9th categories.

[0067] In the ninth group of classification, this type of mode specifically includes 8 modes, 1 mode for each cross-section. First, determine the longitudinal position of the excitation, that is, at which cross-sectional position the transducer is installed. Since for the currently selected mode, the energy distribution of the rail head, rail waist, and rail bottom is relatively uniform, when determining the cross-section, directly according to the total energy distribution of all cross-section modes in this classification, select the cross-section with the highest vibration energy distribution. Therefore, it is selected to install the transducer at cross-section 7 for excitation. By plotting the total vibration energy of the modes at each cross-sectional position and the energy diagrams of the rail head, rail waist, and rail bottom, as Figure 9 shown. It can be intuitively seen that cross-section 7 selected according to the above principle is significantly better than the vibration modes of other cross-sections in terms of both total energy and energy of each part. Its modal shape is as Figure 10 shown, and it is indeed a cross-section suitable for installing the transducer for excitation.

[0068] For cross-section 7, determine the optimal excitation point. When exciting at this point, the guided wave signal excited contains the mode of the ninth group of classification and has a relatively large proportion. The commonly used method is to select the maximum vibration position of the mode as the excitation point.

[0069] A physical model of the switch rail is established in ANSYS for simulation experiments. At the same time, on the 12.4-meter-long switch rail in the laboratory, in the section from 6663 to 11880 mm from the tip, that is, cross-section 7 section, a probe is installed for excitation and acquisition experiments. The simulation and experimental results show that when applying the excitation signal at the optimal excitation point determined in the present invention, the energy of the received guided wave signal is the largest, and the selected optimal mode can be excited. The vibration energy distribution of this mode in the rail head, rail waist, and rail bottom is relatively uniform. The guided wave propagates in the switch rail in the current mode, and has good consistency for the reflected echoes of cracks in each part of the switch rail. Therefore, it has good detection sensitivity for defects at any position of the switch rail cross-section and can meet the requirements of switch rail defect monitoring.

[0070] In the second embodiment, a device for selecting ultrasonic guided wave modes in a switch rail is used to select the modes of the above switch rail. The device includes a memory and a processor, and a computer program capable of being loaded and executed by the processor for any of the above methods is stored on the memory.

[0071] In the third embodiment, a computer-readable storage medium is adopted, which stores a computer program capable of being loaded and executed by a processor to perform any of the above methods. After inputting the above-mentioned switch rail-related information, the program can automatically perform modal classification and selection on it.

[0072] In summary, the present invention classifies the ultrasonic guided wave modes in the switch rail, and the selected modes are used for switch rail crack detection. Taking the displacement of the modal vibration nodes in space and the phase velocity of each mode as feature vectors, calculating the vibration energy of the mode as a feature index, using a clustering algorithm to classify all cross-sections, dividing the modes with similar vibration shapes into the same category, eliminating the classification results that do not cover all cross-sections, and retaining the classification results that can cover all cross-sections of the entire switch rail. Each classification combination has similar vibration characteristics, and the modal classification combination for detecting defects in specific regions of the switch rail cross-section can be selected according to these characteristics. Based on this, the present invention selects a class of modes with significant vibrations at the rail head, rail web, and rail base positions. The vibration shape of this mode covers the entire cross-section of the switch rail, and has high detection sensitivity for cracks at any position of the switch rail. Further, based on the vibration energy feature index, according to the vibration shape characteristics of this mode, combined with the position requirements of on-site sensor installation, the optimal excitation position is determined based on the extreme points of energy distribution. Excitation at this node has a better effect than excitation at other nodes, that is, the guided wave energy value at this node is larger, and the proportion of the selected mode in the overall guided wave component will also increase, which can meet the requirements of switch rail defect monitoring. Therefore, the method of the present invention can achieve rapid selection of switch rail detection modes and improve detection efficiency.

[0073] Through the description of the above embodiments, those skilled in the art can clearly understand that the method of the present invention can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by dedicated hardware including application-specific integrated circuits, dedicated CPUs, dedicated memories, dedicated components, etc. Generally, functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structures for implementing the same function can also be various, such as analog circuits, digital circuits, or dedicated circuits, etc. However, in more cases for the present disclosure, software program implementation is a better implementation manner.

[0074] Although the embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the above specific embodiments and application fields. The above specific embodiments are merely illustrative and guiding, rather than restrictive. Those of ordinary skill in the art can also make many forms under the inspiration of this specification and without departing from the scope protected by the claims of the present invention, and these all belong to the scope of protection of the present invention.

Claims

1. A method for classifying and selecting ultrasonic guided wave modes in a switch rail, wherein the switch rail has a cross-section with a changing area, Characterized in that, classify the ultrasonic guided wave modes of different cross-sections of the switch rail according to the vibration shape characteristics; and based on the classification results, select a class of modes that can cover each cross-section as the detection mode, and the detection mode can be used to detect cracks at any position of the switch rail; The vibration shape characteristics include the displacement of the modal vibration nodes in space and the phase velocity of each mode.

2. The method according to claim 1, Characterized in that, The different cross-sections of the switch rail are determined as follows: S100. Set the cross-sectional area change range, divide the switch rail into several sections according to the range, and regard the cross-section of the switch rail in each section as an equal cross-section; S200. Obtain a cross-section within each section.

3. The method according to claim 1, Characterized in that, The cross-sectional area change range is 0 - 5%.

4. The method according to claim 1, Characterized in that: The vibration shape characteristics are obtained from the dispersion curves of each cross-section, and the dispersion curves are obtained by finite element discretization of each cross-section and calculated by the semi-analytical finite element method.

5. The method according to claim 2, Characterized in that, The cross-section obtained in S200 is located at the center point position of each section.

6. The method according to claim 1, Characterized in that, The classification method used includes one or a combination of the following methods: clustering algorithm, neural network, random forest and other classification methods.

7. The method according to claim 1, Characterized in that, The detection mode is selected through the following steps: S401. Sort the classification result modes according to the energy mean value, and select N modes with large energy mean values as the candidate mode classes, where N is a set value; S402. Select the one with the smallest energy variance from the candidate mode classes as the detection mode.

8. The method according to claim 1, Characterized in that, The excitation point of the detection mode is determined through the following steps: S411. According to the detection mode, select the cross-section with the highest energy value among all cross-sections where transducers can be installed; S412. For the selected cross-section, select the maximum vibration position of the mode as the excitation point.

9. An apparatus for classifying and selecting ultrasonic guided wave modes in a switch rail, Characterized in that: It includes a memory and a processor, and a computer program capable of being loaded and executed by the processor, such as any one of the methods in claims 1 to 8, is stored on the memory.

10. A computer-readable storage medium, Characterized in that: A computer program capable of being loaded and executed by the processor, such as any one of the methods in claims 1 to 8, is stored.

Images