Turnout frog defect detection method, device and equipment

Abstract

The invention provides a turnout frog defect detection method, device and equipment, which are applied to the field of turnout frog detection, and the method comprises the following steps: in response to a sound signal generated when a detection hammer knocks a target sampling point of a turnout frog, extracting a plurality of hammering signals in a continuous time sequence from the sound signal; based on the pulse energy of the hammering signals, effective hammering signals are identified from the hammering signals of the multiple continuous time sequences. The effective hammering signal is used for representing that the pulse energy of the hammering signal of the current time sequence and the pulse energy of the effective hammering signal of the previous time sequence meet a preset detection threshold value. And extracting Mel-frequency cepstrum coefficient (MFCC) features from the effective hammering signals. The MFCC features are used for representing features of the effective hammering signals in a time domain and a frequency domain. The MFCC features are input into the trained defect type recognition model, the defect type of the turnout frog of the target sampling point is obtained, and the problem that the defects of the turnout frog cannot be effectively detected through a traditional turnout frog detection method can be solved.

Description

Technical Field

[0001] The present application relates to the field of turnout frog detection, and in particular, to a method, device and equipment for detecting defects of a turnout frog. Background Art

[0002] As a key conversion device of the railway transportation system, the structural integrity of the turnout frog is directly related to the safety and efficiency of train operation. During the long-term bearing of the wheel-rail impact load, the turnout frog made of cast high manganese steel (CHMS) is prone to generate cracks in stress concentration areas such as welded joints and rail heads, and the crack propagation has the characteristic of concealment. The traditional methods for detecting turnout frogs mainly rely on two traditional methods, namely manual hammering inspection and penetrant inspection, to detect the defects of the turnout frog.

[0003] Among them, the manual hammering inspection mainly judges defects through acoustic feedback. Although the operation is simple, it is limited by the experience threshold of the inspectors and environmental noise interference, and there is a certain missed detection rate. Although the penetrant inspection method can effectively identify surface-opening cracks, it cannot penetrate the material surface layer to detect internal defects. Summary of the Invention

[0004] The purpose of the present application is to provide a method, device and equipment for detecting defects of a turnout frog, which solves the problem that the traditional methods for detecting turnout frogs cannot effectively detect the defects of the turnout frog.

[0005] In a first aspect, an embodiment of the present application provides a method for detecting defects of a turnout frog, which is characterized in that it is applied to a detection terminal, and the detection terminal is communicatively connected to a detection hammer, and the detection hammer includes a microphone, an audio encoder and a communication device. The method includes: in response to a sound signal of the detection hammer hitting a target sampling point of the turnout frog, extracting a plurality of consecutive time-series hammering signals from the sound signal. The sound signal is received by the detection hammer from the turnout frog using the microphone, encoded by the audio encoder, and sent through the communication module. Based on the pulse energy of the hammering signal, an effective hammering signal is identified from the plurality of consecutive time-series hammering signals. The effective hammering signal is used to indicate that the pulse energy of the hammering signal at the current time series satisfies a preset detection threshold compared with the pulse energy of the effective hammering signal at the previous time series. Mel-frequency cepstral coefficient (MFCC) features are extracted from the effective hammering signals. The MFCC features are used to represent the features of the effective hammering signal in the time domain and the frequency domain. The MFCC features are input into a trained defect type recognition model to obtain the defect type of the turnout frog at the target sampling point.

[0006] The defect detection method for turnout frog provided by the embodiment of the present application obtains acoustic signals by using a detection hammer and transmits them to a detection terminal. The detection terminal extracts, detects, and extracts Mel-spectrum features from the collected acoustic signals, then deploys a quantized defect detection neural network using an embedded neural network processor for inference, and finally displays the processed defect detection results on the detection terminal. Without relying on the subjective experience judgment of the detection personnel, it can accurately identify the defect types at the positions where the detection hammer strikes, improving the automation degree of the detection process and the detection efficiency of the defect types of the turnout frog.

[0007] A possible implementation manner is that the defect detection method for turnout frog provided by the embodiment of the present application further includes: for the target hammering signal in any time sequence among multiple consecutive time sequences of hammering signals, extracting the target sub-hammering signal of the target hammering signal within a preset frequency range. Calculating the time-frequency analysis result of the target sub-hammering signal. The time-frequency analysis result is used to represent the hammering signal after the target sub-hammering signal undergoes short-time Fourier transform. Determining the pulse energy of the target hammering signal according to the time-frequency analysis result, the number of time sub-bands of the target sub-hammering signal within the frequency range, and the number of frequency sub-bands of the target sub-hammering signal within the frequency range.

[0008] A possible implementation manner, determining the pulse energy of the target hammering signal according to the time-frequency analysis result, the number of time sub-bands of the target sub-hammering signal within the frequency range, and the number of frequency sub-bands of the target sub-hammering signal within the frequency range, includes: Wherein, X(τ,v) is the time-frequency analysis result, τ is time, v is frequency, T is the number of time sub-bands, F is the number of frequency sub-bands, F1 is the lower frequency limit of the preset frequency range, and F2 is the upper frequency limit of the preset frequency range.

[0009] A possible implementation manner, based on the pulse energy of the hammering signal, identifying the effective hammering signal from multiple consecutive time sequences of hammering signals, includes: obtaining the pulse energy of the effective hammering signal in the previous time sequence of the hammering signal in the current time sequence. If the difference between the logarithm of the pulse energy of the hammering signal in the current time sequence and the logarithm of the pulse energy of the effective hammering signal in the previous time sequence is less than a preset detection threshold, determining that the hammering signal in the current time sequence is an effective hammering signal.

[0010] A possible implementation manner, extracting Mel-frequency cepstral coefficient MFCC features from the effective hammering signal, includes: inputting the effective hammering signal into a Mel filter bank to obtain the Mel spectrum of the effective hammering signal. Performing logarithmic transformation and discrete cosine transformation on the Mel spectrum of the effective hammering signal, and extracting MFCC features from the Mel spectrum.

[0011] A possible implementation manner, performing logarithmic transformation and discrete cosine transformation on the Mel spectrum of the effective hammering signal, and extracting MFCC features from the Mel spectrum, includes: XMFCC (τ, v m ) = DCT(log(Mel(|X(τ, v)| 2 )), τ = 1, 2, …, N τ,MFCC , v m = 1, 2, …, N v,MFCC . Among them, Mel(|X(τ, v)| 2 ) is the Mel spectrum of the effective hammering signal. log(Mel(|X(τ, v)| 2 )) is used to represent the result of logarithmic transformation. DCT(log(Mel(|X(τ, v)| 2 ))) is used to represent the result of discrete cosine transform.

[0012] A possible implementation is to input the MFCC features into the trained defect type recognition model to obtain the defect type of the turnout frog at the target sampling point, including: inputting the MFCC features into a 2D convolutional network to obtain the voiceprint features of the effective hammering signal. Inputting the voiceprint features into a multi-layer perceptron network classifier to determine the defect classification result of the turnout frog at the target sampling point. The defect classification result includes at least one defect type and the probability corresponding to the defect type. Determine the defect type with the highest probability in the defect classification result as the defect type of the target sampling point.

[0013] In a second aspect, an embodiment of the present application provides a defect detection device for a turnout frog, which is characterized in that it is applied to a detection terminal, and the detection terminal is communicatively connected to a detection hammer. The detection hammer includes a microphone, an audio encoder, and a communication device. The device includes: an extraction module, an identification module, and a processing module.

[0014] Among them, the extraction module is used to extract a plurality of consecutive time-series hammering signals from the sound signal in response to the sound signal of the detection hammer hitting the target sampling point of the turnout frog. The sound signal is received by the detection hammer using the microphone, encoded by the audio encoder, and sent through the communication module.

[0015] The identification module is used to identify an effective hammering signal from a plurality of consecutive time-series hammering signals based on the pulse energy of the hammering signal. The effective hammering signal is used to represent that the pulse energy of the hammering signal in the current time series satisfies a preset detection threshold compared with the pulse energy of the effective hammering signal in the previous time series.

[0016] The extraction module is further used to extract Mel Frequency Cepstral Coefficient (MFCC) features from the effective hammering signal. The MFCC features are used to represent the features of the effective hammering signal in the time domain and the frequency domain.

[0017] The processing module is used to input the MFCC features into the trained defect type recognition model to obtain the defect type of the turnout frog at the target sampling point.

[0018] In a third aspect, the present application provides a defect detection device for a turnout frog, and the defect detection device for the turnout frog has a function of implementing the defect detection method for the turnout frog according to the first aspect or any possible implementation manner thereof. This function can be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions.

[0019] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, and instructions are stored in the computer-readable storage medium. When the instructions run on a computer, the computer can execute the defect detection method for the turnout frog according to the first aspect or any possible implementation manner thereof.

[0020] In a fifth aspect, an embodiment of the present application provides a computer program product containing instructions. When the computer program product runs on a computer, the computer can execute the defect detection method for the turnout frog according to the first aspect or any possible implementation manner thereof.

[0021] Among them, for the technical effects brought by any of the design manners in the second aspect to the fifth aspect, reference can be made to the technical effects brought by different possible implementation manners in the first aspect, which will not be elaborated here. Description of the Drawings

[0022] In order to more clearly illustrate the specific implementation manners of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings required to be used in the description of the specific implementation manners or the prior art. Obviously, the drawings in the following description are some implementation manners of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0023] Figure 1 It is a schematic structural diagram of a defect detection system for a turnout frog provided by an embodiment of the present application;

[0024] Figure 2 It is a schematic structural diagram of a detection hammer provided by an embodiment of the present application;

[0025] Figure 3 It is a schematic structural diagram of a detection terminal provided by an embodiment of the present application;

[0026] Figure 4 It is a schematic flowchart of a defect detection method for a turnout frog provided by an embodiment of the present application;

[0027] Figure 5 It is a specific example diagram of a defect detection method for a turnout frog provided by an embodiment of the present application;

[0028] Figure 6 This is a schematic structural diagram of a defect detection device for a turnout frog provided by an embodiment of the present application;

[0029] Figure 7 This is another schematic structural diagram of a defect detection system for a turnout frog provided by an embodiment of the present application. Detailed implementation manners

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some but not all of the embodiments of the present application. Usually, the components of the embodiments of the present application described and illustrated herein can be arranged and designed in various different configurations.

[0031] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0032] Currently, when detecting the defects of turnout frogs, it mainly relies on the hammering detection and penetrant inspection by inspection personnel. Among them, the traditional hammering inspection method involves inspection personnel using a hammer to strike the surface of the rail, and judging whether there are cracks or cavities in the rail based on the differences in the sounds emitted from different parts. A normal rail should emit a clear sound, while a place with cracks or damage will emit a dull or abnormal sound. This method has the advantages of simplicity, easy operation, and low cost. However, this hammering detection method overly relies on the experience and judgment ability of inspection personnel. Even experienced inspectors may miss some problems due to factors such as environmental noise or fatigue. In addition, the detection has great limitations. The hammering inspection is mainly used to identify surface defects and cannot effectively detect cracks or deep damage inside the rail. Due to the subjectivity of the method, it is impossible to quantitatively analyze the size, depth, or impact degree of the defects, so it is difficult to provide a scientific basis for whether repair or replacement is needed.

[0033] The traditional penetrant testing method is a non-destructive testing method commonly used to detect surface cracks. By applying a penetrant (usually a dye or fluorescent agent) to the surface of the rail, allowing it to penetrate into the cracks, then removing the excess penetrant on the surface, and applying a developer, the position and shape of the cracks are revealed by the visible dye or fluorescence. Penetrant testing can effectively detect surface cracks, especially when identifying surface defects, and has simple equipment, low cost, and relatively simple operation. However, this method also has some limitations: First, it is limited to surface defects. Penetrant testing can only detect cases where the crack has extended to the surface and cannot detect internal cracks or deep damage, so it cannot identify early damage. Second, it has high requirements for surface cleanliness. If there is oil, rust, or dirt on the track surface, it may affect the penetrability of the penetrant and result in missed detections. Third, it cannot perform quantitative analysis. Although penetrant testing can detect cracks, it cannot evaluate the size, depth, and specific impact degree of the cracks and lacks a comprehensive assessment of the crack development trend. Fourth, it is restricted by environmental factors. In special environments such as high temperature, low temperature, or humidity, the effect of penetrant testing may be affected, thus affecting its accuracy and reliability.

[0034] Based on this, the embodiment of this application provides a method for detecting defects in turnout frogs, which is applied to a detection terminal. The detection terminal is communicatively connected to a detection hammer. The detection hammer includes a microphone, an audio encoder, and a communication device. The method includes extracting a plurality of consecutive time-series hammering signals from the acoustic signal in response to the acoustic signal of the detection hammer hitting a target sampling point of the turnout frog. The acoustic signal is received by the detection hammer using the microphone, encoded by the audio encoder, and sent through the communication module. Based on the pulse energy of the hammering signals, valid hammering signals are identified from the plurality of consecutive time-series hammering signals. The valid hammering signal is used to indicate that the pulse energy of the hammering signal at the current time series meets a preset detection threshold compared with the pulse energy of the valid hammering signal at the previous time series. The Mel Frequency Cepstral Coefficient (MFCC) features are extracted from the valid hammering signals. The MFCC features are used to represent the characteristics of the valid hammering signals in the time domain and frequency domain. The MFCC features are input into a trained defect type recognition model to obtain the defect type of the turnout frog at the target sampling point.

[0035] The method for detecting defects in turnout frogs provided by the embodiment of this application obtains an acoustic signal by using a detection hammer and transmits it to the detection terminal. The detection terminal extracts, detects, and extracts Mel spectrum features from the collected acoustic signal, and then uses an embedded neural network processor to deploy a quantized defect detection neural network for inference. Finally, the processed defect detection result is displayed on the detection terminal, which can accurately identify the defect type at the point where the detection hammer hits without relying on the subjective experience judgment of the detection personnel, improving the automation degree of the detection process and the detection efficiency of the defect type of the turnout frog.

[0036] The implementation manners of the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0037] On the one hand, the embodiments of the present application provide a defect detection system for a turnout frog, as Figure 1 shown. The defect detection system 100 of the turnout frog may include a detection hammer 101 and a detection terminal 102.

[0038] Among them, the detection hammer 101 is used to collect the sound signal generated by the user knocking on the turnout frog, process the sound signal, and transmit it to the detection terminal 102.

[0039] Exemplarily, as Figure 2 shown, the detection hammer 101 may include a microphone, an audio encoder, a communication device, a hammer head, a hammer handle, a charging module, and a lithium battery. The user can hold the hammer handle and knock on the target sampling point of the turnout frog. The hammer handle collides with the turnout frog to generate a sound signal. The microphone collects the sound signal, and the audio encoder preprocesses the sound signal, converts the sound signal into digital data in a compressed or specific format, and sends the digital data to the detection terminal 102. The charging module and the lithium battery are used to provide power for the detection hammer 101.

[0040] For example, the microphone may be an acoustic sensor, the audio encoder may be a Codec audio encoder, and the communication device may be a Bluetooth module.

[0041] The detection terminal 102 is used to receive the sound signal sent by the detection hammer 101, and process the sound signal by using the defect detection method of the turnout frog provided by the embodiments of the present application to obtain the defect type of the turnout frog.

[0042] Exemplarily, as Figure 3 shown, the detection terminal 102 may include a processor, a communication module, a touch screen, a lithium battery, and a charging module. The detection terminal 102 can receive the sound signal sent by the detection hammer 101 through the communication module. The processor processes the sound signal by using the defect detection method of the turnout frog provided by the embodiments of the present application to obtain the defect type of the turnout frog, and displays the defect type of the turnout frog to the user through the touch screen.

[0043] For example, the processor may be an RK3588 main board, and the communication module may be Bluetooth.

[0044] It should be noted that the detection hammer 101 and the detection terminal 102 may also be connected in a wired manner, and the present application does not limit this. To implement the technical solutions involved in the present application, the number of the detection hammer 101 and the detection terminal 102 may be one or more, and the present application does not limit this.

[0045] On the one hand, an embodiment of the present application provides a method for detecting defects in a turnout frog, and this method can be executed by the detection device 102 in the turnout frog defect detection system 100 shown in Figure 1 As shown in Figure 4 This method may include the following steps.

[0046] S401, in response to the sound signal when the detection hammer strikes the target sampling point of the turnout frog, extract multiple consecutive time-sequence hammering signals from the sound signal.

[0047] Among them, the sound signal is received by the detection hammer using a microphone, encoded by an audio encoder, and sent through a communication module.

[0048] Exemplarily, the user uses the detection hammer to strike the turnout frog, the detection hammer collides with the turnout frog to generate a sound signal, the detection hammer uses a microphone to receive the sound signal, the detection hammer uses an audio encoder to compress the sound signal, and uploads the compressed sound signal to the detection terminal in real time through the communication module.

[0049] After receiving the sound signal, the detection terminal extracts multiple consecutive time-sequence hammering signals from the sound signal in the way of a time window.

[0050] Further, after the detection terminal extracts multiple consecutive time-sequence hammering signals from the sound signal, the detection terminal determines the pulse energy corresponding to the hammering signal in each time sequence.

[0051] A possible implementation method is as follows: for the target hammering signal in any time sequence among multiple consecutive time-sequence hammering signals, extract the target sub-hammering signal of the target hammering signal within a preset frequency range. Calculate the time-frequency analysis result of the target sub-hammering signal. The time-frequency analysis result is used to represent the hammering signal after the short-time Fourier transform of the target sub-hammering signal. According to the time-frequency analysis result, the number of time sub-bands of the target sub-hammering signal within the frequency range, and the number of frequency sub-bands of the target sub-hammering signal within the frequency range, determine the pulse energy of the target hammering signal.

[0052] Specifically, through the following formula, first calculate the time-frequency analysis result of the target sub-hammering signal to obtain the target sub-hammering signal X(τ,v) after the short-time Fourier transform, and calculate the square modulus |X(τ,v)| of the time-frequency analysis result based on this X(τ,v) 2 to obtain the energy density of the target sub-hammering signal at each time and frequency, and use the number of time sub-bands and frequency sub-bands to perform double integration and summation on time and frequency to determine the pulse energy of the target hammering signal.

[0053]

[0054] Among them, X(τ, v) is the time-frequency analysis result, τ is time, v is frequency, T is the number of time sub-bands, F is the number of frequency sub-bands, F1 is the lower limit of the frequency of the preset frequency range, and F2 is the upper limit of the frequency of the preset frequency range.

[0055] This process can effectively detect the signals with instantaneous changes in the pulse energy of the hammering signal by detecting the pulse energy of the hammering signal. Furthermore, by presetting the signals within a specific frequency range, the high-frequency resonance characteristics of the metal cracks in the turnout frog can be effectively screened out from all the sound signals of the turnout frog. For example, the pseudo-impulse signals generated by the vibration when the train passes or the interference signals generated by human walking can be screened out.

[0056] It should be noted that the upper limit F2 and the lower limit F1 of the preset frequency range can be determined based on the frequency bands of historical turnout frogs, and this application does not limit this.

[0057] S402. Based on the pulse energy of the hammering signal, identify the effective hammering signal from multiple consecutive time-sequence hammering signals.

[0058] Among them, the effective hammering signal is used to indicate that the pulse energy of the hammering signal at the current time sequence satisfies the preset detection threshold compared with the pulse energy of the effective hammering signal at the previous time sequence.

[0059] A possible implementation manner is that after determining the pulse energy of the hammering signal at each time sequence by using the steps in S401, obtain the pulse energy of the effective hammering signal at the previous time sequence of the hammering signal at the current time sequence. If the difference between the logarithm of the pulse energy of the hammering signal at the current time sequence and the logarithm of the pulse energy of the effective hammering signal at the previous time sequence is less than the preset detection threshold, determine that the hammering signal at the current time sequence is an effective hammering signal.

[0060] Exemplarily, the following formula is used to test the effectiveness of the hammering signal.

[0061] logE k -logE k-1 >γ.

[0062] Among them, γ is the detection threshold of the hammering signal. This detection threshold γ can be calibrated through experiments.

[0063] This process can screen out abnormal signals from multiple time-sequence hammering signals, such as abnormal signals caused by hitting foreign objects or deviation of the target sampling point during hitting, so as to ensure the effectiveness of the hammering signal acquisition during the defect detection process of the turnout frog.

[0064] Furthermore, if the detection terminal detects effective hammering signals within multiple consecutive time sequences, the detection threshold γ can be updated in real time through multiple effective hammering signals.

[0065] S403. Extract the Mel Frequency Cepstral Coefficient (MFCC) features from the effective hammering signal.

[0066] Among them, the MFCC features are used to represent the characteristics of the effective hammering signal in the time domain and frequency domain.

[0067] A possible implementation is to input the effective hammering signal into the Mel filter bank to obtain the Mel spectrum of the effective hammering signal. Perform logarithmic transformation and discrete cosine transformation on the Mel spectrum of the effective hammering signal, and extract the MFCC features from the Mel spectrum.

[0068] Specifically, the MFCC features are extracted through the following formula.

[0069] X MFCC (τ, v m ) = DCT(log(Mel(|X(τ, v)| 2 ))).

[0070] Among them, τ = 1, 2, …, N τ,MFCC , v m = 1, 2, …, N v,MFCC。

[0071] Among them, Mel(|X(τ, v)| 2 ) is the Mel spectrum of the effective hammering signal. log(Mel(|X(τ, v)| 2 )) is used to represent the result of the logarithmic transformation. DCT(log(Mel(|X(τ, v)| 2 ))) is used to represent the result of the discrete cosine transformation.

[0072] In this process, by passing the effective hammering signal through the Mel filter bank to obtain the Mel spectrum, and then through logarithmic transformation and discrete cosine transformation, the correlation in the feature vector can be removed, thereby extracting the MFCC features in the effective primary signal. Furthermore, through feature normalization, robust MFCC features are obtained.

[0073] S404. Input the MFCC features into the trained defect type recognition model to obtain the defect type of the turnout frog at the target sampling point.

[0074] A possible implementation is to input the MFCC features into a 2D convolutional network to obtain the voiceprint features of the effective hammering signal. Input the voiceprint features into a multi-layer perceptron network classifier to determine the defect classification result of the turnout frog at the target sampling point. The defect classification result includes at least one defect type and the probability corresponding to the defect type. Determine the defect type with the highest probability in the defect classification result as the defect type of the target sampling point.

[0075] Exemplarily, such as Figure 5As shown, the convolutional network uses a 2D convolutional network with a convolutional kernel size of 2*2, and the number of convolutional layers of the 2D convolutional network is N layer .

[0076] The MFCC features are input into the 2D convolutional network as a single-channel image. After N layer layers of convolution by the 2D convolutional network and processed by the rectified linear unit (ReLU) activation function, N Feats -dimensional voiceprint features of the hammering signal are extracted.

[0077] Then, the N Feats -dimensional voiceprint features are input into a multi-layer perceptron (MLP) classifier, and multi-classification probabilities are output through the Softmax layer, such as the probabilities of normal, crack, casting defect, wear, looseness, etc. The output result can support multi-label determination, such as the coexistence of cracks and wear, or the coexistence of cracks and corrosion, etc.

[0078] The defect detection method for turnout frog provided in the embodiment of the present application obtains a sound signal by using a detection hammer and transmits it to a detection terminal. The detection terminal extracts, detects, and extracts Mel-spectrum features from the collected sound signal, and then uses an embedded neural network processor to deploy the quantized defect detection neural network for inference. Finally, the processed defect detection result is displayed on the detection terminal. Without relying on the subjective experience judgment of the detection personnel, the defect type at the point where the detection hammer strikes can be accurately identified, improving the automation degree of the detection process and the detection efficiency of the defect type of the turnout frog. At the same time, the defect detection method for turnout frog provided in the embodiment of the present application is not affected by environmental factors such as temperature and humidity, and can detect the defect problems of the steel rail comprehensively and without dead angles, ensuring the comprehensiveness and accuracy of the detection result.

[0079] The above mainly introduces the solution provided in the embodiment of the present application from the perspective of the working principle of the device. It can be understood that in order to implement the above functions, the defect detection device for turnout frog includes the corresponding hardware structure and / or software module for executing each function. Those skilled in the art should easily realize that, combining the algorithm steps of each example described in the embodiments disclosed in this article, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the way of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0080] The embodiments of the present application can divide the functional modules of the turnout frog defect detection device according to the above method examples. For example, each functional module can be corresponding to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module.

[0081] It should be noted that the division of modules in the embodiments of the present application is illustrative, only a logical function division, and there may be other division methods in actual implementation. In the case of dividing each functional module corresponding to each function, Figure 6 FIG. shows a possible schematic composition diagram of the turnout frog defect detection device involved in the above and embodiments. As Figure 6 shown, the turnout frog defect detection device 600 may include: an extraction module 601, an identification module 602, and a processing module 603.

[0082] Among them, the extraction module 601 is used to support the turnout frog defect detection device 600 to execute Figure 1 S401 and S403 in the schematic turnout frog defect detection method.

[0083] The identification module 602 is used to support the turnout frog defect detection device 600 to execute Figure 1 S402 in the schematic turnout frog defect detection method.

[0084] The processing module 603 is used to support the turnout frog defect detection device 600 to execute Figure 1 S404 in the schematic turnout frog defect detection method.

[0085] A possible implementation manner is that the device can also be used to extract, for any target hammering signal in any time sequence among a plurality of consecutive time-sequence hammering signals, a target sub-hammering signal within a preset frequency range of the target hammering signal, calculate the time-frequency analysis result of the target sub-hammering signal, where the time-frequency analysis result is used to represent the hammering signal after the short-time Fourier transform of the target sub-hammering signal, and determine the pulse energy of the target hammering signal according to the time-frequency analysis result, the number of time sub-bands of the target sub-hammering signal within the frequency range, and the number of frequency sub-bands of the target sub-hammering signal within the frequency range.

[0086] A possible implementation manner is that the device can also be used to determine the pulse energy of the target hammering signal according to the time-frequency analysis result, the number of time sub-bands of the target sub-hammering signal within the frequency range, and the number of frequency sub-bands of the target sub-hammering signal within the frequency range, including: Among them, X(τ, v) is the time-frequency analysis result, τ is time, v is frequency, T is the number of time sub-bands, F is the number of frequency sub-bands, F1 is the lower limit of the frequency in the preset frequency range, and F2 is the upper limit of the frequency in the preset frequency range.

[0087] In a possible implementation, the device can also be used to obtain the pulse energy of the effective hammering signal in the previous time sequence of the hammering signal in the current time sequence. If the difference between the logarithm of the pulse energy of the hammering signal in the current time sequence and the logarithm of the pulse energy of the effective hammering signal in the previous time sequence is less than a preset detection threshold, it is determined that the hammering signal in the current time sequence is an effective hammering signal.

[0088] In a possible implementation, the device can also be used to input the effective hammering signal into a Mel filter bank to obtain the Mel spectrum of the effective hammering signal. Perform logarithmic transformation and discrete cosine transformation on the Mel spectrum of the effective hammering signal, and extract MFCC features from the Mel spectrum.

[0089] In a possible implementation, the device can also be used to perform logarithmic transformation and discrete cosine transformation on the Mel spectrum of the effective hammering signal, and extract MFCC features from the Mel spectrum, including: X MFCC (τ, v m ) = DCT(log(Mel(|X(τ, v)| 2 ))), τ = 1, 2, …, N τ,MFCC , v m =

[0090] 1, 2, …, N v,MFCC . Among them, Mel(|X(τ, v)| 2 ) is the Mel spectrum of the effective hammering signal. log(Mel(|X(τ, v)| 2 )) is used to represent the result of the logarithmic transformation. DCT(log(Mel(|X(τ, v)| 2 ))) is used to represent the result of the discrete cosine transformation.

[0091] In a possible implementation, the device can also be used to input the MFCC features into a 2D convolutional network to obtain the voiceprint features of the effective hammering signal. Input the voiceprint features into a multi-layer perceptron network classifier to determine the defect classification result of the turnout frog at the target sampling point. The defect classification result includes at least one defect type and the probability corresponding to the defect type. Determine the defect type with the highest probability in the defect classification result as the defect type of the target sampling point.

[0092] It should be noted that all relevant contents of the steps involved in the above method embodiments can be cited in the function descriptions of the corresponding functional modules, and will not be elaborated here.

[0093] The defect detection device 600 for the turnout frog provided by the embodiment of the present application is used to execute the above-mentioned Figure 4 shown defect detection method for the turnout frog, so the same effect as the above-mentioned defect detection method for the turnout frog can be achieved.

[0094] The embodiment of the present application also provides a defect detection device for the turnout frog. The defect detection device for the turnout frog can be the Figure 1 shown detection terminal 102 in the foregoing embodiment. The defect detection device for the turnout frog can execute the defect detection method and related steps for the turnout frog in the above method embodiment.

[0095] The embodiment of the present application also provides a computer-readable storage medium, on which instructions are stored. When the instructions are executed, the defect detection method and related steps for the turnout frog in the above method embodiment are executed.

[0096] The embodiment of the present application also provides a computer program product. When the computer program product runs on a computer, the computer is enabled to execute the defect detection method and related steps for the turnout frog in the above method embodiment.

[0097] In some embodiments, the method shown in the present application can be implemented as computer program instructions encoded in a computer-readable storage medium in a machine-readable format or encoded in other non-transitory media or articles.

[0098] The embodiment of the present application also provides a defect detection system 100 for the turnout frog, as Figure 7 shown. The defect detection system 100 for the turnout frog includes at least one processor 701 and at least one interface circuit 702.

[0099] As an example, when the defect detection system 100 for the turnout frog includes one processor and one interface circuit, the one processor can be the Figure 7 processor 701 shown by the solid line box (or the processor 701 shown by the dashed line box) in Figure 7 , and the one interface circuit can be the Figure 7 interface circuit 702 shown by the solid line box (or the interface circuit 702 shown by the dashed line box) in Figure 7 . When the defect detection system 100 for the turnout frog includes two processors and two interface circuits, the two processors include the Figure 7 processor 701 shown by the solid line box and the processor 701 shown by the dashed line box in Figure 7 , and the two interface circuits include the Figure 7 interface circuit 702 shown by the solid line box and the interface circuit 702 shown by the dashed line box in Figure 7 . There is no limitation on this.

[0100] The processor 701 and the interface circuit 702 can be interconnected by a line. For example, the interface circuit 702 can be used to receive signals. Also, for example, the interface circuit 702 can be used to send signals to other devices (such as the processor 701). By way of example, the interface circuit 702 can read the computer instructions stored in the memory and send the computer instructions to the processor 701. The processor 701 executes the instructions and, in combination with the input / output devices, implements each step in the above embodiments, such as implementing Figure 4 or each step performed in the method embodiments shown in FIGS. 5. Of course, the chip system may also include other discrete devices, and the embodiments of the present application do not make specific limitations thereto.

[0101] From the description of the above embodiments, those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above division of each functional module is used as an example. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.

[0102] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be in an electrical, mechanical or other form.

[0103] The units described as separate components may or may not be physically separated. The components displayed as units may be one physical unit or multiple physical units, that is, they can be located in one place, or they can be distributed to multiple different places. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0104] In addition, each functional unit in the various embodiments of the present application can be integrated in one processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0105] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiments of the present application, in essence, or the part that makes a contribution, or all or part of the technical solution, can be embodied in the form of a software product. The software product is stored in a storage medium and includes several instructions for causing a device (which can be a single-chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods described in the embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0106] The above content is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claimed rights.

Claims

1. A defect detection method for a turnout frog, characterized in that Applied to a detection terminal, the detection terminal is communicatively connected to a detection hammer, and the detection hammer includes a microphone, an audio encoder, and a communication device; the method includes: In response to a sound signal of the detection hammer hitting a target sampling point of a turnout frog, extracting a plurality of consecutive time-series hammering signals from the sound signal; the sound signal is received by the detection hammer using the microphone, encoded by the audio encoder, and sent through the communication module; Based on the pulse energy of the hammering signals, identifying valid hammering signals from the plurality of consecutive time-series hammering signals; the valid hammering signals are used to indicate that the pulse energy of the hammering signal in the current time series satisfies a preset detection threshold compared to the pulse energy of the valid hammering signal in the previous time series; Extracting Mel Frequency Cepstral Coefficient (MFCC) features from the valid hammering signals; the MFCC features are used to represent the features of the valid hammering signals in the time domain and the frequency domain; Inputting the MFCC features into a trained defect type recognition model to obtain the defect type of the turnout frog at the target sampling point.

2. The method according to claim 1, characterized in that, The method further includes: For a target hammering signal in any time series among the plurality of consecutive time-series hammering signals, extracting a target sub-hammering signal of the target hammering signal within a preset frequency range; Calculating the time-frequency analysis result of the target sub-hammering signal; the time-frequency analysis result is used to represent the hammering signal after the short-time Fourier transform of the target sub-hammering signal; Determining the pulse energy of the target hammering signal according to the time-frequency analysis result, the number of time sub-bands of the target sub-hammering signal within the frequency range, and the number of frequency sub-bands of the target sub-hammering signal within the frequency range.

3. The method according to claim 2, wherein The determining the pulse energy of the target hammering signal according to the time-frequency analysis result, the number of time sub-bands of the target sub-hammering signal within the frequency range, and the number of frequency sub-bands of the target sub-hammering signal within the frequency range includes: Where X(τ, v) is the time-frequency analysis result, τ is time, v is frequency, T is the number of time sub-bands, F is the number of frequency sub-bands, F1 is the lower frequency limit of the preset frequency range, and F2 is the upper frequency limit of the preset frequency range.

4. The method according to claim 1, characterized in that The identifying valid hammering signals from the plurality of consecutive time-series hammering signals based on the pulse energy of the hammering signals includes: Obtaining the pulse energy of the valid hammering signal in the previous time series of the hammering signal in the current time series; If the difference between the logarithm of the pulse energy of the hammering signal in the current time series and the logarithm of the pulse energy of the valid hammering signal in the previous time series is less than the preset detection threshold, determining that the hammering signal in the current time series is the valid hammering signal.

5. The method according to claim 1, characterized in that, The extracting Mel Frequency Cepstral Coefficient (MFCC) features from the valid hammering signals includes: Inputting the valid hammering signal into a Mel filter bank to obtain the Mel spectrum of the valid hammering signal; Performing a logarithmic transformation and a discrete cosine transformation on the Mel spectrum of the valid hammering signal, and extracting the MFCC features from the Mel spectrum.

6. The method according to claim 5, wherein Performing logarithmic transformation and discrete cosine transformation on the Mel spectrum of the effective hammering signal, and extracting the MFCC features from the Mel spectrum, including: X MFCC (τ, v m ) = DCT(log(Mel(|X(τ, v)| 2 ))), τ = 1, 2, …, N τ,MFCC , v m = 1, 2, …, N v,MFCC ; Among them, Mel(|X(τ,v)| 2 ) is the Mel spectrum of the effective hammering signal; log9Mel(|X(τ,v)| 2 )) is used to represent the result of the logarithmic transformation; DCT(log(Mel(|X(τ,v)| 2 ))) is used to represent the result of the discrete cosine transform.

7. The method according to claim 1, wherein Inputting the MFCC features into a trained defect type recognition model to obtain the defect type of the turnout frog at the target sampling point, including: Inputting the MFCC features into a 2D convolutional network to obtain the voiceprint features of the effective hammering signal; Inputting the voiceprint features into a multi-layer perceptron network classifier to determine the defect classification result of the turnout frog at the target sampling point; the defect classification result includes at least one defect type and the probability corresponding to the defect type; Determining the defect type with the highest probability in the defect classification result as the defect type of the target sampling point.

8. A defect detection device for a turnout frog, characterized in that Applied to a detection terminal, the detection terminal is communicatively connected to a detection hammer, and the detection hammer includes a microphone, an audio encoder, and a communication device; the device includes: An extraction module, configured to, in response to a sound signal of the detection hammer hitting a target sampling point of the turnout frog, extract a plurality of consecutive time-series hammering signals from the sound signal; the sound signal is received by the detection hammer using the microphone, encoded by the audio encoder, and sent through the communication module; An identification module, configured to identify an effective hammering signal from a plurality of consecutive time-series hammering signals based on the pulse energy of the hammering signal; the effective hammering signal is used to indicate that the pulse energy of the hammering signal at the current time series satisfies a preset detection threshold with the pulse energy of the effective hammering signal at the previous time series; The extraction module is further configured to extract Mel Frequency Cepstral Coefficient (MFCC) features from the effective hammering signal; the MFCC features are used to represent the features of the effective hammering signal in the time domain and the frequency domain; A processing module, configured to input the MFCC features into a trained defect type recognition model to obtain the defect type of the turnout frog at the target sampling point.

9. A defect detection device for a turnout frog, characterized in that, The defect detection device for the turnout frog includes a processor and a memory, the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the defect detection method for the turnout frog according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions cause the processor to implement the defect detection method for the turnout frog according to any one of claims 1 to 7.

Images