Railway turnout health state analysis method, device, equipment, medium and program

Abstract

The invention relates to a railway turnout health state analysis method and device, equipment, a storage medium and a computer program, and the method comprises the steps: collecting the operation process data of a railway turnout in real time through a preset sensor, carrying out the time domain analysis of the operation process data, obtaining a time domain component, carrying out the frequency domain analysis of the time domain component, and obtaining the health state of the railway turnout. And obtaining a frequency domain component, calculating a power density spectrum according to the frequency domain component to obtain operation data characteristics, and outputting prediction probabilities corresponding to different health levels of the railway turnout according to the operation data characteristics by using a pre-trained railway turnout state analysis model.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a railway turnout health status analysis method, device, equipment, storage medium and computer program. Background Art

[0002] Turnouts are key facilities for realizing train switching. Their poor condition may lead to poor switching of switch machines and loose fit of point rails, which may lead to turnout loss of gauge, switch squeezing and other incidents, affecting operational safety.

[0003] Existing status assessment and analysis of turnouts mostly involves manual on-site data monitoring and related analysis to assess the health status of railway turnouts. This method is costly and inefficient. Summary of the invention

[0004] The present invention provides a railway turnout health status analysis method, device, equipment, storage medium and computer program to solve the problem of high cost and low efficiency of existing railway turnout status analysis methods.

[0005] In a first aspect, the present disclosure provides a method for analyzing the health status of a railway turnout, comprising:

[0006] Use preset sensors to collect the operation process data of railway turnouts in real time;

[0007] Performing time domain analysis on the operation process data to obtain a time domain component;

[0008] Performing frequency domain analysis on the time domain component to obtain a frequency domain component;

[0009] Calculate the power density spectrum according to the frequency domain components to obtain the operation data characteristics;

[0010] Outputting prediction probabilities corresponding to different health levels of railway turnouts according to the operation data characteristics using a pre-trained railway turnout state analysis model;

[0011] The health level with the highest predicted probability is selected as the final health status analysis result of the railway turnout.

[0012] In some embodiments, the real-time collection of railway turnout operation process data using a preset sensor includes:

[0013] Using sensors preset in railway turnouts to obtain turnout position information, turnout acceleration change information, turnout pressure information, and turnout indicating rod position change value in real time, and obtain a turnout operation information set;

[0014] De-noising the turnout operation information set to obtain a de-noised operation information set;

[0015] The denoised operation information set is normalized using the following formula to obtain the operation process data:

[0016]

[0017] Wherein, X is the operation process data, F is the number of samples of the denoised operation information set, and x i represents the i-th data in the denoising operation information set, x max Indicates the maximum value in the denoising operation information set.

[0018] In some embodiments, performing time domain analysis on the operation process data to obtain the time domain component includes:

[0019] The following formula is used to perform time domain analysis on the operation process data:

[0020]

[0021] Wherein, R is the time domain component, N is the number of samples of the running process data, and x i Represents the vibration amplitude value of the i-th sample in the running process data.

[0022] In some embodiments, performing frequency domain analysis on the time domain component to obtain the frequency domain component includes:

[0023]

[0024] Wherein, X[k] is the complex number representation of the frequency domain component, M is the number of sample points of the time domain component, and R n Represents the amplitude of the nth signal in the time domain component.

[0025] In some embodiments, the method further includes: before outputting the prediction probabilities corresponding to different health levels of the railway turnouts according to the operation data characteristics using the pre-trained railway turnout state analysis model, the method further includes:

[0026] Obtain a training feature set and a test feature set with real health level annotations;

[0027] Establish and preset a neural network model with health level as output;

[0028] Outputting the prediction results of the health level according to the training feature set using the neural network model, taking the health level with the highest probability in the prediction results as the prediction output, and obtaining a prediction set;

[0029] The loss value of each prediction output of the prediction set and its corresponding health level label is calculated using the following formula to obtain a loss value set:

[0030]

[0031] Among them, L i is the i-th loss value in the loss value set, Describe the i-th prediction output in the prediction set, Y i Describe the true health level label corresponding to the i-th prediction output in the prediction set;

[0032] The weight parameters of the neural network model are updated according to the loss value set and the preset learning rate using the following formula:

[0033]

[0034] Among them, W * is the weight parameter of the neural network model after updating, W is the weight parameter of the neural network model before updating, μ is the learning rate, D is the number of samples of the loss value set, L i is the i-th loss value in the loss value set;

[0035] When the preset number of iterations is reached, the regression rate test of the neural network model is performed using the test feature set;

[0036] If the neural network model fails the regression rate test, adjusting the learning rate of the neural network model, returning to use the neural network model of last month to predict the health level according to the training feature set, taking the health level with the highest probability in the prediction results as the prediction output, and obtaining the prediction set;

[0037] If the neural network model passes the regression rate test, it is confirmed that the railway switch state analysis model is obtained.

[0038] In some embodiments, the method of using a pre-trained railway turnout state analysis model to output prediction probabilities corresponding to different health levels of railway turnouts according to the operation data characteristics includes:

[0039] Using the convolution layer of the railway turnout state analysis model to extract local features of the operation data features to obtain convolution features;

[0040] Using the pooling layer of the railway turnout state analysis model to extract the local average value of the convolution feature, so as to obtain the pooling feature;

[0041] Using the activation layer of the railway turnout state analysis model to add nonlinear factors to the pooling feature, so as to obtain an activation feature;

[0042] The fully connected layer of the railway turnout state analysis model is used to output the predicted probability corresponding to each health level according to the activation features and the weight parameters and bias items of the railway turnout state analysis model.

[0043] In a second aspect, the present disclosure provides a railway turnout health status analysis device, comprising:

[0044] A data acquisition module is used to collect the operation process data of the railway turnout in real time using preset sensors;

[0045] A data analysis module, configured to perform time domain analysis on the operation process data to obtain time domain components, perform frequency domain analysis on the time domain components to obtain frequency domain components, and calculate a power density spectrum according to the frequency domain components to obtain operation data characteristics;

[0046] A probability prediction module, used to output prediction probabilities corresponding to different health levels of railway turnouts according to the operation data characteristics using a pre-trained railway turnout state analysis model;

[0047] The result confirmation module selects the health level with the highest prediction probability as the final health status analysis result of the railway turnout.

[0048] In a third aspect, the present disclosure provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the above aspects.

[0049] In a fourth aspect, the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method described in the above aspects when executed by a processor.

[0050] In a fifth aspect, the present disclosure provides a computer program product, including a computer program / instructions, which implements the steps of the method described in the above aspects when the computer program is executed by a processor.

[0051] The present disclosure provides a railway turnout health status analysis method, device, equipment, storage medium and computer program, which collects the operation process data of the railway turnout in real time by using a preset sensor, performs time domain analysis on the operation process data to obtain time domain components, performs frequency domain analysis on the time domain components to obtain frequency domain components, calculates the power density spectrum according to the frequency domain components to obtain the operation data characteristics, uses a pre-trained railway turnout state analysis model to output the prediction probabilities corresponding to different health levels of the railway turnout according to the operation data characteristics, and selects the health level with the highest prediction probability as the final health status analysis result of the railway turnout. This solves the problem of high cost and low efficiency of the existing railway turnout state analysis method, and improves the efficiency of health status analysis of railway turnouts.

[0052] 1. De-noising is performed on the turnout operation information set, and the technical effect is to improve the quality of the data and the accuracy of subsequent data analysis.

[0053] 2. The denoising operation information set is normalized, and the technical effect is to improve the efficiency of data analysis.

[0054] 3. Using the pre-trained railway turnout status analysis model, the prediction probabilities corresponding to different health levels of the railway turnout are output according to the operating data characteristics. The technical effect is to improve the efficiency of health status analysis of the railway turnout. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The present disclosure will be described in more detail below based on embodiments and with reference to the accompanying drawings:

[0056] Figure 1 A schematic diagram of a flow chart of a railway turnout health status analysis method provided in an embodiment of the present disclosure;

[0057] Figure 2 A schematic diagram of a flow chart for outputting prediction probabilities corresponding to different health levels of railway turnouts provided in an embodiment of the present disclosure;

[0058] Figure 3 A functional module diagram of a railway turnout health status analysis device provided for an embodiment of the present disclosure. In the accompanying drawings, the same components are marked with the same reference numerals, and the drawings are not drawn according to the actual scale. DETAILED DESCRIPTION

[0059] In order to enable those skilled in the art to better understand the technical solution of the present disclosure, and to fully understand and implement how the present disclosure applies technical means to solve technical problems and achieve the corresponding technical effects, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only embodiments of a part of the present disclosure, not all of the embodiments. The embodiments of the present disclosure and the various features in the embodiments can be combined with each other without conflict, and the technical solutions formed are all within the scope of protection of the present disclosure. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present disclosure.

[0060] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0061] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0062] Embodiment 1

[0063] Figure 1 The following is a flow chart of a method for analyzing the health status of a railway turnout provided in an embodiment of the present disclosure. Figure 1 As shown, a smart device control method includes:

[0064] S1. Use preset sensors to collect the operation process data of railway turnouts in real time.

[0065] In the embodiment of the present invention, the operation process data of the railway turnout is collected in real time by using preset sensors, which is achieved by deploying position sensors, acceleration sensors, pressure sensors and displacement sensors at key parts of the railway turnout. The relevant data of the railway turnout is obtained in real time by the deployed sensors.

[0066] In detail, the position sensor is used to obtain the position information of the turnout in real time, the acceleration sensor is used to obtain the acceleration change of the turnout in real time, the pressure sensor is used to obtain the pressure data of the turnout in real time, and the displacement sensor is used to obtain the displacement change value of the turnout indicating rod in real time.

[0067] In the embodiment of the present invention, the real-time collection of the operation process data of the railway turnout by using a preset sensor includes:

[0068] Using sensors preset in railway turnouts to obtain turnout position information, turnout acceleration change information, turnout pressure information, and turnout indicating rod position change value in real time, and obtain a turnout operation information set;

[0069] De-noising the turnout operation information set to obtain a de-noised operation information set;

[0070] The denoised operation information set is normalized using the following formula to obtain the operation process data:

[0071]

[0072] Wherein, X is the operation process data, F is the number of samples of the denoised operation information set, and x i represents the i-th data in the denoising operation information set, x max Indicates the maximum value in the denoising operation information set.

[0073] In detail, the denoising process of the turnout operation information set refers to the process of removing or weakening those irrelevant, unexpected noises or interference signals introduced due to measurement errors, transmission errors, system defects, etc. from the turnout operation information set. These noises may exist in various forms, such as outliers, random errors, duplicate data, missing values, etc., which will interfere with the authenticity and accuracy of the data, thereby affecting the effects of subsequent steps such as data analysis, data mining and machine learning.

[0074] In detail, the denoising process for the turnout operation information set may be performed by wavelet transform.

[0075] In the embodiment of the present invention, by using preset sensors to collect the operation process data of the railway turnout in real time, the efficiency and accuracy of the subsequent health status analysis of the railway turnout are improved.

[0076] S2. Perform time domain analysis on the operation process data to obtain time domain components.

[0077] In the embodiment of the present invention, the time domain analysis of the operation process data is to analyze the characteristics of the operation process data in the time dimension, which can reveal the periodic characteristics of the operation process data.

[0078] In the embodiment of the present invention, performing time domain analysis on the operation process data to obtain the time domain component includes:

[0079] The following formula is used to perform time domain analysis on the operation process data:

[0080]

[0081] Wherein, R is the time domain component, N is the number of samples of the running process data, and x i Represents the vibration amplitude value of the i-th sample in the running process data.

[0082] In detail, the vibration amplitude value of the i-th sample refers to the difference between the value of the i-th sample in the operation process data and the average value of the operation process data.

[0083] In the embodiment of the present invention, by performing time domain analysis on the operation process data, the efficiency of subsequent frequency domain analysis on the time domain components is improved.

[0084] S3. Perform frequency domain analysis on the time domain component to obtain a frequency domain component.

[0085] In the embodiment of the present invention, the frequency domain analysis of the time domain component is to analyze the characteristics of the time domain component in the frequency dimension, which can reveal the frequency distribution of data.

[0086] In the embodiment of the present invention, performing frequency domain analysis on the time domain component to obtain the frequency domain component includes:

[0087]

[0088] Wherein, X[k] is the complex number representation of the frequency domain component, M is the number of sample points of the time domain component, and R n Represents the amplitude of the nth signal in the time domain component.

[0089] S4. Calculate the power density spectrum according to the frequency domain components to obtain the operation data characteristics.

[0090] In the embodiment of the present invention, the power density spectrum is a function that describes the power distribution of a signal or a time series at different frequencies. It is an estimation of the signal power spectrum and is usually used to analyze the frequency components of a signal.

[0091] In the embodiment of the present invention, the step of calculating the power density spectrum according to the frequency domain component to obtain the operating data feature includes:

[0092] The power density spectrum is calculated using the following formula:

[0093]

[0094] Wherein, P is the power density spectrum, H represents the number of sample points of the frequency domain component, and X k represents the amplitude of the tth component in the frequency domain component.

[0095] In the embodiment of the present invention, the operating data characteristics are obtained by calculating the power density spectrum according to the frequency domain components, thereby improving the efficiency of subsequent health level analysis of the railway turnout according to the operating data characteristics.

[0096] S5. Using a pre-trained railway turnout state analysis model, output prediction probabilities corresponding to different health levels of the railway turnout according to the operation data characteristics.

[0097] In an embodiment of the present invention, the railway switch state analysis model is a neural network model, including but not limited to a convolutional neural network model and a recurrent neural network model.

[0098] In the embodiment of the present invention, the different health levels are pre-set health level divisions, specifically including three levels: failure, sub-health, and health.

[0099] In the embodiment of the present invention, the method further includes: before outputting the prediction probabilities corresponding to different health levels of the railway turnouts according to the operation data characteristics using the pre-trained railway turnout state analysis model, the method further includes:

[0100] Obtain training feature sets and test feature sets with true health level annotations;

[0101] Establish and preset a neural network model with health level as output;

[0102] Outputting the prediction results of the health level according to the training feature set using the neural network model, taking the health level with the highest probability in the prediction results as the prediction output, and obtaining a prediction set;

[0103] The loss value of each prediction output of the prediction set and its corresponding health level label is calculated using the following formula to obtain a loss value set:

[0104]

[0105] Among them, L i is the i-th loss value in the loss value set, Describe the i-th prediction output in the prediction set, Y i Describe the true health level label corresponding to the i-th prediction output in the prediction set;

[0106] The weight parameters of the neural network model are updated according to the loss value set and the preset learning rate using the following formula:

[0107]

[0108] Among them, W * is the weight parameter of the neural network model after updating, W is the weight parameter of the neural network model before updating, μ is the learning rate, D is the number of samples of the loss value set, L i is the i-th loss value in the loss value set;

[0109] When the preset number of iterations is reached, the regression rate test of the neural network model is performed using the test feature set;

[0110] If the neural network model fails the regression rate test, adjusting the learning rate of the neural network model, returning to use the neural network model of last month to predict the health level according to the training feature set, taking the health level with the highest probability in the prediction results as the prediction output, and obtaining the prediction set;

[0111] If the neural network model passes the regression rate test, it is confirmed that the railway switch state analysis model is obtained.

[0112] In detail, the training feature set and the test feature set with real health level annotations are obtained by acquiring historical operation data of railway turnouts, extracting features to obtain feature sets, and dividing the feature sets into training feature sets and test feature sets.

[0113] In detail, the loss value is an indicator for measuring the difference between the predicted result and the true result.

[0114] In detail, the weight parameter is a parameter that determines the connection strength between each neural unit in the neural network model.

[0115] In detail, the learning rate is a parameter that determines the step size of updating the weight parameters.

[0116] In the embodiment of the present invention, refer to Figure 2 As shown, a flow chart of outputting prediction probabilities corresponding to different health levels of railway turnouts is provided in an embodiment of the present invention, wherein the prediction probabilities corresponding to different health levels of railway turnouts are outputted according to the operation data characteristics using a pre-trained railway turnout state analysis model, including:

[0117] S21, using the convolution layer of the railway turnout state analysis model to perform local feature extraction on the operation data features to obtain convolution features;

[0118] S22, using the pooling layer of the railway turnout state analysis model to extract the local average value of the convolution feature to obtain a pooling feature;

[0119] S23, using the activation layer of the railway turnout state analysis model to add nonlinear factors to the pooling feature to obtain an activation feature;

[0120] S24. Utilize the fully connected layer of the railway turnout state analysis model to output the prediction probability corresponding to each health level according to the activation features and the weight parameters and bias items of the railway turnout state analysis model.

[0121] In detail, the convolution layer of the railway turnout state analysis model is used to extract local features of the operation data features to obtain convolution features, which is to use a preset convolution kernel to slide on the operation data features according to a preset step size, calculate the dot product of the convolution kernel and the coverage area data after each sliding, and sum all dot products to obtain the convolution features. The convolution kernel is a matrix with a size much smaller than the input data.

[0122] In detail, the pooling layer of the railway switch state analysis model is used to extract the local average of the convolution feature to obtain the pooling feature, and a preset rectangular window is used to slide on the operation data feature according to a preset step size, and the average value of the area covered by the rectangular window after each sliding is calculated. The pooling feature is obtained by summing up all the average values.

[0123] In detail, the activation layer of the railway turnout state analysis model is used to add nonlinear factors to the pooling features to obtain the activation features, which is to add nonlinear factors to the pooling features using a pre-set activation function. Common activation functions include Sigmoid function, ReLU function and Softmax function.

[0124] In detail, the bias term is a constant added to the input of the activation function of the neural network neuron, which provides the model with the ability to translate, that is, it can adjust the position of the output of the activation function on the y-axis, thereby improving the flexibility of the model.

[0125] In the embodiment of the present invention, by using a pre-trained railway turnout state analysis model to output prediction probabilities corresponding to different health levels of the railway turnout according to the operating data characteristics, the efficiency of obtaining the final health state analysis result of the railway turnout is improved.

[0126] S6. Select the health level with the highest prediction probability as the final health status analysis result of the railway turnout.

[0127] In the embodiment of the present invention, since the railway turnout state analysis model outputs the probability corresponding to each health level of the railway turnout, it is necessary to select the health level with the highest predicted probability as the final health state analysis result.

[0128] In the embodiment of the present invention, after selecting the health level with the highest prediction probability as the final health status analysis result of the railway turnout, the railway turnout can be manually inspected to analyze the true health status of the railway turnout to ensure the safe operation of the railway turnout.

[0129] In an embodiment of the present invention, after the railway turnout status analysis model outputs the predicted probabilities corresponding to different health levels of the railway turnout, the efficiency of health status analysis of the railway turnout is improved by selecting the health level with the highest predicted probability as the final health status analysis result of the railway turnout.

[0130] Embodiment 2

[0131] like Figure 3 , which is a functional module diagram of a railway turnout health status analysis device provided in this embodiment.

[0132] The railway turnout health status analysis device 100 of the present invention can be installed in an electronic device. According to the functions to be implemented, the railway turnout health status analysis device 100 can include a data acquisition module 101, a data analysis module 102, a probability prediction module 103 and a result confirmation module 104. The module of the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0133] In this embodiment, the functions of each module / unit are as follows:

[0134] The data acquisition module 101 is used to collect the operation process data of the railway turnout in real time using a preset sensor;

[0135] The data analysis module 102 is used to perform time domain analysis on the operation process data to obtain time domain components, perform frequency domain analysis on the time domain components to obtain frequency domain components, and calculate power density spectrum according to the frequency domain components to obtain operation data characteristics;

[0136] The probability prediction module 103 is used to output the prediction probabilities corresponding to different health levels of railway turnouts according to the operation data characteristics using a pre-trained railway turnout state analysis model;

[0137] The result confirmation module 104 selects the health level with the highest prediction probability as the final health status analysis result of the railway turnout.

[0138] Embodiment 3

[0139] Based on the above embodiments, this embodiment provides an application example.

[0140] The present invention provides a method for analyzing the health status of heavy-duty railway turnouts to solve the problem in the prior art that turnouts are key facilities for realizing train switching, and their poor status may lead to poor switching of switch machines and loose contact of point rails, thereby causing turnout loss and squeezed turnouts, affecting operational safety. Therefore, it is necessary to provide a method for analyzing the health status of turnouts, giving early warnings for poor turnouts, and prompting operating personnel to repair them in a timely manner.

[0141] In order to achieve the above object, the present invention provides the following technical solutions:

[0142] According to a first aspect of the present invention, a method for analyzing the health status of a heavy-haul railway turnout comprises the following steps:

[0143] S100, data collection, it is necessary to collect data during the operation of the turnout, including action power curve, current data, etc. These data are usually obtained through a microcomputer monitoring system and are collected in real time or regularly;

[0144] S200, feature extraction, by analyzing the action mechanism of the turnout equipment, the turnout power curve is divided into different action processes, and time domain parameters are extracted from them;

[0145] S300, health level classification, in order to better carry out preventive maintenance, the status of the turnout is divided into different levels:

[0146] S400, model establishment and analysis, using support vector description or self-organizing map neural network machine learning algorithm to establish an analysis model of the health status of the turnout. The model can analyze the health status of the turnout according to the extracted characteristic parameters and predict possible faults in the future, specifically:

[0147] S410, feature extraction, establishing data model;

[0148] S420, training the model and classifying the health status;

[0149] S430, defining a deviation amount and optimizing the determination parameters;

[0150] S440, judging the health status of the turnout;

[0151] S500, field verification.

[0152] Furthermore, the step S100, data collection, specifically includes:

[0153] Step 1: Sensor deployment: various sensors are installed at key locations of the turnout, including: position sensors for real-time monitoring of the position of the turnout, acceleration sensors for monitoring the acceleration changes of the turnout during its movement, pressure sensors for monitoring the pressure on the turnout during its movement, and displacement sensors for collecting the displacement change values ​​of the turnout indicating rods;

[0154] Step 2: Data collection and storage: The signals measured by the sensors are collected through the data acquisition system. The collected data needs to be stored in a database or data warehouse for subsequent analysis.

[0155] Step 3: Data preprocessing. Before the data enters the analysis process, preprocessing is usually required, including denoising, filtering, normalization and other steps to improve data quality and reduce interference factors in the analysis process.

[0156] Further, the feature extraction in S200 specifically includes:

[0157] S210, time domain analysis, according to the characteristics of the turnout action power curve, its time domain characteristic parameters can be extracted in three sections, and the calculation formula is:

[0158]

[0159] Among them, x i is the vibration amplitude of the ith sample point, and N is the total number of sample points.

[0160] Furthermore, the S200, feature extraction specifically includes:

[0161] S220, frequency domain analysis, its FFT conversion calculation formula is:

[0162]

[0163] Where (x[n]) is the vibration signal in the time domain, and (X[k]) is the complex representation in the frequency domain;

[0164] Its power spectral density expression is:

[0165]

[0166] Where |X[k]| represents the amplitude of the (k)th frequency component in the frequency domain.

[0167] Furthermore, the S300 health level classification specifically includes three levels: failure, sub-health and health, so as to facilitate classification and priority sorting of maintenance and repair work of turnouts.

[0168] Furthermore, the S400, model building analysis specifically includes:

[0169] S410, feature extraction, building a data model, assuming that the original data is X, after feature extraction, the feature set F = {f1, f2, ..., f n}, where n is the number of features;

[0170] Furthermore, the S400, model building analysis specifically includes:

[0171] S420, train the model to classify the health status, and train the model to learn the mapping from the feature set F to the health status label Y, wherein the forward propagation process of the model can be expressed as:

[0172] h t ,C t =LSTM(h t-1 ,C t-1 ,x t ; W)

[0173] Furthermore, the S400, model building analysis specifically includes:

[0174] S430, define the deviation, optimize the judgment parameters, and define the loss function L to measure the predicted health status The difference between the actual health status Y is expressed as:

[0175]

[0176] The model is trained by minimizing the loss function L, that is, finding the optimal parameter W*, the result is:

[0177]

[0178] Furthermore, the S400, model building analysis specifically includes:

[0179] S440, judge the health status of the turnout according to the output of the model Select the health status category with the highest probability as the judgment result:

[0180]

[0181] Further, the S500, on-site verification specifically includes:

[0182] Step 1: On-site observation and analysis: On-site observation and analysis of the prediction results;

[0183] Step 2: Manual inspection: Maintenance personnel can check specific parts according to the prompts of the automation system to ensure that no possible problems are missed;

[0184] Step 3: Feedback loop, continuously collect new operating data, and iteratively train and optimize the model.

[0185] Embodiment 4

[0186] On the basis of the above embodiments, this embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the above embodiments.

[0187] In some implementations of this embodiment, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that when the computer program is executed by a processor, the steps of the method described in the above embodiment are implemented.

[0188] In some implementations of this embodiment, a computer program product is provided, including a computer program / instructions, characterized in that when the computer program is executed by a processor, the steps of the method described in the above embodiment are implemented.

[0189] The processor may include, but is not limited to, one or more processors or microprocessors. Each processor may be an application specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a controller, a microcontroller, a microprocessor or other electronic components to execute the method in the above embodiments.

[0190] The computer-readable storage medium may be implemented by any type of volatile or non-volatile storage device or a combination thereof, and the computer-readable storage medium may include but is not limited to, for example, random access memory (RAM), read-only memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, computer storage media (e.g., hard disk, floppy disk, solid-state drive, removable disk, CD-ROM, DVD-ROM, Blu-ray disc, etc.).

[0191] The computer-readable storage medium may also store at least one computer executable program / instruction, which may be, for example, a computer-readable instruction. The computer-readable storage medium includes, but is not limited to, for example, a volatile memory and / or a non-volatile memory. The volatile memory may include, for example, a random access memory (RAM) and / or a cache memory (cache), etc. The computer-readable storage medium may include, for example, a read-only memory (ROM), a hard disk, a flash memory, etc. For example, a non-transitory computer-readable storage medium may be connected to a computing device such as a computer, and then, when the computing device runs the computer-readable instructions stored on the computer-readable storage medium, the various methods described above may be performed.

[0192] In addition, the computer device may also include (but not limited to) a data bus, an input / output (I / O) bus, a display, and input / output devices (eg, keyboard, mouse, speaker, etc.), etc.

[0193] The processor may communicate with external devices via an I / O bus via a wired or wireless network.

[0194] In one embodiment, the at least one computer executable instruction may also be compiled into or constitute a software product / computer program product, wherein one or more computer executable instructions are executed by a processor to perform the various functions and / or method steps in the embodiments described in the present technology.

[0195] In the embodiments provided in the present disclosure, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the above-mentioned module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0196] It should be noted that in the present disclosure, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element limited by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0197] Although the embodiments disclosed in the present disclosure are as above, the above contents are only embodiments adopted for facilitating the understanding of the present disclosure and are not intended to limit the present disclosure. Any technician in the technical field to which the present disclosure belongs can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in the present disclosure, but the scope of patent protection of the present disclosure shall still be subject to the scope defined in the attached claims.

Claims

1. A railway turnout health status analysis method, characterized in that: include: Use preset sensors to collect the operation process data of railway turnouts in real time; Performing time domain analysis on the operation process data to obtain a time domain component; Performing frequency domain analysis on the time domain component to obtain a frequency domain component; Calculate the power density spectrum according to the frequency domain components to obtain the operation data characteristics; Outputting prediction probabilities corresponding to different health levels of railway turnouts according to the operation data characteristics using a pre-trained railway turnout state analysis model; The health level with the highest predicted probability is selected as the final health status analysis result of the railway turnout.

2. The method according to claim 1, characterized in that The method of collecting the operation process data of the railway turnout in real time by using the preset sensors includes: Using sensors preset in railway turnouts to obtain turnout position information, turnout acceleration change information, turnout pressure information, and turnout indicating rod position change value in real time, and obtain a turnout operation information set; De-noising the turnout operation information set to obtain a de-noised operation information set; The denoised operation information set is normalized using the following formula to obtain the operation process data: Wherein, X is the operation process data, F is the number of samples of the denoised operation information set, and x i represents the i-th data in the denoising operation information set, x max Indicates the maximum value in the denoising operation information set.

3. The method according to claim 1, characterized in that: The performing time domain analysis on the operation process data to obtain the time domain component includes: The following formula is used to perform time domain analysis on the operation process data: Wherein, R is the time domain component, N is the number of samples of the running process data, and x i Represents the vibration amplitude value of the i-th sample in the running process data.

4. The method according to claim 3, characterized in that The performing frequency domain analysis on the time domain component to obtain the frequency domain component includes: Wherein, X[k] is the complex number representation of the frequency domain component, M is the number of sample points of the time domain component, and R n Represents the amplitude of the nth signal in the time domain component.

5. The method according to claim 1, characterized in that Prior to the method using the pre-trained railway turnout state analysis model to output the prediction probabilities corresponding to different health levels of the railway turnout according to the operation data characteristics, the method further includes: Obtain training feature sets and test feature sets with true health level annotations; Establish and preset a neural network model with health level as output; Outputting the prediction results of the health level according to the training feature set using the neural network model, taking the health level with the highest probability in the prediction results as the prediction output, and obtaining a prediction set; The loss value of each prediction output of the prediction set and its corresponding health level label is calculated using the following formula to obtain a loss value set: Among them, L i is the i-th loss value in the loss value set, Describe the i-th prediction output in the prediction set, Y i Describe the true health level label corresponding to the i-th prediction output in the prediction set; The weight parameters of the neural network model are updated according to the loss value set and the preset learning rate using the following formula: Among them, W * is the weight parameter of the neural network model after updating, W is the weight parameter of the neural network model before updating, μ is the learning rate, D is the number of samples of the loss value set, L i is the i-th loss value in the loss value set; When the preset number of iterations is reached, the regression rate test of the neural network model is performed using the test feature set; If the neural network model fails the regression rate test, adjusting the learning rate of the neural network model, returning to use the neural network model of last month to predict the health level according to the training feature set, taking the health level with the highest probability in the prediction results as the prediction output, and obtaining the prediction set; If the neural network model passes the regression rate test, it is confirmed that the railway switch state analysis model is obtained.

6. The method according to claim 5, characterized in that The method of using the pre-trained railway turnout state analysis model to output the prediction probabilities corresponding to different health levels of the railway turnout according to the operation data characteristics includes: Using the convolution layer of the railway turnout state analysis model to extract local features of the operation data features to obtain convolution features; Using the pooling layer of the railway turnout state analysis model to extract the local average value of the convolution feature, so as to obtain the pooling feature; Using the activation layer of the railway turnout state analysis model to add nonlinear factors to the pooling feature, so as to obtain an activation feature; The fully connected layer of the railway turnout state analysis model is used to output the predicted probability corresponding to each health level according to the activation features and the weight parameters and bias items of the railway turnout state analysis model.

7. A railway turnout health status analysis device, characterized in that: include: A data acquisition module is used to collect the operation process data of the railway turnout in real time using preset sensors; A data analysis module, configured to perform time domain analysis on the operation process data to obtain time domain components, perform frequency domain analysis on the time domain components to obtain frequency domain components, and calculate a power density spectrum according to the frequency domain components to obtain operation data characteristics; A probability prediction module, used to output the prediction probabilities corresponding to different health levels of railway turnouts according to the operation data characteristics using a pre-trained railway turnout state analysis model; The result confirmation module selects the health level with the highest prediction probability as the final health status analysis result of the railway turnout.

8. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images