CN111896625B - Rail damage real-time monitoring method and monitoring system thereof - Google Patents
Rail damage real-time monitoring method and monitoring system thereof Download PDFInfo
- Publication number
- CN111896625B CN111896625B CN202010826057.2A CN202010826057A CN111896625B CN 111896625 B CN111896625 B CN 111896625B CN 202010826057 A CN202010826057 A CN 202010826057A CN 111896625 B CN111896625 B CN 111896625B
- Authority
- CN
- China
- Prior art keywords
- damage
- real
- data
- rail
- monitoring
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/04—Analysing solids
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61K—AUXILIARY EQUIPMENT SPECIALLY ADAPTED FOR RAILWAYS, NOT OTHERWISE PROVIDED FOR
- B61K9/00—Railway vehicle profile gauges; Detecting or indicating overheating of components; Apparatus on locomotives or cars to indicate bad track sections; General design of track recording vehicles
- B61K9/08—Measuring installations for surveying permanent way
- B61K9/10—Measuring installations for surveying permanent way for detecting cracks in rails or welds thereof
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61L—GUIDING RAILWAY TRAFFIC; ENSURING THE SAFETY OF RAILWAY TRAFFIC
- B61L23/00—Control, warning, or like safety means along the route or between vehicles or vehicle trains
- B61L23/04—Control, warning, or like safety means along the route or between vehicles or vehicle trains for monitoring the mechanical state of the route
- B61L23/042—Track changes detection
- B61L23/044—Broken rails
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
- G01N29/4481—Neural networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2291/00—Indexing codes associated with group G01N29/00
- G01N2291/02—Indexing codes associated with the analysed material
- G01N2291/023—Solids
- G01N2291/0234—Metals, e.g. steel
Abstract
The invention discloses a real-time monitoring method for rail damage, which comprises the steps of arranging monitoring systems on two sides of a rail; acquiring real-time monitoring data obtained by a monitoring system; performing data processing on the real-time monitoring data to obtain a processed damage picture; the neural network architecture is adopted to identify the damage-treated picture so as to obtain an identification result; and carrying out real-time early warning on the rail damage according to the identification result and realizing real-time monitoring on the rail damage. The invention also discloses a monitoring system for realizing the rail damage real-time monitoring method. The invention monitors the guide rail in real time through the ultrasonic transducer and collects and processes the data in real time, thereby realizing the real-time monitoring and damage positioning of the structural state of the full life cycle of the steel rail in the service period, sending out early warning information, improving the accuracy and efficiency of the detection of the damage of the steel rail, reducing maintenance cost, improving the driving safety of the train, and having high reliability, good real-time performance and higher accuracy.
Description
Technical Field
The invention belongs to the field of steel rail monitoring, and particularly relates to a steel rail damage real-time monitoring method and a monitoring system thereof.
Background
Along with the development of economic technology and the improvement of living standard of people, rail transit is widely applied to the production and living of people, and brings endless convenience to the production and living of people.
In a rail transit system, one of the threats of safe running of a train and normal running of the rail transit system is rail breakage. Rail breakage can be caused by nuclear damage and weld quality, if rail breakage is light on a railway transportation line network, transportation delay is caused, and serious driving accidents such as derailment and overturning of a train can be caused, so that extremely high casualties and property loss are caused. Therefore, real-time monitoring of the full life cycle of the steel rail in the service period becomes an urgent problem to be solved.
Currently, rail flaw detection management rules of railway departments are compiled based on ultrasonic detection technology, and manual flaw detection is mainly used, and flaw detection vehicles are auxiliary. The steel rail flaw detection method has the advantages that the steel rail flaw detection method is started once in a period, more leakage detection and false detection are performed, and the state of the steel rail between two flaw detection cannot be known. The detection blind areas exist on the surface of the top surface of the steel rail and near-surface fatigue damage, the flaw detection efficiency is low, and the real-time monitoring of the health state of the steel rail and the change of the health state cannot be realized. With the development of high-speed railways, the traditional nondestructive testing method cannot meet the requirement of real-time monitoring under new situation.
Disclosure of Invention
The invention aims to provide a rail damage real-time monitoring method with high reliability, good real-time performance and high accuracy.
The second object of the invention is to provide a monitoring system for realizing the rail damage real-time monitoring method.
The invention provides a real-time monitoring method for rail damage, which comprises the following steps:
s1, arranging monitoring systems on two sides of a steel rail;
s2, acquiring real-time monitoring data obtained by a monitoring system;
s3, carrying out data processing on the real-time monitoring data obtained in the step S2, so as to obtain a processed injury picture;
s4, identifying the damage processing picture obtained in the step S3 by adopting a neural network architecture, so as to obtain an identification result;
s5, carrying out real-time early warning on the damage of the steel rail according to the identification result obtained in the step S4, thereby realizing real-time monitoring of the damage of the steel rail.
The monitoring systems are arranged on two sides of the steel rail in the step S1, and specifically comprise an ultrasonic transducer, a signal receiving device and a control module. Arranging a pair of ultrasonic transducers at intervals of a set distance on two sides of a steel rail, so that the whole detection section is divided into a plurality of small sections for detection; the data output end of the ultrasonic transducer is connected with a signal receiving device, and the data output end of the signal receiving device is connected with a control module; the ultrasonic transducer is used for generating ultrasonic guided waves and receiving reflected waves; the signal receiving device is used for receiving the reflected wave received by the ultrasonic transducer and uploading the reflected wave to the control module; the control module is used for monitoring the damage of the steel rail in real time and carrying out early warning according to the received data.
The step S2 is to obtain the real-time monitoring data obtained by the monitoring system, specifically, the detection channel and the monitoring frequency are set through the control module, the ultrasonic transducer completes the excitation and the reception of the ultrasonic guided wave, and the signal receiving device performs digital-to-analog conversion and uploads the digital-to-analog conversion to the control module.
When ultrasonic guided waves are generated, the guided waves of bending modes are selected, the dispersion characteristics of the guided waves are analyzed by adopting a semi-analytic finite element method, the sensitivity of the specific guided waves to damage of different areas in the section of the steel rail is judged through a calculated section wave structure diagram of the steel rail, and the receiving position is arranged at the position with larger guided wave displacement in the steel rail.
The excitation position is calculated by the following steps:
A. listing the positions of all external excitable nodes of the cross section of the steel rail, wherein each node has three degrees of freedom;
B. exciting by adopting a selected guided wave mode, generating three degrees of freedom displacement vectors of each node position, and drawing a graph of the three displacement vectors;
C. based on a statistical rule, removing smaller extreme points smaller than a set value;
D. and finally, according to the residual local extreme points, taking the maximum amplitude position of 3 degrees of freedom as the optimal excitation point of the mode.
The guided wave excitation frequency and the analysis frequency are determined by the following steps:
a. selecting upper and lower limits of ultrasonic guided wave frequency, and separating out frequency measuring points according to the same interval;
b. using the measuring point frequency as excitation frequency to perform guided wave detection;
c. performing time-frequency analysis on the signals received by each measuring point;
d. drawing a guided wave time-frequency analysis chart corresponding to the excitation frequency;
e. and selecting the final guided wave excitation frequency and analysis frequency.
And step S3, performing data processing on the real-time monitoring data acquired in the step S2 to obtain a processed injury picture, wherein the data processing is performed by adopting the following steps:
(1) Threshold comparison is carried out on the received real-time monitoring data: when the real-time monitoring data is larger than the set gate threshold value, generating B display data from the monitoring data;
(2) And (3) denoising and dividing the B display data obtained in the step (1):
noise reduction: ignoring small damage of less than N reflection echo points, and shielding continuous waveforms with single-point clutter and inconsistent angles by adopting an 8-neighborhood-point continuity judging method; n is a natural number;
segmentation: and dividing the ultrasonic guided wave echo points close to the position on the B display image into one picture by using a DBSCAN algorithm, thereby ensuring that each picture only contains 1 injury.
The method is characterized in that the ultrasonic guided wave echo points close to each other on the B display image are segmented into one image by using a DBSCAN algorithm, so that each image is ensured to contain only 1 injury, and the method specifically comprises the following steps:
1) Finding core points and forming temporary clustering clusters:
scanning all sample points: if the number of the points within the radius R of a certain sample point is larger than or equal to a set threshold value, taking the point as a core point, and forming a corresponding temporary cluster by directly taking the density of the core point;
2) Merging the temporary cluster clusters to obtain a cluster:
for each temporary cluster, check if the point therein is a core point: if yes, combining the temporary cluster corresponding to the point with the current temporary cluster to obtain a new temporary cluster; repeating the step until each point in the current temporary cluster is not a core point or the point with direct density is in the temporary cluster, and converting the temporary cluster into a cluster;
3) Repeating the step 2) until all temporary cluster clusters are combined.
The neural network architecture described in step S4 is specifically an AlexNet convolutional neural network architecture: the network architecture comprises 10 layers in total, including 1 input layer, 3 convolution layers, 3 pooling layers, 2 full connection layers and 1 output layer; the input layer is B display data subjected to data preprocessing; adopting a smote algorithm to obtain an oversampling balance data set, wherein the size of the processed data is 128 x 16 pixels; the input data is slid out from the whole detection file, and the sliding step length is 42 pixels; the convolution layer performs feature extraction on input data through convolution operation; the depth after passing through three convolution layers is 100; the pooling layer adopts a mean pooling method; the activation functions all select the Relu function; the fully connected layer classifies the input data using a softmax function, wherein the first fully connected layer is 2048 neurons and the second fully connected layer contains 1024 neurons; the output layer classifies the damage condition into 8 types; the output layer is the probability of each damage type; randomly selecting 80% of various B-display data sample sets as training sets, and the remaining 20% as test sets; during the training process, early Stopping and dropout methods are adopted to prevent overfitting; the dropout method is used in two full-connection layers, and connection is closed according to 50% probability in the training process; the training learning rate is set to be 0.01, the accuracy of the test set is calculated every 5 rounds, and iteration is stopped after the accuracy is not improved after 15 rounds of continuous calculation.
And step S5, carrying out real-time early warning on the steel rail damage according to the identification result obtained in the step S4, specifically, according to the steel rail damage identification result obtained in the step S4, a control module of the detection system combines the damage type, the severity and the development rule, and gives maintenance reminding and maintenance advice to the damage at all positions, thereby achieving the purposes of timely finding the damage, treating the damage and carrying out maintenance at the initial stage of the damage.
The invention also provides a monitoring system for realizing the rail damage real-time monitoring method, which specifically comprises an ultrasonic transducer, a signal receiving device and a control module; arranging a pair of ultrasonic transducers at intervals of a set distance on two sides of a steel rail, so that the whole detection section is divided into a plurality of small sections for detection; the data output end of the ultrasonic transducer is connected with a signal receiving device, and the data output end of the signal receiving device is connected with a control module; the ultrasonic transducer is used for generating ultrasonic guided waves and receiving reflected waves; the signal receiving device is used for receiving the reflected wave received by the ultrasonic transducer and uploading the reflected wave to the control module; the control module is used for monitoring the damage of the steel rail in real time and carrying out early warning according to the received data.
According to the rail damage real-time monitoring method and the rail damage real-time monitoring system, the rail is monitored in real time through the ultrasonic transducer, and the data are collected and processed in real time, so that the real-time monitoring and damage positioning of the rail in-service full life cycle structure state are realized, early warning information is sent, the rail damage detection accuracy and efficiency are improved, the maintenance cost is reduced, the train driving safety is improved, and the reliability is high, the real-time performance is good, and the accuracy is higher.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
Fig. 2 is a functional block diagram of the system of the present invention.
Fig. 3 is an initial interface schematic diagram of data analysis software in a preferred embodiment of a method for monitoring rail damage in real time provided by the system of the present invention.
Detailed Description
A schematic process flow diagram of the method of the present invention is shown in fig. 1: the invention provides a real-time monitoring method for rail damage, which comprises the following steps:
s1, arranging monitoring systems on two sides of a steel rail;
s2, acquiring real-time monitoring data obtained by a monitoring system;
the method is characterized in that when in specific implementation:
when ultrasonic guided waves are generated, the guided waves of bending modes are selected, the dispersion characteristics of the guided waves are analyzed by adopting a semi-analytic finite element method, the sensitivity of the specific guided waves to damage of different areas in the section of the steel rail is judged through a calculated section wave structure diagram of the steel rail, and the receiving position is arranged at the position with larger guided wave displacement in the steel rail: the semi-analytic finite element method is a calculation method for solving the problem of frequency in a complex cross section waveguide medium; when the method is used for calculating the dispersion solution, only finite element dispersion is needed to be carried out on the cross section of the waveguide medium, and the displacement along the propagation direction of the waveguide medium is represented by a simple harmonic vibration mode; when solving by a semi-analytic finite element method, firstly, carrying out finite element grid division on the section of a steel rail, and dispersing the section of the plate by a one-dimensional three-node unit; according to the Hamiltonian principle, strain energy and potential energy of any point in a waveguide medium are brought into a Hamiltonian formula, a wave equation of guided waves is deduced, and then the wave equation is converted into a standard characteristic equation form for solving, so that the relation between wave numbers and frequencies and displacement distribution of any node of a steel rail are obtained, and a dispersion characteristic curve is drawn; then, obtaining displacement distribution of any node of the section of the steel rail according to the semi-analytic finite element method to obtain a wave structure diagram; the wave structure represents the distribution of particle displacement energy along the depth direction in the wave propagation process, can intuitively present the propagation characteristics of different frequency and different mode guide waves in the steel rail, and distinguishes damage according to the mode shape; finally, the distribution of the guided wave in the section of the steel rail shows obvious regionalization, and the distribution is influenced by the frequency of the guided wave; the wave structure diagram is obtained through the calculation, the guided wave displacement of different positions in the steel rail is analyzed, and the larger the displacement is, the higher the sensitivity of the guided wave to the damage of the area of the section of the steel rail is;
the excitation position is calculated by the following steps:
A. listing the positions of all external excitable nodes of the cross section of the steel rail, wherein each node has three degrees of freedom;
B. exciting by adopting a selected guided wave mode, generating three degrees of freedom displacement vectors of each node position, and drawing a graph of the three displacement vectors;
the method comprises the following steps: firstly, listing the positions of all external excitable nodes of the section of the steel rail, and determining the positions as characteristic points of all modes; setting the total number of all peripheral nodes of the section of the steel rail as P, wherein each node comprises three degrees of freedom in the three directions and corresponds to displacement in the three directions; the displacement of each degree of freedom is respectively noted as xi, yi, zi; starting from the left side of the rail bottom, sequentially numbering and listing the characteristic points along the clockwise direction, and obtaining displacement vectors of three degrees of freedom of each characteristic point after excitation of a selected vibration mode is applied: x= [ X1, X2, …, xp ], y= [ Y1, Y2, …, yp ], z= [ Z1, Z2, …, zp ];
C. based on a statistical rule, removing smaller extreme points smaller than a set value;
D. finally, according to the residual local extreme points, taking the maximum amplitude position of 3 degrees of freedom as the optimal excitation point of the mode;
the method comprises the following steps: the basic idea of the optimal excitation position positioning algorithm is to excite the maximum amplitude position of the three degrees of freedom of the vibration mode, and the maximum amplitude position is required to be searched for excitation according to the actual condition of the vibration curve because the vibration curves of different modes are different;
the guided wave excitation frequency and the analysis frequency are determined by the following steps:
a. selecting upper and lower limits of ultrasonic guided wave frequency, and separating out frequency measuring points according to the same interval;
b. using the measuring point frequency as excitation frequency to perform guided wave detection;
c. performing time-frequency analysis on the signals received by each measuring point;
d. drawing a guided wave time-frequency analysis chart of the excitation frequency corresponding to the selected frequency;
e. selecting final guided wave excitation frequency and analysis frequency;
the method comprises the following steps: the ultrasonic guided wave detection frequency range which is generally considered to be applicable to the steel rail is 30-100kHz; in the upper and lower limit frequency ranges, measuring points are separated at equal intervals at 2 kHz; using the measuring points as excitation frequencies to perform guided wave detection; performing wavelet transformation on the received signals of each measuring point to obtain a time-frequency analysis result of the guided wave received signals of each measuring point; the amplitudes of the corresponding time points and the frequency points in all the time-frequency analysis results are respectively added to obtain a guided wave time-frequency analysis chart with excitation frequency corresponding to the selected frequency range in the guided wave detection, and the chart can intuitively show the distribution, the intensity and the dispersion characteristics of different guided wave modes in each frequency band; according to the time-frequency analysis chart, the frequency with the largest amplitude near zero time in the chart is taken as the guided wave excitation frequency, and the frequency range with larger amplitude at the receiving time is taken as the guided wave analysis frequency band, so that the guided wave excitation and analysis frequency which accords with the principles of small modal number, easy identification, ideal amplitude and the like is obtained;
s3, carrying out data processing on the real-time monitoring data obtained in the step S2, so as to obtain a processed injury picture; the method specifically comprises the following steps of:
(1) Threshold comparison is carried out on the received real-time monitoring data: when the real-time monitoring data is larger than the set gate threshold value, generating B display data from the monitoring data;
(2) And (3) denoising and dividing the B display data obtained in the step (1):
noise reduction: ignoring small damage of less than N reflection echo points, and shielding continuous waveforms with single-point clutter and inconsistent angles by adopting an 8-neighborhood-point continuity judging method; n is a natural number;
segmentation: dividing ultrasonic guided wave echo points close to each other on a B display image into a picture by using a DBSCAN algorithm, thereby ensuring that each picture only contains 1 injury; the method comprises the following steps:
1) Finding core points and forming temporary clustering clusters:
scanning all sample points: if the number of the points within the radius R of a certain sample point is larger than or equal to a set threshold value, taking the point as a core point, and forming a corresponding temporary cluster by directly taking the density of the core point;
2) Merging the temporary cluster clusters to obtain a cluster:
for each temporary cluster, check if the point therein is a core point: if yes, combining the temporary cluster corresponding to the point with the current temporary cluster to obtain a new temporary cluster; repeating the step until each point in the current temporary cluster is not a core point or the point with direct density is in the temporary cluster, and converting the temporary cluster into a cluster;
3) Repeating the step 2) until all temporary cluster clusters are combined;
s4, identifying the damage processing picture obtained in the step S3 by adopting a neural network architecture, so as to obtain an identification result; specifically, an AlexNet convolutional neural network architecture is adopted: the network architecture comprises 10 layers in total, including 1 input layer, 3 convolution layers, 3 pooling layers, 2 full connection layers and 1 output layer; the input layer is B display data subjected to data preprocessing; adopting a smote algorithm to obtain an oversampling balance data set, wherein the size of the processed data is 128 x 16 pixels; the input data is slid out from the whole detection file, and the sliding step length is 42 pixels; the convolution layer performs feature extraction on input data through convolution operation; the depth after passing through three convolution layers is 100; the pooling layer adopts a mean pooling method; the activation functions all select the Relu function; the fully connected layer classifies the input data using a softmax function, wherein the first fully connected layer is 2048 neurons and the second fully connected layer contains 1024 neurons; the output layer divides the damage condition into 8 types (no damage (including normal screw holes and wire holes), rail head nuclear damage, rail waist oblique cracks, rail waist separation, rail bottom crescent damage, rail joint damage, screw hole cracks and other damage, wherein the output layer is the probability of each damage type); the output layer is the probability of each damage type; randomly selecting 80% of various B-display data sample sets as training sets, and the remaining 20% as test sets; during the training process, early Stopping and dropout methods are adopted to prevent overfitting; the dropout method is used in two full-connection layers, and connection is closed according to 50% probability in the training process; training learning rate is set to be 0.01, the accuracy of the test set is calculated every 5 rounds, and iteration is stopped after the accuracy is not improved after 15 rounds of continuous calculation;
s5, carrying out real-time early warning on the damage of the steel rail according to the identification result obtained in the step S4, thereby realizing real-time monitoring of the damage of the steel rail.
A functional block diagram of the system of the present invention is shown in fig. 2: the monitoring system for realizing the rail damage real-time monitoring method provided by the invention specifically comprises an ultrasonic transducer, a signal receiving device and a control module; arranging a pair of ultrasonic transducers at intervals of a set distance on two sides of a steel rail, so that the whole detection section is divided into a plurality of small sections for detection; the data output end of the ultrasonic transducer is connected with a signal receiving device, and the data output end of the signal receiving device is connected with a control module; the ultrasonic transducer is used for generating ultrasonic guided waves and receiving reflected waves; the signal receiving device is used for receiving the reflected wave received by the ultrasonic transducer and uploading the reflected wave to the control module; the control module is used for monitoring the damage of the steel rail in real time and carrying out early warning according to the received data.
FIG. 3 is a schematic diagram of an initial interface of data analysis software of the system of the present invention: the left side bar displays real-time monitoring and early warning results, and can click and check the detailed condition of the injury for each piece of early warning information; the method comprises the steps of realizing the check of a specific analysis object through a menu bar 'loading data', wherein an analysis result is displayed in each sub-window of a page and comprises data information (probe number and detection time), a dispersion curve, an excitation response result, a damage processing picture obtained by B display data, a damage discrimination result obtained by a neural network and damage early warning information obtained according to the analysis result; the menu acquisition parameter setting is used for setting the acquisition frequency, the channel and other damage data acquisition schemes; the menu data identification parameter is used for setting parameter values in the identification calculation of the damage data such as B display data generation, damage picture processing, neural network algorithm and the like.
The invention provides a rail damage real-time monitoring method based on ultrasonic guided waves, which comprises the following steps: firstly, arranging monitoring systems on two sides of a steel rail, wherein the monitoring systems comprise ultrasonic transducers, signal receiving devices and control modules; then, acquiring real-time monitoring data obtained by a monitoring system; then, performing data processing on the acquired real-time monitoring data to obtain a processed injury picture; then adopting a neural network architecture to identify the damage-treated picture, and obtaining an identification result; finally, the rail damage is early-warned in real time according to the identification result, so that the rail damage is monitored in real time, the practical problem of the current railway industry is solved, scientific guarantee is provided for railway safety operation, the core competitiveness of the high-speed railway technology in China is improved, and the continuous guiding effect of the high-speed railway in China in internationally is promoted.
Furthermore, the invention formulates ultrasonic guided wave excitation and receiving schemes for flaw detection from three aspects of guided wave propagation characteristic analysis, guided wave excitation and receiving frequency determination and optimal excitation position positioning analysis, provides real, effective, objective and large amount of data support for realizing all-weather automatic monitoring of the steel rail, greatly reduces the labor intensity of a line flaw detector, and avoids the problems of untimely detection and flaw detection blind areas, leakage detection and false detection of an ultrasonic detection device existing in the current period detection.
The rail section wave structure diagram obtained by calculation by the method has the advantage of good intuitiveness, is convenient for analyzing the propagation characteristics of guided waves with different frequencies and modes in the rail, and is very suitable for evaluating the sensitivity of the guided waves to damage of different areas in the rail section; the wavelets are transformed and analyzed to optimize the guided wave excitation frequency and analysis frequency which have the advantages of less mode quantity, easy identification and ideal amplitude, thereby having the characteristics of convenience and intuitiveness and effectively solving the problem of mismatch between the theoretical analysis based on the dispersion curve and the actual guided wave detection when the guided wave excitation frequency is selected; the proper ultrasonic guided wave probe installation position can be calculated through the optimal excitation position positioning algorithm, so that the deviation degree between the optimal excitation position and the actual installation position is smaller; and further, the rail guided wave mode which is weak in dispersion, small in attenuation, sensitive to damage and good in identification degree under different environments is determined, the detection capability of small defects of the rail is effectively improved, the rationality and the effectiveness of the design of rail flaw detection transducer, guided wave excitation and receiving schemes are ensured, and the method is high in practicability.
In addition, compared with the prior art, the invention adopts the neural network architecture to design the rail damage data analysis method, inputs the ultrasonic guided wave B display data into the recognition system, utilizes the DBSCAN algorithm to process the damage picture, utilizes the deep learning method to learn the rule from a large amount of data, achieves the purpose of accurately and rapidly extracting the data characteristics, thereby realizing the comprehensive and precise recognition of the rail damage position and the damage type, achieving the detection effect with high efficiency and low cost, and simultaneously providing technical support for establishing the characteristic database, thereby providing more accurate, objective and effective basis for fault investigation and state maintenance. The method converts the problem of 'object detection' into the problem of 'classification', effectively improves the intuitiveness of the detection result, has good robustness and generalization capability, and has better detection effect and efficiency than the traditional image processing method and manual identification.
Compared with the prior art, the rail damage real-time monitoring and early warning system provided by the invention sends early warning information on the basis of real-time monitoring of damage, ensures that an alarm can be uploaded to a monitoring center and related personnel at the first time when the early-stage representation of rail breakage occurs, is convenient for management personnel to take measures in time, avoids driving safety accidents and life and property losses, and reduces maintenance cost.
Claims (4)
1. A rail damage real-time monitoring method comprises the following steps:
s1, arranging monitoring systems on two sides of a steel rail; specifically, a pair of ultrasonic transducers are distributed at intervals of a set distance on two sides of a steel rail, so that the whole detection section is divided into a plurality of small sections for detection; the data output end of the ultrasonic transducer is connected with a signal receiving device, and the data output end of the signal receiving device is connected with a control module; the ultrasonic transducer is used for generating ultrasonic guided waves and receiving reflected waves; the signal receiving device is used for receiving the reflected wave received by the ultrasonic transducer and uploading the reflected wave to the control module; the control module is used for monitoring the damage of the steel rail in real time and carrying out early warning according to the received data;
in specific implementation, the excitation position is calculated by adopting the following steps:
A. listing the positions of all external excitable nodes of the cross section of the steel rail, wherein each node has three degrees of freedom;
B. exciting by adopting a selected guided wave mode, generating three degrees of freedom displacement vectors of each node position, and drawing a graph of the three displacement vectors;
C. based on a statistical rule, removing smaller extreme points smaller than a set value;
D. finally, according to the residual local extreme points, taking the maximum amplitude position of 3 degrees of freedom as the optimal excitation point of the mode;
the guided wave excitation frequency and the analysis frequency are determined by the following steps:
a. selecting upper and lower limits of ultrasonic guided wave frequency, and separating out frequency measuring points according to the same interval;
b. using the measuring point frequency as excitation frequency to perform guided wave detection;
c. performing time-frequency analysis on the signals received by each measuring point;
d. drawing a guided wave time-frequency analysis chart of the excitation frequency corresponding to the selected frequency;
e. selecting final guided wave excitation frequency and analysis frequency;
s2, acquiring real-time monitoring data obtained by a monitoring system;
s3, carrying out data processing on the real-time monitoring data obtained in the step S2, so as to obtain a processed injury picture; the method specifically comprises the following steps of:
(1) Threshold comparison is carried out on the received real-time monitoring data: when the real-time monitoring data is larger than the set gate threshold value, generating B display data from the monitoring data;
(2) And (3) denoising and dividing the B display data obtained in the step (1):
noise reduction: ignoring small damage of less than N reflection echo points, and shielding continuous waveforms with single-point clutter and inconsistent angles by adopting an 8-neighborhood-point continuity judging method; n is a natural number;
segmentation: dividing ultrasonic guided wave echo points close to each other on a B display image into a picture by using a DBSCAN algorithm, thereby ensuring that each picture only contains 1 injury;
s4, identifying the damage processing picture obtained in the step S3 by adopting a neural network architecture, so as to obtain an identification result; the neural network architecture specifically adopts an AlexNet convolutional neural network architecture: the network architecture comprises 10 layers in total, including 1 input layer, 3 convolution layers, 3 pooling layers, 2 full connection layers and 1 output layer; the input layer is B display data subjected to data preprocessing; adopting a smote algorithm to obtain an oversampling balance data set, wherein the size of the processed data is 128 x 16 pixels; the input data is slid out from the whole detection file, and the sliding step length is 42 pixels; the convolution layer performs feature extraction on input data through convolution operation; the depth after passing through three convolution layers is 100; the pooling layer adopts a mean pooling method; the activation functions all select the Relu function; the fully connected layer classifies the input data using a softmax function, wherein the first fully connected layer is 2048 neurons and the second fully connected layer contains 1024 neurons; the output layer classifies the damage condition into 8 types; the output layer is the probability of each damage type; randomly selecting 80% of various B-display data sample sets as training sets, and the remaining 20% as test sets; during the training process, early Stopping and dropout methods are adopted to prevent overfitting; the dropout method is used in two full-connection layers, and connection is closed according to 50% probability in the training process; training learning rate is set to be 0.01, the accuracy of the test set is calculated every 5 rounds, and iteration is stopped after the accuracy is not improved after 15 rounds of continuous calculation;
and S5, carrying out real-time early warning on the rail damage according to the identification result obtained in the step S4, thereby realizing real-time monitoring of the rail damage.
2. The method for monitoring the damage of the steel rail in real time according to claim 1, wherein when ultrasonic guided waves are generated, the guided waves of bending modes are selected, the dispersion characteristics of the guided waves are analyzed by adopting a semi-analytic finite element method, the sensitivity of the specific guided waves to the damage of different areas in the section of the steel rail is judged through a calculated section wave structure diagram of the steel rail, and the receiving position is set at the position with larger guided wave displacement in the steel rail.
3. The method for monitoring the damage of the steel rail in real time according to claim 1, wherein the method is characterized in that ultrasonic guided wave echo points close to each other on the B display image are segmented into one picture by using a DBSCAN algorithm, so that each picture is ensured to contain only 1 damage, and the method specifically comprises the following steps:
1) Finding core points and forming temporary clustering clusters:
scanning all sample points: if the number of the points within the radius R of a certain sample point is larger than or equal to a set threshold value, taking the point as a core point, and forming a corresponding temporary cluster by directly taking the density of the core point;
2) Merging the temporary cluster clusters to obtain a cluster:
for each temporary cluster, check if the point therein is a core point: if yes, combining the temporary cluster corresponding to the point with the current temporary cluster to obtain a new temporary cluster; repeating the step until each point in the current temporary cluster is not a core point or the point with direct density is in the temporary cluster, and converting the temporary cluster into a cluster;
3) Repeating the step 2) until all temporary cluster clusters are combined.
4. A monitoring system for realizing the rail damage real-time monitoring method according to one of claims 1 to 3, which is characterized by comprising an ultrasonic transducer, a signal receiving device and a control module; arranging a pair of ultrasonic transducers at intervals of a set distance on two sides of a steel rail, so that the whole detection section is divided into a plurality of small sections for detection; the data output end of the ultrasonic transducer is connected with a signal receiving device, and the data output end of the signal receiving device is connected with a control module; the ultrasonic transducer is used for generating ultrasonic guided waves and receiving reflected waves; the signal receiving device is used for receiving the reflected wave received by the ultrasonic transducer and uploading the reflected wave to the control module; the control module is used for monitoring the damage of the steel rail in real time and carrying out early warning according to the received data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010826057.2A CN111896625B (en) | 2020-08-17 | 2020-08-17 | Rail damage real-time monitoring method and monitoring system thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010826057.2A CN111896625B (en) | 2020-08-17 | 2020-08-17 | Rail damage real-time monitoring method and monitoring system thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111896625A CN111896625A (en) | 2020-11-06 |
| CN111896625B true CN111896625B (en) | 2023-07-14 |
Family
ID=73230436
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010826057.2A Active CN111896625B (en) | 2020-08-17 | 2020-08-17 | Rail damage real-time monitoring method and monitoring system thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111896625B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112464524B (en) * | 2020-11-07 | 2023-04-07 | 西南交通大学 | Method for determining guided wave propagation characteristics of turnout variable cross-section steel rail |
| CN112465684A (en) * | 2020-11-09 | 2021-03-09 | 覃强 | Full life cycle management system of track |
| CN112649513A (en) * | 2020-12-30 | 2021-04-13 | 天津精益铁安机电技术有限公司 | Railway artificial intelligence damage judging method based on image recognition |
| CN113239492B (en) * | 2021-04-09 | 2022-08-05 | 山东建筑大学 | Tower crane body steel structure damage positioning and monitoring method |
| CN113252789B (en) * | 2021-06-11 | 2022-03-08 | 东莞理工学院 | Nonlinear ultrasonic harmonic detection method for rail joint screw hole cracks |
| CN113533511B (en) * | 2021-06-29 | 2022-09-16 | 北京交通大学 | Steel rail weld joint monitoring method based on deep learning network model |
| CN113884567B (en) * | 2021-06-29 | 2023-12-08 | 北京交通大学 | Rail weld damage detection method and device based on ultrasonic Lamb wave |
| CN114062490B (en) * | 2021-06-29 | 2023-12-08 | 北京交通大学 | GAN-based rail weld Lamb wave modal decomposition and crack damage monitoring method |
| CN113640308B (en) * | 2021-08-31 | 2024-03-29 | 夏冰心 | Rail anomaly monitoring system based on machine vision |
| CN113447572B (en) * | 2021-09-01 | 2021-11-30 | 深圳市信润富联数字科技有限公司 | Steel rail flaw detection method, electronic device, steel rail flaw detection vehicle and readable storage medium |
| CN113808095A (en) * | 2021-09-13 | 2021-12-17 | 合肥德泰科通测控技术有限公司 | Big data-based intelligent damage identification and analysis system for railway steel rails |
| CN113740433B (en) * | 2021-09-18 | 2023-01-06 | 哈尔滨工业大学 | Device and method for detecting large lifting distance of SH guided wave EMAT of rail bottom crack of steel rail |
| CN113791034B (en) * | 2021-09-30 | 2022-09-06 | 合肥德泰轨交数据有限公司 | Sample collection and classification supervision system for steel rail flaw detection |
| CN114802344A (en) * | 2022-04-30 | 2022-07-29 | 西安市轨道交通集团有限公司 | Ultrasonic wave and carrier broken rail monitoring system |
Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE2500286A1 (en) * | 1975-01-04 | 1976-07-15 | Ruhrkohle Ag | Powered mine support shield units - shifting cylinder channel has floor clearance pusher to remove fine coal |
| JP2002090308A (en) * | 2000-09-19 | 2002-03-27 | Nippon Denro Kk | Evaluation system for degree of surface degradation of steel using image processing |
| US6846270B1 (en) * | 1999-02-25 | 2005-01-25 | Grant Etnyre | Method and apparatus for assisting or resisting postures or movements related to the joints of humans or devices |
| JP2005235258A (en) * | 2004-02-17 | 2005-09-02 | Fuji Electric Holdings Co Ltd | Inspection method of surface of magnetic recording medium |
| JP2007004390A (en) * | 2005-06-22 | 2007-01-11 | Ntt Docomo Inc | Sensor network system, cluster formation method, and sensor node |
| CN101571519A (en) * | 2009-06-16 | 2009-11-04 | 北京理工大学 | Ultrasonic guided wave detection technology for quantifying defects of composite laminated plate |
| CN102175768A (en) * | 2011-02-22 | 2011-09-07 | 哈尔滨工业大学 | Method and device for detecting defects and failures of high-speed rail based on vibration signals |
| WO2013097153A1 (en) * | 2011-12-27 | 2013-07-04 | 华南理工大学 | Radio ultrasound wave probe assembly for crawler belt-type steel rail test and test method thereof |
| CN104020221A (en) * | 2014-05-30 | 2014-09-03 | 杨媛 | Real-time broken-rail detecting and positioning system based on ultrasonic guided waves |
| CN104297346A (en) * | 2014-09-11 | 2015-01-21 | 天津大学 | Nondestructive detection system of sheet metal by ultrasonic planar guided-wave and detection method thereof |
| CN205176149U (en) * | 2015-11-10 | 2016-04-20 | 台州红炭电子科技有限公司 | RFID antenna performance testing arrangement on machine is got ready to RFID antenna |
| CN108956787A (en) * | 2018-06-13 | 2018-12-07 | 西安理工大学 | A kind of rail failure detection method neural network based |
| CN109543303A (en) * | 2018-11-22 | 2019-03-29 | 华北水利水电大学 | A method of the Damage Assessment Method to be perturbed based on class curvature of the flexibility difference matrix and frequency |
| CN109649432A (en) * | 2019-01-23 | 2019-04-19 | 浙江大学 | Cloud platform rail integrity monitoring systems and method based on guided wave technology |
| WO2019201176A1 (en) * | 2018-04-17 | 2019-10-24 | 江苏必得科技股份有限公司 | Method and device for predicting crack damage of train component |
| WO2019201178A1 (en) * | 2018-04-17 | 2019-10-24 | 江苏必得科技股份有限公司 | Train component crack damage detection method and system based on lamb wave imaging |
| WO2020156348A1 (en) * | 2019-01-31 | 2020-08-06 | 青岛理工大学 | Structural damage identification method based on ensemble empirical mode decomposition and convolution neural network |
| AU2020101234A4 (en) * | 2020-07-02 | 2020-08-06 | Hunan University Of Science And Technology | An Optimization Method for Excitation Parameters of Ultrasonic Infrared Thermography Crack Nondestructive Testing |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2003267144A1 (en) * | 2002-09-09 | 2004-03-29 | Zygo Corporation | Measurement and compensation of errors in interferometrs |
| US7512503B2 (en) * | 2003-05-12 | 2009-03-31 | Simmonds Precision Products, Inc. | Wire fault detection |
| US7463771B2 (en) * | 2005-09-08 | 2008-12-09 | Kun Shan University | Method for retrieving original intact characteristics of heavily polluted images and its image processing |
| CN103026216B (en) * | 2010-06-08 | 2016-09-07 | Dcg系统有限公司 | Three Dimensional Thermal point location |
-
2020
- 2020-08-17 CN CN202010826057.2A patent/CN111896625B/en active Active
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE2500286A1 (en) * | 1975-01-04 | 1976-07-15 | Ruhrkohle Ag | Powered mine support shield units - shifting cylinder channel has floor clearance pusher to remove fine coal |
| US6846270B1 (en) * | 1999-02-25 | 2005-01-25 | Grant Etnyre | Method and apparatus for assisting or resisting postures or movements related to the joints of humans or devices |
| JP2002090308A (en) * | 2000-09-19 | 2002-03-27 | Nippon Denro Kk | Evaluation system for degree of surface degradation of steel using image processing |
| JP2005235258A (en) * | 2004-02-17 | 2005-09-02 | Fuji Electric Holdings Co Ltd | Inspection method of surface of magnetic recording medium |
| JP2007004390A (en) * | 2005-06-22 | 2007-01-11 | Ntt Docomo Inc | Sensor network system, cluster formation method, and sensor node |
| CN101571519A (en) * | 2009-06-16 | 2009-11-04 | 北京理工大学 | Ultrasonic guided wave detection technology for quantifying defects of composite laminated plate |
| CN102175768A (en) * | 2011-02-22 | 2011-09-07 | 哈尔滨工业大学 | Method and device for detecting defects and failures of high-speed rail based on vibration signals |
| WO2013097153A1 (en) * | 2011-12-27 | 2013-07-04 | 华南理工大学 | Radio ultrasound wave probe assembly for crawler belt-type steel rail test and test method thereof |
| CN104020221A (en) * | 2014-05-30 | 2014-09-03 | 杨媛 | Real-time broken-rail detecting and positioning system based on ultrasonic guided waves |
| CN104297346A (en) * | 2014-09-11 | 2015-01-21 | 天津大学 | Nondestructive detection system of sheet metal by ultrasonic planar guided-wave and detection method thereof |
| CN205176149U (en) * | 2015-11-10 | 2016-04-20 | 台州红炭电子科技有限公司 | RFID antenna performance testing arrangement on machine is got ready to RFID antenna |
| WO2019201176A1 (en) * | 2018-04-17 | 2019-10-24 | 江苏必得科技股份有限公司 | Method and device for predicting crack damage of train component |
| WO2019201178A1 (en) * | 2018-04-17 | 2019-10-24 | 江苏必得科技股份有限公司 | Train component crack damage detection method and system based on lamb wave imaging |
| CN108956787A (en) * | 2018-06-13 | 2018-12-07 | 西安理工大学 | A kind of rail failure detection method neural network based |
| CN109543303A (en) * | 2018-11-22 | 2019-03-29 | 华北水利水电大学 | A method of the Damage Assessment Method to be perturbed based on class curvature of the flexibility difference matrix and frequency |
| CN109649432A (en) * | 2019-01-23 | 2019-04-19 | 浙江大学 | Cloud platform rail integrity monitoring systems and method based on guided wave technology |
| WO2020156348A1 (en) * | 2019-01-31 | 2020-08-06 | 青岛理工大学 | Structural damage identification method based on ensemble empirical mode decomposition and convolution neural network |
| AU2020101234A4 (en) * | 2020-07-02 | 2020-08-06 | Hunan University Of Science And Technology | An Optimization Method for Excitation Parameters of Ultrasonic Infrared Thermography Crack Nondestructive Testing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111896625A (en) | 2020-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111896625B (en) | Rail damage real-time monitoring method and monitoring system thereof | |
| CN110389170B (en) | Train component crack damage detection method and system based on Lamb wave imaging | |
| CN108569607B (en) | Elevator fault early warning method based on bidirectional gating cyclic neural network | |
| CN111637964B (en) | Rail corrugation identification method | |
| CN111623868B (en) | Convolutional neural network construction method for rail corrugation identification | |
| CN101660969B (en) | Intelligent fault diagnosis method for gear box | |
| CN109649432B (en) | System and method for monitoring integrity of steel rail of cloud platform based on guided wave technology | |
| CN111767897B (en) | Rail crack defect identification method based on support vector machine | |
| CN1743839A (en) | Structure defect ultrasonic on-line intelligent identifying system and identifying method | |
| CN102735747A (en) | Defect quantitative identification method of high-speed magnetic flux leakage inspection of high-speed railway rails | |
| CN106198749A (en) | A kind of data fusion method of multiple sensor based on Metal Crack monitoring | |
| CN104949840A (en) | Diesel engine fault diagnosis method based on vibration analysis | |
| CN114166943A (en) | Active rail damage monitoring method and terminal based on nonlinear ultrasonic waves | |
| CN110533115B (en) | Quantitative evaluation method for transmission characteristics of track circuit based on variational modal decomposition | |
| CN114942436A (en) | Tunnel lining disease identification method and system based on geological radar and deep learning | |
| CN2809640Y (en) | Ultrasonic online intelligent recognition system for structural defect | |
| CN114660180A (en) | Sound emission and 1D CNNs-based light-weight health monitoring method and system for medium and small bridges | |
| CN106645399B (en) | A kind of damage Detection of Smart Composite Structure appraisal procedure and system | |
| CN102879475B (en) | Identification system for fatigue damage state of 16 manganese steel bearing member based on PCA and TDF | |
| US20230024104A1 (en) | Identification of false transformer humming using machine learning | |
| CN115270875A (en) | Diaphragm pump running state monitoring method based on deep learning | |
| CN115310499A (en) | Industrial equipment fault diagnosis system and method based on data fusion | |
| CN110940732B (en) | Steel rail structure corrosion online monitoring and evaluation method | |
| CN110579534B (en) | Non-baseline detection and positioning method for defects of steel plate with welding seam based on reciprocity damage | |
| CN113640394A (en) | Fatigue crack identification method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |