CN113945700A - Track damage detection method, device, equipment and storage medium - Google Patents
Track damage detection method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN113945700A CN113945700A CN202111209783.0A CN202111209783A CN113945700A CN 113945700 A CN113945700 A CN 113945700A CN 202111209783 A CN202111209783 A CN 202111209783A CN 113945700 A CN113945700 A CN 113945700A
- Authority
- CN
- China
- Prior art keywords
- track
- signal
- detection
- detected
- data
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/20—Metals
- G01N33/204—Structure thereof, e.g. crystal structure
- G01N33/2045—Defects
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Chemical & Material Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Evolutionary Biology (AREA)
- Mechanical Engineering (AREA)
- Bioinformatics & Computational Biology (AREA)
- Crystallography & Structural Chemistry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Food Science & Technology (AREA)
- Medicinal Chemistry (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Machines For Laying And Maintaining Railways (AREA)
Abstract
The application discloses a rail damage detection method, a device, equipment and a storage medium, comprising the following steps: acquiring a current track detection signal obtained after a track is detected in the running process of a train; preprocessing the current track detection signal to obtain track data to be detected; inputting the track data to be detected into a trained track detection model, and acquiring a track state type label which is output by the track detection model and corresponds to the track data to be detected; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set; and determining whether the track has track damage and a corresponding track damage type at present based on the track state type label. According to the rail damage detection method and device, real-time detection is carried out on the rail state through the pre-established rail detection model, multiple fault modes can be detected simultaneously, and the efficiency and accuracy of rail damage detection are improved.
Description
Technical Field
The present disclosure relates to the field of track detection technologies, and in particular, to a method, an apparatus, a device, and a storage medium for detecting track damage.
Background
At present, along with the continuous increase of rail train traffic volume such as subway, high-speed railway, ordinary train to and the change of locomotive condition, orbital damage problem becomes more and more outstanding, and the track damage can cause rail train to have various potential safety hazards in the operation process, can influence rail train's normal operating and take the comfort level, threatens passenger's life safety even, consequently has important meaning to rail damage's detection.
The rail damage forms are various, and the common damage forms include uneven joints, cracks inside the rail, surface peeling of the rail, rail corrugation, bolt fracture, rail deformation and the like. In the prior art, professional instruments are generally adopted to check rail damage in a subentry manner, and if the unevenness of a welding line is checked, a level gauge and a horizontal feeler gauge are used; inspecting internal cavities and internal cracks by using a flaw detector; visual inspection or images are used for inspecting the surface stripping of the track; when the rail corrugation is inspected, a special wavelength depth detector is used, and certainly, a method for applying a comprehensive detection vehicle is available at present, but the method is low in detection frequency, long in detection time and high in equipment price, so that the method cannot be widely applied to actual detection.
At present, the rail state detection technology at home and abroad is mainly divided into static detection and dynamic detection. Wherein, static detection is mainly through the manual work, light-duty measurement dolly detects the track, and dynamic detection mainly detects the track through equipment such as track detection car, comprehensive detection car, this kind of dynamic detection can not accomplish to carry out real-time detection to the track circuit, and also need occupy the line resource when detecting at every turn, also will influence the operating efficiency of circuit to a certain extent, the key is that current detection method detects the rate of accuracy not high, need the artificial experience to check the scene repeatedly, and there is the omission easily to some specific fault modes.
In summary, the conventional rail damage detection technology mainly depends on human subjectivity and uncertainty of instruments, the identification accuracy is low, and a detection method can usually identify only one fault mode, which results in low identification efficiency. Therefore, how to detect multiple fault modes is to improve the efficiency and accuracy of track fault detection.
Disclosure of Invention
In view of this, an object of the present application is to provide a method, an apparatus, a device and a storage medium for detecting rail damage, which can detect multiple failure modes simultaneously, improve efficiency and accuracy of rail failure detection, and have good repeatability of detection results. The specific scheme is as follows:
in a first aspect, the application discloses a rail damage detection method, including:
acquiring a current track detection signal obtained after a track is detected in the running process of a train;
preprocessing the current track detection signal to obtain track data to be detected;
inputting the track data to be detected into a trained track detection model, and acquiring a track state type label which is output by the track detection model and corresponds to the track data to be detected; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, wherein the training set comprises historical track data of a damaged track and a corresponding track damage type label;
and determining whether the track has track damage and a corresponding track damage type at present based on the track state type label.
Optionally, the detecting whether the current track detection signal is an abnormal signal subjected to external interference includes:
eliminating the direct current component of the current track detection signal to obtain a processed signal;
detecting whether any condition of positive and negative signal asymmetry condition, signal amplitude abnormity condition, signal amplitude limiting condition and signal jumping condition exists in the processed signal;
if so, judging that the current track detection signal is an abnormal signal subjected to external interference.
Optionally, the removing the dc component of the current track detection signal to obtain a processed signal includes:
calculating the mean value of the current track detection signal to obtain the direct current component of the current track detection signal;
and subtracting the direct current component from the current track detection signal to obtain a processed signal.
Optionally, the detecting whether the processed signal has any of a positive-negative signal asymmetry condition, a signal amplitude abnormality condition, a signal amplitude limiting condition, and a signal jump condition includes:
carrying out segmentation processing on the processed signal to obtain a plurality of signal segments, and calculating a positive maximum value and a negative maximum value corresponding to each signal segment; calculating the average value of all the positive maximum values to obtain a positive average value and calculating the average value of all the negative maximum values to obtain a negative average value; judging whether the difference value between the absolute values of the positive average value and the negative average value exceeds a preset difference value range, and if so, judging that the processed signal has a positive and negative signal asymmetry condition;
and/or judging whether the maximum absolute value of the amplitude in the processed signal is smaller than a first preset threshold value, and if so, judging that the processed signal has a signal amplitude abnormal condition;
and/or judging that the processed signal has a signal amplitude limiting condition when the proportion that the absolute value of the amplitude in the processed signal is greater than a second preset threshold value exceeds a first preset proportion threshold value;
and/or, carrying out segmentation processing on the processed signal to obtain a plurality of signal segments, and calculating the variance corresponding to each signal segment; and if the proportion of the variance exceeding a preset variance threshold value in all the signal segments is larger than a second preset proportion threshold value, judging that the processed signal has a signal jump condition.
Optionally, the preprocessing the current track detection signal to obtain track data to be detected includes:
and processing the normal signals which are not interfered by the outside according to a preset signal component fusion rule so as to fuse the signal components of the same type collected by different sensors in the current track detection signal to obtain the data of the track to be detected.
Optionally, the signal component fusion rule is to fuse the same type of data acquired by the sensors on the coaxial left and right wheels of the rail vehicle; and/or fusing data collected by sensors on front and rear wheels of the rail vehicle.
Optionally, the training set includes historical track data of damaged tracks and corresponding track damage type labels; wherein the historical orbit data is the preprocessed data.
Optionally, after determining whether there is a track damage in the current track and a corresponding track damage type based on the track status type tag, the method further includes:
if the current track has track damage, determining a track position corresponding to the current track detection signal, and sending the track position to a preset user terminal;
acquiring a track state type comparison result fed back by the preset user terminal, and using the track data to be detected and a track state type obtained after field on-site inspection as a training sample of the track detection model so as to retrain the track detection model; and the track state type comparison result is a result of comparing whether the track damage type obtained by the track detection model is consistent with the track state type obtained after the on-site inspection.
Optionally, the orbit detection model is a neural network model satisfying any parameter conditions of using 2 to 4 hidden layers, performing data normalization by z-score, using a unipolar sigmoid function as an activation function, having a learning rate of 0.001 to 0.1, a weight momentum factor of 0.5 to 0.9, and a sparse target value of 0.05 to 0.1.
Optionally, in the process of training the initial detection model constructed based on the machine learning algorithm by using the training set, the method further includes:
and setting the node existence probability in the initial detection model by adopting a random inactivation regularization method.
Optionally, the setting, by using a random inactivation regularization method, a node existence probability in the initial detection model includes:
determining a node existence probability threshold of a target layer in the initial detection model;
randomly distributing corresponding existence probability to each node of the target layer;
and respectively judging whether the existence probability corresponding to each node in the target layer is greater than or equal to the node existence probability threshold, if so, reserving the corresponding node in the target layer, and if not, deleting the corresponding node from the target layer.
In a second aspect, the present application discloses a rail flaw detection device, comprising:
the signal acquisition module is used for acquiring a current track detection signal obtained after a track is detected in the running process of the train;
the preprocessing module is used for preprocessing the current track detection signal to obtain to-be-detected track data;
the model detection module is used for inputting the to-be-detected track data into a trained track detection model and acquiring a track state type label which is output by the track detection model and corresponds to the to-be-detected track data; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, wherein the training set comprises historical track data of a damaged track and a corresponding track damage type label;
and the damage determining module is used for determining whether the track has track damage and a corresponding track damage type based on the track state type label.
In a third aspect, the present application discloses an electronic device comprising a processor and a memory; wherein the processor implements the aforementioned rail flaw detection method when executing the computer program stored in the memory.
In a fourth aspect, the present application discloses a computer readable storage medium for storing a computer program; wherein the computer program realizes the aforementioned rail damage detection method when executed by a processor.
According to the method, a current track detection signal obtained after a track is detected in the running process of a train is obtained, then the current track detection signal is preprocessed to obtain track data to be detected, the track data to be detected is input into a trained track detection model, and a track state type label which is output by the track detection model and corresponds to the track data to be detected is obtained; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, the training set comprises historical track data of damaged tracks and corresponding track damage type labels, and finally whether track damage exists in the current tracks or not and the corresponding track damage types are determined based on the track state type labels. Therefore, the rail state is detected through the rail damage detection model which is created in advance based on the machine learning algorithm, the subjectivity of people and the uncertainty of instruments are eliminated, multiple fault modes can be detected simultaneously, the efficiency and the accuracy of rail fault detection are improved, and the detection result has good repeatability.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a flow chart of a rail flaw detection method disclosed herein;
FIG. 2 is a flow chart of a particular trajectory detection model retraining method disclosed herein;
FIG. 3 is a schematic diagram of actual orbit training set sample information disclosed in the present application;
FIG. 4 is a schematic diagram of sample information of an actual track test set disclosed in the present application;
FIG. 5 is a sample test result diagram of a specific test set disclosed herein;
FIG. 6 is a schematic illustration of a field in-situ test result for a weld joint failure as disclosed herein;
FIG. 7 is a flow chart of a specific rail flaw detection method disclosed herein;
FIG. 8 is a schematic structural diagram of a rail flaw detection apparatus disclosed in the present application;
fig. 9 is a block diagram of an electronic device disclosed in the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The embodiment of the application discloses a rail damage detection method, which is shown in fig. 1 and comprises the following steps:
step S11: and acquiring a current track detection signal obtained after the track is detected in the running process of the train.
In this embodiment, first, a sensor installed on a rail train is required to collect a rail signal detected during a current train running process, so as to obtain a current rail detection signal. It is understood that, due to the different functions of the different sensors and the different positions of the rail train, there are differences in the types of signals included in the current track detection signal, which includes, but is not limited to, image data detected by the image sensor, sound data detected by the sound sensor, vibration acceleration data and vibration speed data collected by the vibration sensor, and the like.
Step S12: and preprocessing the current track detection signal to obtain the track data to be detected.
In this embodiment, after acquiring a current track detection signal obtained after detecting a track in a running process of a train, corresponding preprocessing is usually performed on the acquired current track detection signal to prevent interference outside the track in an acquisition process. For example, due to interference and faults of the track signals acquired by the detection instrument, some track signals may be abnormal, and the real situation of the track cannot be well reflected, so that the track signals with the abnormality need to be correspondingly preprocessed, for example, the identified abnormal track signals can be removed, and the track data to be detected can be obtained.
In this embodiment, the preprocessing the current track detection signal to obtain the to-be-detected track data may specifically include: and processing the normal signals which are not interfered by the outside according to a preset signal component fusion rule so as to fuse the signal components of the same type collected by different sensors in the current track detection signal to obtain the data of the track to be detected. The signal component fusion rule is that the same type data collected by the sensors on the coaxial left and right wheels of the rail vehicle are fused; and/or fusing data collected by sensors on front and rear wheels of the rail vehicle. It can be understood that one sensor may have a fault, data is lost, and data is limited, which results in inaccurate judgment, so that after a current track detection signal obtained after a track is detected in a running process of a train is obtained, if the current track detection signal is detected to be a normal signal without external interference, the normal signal without external interference can be processed according to a rule that data of the same type collected by sensors on the left and right coaxial wheels of the rail vehicle are fused, and/or data collected by sensors on the front and rear wheels of the rail vehicle are fused. The sensor includes, but is not limited to, an image sensor, a sound sensor, a vibration sensor, and the like. For example, the sound data collected by 2 sensors on the same shaft on a certain wheel pair of the same railway vehicle are fused, that is, the sound data collected by the sensors on the left and right wheels are fused; and/or fusing the data acquired by the sensors on the front wheel and the rear wheel of the same railway vehicle, namely the sensors on different parts of the same vehicle, so as to obtain the data of the railway to be detected.
Step S13: inputting the track data to be detected into a trained track detection model, and acquiring a track state type label which is output by the track detection model and corresponds to the track data to be detected; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, wherein the training set comprises historical track data of a damaged track and a corresponding track damage type label.
In this embodiment, after the current track detection signal is preprocessed to obtain track data to be detected, the track data to be detected is further input to a track detection model obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set in advance, and a track state type tag corresponding to the track data to be detected and output by the track detection model is obtained. The track detection model is a model obtained by acquiring historical track detection signals obtained after a track is detected in the running process of a train, then preprocessing the acquired signal data, and inputting a training set obtained after preprocessing into an initial detection model which is constructed in advance based on a machine learning algorithm for training. The machine learning Algorithm includes, but is not limited to, DNN (Deep Neural Networks), SVM (Support Vector Machines), GA (Genetic Algorithm), and the like.
In this embodiment, the training set may specifically include historical track data of damaged tracks and corresponding track damage type labels; wherein the historical orbit data is the preprocessed data. Specifically, the training set includes historical track data of a damaged track and a track damage type label obtained by manually labeling the historical track data of the damaged track, for example, when model training is performed on damage with unsmooth welded joints, first, data such as vibration and impact at the axle box position needs to be collected by a sensor installed above an axle box body of a track train to obtain historical track data including impact, vibration, speed information and kilometer sign information, and the historical track data is classified to obtain four types of data sets of unsmooth welded joints, rail joints, turnouts and normal welded joints, and then, the four types of data sets are labeled by manually labeling to obtain a digital label 1 representing the unsmooth welded joints, a digital label 2 representing the rail joints, and a digital label 3 representing the turnouts, And inputting the four training sets of the digital labels 4 representing the normal welding joints into an initial detection model constructed in advance based on a deep neural network for training, and taking the trained model as a track detection model. It should be noted how many classes of labels are in the training set, and how many types of track states are detected in the subsequent model detection process. For example, only 20 fault sample data of uneven weld joints and 30 fault sample data of orbital corrugation are in the training set, and the corresponding orbit detection model can only identify two damage types, namely uneven weld joints and orbital corrugation. It should be noted that the historical track data is obtained by performing the same operation as the preprocessing on the historical track detection signals, that is, removing the abnormal signals and fusing the abnormal signals.
In this embodiment, the orbit detection model is a neural network model that satisfies any parameter conditions of using three hidden layers, performing data normalization by z-score, using a unipolar sigmoid function as an activation function, having a learning rate of 0.001 to 0.1, a weight momentum factor of 0.5 to 0.9, and a sparse target value of 0.05 to 0.1. It can be understood that the initial detection model constructed based on the deep neural network plays a decisive role in the accuracy of subsequent track state identification, so that the accuracy of final track state detection can be ensured only by selecting appropriate parameters when the initial detection model is constructed, and the detection efficiency is improved. Preferably, a neural network model with three hidden layers, data normalization by z-score, unipolar Sigmoid function as an activation function (i.e., Sigmoid function), Learning Rate of 0.001 (i.e., Learning Rate of 0.001), weight momentum factor of 0.5 (i.e., monentum of 0.5), and sparse target value of 0.05 is used.
In addition, in the process of training the initial detection model constructed based on the machine learning algorithm by using the training set, in order to prevent an overfitting phenomenon caused by insufficient sample set data, the method may specifically include: and setting the node existence probability in the initial detection model by adopting a random inactivation regularization method. I.e., by a random deactivation (i.e., Dropout) regularization method, such that some nodes in the model are randomly closed during each iteration of the initial detection model, thereby reducing the complexity of the network model. The specific process can comprise the following steps: determining a node existence probability threshold of a target layer in the initial detection model; randomly distributing corresponding existence probability to each node of the target layer; and respectively judging whether the existence probability corresponding to each node in the target layer is greater than or equal to the node existence probability threshold, if so, reserving the corresponding node in the target layer, and if not, deleting the corresponding node from the target layer. For example, for the node of the k-th layer in the initial detection model, a number keep _ prob belonging to (0, 1) is selected to indicate the existence probability of each node, a number d belonging to (0, 1) is randomly allocated to all nodes of the k-th layer, if d is less than or equal to keep _ prob, the node is saved, if d is greater than keep _ prob, the node is deleted, namely, the output value representing the forward propagation of the node in the iteration is 0, for the node under saving, the new output value Z is the original output value A divided by keep _ prob, and when the node under saving is in the reverse propagation, the new output value A dA of the node under saving is divided by the keep _ prob.
Step S14: and determining whether the track has track damage and a corresponding track damage type at present based on the track state type label.
In this embodiment, after the track data to be detected is input to the trained track detection model and the track state type tag corresponding to the track data to be detected and output by the track detection model is obtained, it may be determined whether the track has track damage and a corresponding track damage type through the track state type tag. It can be understood that, before the initial detection model training is performed, the track state type labels corresponding to the historical track data, which are classified in a manual labeling manner, have been acquired, so that whether the track has track damage and a corresponding track damage type can be directly judged according to the correspondence between the output labels and the track states. For example, when a digital label with 1 is output, the rail fault type can be directly determined to be a fault of welding irregularity; when the digital label of 2 is output, the track fault type can be directly determined to be the fault of the internal crack through the track state type corresponding to the digital label 2.
Further, as shown in fig. 2, after determining whether a track damage exists in the current track and a corresponding track damage type based on the track state type tag, the method may specifically include:
step S21: if the current track has track damage, determining a track position corresponding to the current track detection signal, and sending the track position to a preset user terminal;
step S22: acquiring a track state type comparison result fed back by the preset user terminal, and using the track data to be detected and a track state type obtained after field on-site inspection as a training sample of the track detection model so as to retrain the track detection model; and the track state type comparison result is a result of comparing whether the track damage type obtained by the track detection model is consistent with the track state type obtained after the on-site inspection.
In this embodiment, if it is determined that there is a track damage in the current track through the track state type tag, a track position corresponding to the current track detection signal may be further determined, and the determined position information is sent to a preset user terminal, so that a user may obtain a specific position corresponding to the track damage through the user terminal, perform field and field inspection on the damage in the position in a manual confirmation manner, input a track state type obtained after the field and field inspection to the user terminal, compare the track damage type obtained by the track detection model with a track state type obtained after the field and determine whether the track state types are consistent with each other, and obtain a track state type comparison result. If the track state type comparison result is inconsistent, the track data to be detected and the track state type obtained after the on-site inspection can be used as training samples of the track detection model, and the track detection model is retrained to obtain an updated track detection model; if the track state type comparison result is consistent, the track data to be detected and the track state type obtained after the on-site inspection, namely the track damage type obtained by the track detection model, can be used as a training sample of the track detection model, the track detection model is retrained, an updated track detection model is obtained, and the accuracy of the track detection model detection can be continuously improved through the training.
In a specific embodiment, historical actual operation monitoring data of a subway line is collected by adopting a rotating speed with a rotating speed value of 400, historical track detection signals containing impact, vibration and speed information and kilometer sign information are obtained, the collected historical track detection signals are classified and labeled in a manual labeling mode, four data sets of unsmooth welding joints, rail joints, turnouts and normal welding joints are obtained, the data sets are preprocessed and grouped, a training set shown in a figure 3 and a test set shown in a figure 4 are obtained, and the training and testing are used for training and testing of a subsequent initial detection model constructed based on a deep neural network. It is noted that the criterion for the uneven damage of the welded joint is set to be greater than or equal to 0.3mm, according to the specific application requirements.
Further, the training set is input into an initial detection model which is constructed in advance based on a deep neural network for training, the trained initial detection model is verified by using the test set, a track state type label corresponding to the test set is output, the track state type label is counted, a test set result shown in fig. 5 is obtained, and the accuracy rate of detection of the trained initial detection model exceeds 90% through calculation. In order to further verify the effectiveness of the initial detection model, two intervals are selected from the final test result to carry out on-site investigation verification, and the intervals output 12 positions where the welding joint is considered to have the unsmooth damage. The 12 positions were inspected on site to obtain inspection results as shown in fig. 6, and the initial detection model after training was found to have an accuracy of 83.3% and a false alarm rate of 16.7% by counting the inspection results. After the verification set is verified, the training set data corresponding to the detection error and the track state type obtained after field on-site inspection can be used as the training sample of the track detection model, and the track detection model can be retrained. For example, the training set data corresponding to the error prediction results with numbers 11 and 12 in fig. 6 and the track state type obtained after field and field inspection are used as the training samples of the track detection model, and the track detection model is retrained.
As can be seen, in the embodiment of the application, a current track detection signal obtained after a track is detected in a running process of a train is obtained, the current track detection signal is preprocessed to obtain track data to be detected, the track data to be detected is input into a trained track detection model, and a track state type label corresponding to the track data to be detected and output by the track detection model is obtained; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, the training set comprises historical track data of damaged tracks and corresponding track damage type labels, and finally whether track damage exists in the current tracks or not and the corresponding track damage types are determined based on the track state type labels. Therefore, the rail state is detected through the rail damage detection model which is created in advance based on the machine learning algorithm, the subjectivity of people and the uncertainty of instruments are avoided, multiple fault modes can be detected simultaneously, the efficiency and the accuracy of rail fault detection are improved, and the detection result has good repeatability.
The embodiment of the application discloses a specific rail damage detection method, and as shown in fig. 7, the method includes:
step S31: and acquiring a current track detection signal obtained after the track is detected in the running process of the train.
Step S32: and detecting whether the current track detection signal is an abnormal signal interfered by the outside.
In this embodiment, after the current track detection signal obtained by detecting the track in the running process of the train is obtained, the current track detection signal needs to be further detected to determine whether there is an abnormal signal interfered by the outside. For example, some track signals may have abnormalities due to interference and faults of the track signals collected by the detection instrument, and the track signals with the abnormalities generally have specific characteristics, such as signal discontinuity, asymmetry of positive and negative signals, and the like.
In this embodiment, the detecting whether the current track detection signal is an abnormal signal that is interfered by the outside specifically may include: eliminating the direct current component of the current track detection signal to obtain a processed signal; detecting whether any condition of positive and negative signal asymmetry condition, signal amplitude abnormity condition, signal amplitude limiting condition and signal jumping condition exists in the processed signal; if so, judging that the current track detection signal is an abnormal signal subjected to external interference. In this embodiment, after a current track detection signal obtained after a track is detected in a running process of a train is obtained, the current track detection signal needs to be standardized first, and a direct current component in the current track detection signal is removed to obtain a processed signal. Further, the processed signal is detected, whether any one of a positive-negative signal asymmetry condition, a signal amplitude abnormal condition, a signal amplitude limiting condition and a signal jumping condition exists in the processed signal is judged, and if any one of the positive-negative signal asymmetry condition, the signal amplitude abnormal condition, the signal amplitude limiting condition and the signal jumping condition exists in the processed signal, the current track detection signal is judged to be an abnormal signal interfered by a detection instrument, namely an abnormal signal detected by external interference.
Specifically, the removing the dc component of the current track detection signal to obtain a processed signal may include: calculating the mean value of the current track detection signal to obtain the direct current component of the current track detection signal; and subtracting the direct current component from the current track detection signal to obtain a processed signal. In this embodiment, the mean value of the current track detection signal, that is, the dc component of the current track detection signal, may be calculated by a mean function.
In addition, the detecting whether the processed signal has any of a positive-negative signal asymmetry condition, a signal amplitude abnormality condition, a signal amplitude limiting condition, and a signal jump condition may specifically include: carrying out segmentation processing on the processed signal to obtain a plurality of signal segments, and calculating a positive maximum value and a negative maximum value corresponding to each signal segment; calculating the average value of all the positive maximum values to obtain a positive average value and calculating the average value of all the negative maximum values to obtain a negative average value; judging whether the difference value between the absolute values of the positive average value and the negative average value exceeds a preset difference value range, and if so, judging that the processed signal has a positive and negative signal asymmetry condition; and/or judging whether the maximum absolute value of the amplitude in the processed signal is smaller than a first preset threshold value, and if so, judging that the processed signal has a signal amplitude abnormal condition; and/or judging that the processed signal has a signal amplitude limiting condition when the proportion that the absolute value of the amplitude in the processed signal is greater than a second preset threshold value exceeds a first preset proportion threshold value; and/or, carrying out segmentation processing on the processed signal to obtain a plurality of signal segments, and calculating the variance corresponding to each signal segment; and if the proportion of the variance exceeding a preset variance threshold value in all the signal segments is larger than a second preset proportion threshold value, judging that the processed signal has a signal jump condition. For example, in a first specific embodiment, the specific steps of detecting whether the processed signal has positive and negative signal asymmetry are as follows: dividing the processed signals into m sections, calculating the positive maximum value and the negative maximum value of each section, averaging the m positive maximum values to obtain an average value A, averaging the m negative maximum values, and then taking an absolute value to obtain an average value B, if the difference value of A and B exceeds a preset difference value range, judging that the processed signals are in asymmetric distribution, and judging the corresponding current track detection signals as abnormal signals subjected to external interference; and/or, in a second specific embodiment, the specific steps of detecting the abnormal condition of the signal amplitude are: if the maximum absolute value of the amplitude in the processed signals is smaller than a first preset threshold, the processed signals are considered to have abnormal signal amplitude, and the corresponding current track detection signals are judged to be abnormal signals subjected to external interference; and/or, in a third specific embodiment, the specific steps of detecting the signal clipping condition are: if the ratio of the absolute value of the amplitude in the processed signal larger than the second preset threshold exceeds the first preset ratio threshold, judging that the processed signal has a signal amplitude limiting condition, and judging the corresponding current track detection signal as an abnormal signal subjected to external interference; and/or, in a fourth specific embodiment, the specific step of detecting the signal transition situation is: and performing segmentation processing on the processed signal to obtain a plurality of signal segments, respectively calculating the variance corresponding to each signal segment, comparing the variance in all the signal segments with a preset variance threshold, if the occupation ratio of the variance exceeding the preset variance threshold is larger than a second preset proportion threshold, judging that the processed signal has signal jump, and judging that the corresponding current track detection signal is an abnormal signal subjected to external interference.
It should be noted that, in this embodiment, the method for determining that the current track detection signal is an abnormal signal subjected to external interference includes, but is not limited to, detecting whether the processed signal has a positive-negative signal asymmetry condition, a signal amplitude abnormal condition, a signal amplitude limiting condition, a signal jump condition, and the like.
Step S33: and if the current track detection signal is detected to be an abnormal signal subjected to external interference, rejecting the current track detection signal, and re-executing the step of acquiring the current track detection signal obtained after the track is detected in the running process of the train, and if the current track detection signal is detected to be a normal signal not subjected to external interference, determining the data of the track to be detected based on the current track detection signal.
In this embodiment, after detecting whether the current track detection signal is an abnormal signal that is interfered by the outside world, if it is detected that the current track detection signal is determined to be the abnormal signal that is interfered by the outside world, the current track detection signal is rejected, and the step of obtaining the current track detection signal obtained after detecting the track in the running process of the train is executed again.
Step S34: inputting the track data to be detected into a trained track detection model, and acquiring a track state type label which is output by the track detection model and corresponds to the track data to be detected; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, wherein the training set comprises historical track data of a damaged track and a corresponding track damage type label.
Step S35: and determining whether the track has track damage and a corresponding track damage type at present based on the track state type label.
For more specific processing procedures of the steps S31, S34, and S35, reference may be made to corresponding contents disclosed in the foregoing embodiments, and details are not repeated here.
It can be seen that, this application embodiment detects after obtaining the current track detection signal that obtains after surveying the track in-process that the train goes current track detection signal is for receiving external disturbance's abnormal signal, if detects current track detection signal is for receiving external disturbance's abnormal signal, then rejects current track detection signal, through the rejection to abnormal signal, can guarantee the validity of the current track detection signal of input, and then ensures the exactness of testing result, has improved track fault detection's the degree of accuracy.
Correspondingly, the embodiment of the present application further discloses a rail damage detection device, as shown in fig. 8, the device includes:
the signal acquisition module 11 is configured to acquire a current track detection signal obtained after a track is detected in a running process of a train;
the preprocessing module 12 is configured to preprocess the current track detection signal to obtain to-be-detected track data;
the model detection module 13 is configured to input the to-be-detected track data to a trained track detection model, and acquire a track state type tag corresponding to the to-be-detected track data output by the track detection model; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, wherein the training set comprises historical track data of a damaged track and a corresponding track damage type label;
and a damage determining module 14, configured to determine whether there is a track damage in the track and a corresponding track damage type based on the track status type tag.
For the specific work flow of each module, reference may be made to corresponding content disclosed in the foregoing embodiments, and details are not repeated here.
In the embodiment of the application, a current track detection signal obtained after a track is detected in the running process of a train is obtained, the current track detection signal is preprocessed to obtain track data to be detected, the track data to be detected is input into a trained track detection model, and a track state type label which is output by the track detection model and corresponds to the track data to be detected is obtained; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, the training set comprises historical track data of damaged tracks and corresponding track damage type labels, and finally whether track damage exists in the current tracks or not and the corresponding track damage types are determined based on the track state type labels. Therefore, the rail state is detected through the rail damage detection model which is created in advance based on the machine learning algorithm, the subjectivity of people and the uncertainty of instruments are avoided, multiple fault modes can be detected simultaneously, the efficiency and the accuracy of rail fault detection are improved, and the detection result has good repeatability.
In some specific embodiments, the preprocessing module 12 may specifically include:
the first abnormal signal detection unit is used for detecting whether the current track detection signal is an abnormal signal interfered by the outside;
a first rejecting unit, configured to reject the current track detection signal if it is detected that the current track detection signal is an abnormal signal that is interfered by the outside, and to re-execute the step of acquiring the current track detection signal obtained after detecting the track in the running process of the train;
and the data determining unit is used for determining the track data to be detected based on the current track detection signal if the current track detection signal is detected to be a normal signal which is not interfered by the outside.
In some specific embodiments, the first abnormal signal detecting unit may specifically include:
the second eliminating unit is used for eliminating the direct current component of the current track detection signal to obtain a processed signal;
the second abnormal signal detection unit is used for detecting whether the processed signal has any one of a positive and negative signal asymmetry condition, a signal amplitude abnormal condition, a signal amplitude limiting condition and a signal jumping condition;
and the first abnormal signal determining unit is used for judging that the current track detection signal is an abnormal signal subjected to external interference if any one of the conditions of asymmetry of positive and negative signals, abnormal signal amplitude, signal amplitude limiting and signal jumping exists in the processed signal.
In some specific embodiments, the second rejecting unit may specifically include:
the mean value calculating unit is used for calculating the mean value of the current track detection signal so as to obtain the direct current component of the current track detection signal;
and the direct current component removing unit is used for subtracting the direct current component from the current track detection signal to obtain a processed signal.
In some specific embodiments, the first abnormal signal determining unit may specifically include:
the second abnormal signal determining unit is used for carrying out segmentation processing on the processed signal to obtain a plurality of signal segments and calculating a positive maximum value and a negative maximum value corresponding to each signal segment; calculating the average value of all the positive maximum values to obtain a positive average value and calculating the average value of all the negative maximum values to obtain a negative average value; judging whether the difference value between the absolute values of the positive average value and the negative average value exceeds a preset difference value range, and if so, judging that the processed signal has a positive and negative signal asymmetry condition;
and/or a third abnormal signal determining unit, configured to determine whether a maximum absolute value of an amplitude in the processed signal is smaller than a first preset threshold, and if so, determine that a signal amplitude abnormality exists in the processed signal;
and/or the fourth abnormal signal determining unit is used for judging that the processed signal has a signal amplitude limiting condition when the ratio of the absolute value of the amplitude of the processed signal greater than the second preset threshold value exceeds the first preset ratio threshold value;
and/or a fifth abnormal signal determining unit, configured to perform segmentation processing on the processed signal to obtain a plurality of signal segments, and calculate a variance corresponding to each signal segment; and if the proportion of the variance exceeding a preset variance threshold value in all the signal segments is larger than a second preset proportion threshold value, judging that the processed signal has a signal jump condition.
In some specific embodiments, the preprocessing module 12 may specifically include:
and the fusion unit is used for processing the normal signals which are not interfered by the outside according to a preset signal component fusion rule so as to fuse the signal components of the same type collected by different sensors in the current track detection signal and obtain the data of the track to be detected.
In some specific embodiments, the signal component fusion rule is to fuse the same type of data collected by the sensors on the coaxial left and right wheels of the rail vehicle; and/or fusing data collected by sensors on front and rear wheels of the rail vehicle.
In some embodiments, the training set includes historical track data for damaged tracks and corresponding track damage type labels; wherein the historical orbit data is the preprocessed data.
In some specific embodiments, after the damage determining module 14, the method may further include:
the position determining unit is used for determining a track position corresponding to the current track detection signal and sending the track position to a preset user terminal if the track is damaged;
the comparison result acquisition unit is used for acquiring a comparison result of the track state types fed back by the preset user terminal, and using the track data to be detected and the track state types obtained after field on-site inspection as training samples of the track detection model so as to retrain the track detection model; and the track state type comparison result is a result of comparing whether the track damage type obtained by the track detection model is consistent with the track state type obtained after the on-site inspection.
In some embodiments, the orbit detection model is a neural network model satisfying any parameter condition of using 2 to 4 hidden layers, performing data normalization by z-score, using a unipolar sigmoid function as an activation function, having a learning rate of 0.001 to 0.1, a weighted momentum factor of 0.5 to 0.9, and a sparseness target value of 0.05 to 0.1.
In some specific embodiments, in the process of training the initial detection model constructed based on the machine learning algorithm by using the training set, the method may further include:
and the existence probability setting unit is used for setting the existence probability of the nodes in the initial detection model by adopting a random inactivation regularization method.
In some specific embodiments, the existence probability setting unit may specifically include:
a probability threshold existence determining unit, configured to determine a probability threshold of existence of a node of a target layer in the initial detection model;
the existence probability distribution unit is used for randomly distributing corresponding existence probabilities for each node of the target layer;
and the judging unit is used for respectively judging whether the existence probability corresponding to each node in the target layer is greater than or equal to the node existence probability threshold value, if so, reserving the corresponding node in the target layer, and if not, deleting the corresponding node from the target layer.
Further, an electronic device is disclosed in the embodiments of the present application, and fig. 9 is a block diagram of an electronic device 20 according to an exemplary embodiment, which should not be construed as limiting the scope of the application.
Fig. 9 is a schematic structural diagram of an electronic device 20 according to an embodiment of the present disclosure. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input output interface 25, and a communication bus 26. The memory 22 is used for storing a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the track damage detection method disclosed in any of the foregoing embodiments. In addition, the electronic device 20 in the present embodiment may be specifically an electronic computer.
In this embodiment, the power supply 23 is configured to provide a working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and an external device, and a communication protocol followed by the communication interface is any communication protocol applicable to the technical solution of the present application, and is not specifically limited herein; the input/output interface 25 is configured to obtain external input data or output data to the outside, and a specific interface type thereof may be selected according to specific application requirements, which is not specifically limited herein.
In addition, the storage 22 is used as a carrier for resource storage, and may be a read-only memory, a random access memory, a magnetic disk or an optical disk, etc., and the resources stored thereon may include an operating system 221, a computer program 222, etc., and the storage manner may be a transient storage or a permanent storage.
The operating system 221 is used for managing and controlling each hardware device on the electronic device 20 and the computer program 222, and may be Windows Server, Netware, Unix, Linux, or the like. The computer program 222 may further include a computer program that can be used to perform other specific tasks in addition to the computer program that can be used to perform the rail flaw detection method disclosed in any of the foregoing embodiments and executed by the electronic device 20.
Further, the present application also discloses a computer-readable storage medium for storing a computer program; wherein the computer program when executed by a processor implements the rail damage detection method as disclosed above. For the specific steps of the method, reference may be made to the corresponding contents disclosed in the foregoing embodiments, which are not described herein again.
The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The track damage detection method, apparatus, device and storage medium provided by the present application are introduced in detail above, and a specific example is applied in the present application to explain the principle and the implementation of the present application, and the description of the above embodiment is only used to help understand the method and the core idea of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Claims (15)
1. A rail flaw detection method is characterized by comprising the following steps:
acquiring a current track detection signal obtained after a track is detected in the running process of a train;
preprocessing the current track detection signal to obtain track data to be detected;
inputting the track data to be detected into a trained track detection model, and acquiring a track state type label which is output by the track detection model and corresponds to the track data to be detected; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, wherein the training set comprises historical track data of a damaged track and a corresponding track damage type label;
and determining whether the track has track damage and a corresponding track damage type at present based on the track state type label.
2. The method according to claim 1, wherein the preprocessing the current track detection signal to obtain the track data to be detected comprises:
detecting whether the current track detection signal is an abnormal signal interfered by the outside;
and if the current track detection signal is detected to be an abnormal signal subjected to external interference, rejecting the current track detection signal, and re-executing the step of acquiring the current track detection signal obtained after the track is detected in the running process of the train, and if the current track detection signal is detected to be a normal signal not subjected to external interference, determining the data of the track to be detected based on the current track detection signal.
3. The method according to claim 2, wherein the detecting whether the current track detection signal is an abnormal signal subjected to external interference comprises:
eliminating the direct current component of the current track detection signal to obtain a processed signal;
detecting whether any condition of positive and negative signal asymmetry condition, signal amplitude abnormity condition, signal amplitude limiting condition and signal jumping condition exists in the processed signal;
if so, judging that the current track detection signal is an abnormal signal subjected to external interference.
4. The method according to claim 3, wherein the removing the dc component of the current track detection signal to obtain a processed signal comprises:
calculating the mean value of the current track detection signal to obtain the direct current component of the current track detection signal;
and subtracting the direct current component from the current track detection signal to obtain a processed signal.
5. The method according to claim 3, wherein the detecting whether the processed signal has any of a positive-negative signal asymmetry condition, a signal amplitude abnormality condition, a signal amplitude limiting condition, and a signal jump condition comprises:
carrying out segmentation processing on the processed signal to obtain a plurality of signal segments, and calculating a positive maximum value and a negative maximum value corresponding to each signal segment; calculating the average value of all the positive maximum values to obtain a positive average value and calculating the average value of all the negative maximum values to obtain a negative average value; judging whether the difference value between the absolute values of the positive average value and the negative average value exceeds a preset difference value range, and if so, judging that the processed signal has a positive and negative signal asymmetry condition;
and/or judging whether the maximum absolute value of the amplitude in the processed signal is smaller than a first preset threshold value, and if so, judging that the processed signal has a signal amplitude abnormal condition;
and/or judging that the processed signal has a signal amplitude limiting condition when the proportion that the absolute value of the amplitude in the processed signal is greater than a second preset threshold value exceeds a first preset proportion threshold value;
and/or, carrying out segmentation processing on the processed signal to obtain a plurality of signal segments, and calculating the variance corresponding to each signal segment; and if the proportion of the variance exceeding a preset variance threshold value in all the signal segments is larger than a second preset proportion threshold value, judging that the processed signal has a signal jump condition.
6. The method according to claim 2, wherein the preprocessing the current track detection signal to obtain the track data to be detected comprises:
and processing the normal signals which are not interfered by the outside according to a preset signal component fusion rule so as to fuse the signal components of the same type collected by different sensors in the current track detection signal to obtain the data of the track to be detected.
7. The rail damage detection method according to claim 6, wherein the signal component fusion rule is to fuse the same type of data collected by the sensors on the left and right coaxial wheels of the rail vehicle; and/or fusing data collected by sensors on front and rear wheels of the rail vehicle.
8. The method according to claim 7, wherein the training set comprises historical track data of damaged tracks and corresponding track damage type labels; wherein the historical orbit data is the preprocessed data.
9. The method according to any one of claims 1 to 8, wherein after determining whether there is a track damage in the current track and a corresponding track damage type based on the track status type tag, the method further comprises:
if the current track has track damage, determining a track position corresponding to the current track detection signal, and sending the track position to a preset user terminal;
acquiring a track state type comparison result fed back by the preset user terminal, and using the track data to be detected and a track state type obtained after field on-site inspection as a training sample of the track detection model so as to retrain the track detection model; and the track state type comparison result is a result of comparing whether the track damage type obtained by the track detection model is consistent with the track state type obtained after the on-site inspection.
10. The method according to any one of claims 1 to 8, wherein the orbit detection model is a neural network model satisfying any parameter condition of using 2 to 4 hidden layers, performing data normalization by z-score, using a unipolar sigmoid function as an activation function, having a learning rate of 0.001 to 0.1, a weight momentum factor of 0.5 to 0.9, and a sparseness target value of 0.05 to 0.1.
11. The rail damage detection method according to any one of claims 1 to 8, wherein the training of the initial detection model constructed based on the machine learning algorithm by using the training set further comprises:
and setting the node existence probability in the initial detection model by adopting a random inactivation regularization method.
12. The rail damage detection method according to claim 11, wherein the setting of the probability of existence of the node in the initial detection model by using a random inactivation regularization method comprises:
determining a node existence probability threshold of a target layer in the initial detection model;
randomly distributing corresponding existence probability to each node of the target layer;
and respectively judging whether the existence probability corresponding to each node in the target layer is greater than or equal to the node existence probability threshold, if so, reserving the corresponding node in the target layer, and if not, deleting the corresponding node from the target layer.
13. A rail flaw detection device, comprising:
the signal acquisition module is used for acquiring a current track detection signal obtained after a track is detected in the running process of the train;
the preprocessing module is used for preprocessing the current track detection signal to obtain to-be-detected track data;
the model detection module is used for inputting the to-be-detected track data into a trained track detection model and acquiring a track state type label which is output by the track detection model and corresponds to the to-be-detected track data; the trained track detection model is obtained by training an initial detection model constructed based on a machine learning algorithm by using a training set, wherein the training set comprises historical track data of a damaged track and a corresponding track damage type label;
and the damage determining module is used for determining whether the track has track damage and a corresponding track damage type based on the track state type label.
14. An electronic device comprising a processor and a memory; wherein the processor implements the method of rail damage detection according to any one of claims 1 to 12 when executing the computer program stored in the memory.
15. A computer-readable storage medium for storing a computer program; wherein the computer program when executed by a processor implements the method of rail impairment detection of any one of claims 1 to 12.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111209783.0A CN113945700A (en) | 2021-10-18 | 2021-10-18 | Track damage detection method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111209783.0A CN113945700A (en) | 2021-10-18 | 2021-10-18 | Track damage detection method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113945700A true CN113945700A (en) | 2022-01-18 |
Family
ID=79331261
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111209783.0A Pending CN113945700A (en) | 2021-10-18 | 2021-10-18 | Track damage detection method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113945700A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114354106A (en) * | 2022-02-09 | 2022-04-15 | 中大检测(湖南)股份有限公司 | Bridge intelligent monitoring, early warning and analyzing system based on RISC-V |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109142524A (en) * | 2018-08-14 | 2019-01-04 | 唐智科技湖南发展有限公司 | A kind of track damage detecting method, device and equipment |
| CN110246134A (en) * | 2019-06-24 | 2019-09-17 | 株洲时代电子技术有限公司 | A kind of rail defects and failures sorter |
| CN112308855A (en) * | 2020-11-27 | 2021-02-02 | 株洲时代电子技术有限公司 | Rail damage recognition model generation device, damage detection device and system |
-
2021
- 2021-10-18 CN CN202111209783.0A patent/CN113945700A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109142524A (en) * | 2018-08-14 | 2019-01-04 | 唐智科技湖南发展有限公司 | A kind of track damage detecting method, device and equipment |
| CN110246134A (en) * | 2019-06-24 | 2019-09-17 | 株洲时代电子技术有限公司 | A kind of rail defects and failures sorter |
| CN112308855A (en) * | 2020-11-27 | 2021-02-02 | 株洲时代电子技术有限公司 | Rail damage recognition model generation device, damage detection device and system |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114354106A (en) * | 2022-02-09 | 2022-04-15 | 中大检测(湖南)股份有限公司 | Bridge intelligent monitoring, early warning and analyzing system based on RISC-V |
| CN114354106B (en) * | 2022-02-09 | 2024-03-15 | 中大智能科技股份有限公司 | RISC-V based bridge intelligent monitoring early warning analysis system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Rizvi et al. | Crack detection in railway track using image processing | |
| CN103593973B (en) | A kind of urban road traffic situation assessment system | |
| CN113562562A (en) | Elevator safety early warning monitoring system and judgment method thereof | |
| CN113878214B (en) | Welding quality real-time detection method and system based on LSTM and residual distribution | |
| Ye et al. | Shock detection of rotating machinery based on activated time-domain images and deep learning: An application to railway wheel flat detection | |
| CN113705412A (en) | Deep learning-based multi-source data fusion track state detection method and device | |
| CN111144485A (en) | Vehicle accident judgment method and system based on xgboost classification algorithm | |
| CN113945700A (en) | Track damage detection method, device, equipment and storage medium | |
| CN112529109A (en) | Unsupervised multi-model-based anomaly detection method and system | |
| CN102622883B (en) | Method and device for judging release of traffic incident | |
| CN113469388B (en) | Maintenance system and method for rail transit vehicle | |
| Hassan et al. | Road crack detection using the particle filter | |
| CN116380496B (en) | Automobile door fatigue endurance test method, system and medium | |
| Zeng et al. | Rail break prediction and cause analysis using imbalanced in-service train data | |
| CN117078585A (en) | Method, device, equipment and medium for detecting defect damage of pipeline | |
| Arjapure et al. | Review on analysis techniques for road pothole detection | |
| CN112330615B (en) | Method and system for monitoring state of high-strength bolt of rotating part | |
| CN104504713A (en) | Automatic trouble identification method based on TEDS (trouble of moving EMU detection system) | |
| Asplund et al. | Assessment of the data quality of wayside wheel profile measurements. | |
| Vincent et al. | A Cognitive Rail Track Breakage Detection System Using Artificial Neural Network | |
| CN111984676A (en) | Method and device for determining damage of internal parts of vehicle | |
| Shadfar et al. | An index for rail weld health assessment in urban metro using in-service train | |
| CN110956143A (en) | Abnormal behavior detection method and device, electronic equipment and storage medium | |
| Van Lint et al. | Toward a robust framework for freeway travel time prediction: Experiments with simple imputation and state-space neural networks | |
| Kondaka et al. | Artfds–advanced railway track fault detection system using machine learning |
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 |