CN114862743A - Method and system for determining rail damage based on convolutional neural network - Google Patents
Method and system for determining rail damage based on convolutional neural network Download PDFInfo
- Publication number
- CN114862743A CN114862743A CN202110150634.5A CN202110150634A CN114862743A CN 114862743 A CN114862743 A CN 114862743A CN 202110150634 A CN202110150634 A CN 202110150634A CN 114862743 A CN114862743 A CN 114862743A
- Authority
- CN
- China
- Prior art keywords
- steel rail
- damage
- rail
- information
- position information
- 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
- 230000006378 damage Effects 0.000 title claims abstract description 194
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 40
- 229910000831 Steel Inorganic materials 0.000 claims abstract description 265
- 239000010959 steel Substances 0.000 claims abstract description 265
- 230000002159 abnormal effect Effects 0.000 claims abstract description 70
- 238000004458 analytical method Methods 0.000 claims abstract description 30
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 239000000523 sample Substances 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 5
- 238000001514 detection method Methods 0.000 abstract description 8
- 238000003466 welding Methods 0.000 description 8
- 238000012545 processing Methods 0.000 description 6
- 238000007689 inspection Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 208000027418 Wounds and injury Diseases 0.000 description 4
- 208000014674 injury Diseases 0.000 description 4
- 230000010354 integration Effects 0.000 description 4
- 239000003832 thermite Substances 0.000 description 4
- 238000012790 confirmation Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 239000011324 bead Substances 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013434 data augmentation Methods 0.000 description 2
- 238000013523 data management Methods 0.000 description 2
- 239000012634 fragment Substances 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000002787 reinforcement Effects 0.000 description 2
- 230000003014 reinforcing effect Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 241000272814 Anser sp. Species 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- 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
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30136—Metal
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Investigating Or Analyzing Materials By The Use Of Magnetic Means (AREA)
Abstract
The invention provides a rail damage determination method based on a convolutional neural network, which comprises the following steps: preprocessing original multi-dimensional time-series steel rail data, analyzing a convolutional neural network, and outputting current steel rail support bit information; after the current steel rail supporting position information is compared with the historical steel rail supporting position information, marking the consistent supporting position information as the same signal mark and marking the abnormal supporting position information as an abnormal signal mark; and after the artificial damage logic summary content and expert experience analysis of the steel rail are carried out on the abnormal support position information marked as the abnormal signal mark, a suspected steel rail damage signal and the corresponding suspected steel rail support position damage information are obtained. The determining system applying the method comprises a current steel rail supporting position information acquiring part, a supporting position information signal marking part and a steel rail suspected damage information analyzing part. The invention can provide more accurate and rapid rail damage information for rail flaw detection.
Description
Technical Field
The invention relates to a rail damage determining method, in particular to a rail damage determining method and a rail damage determining system based on a convolutional neural network.
Background
With the high-speed development of road traffic, the steel rail is used as the foundation of railway operation, the safety and the reliability of train operation are guaranteed, and the steel rail is important work for daily maintenance of the railway. The damage inside the steel rail is also the heavy part of the damage of the steel rail, and is also the part with more manpower investment for ensuring the safe operation of the steel rail.
The common condition of manual flaw detection is that the rail break event occurs due to the missed judgment and the misjudgment caused by artificial unstable factors in the flaw detection process, and the life and property safety of carriage personnel can be endangered if the train can not normally run in a short time at a low rate and if the train can not normally run in a short time at a high rate.
The main form of current manual inspection is through the dolly data acquisition of detecting a flaw, and playback personnel use B to show the image, can know the play ripples characteristics of different probes, and the work experience of reunion oneself detects different injuries. However, the labor investment is large, the efficiency is low, and uncontrollable factors such as missing judgment are too many.
Disclosure of Invention
Aiming at the defects existing in the problems, the invention provides the rail damage determining method and the rail damage determining system based on the convolutional neural network, which can provide more accurate and rapid rail damage information for rail flaw detection and can ensure safe and reliable operation of railways.
In order to achieve the purpose, the invention provides a rail damage determining method based on a convolutional neural network, which comprises the following steps of:
s1, preprocessing the original multi-dimensional time-series steel rail data, analyzing the convolutional neural network, and outputting current steel rail support position information corresponding to the current steel rail structure;
s2, comparing the current steel rail supporting position information with historical steel rail supporting position information, marking consistent supporting position information included in the current supporting position information as the same signal mark, and marking abnormal supporting position information included in the current supporting position information as an abnormal signal mark;
and S3, carrying out logic summary content of artificial damage of the steel rail and expert experience analysis on the abnormal supporting position information marked as the abnormal signal mark to obtain a suspected damage signal of the steel rail and the corresponding suspected damage information of the current supporting position of the steel rail.
The method for determining the rail damage based on the convolutional neural network comprises the following sub-steps in step S1:
s11, training or deducing the original multi-dimensional time-series steel rail data to form structural data of a probe channel;
and S12, analyzing the structural data of the probe channel by the convolutional neural network to output current steel rail support position information corresponding to the current steel rail structural body.
The method for determining the rail damage based on the convolutional neural network comprises the following sub-steps in step S2:
s21, inquiring whether the steel rail structure information database comprises a historical steel rail structure corresponding to the current steel rail structure, and calling historical steel rail support position information corresponding to the historical steel rail structure if the historical steel rail structure corresponding to the current steel rail structure is inquired;
s22, judging whether the current steel rail supporting position information and the historical steel rail supporting position information comprise the same supporting position information;
and S23, marking consistent supporting position information which is contained in the current steel rail supporting position information and is the same as the historical steel rail supporting position information as the same signal mark, and marking abnormal supporting position information which is contained in the current steel rail supporting position information and is different from the historical steel rail supporting position information as an abnormal signal mark.
The method for determining the rail damage based on the convolutional neural network comprises the following sub-steps in step S3:
s31, acquiring logic summary content of the artificial damage of the steel rail;
and S32, carrying out artificial damage logic summary content and expert experience on the abnormal support position information marked as the abnormal signal mark to obtain a suspected damage signal of the steel rail and the corresponding suspected damage information of the current steel rail support position.
The above method for determining a rail damage based on a convolutional neural network further includes step S4, comparing and integrating the current rail structure information at the same damage position with the historical rail structure information at a plurality of different periods, and updating a rail structure damage trend database capable of predicting a rail damage trend.
The method for determining the rail damage based on the convolutional neural network further includes step S5, which is to perform modeling according to the rail damage information to generate a current rail structure model.
The invention also provides a rail damage determining system based on the convolutional neural network, which comprises the following components: the system comprises a current steel rail supporting position information acquisition part, a supporting position information signal marking part and a steel rail suspected damage information analysis part;
the current steel rail support bit information acquisition part is used for outputting current steel rail support bit information corresponding to a current steel rail structure body after preprocessing and convolutional neural network analysis are carried out on original multi-dimensional time-series steel rail data;
the support bit information signal marking part is used for comparing the current steel rail support bit information with historical steel rail support bit information, marking consistent support bit information included in the current support bit information as the same signal mark and marking abnormal support bit information included in the current support bit information as an abnormal signal mark;
and the suspected rail damage information analysis part is used for carrying out artificial damage logic summary content and expert experience analysis on the abnormal support position information marked as the abnormal signal mark to obtain a suspected rail damage signal and the corresponding suspected rail support position damage information.
In the above determining system, the current steel rail support bit information acquiring part includes a convolutional neural network, which is used for analyzing the structured data of the probe channel to output the current steel rail support bit information corresponding to the current steel rail structure.
The determining system further includes a rail structure damage trend database, which is used for storing updated rail damage trend prediction information after comparing and integrating the current rail structure information at the same damage position with the historical rail structure information of a plurality of different periods.
The above determination system further includes a rail structure model formed by modeling based on the rail damage information.
Compared with the prior art, the invention has the following advantages:
the invention can provide more accurate and rapid intelligent steel rail damage information for steel rail flaw detection under the condition of ensuring the efficiency, and can ensure the safety and reliability of railway operation.
Drawings
FIG. 1 is a flow chart of a rail flaw determination method according to a first embodiment of the present invention;
FIG. 2 is a flow chart of a rail flaw determination method according to a second embodiment of the present invention;
FIG. 3 is a flow chart of a method for determining rail damage according to a third embodiment of the present invention;
FIG. 4 is a block diagram of a first embodiment of a rail damage determination system according to the present invention;
FIG. 5 is a block diagram of a second embodiment of a rail damage determination system according to the present invention;
fig. 6 is a block diagram showing a configuration of a rail flaw determination system according to a third embodiment of the present invention.
Detailed Description
As shown in fig. 1, the present embodiment provides a rail damage determining method based on a convolutional neural network, including the following steps:
and S1, preprocessing the original multi-dimensional time-series steel rail data, analyzing the original multi-dimensional time-series steel rail data by a convolutional neural network, and outputting current steel rail support position information corresponding to the current steel rail structure.
In step S1, the following substeps are included:
and S11, training or deducing the original multi-dimensional time-series steel rail data to form the structural data of the probe channel.
In addition, two embodiments are included in sub-step S11.
The first embodiment is as follows:
and training the original multi-dimensional time-series steel rail data to form structural data of the probe channel.
S111, screening original multi-dimensional time-series steel rail data, and marking support bit segments for data types manually.
In step S111, after the original multidimensional time-series rail data is screened, the data types are artificially labeled with 60 step lengths as a sliding window.
The support position fragment mark types comprise: a support position reinforcing position, a support position thermite welding position, a support position base material position, a support position turnout position, a support position switch rail position and a support position noise position.
And S112, carrying out matrix processing on the manually marked support position fragment data, and forming structured data of the probe channel after one-hot coding processing and processing.
The original multi-dimensional time-series steel rail data are steel rail data acquired by a steel rail flaw detection trolley.
Example two:
and S11, after the original multi-dimensional time-series steel rail data are deduced, the structural data of the probe channel are formed.
S111, after original multi-dimensional time-series steel rail data are unfolded, invalid data are filtered out, and the reserved valid data are intercepted in a fixed length mode;
and S112, performing matrix processing on the intercepted fixed-length data, and forming structured data of the probe channel after one-hot coding processing and processing.
The original multi-dimensional time-series steel rail data are steel rail data acquired by a steel rail flaw detection trolley.
And S12, analyzing the structural data of the probe channel by the convolutional neural network to output current steel rail support position information corresponding to the current steel rail structural body.
In step S12, the convolutional neural network is divided into a data augmentation layer, a feature building layer and a target classification layer, and includes the following steps:
and S121, in the data augmentation layer, transforming the data to augment the data volume in different dimensions.
In step S121, three operators, i.e., left offset, right offset, and up and down offset, are added to the data enhancement layer.
The transformation of the data comprises geometric transformation, random expansion, clipping, mirror image turning and the like, and the data size is expanded in different dimensions.
And S122, in the feature construction layer, the structural data of the probe channel sequentially enter a convolution layer and a pooling layer, and goose feature extraction is performed after average pooling is used.
And S123, outputting the category and position coordinates of the support bits by using SoftMax in the target classification layer.
In the target classification layer, the current steel rail supporting position information comprises a label of the supporting position and a category of the supporting position, wherein the label of the supporting position corresponds to the category of the supporting position.
S2, comparing the current steel rail supporting position information with historical steel rail supporting position information, marking consistent supporting position information included in the current supporting position information as the same signal mark, and marking abnormal supporting position information included in the current supporting position information as an abnormal signal mark.
Wherein, in step S2, the method includes the following substeps:
s21, inquiring whether the steel rail structure information database comprises a historical steel rail structure corresponding to the current steel rail structure, and calling historical steel rail support position information corresponding to the historical steel rail structure if the historical steel rail structure corresponding to the current steel rail structure is inquired;
s22, judging whether the current steel rail supporting position information and the historical steel rail supporting position information comprise the same supporting position information;
and S23, marking consistent supporting position information which is contained in the current steel rail supporting position information and is the same as the historical steel rail supporting position information as the same signal mark, and marking abnormal supporting position information which is contained in the current steel rail supporting position information and is different from the historical steel rail supporting position information as an abnormal signal mark.
In step 2, according to the advancing direction of the train on the steel rail, after the current steel rail structural body is inquired in the steel rail structural body information database to obtain a corresponding historical steel rail structural body, calling historical steel rail supporting position information corresponding to the historical steel rail structural body. The historical steel rail supporting position information comprises information such as a welding bead position, a thermite welding reinforcing position, a factory welding position and a joint position of the steel rail. And comparing the information including the welding bead position, the thermite welding reinforcement position, the factory welding position, the joint position and the like of the steel rail in the current steel rail supporting position information with the information included in the historical steel rail supporting position information to determine whether the current steel rail supporting position information output by the convolutional neural network is the same. For example, whether the wave comes out from the 37-degree probe before and after the thermite welding reinforcement position is combined with whether the wave comes out from the 0-degree probe accords with the wave-out logic. After the comparison and confirmation, marking the consistent supporting position information which is contained in the current steel rail supporting position information and is the same as the historical steel rail supporting position information as the same signal mark, and marking the abnormal supporting position information which is contained in the current steel rail supporting position information and is different from the historical steel rail supporting position information as an abnormal signal mark.
And S3, carrying out logic summary content of artificial damage of the steel rail and expert experience analysis on the abnormal supporting position information marked as the abnormal signal mark to obtain a suspected damage signal of the steel rail and the corresponding suspected damage information of the current supporting position of the steel rail.
Wherein, in step S3, the method includes the following substeps:
and S31, acquiring the summary content of the artificial damage logic of the steel rail.
The rail artificial damage logic summary content comprises the damage logic summary content of a rail head part, the damage logic summary content of a rail waist part and the damage logic summary content of a rail bottom part;
the summary contents of the damage logic of the rail head part comprise rail head surface damage, first-line wave generation, double waves and rail head abnormity;
the summary contents of the damage logic of the rail web part comprise rail web upper oblique fracture, rail web lower oblique fracture, rail web horizontal fracture and rail web abnormity;
the summary contents of the damage logic of the rail web part comprise rail bottom transverse cracks, rail bottom longitudinal cracks and rail bottom abnormity.
And S32, carrying out artificial damage logic summary content and expert experience on the abnormal support position information marked as the abnormal signal mark to obtain a suspected damage signal of the steel rail and the corresponding suspected damage information of the current steel rail support position.
Wherein, in step S32, the method includes the following substeps:
s321, analyzing the abnormal support position information by the artificial damage logic summary content of the steel rail to obtain the wave output type of the abnormal support position information;
and S322, logically analyzing the wave output type of the support position information by the expert system to obtain a suspected rail damage signal.
In step S322, the wave-emitting type of the abnormal supporting location information is processed by the relative position determination operator, the supporting location surrounding environment determination operator and the supporting location energy weight operator in the expert experience, so as to obtain the suspected rail damage signal.
The suspected damage signal comprises a part of abnormal signals and residual abnormal signals;
and part of abnormal signals are marked as suspected damage signals and output, and the rest of abnormal signals are marked as artificial damage logic summary data of the steel rail.
In addition, after a large amount of artificial damage logic summary data are obtained, the contents of the artificial damage logic summary of the steel rail can be supplemented, and the contents can also be used as reference data for improving the direction of the positioning analysis method, so that the further improvement in the future is facilitated.
In step 3, the wave output type of the abnormal support bit information includes a normal wave output type and an abnormal wave output type.
As shown in fig. 2, the present embodiment provides a rail damage determining method based on a convolutional neural network, including the following steps:
s1, preprocessing the original multi-dimensional time-series steel rail data, analyzing the convolutional neural network, and outputting current steel rail support position information corresponding to the current steel rail structure;
s2, comparing the current steel rail supporting position information with historical steel rail supporting position information, marking consistent supporting position information included in the current supporting position information as the same signal mark, and marking abnormal supporting position information included in the current supporting position information as an abnormal signal mark;
and S3, carrying out logic summary content of artificial damage of the steel rail and expert experience analysis on the abnormal supporting position information marked as the abnormal signal mark to obtain a suspected damage signal of the steel rail and the corresponding suspected damage information of the current supporting position of the steel rail.
The specific contents of step 1 to step 3 are the same as those of step 1 to step 3 shown in fig. 1, and will not be repeated.
In addition, after step 3 is performed, step 4 is also included.
And S4, comparing and integrating the current steel rail structure information of the same damage position with the historical steel rail structure information of a plurality of different periods, and updating a steel rail structure damage trend database capable of predicting the steel rail damage trend.
After a suspected rail damage signal, the corresponding suspected rail support position damage information of the current rail support position and the current rail structural body corresponding to the suspected rail support position damage information of the current rail are obtained, whether a historical rail structural body corresponding to the current rail structural body is stored in a database of the rail structural body information is judged. After the historical steel rail structural body corresponding to the current steel rail structural body is inquired, historical steel rail supporting position suspected damage information corresponding to the historical steel rail structural body in different historical periods is called, and the suspected damage information of the historical steel rail supporting position is compared with the suspected damage information of the current steel rail supporting position.
During comparison, firstly, intelligently aligning the suspected damage information of the historical steel rail supporting position with the suspected damage information of the current steel rail supporting position; and then, integrating and displaying the suspected damage information of the historical steel rail supporting position and the base information of the suspected damage position, the suspected damage rail type and the like in the current steel rail supporting position. In the integration and display process, the method has great significance for the reexamination of a line inspection worker, mainly embodies the full application of historical data, facilitates data management, improves the inspection accuracy, and lays a foundation for the exploration of the rule of our injury trend.
The suspected damage information of the current steel rail supporting position corresponding to the current steel rail structural body is stored in the steel rail structural body information database, the information stored in the steel rail structural body information database is updated, and the suspected damage information can be used as a summary of the steel rail structural body damage trend rule.
As shown in fig. 3, the present embodiment provides a rail damage determining method based on a convolutional neural network, including the following steps:
s1, preprocessing the original multi-dimensional time-series steel rail data, analyzing the convolutional neural network, and outputting current steel rail support position information corresponding to the current steel rail structure;
s2, comparing the current steel rail supporting position information with historical steel rail supporting position information, marking consistent supporting position information included in the current supporting position information as the same signal mark, and marking abnormal supporting position information included in the current supporting position information as an abnormal signal mark;
and S3, carrying out logic summary content of artificial damage of the steel rail and expert experience analysis on the abnormal supporting position information marked as the abnormal signal mark to obtain a suspected damage signal of the steel rail and the corresponding suspected damage information of the current supporting position of the steel rail.
And S4, comparing and integrating the current steel rail structure information of the same damage position with the historical steel rail structure information of a plurality of different periods, and updating a steel rail structure damage trend database capable of predicting the steel rail damage trend.
The specific contents of steps 1 to 4 are the same as those of steps 1 to 3 shown in fig. 2, and will not be repeated.
In addition, after step 4 is performed, step 5 is also included.
And S5, modeling according to the rail damage information to generate a current rail structural body model.
After suspected damage information of the current steel rail supporting position corresponding to the current steel rail structure is obtained, multi-dimensional information integration is carried out on the suspected damage information of the current steel rail supporting position, data support is provided for the accuracy of intelligent measurement and calculation, more accurate data reference is provided for the hierarchical 'structure' confirmation process, and the accuracy of damage analysis is improved. And according to different basic information of the line, orderly modeling the line from the starting point to the end point, orderly integrating the information to generate a line model, and storing modeling data.
As shown in fig. 4, the present embodiment provides a rail damage determination system based on a convolutional neural network, including: the device comprises a current steel rail supporting position information acquisition part, a supporting position information signal marking part and a steel rail suspected damage information analysis part.
And the current steel rail support bit information acquisition part is used for preprocessing original multi-dimensional time-series steel rail data and analyzing a convolutional neural network so as to output current steel rail support bit information corresponding to the current steel rail structural body.
The current steel rail supporting position information acquisition part comprises a preprocessing device and a convolution neural network.
The preprocessing device is used for preprocessing the original multi-dimensional time-series steel rail data to form the structural data of the probe channel.
The method comprises the steps of training or deducing original multi-dimensional time-series steel rail data to form structural data of a probe channel.
The original multi-dimensional time-series steel rail data are steel rail data acquired by the steel rail flaw detection trolley.
The convolutional neural network is used for analyzing the structural data of the probe channel so as to output current steel rail support bit information corresponding to the current steel rail structural body.
The support bit information signal marking part is used for comparing the current steel rail support bit information with the historical steel rail support bit information, marking the consistent support bit information included in the current support bit information as the same signal mark, and marking the abnormal support bit information included in the current support bit information as the abnormal signal mark.
The support bit information signal marking part comprises a steel rail structure body information database and a signal marking device.
The rail structure information database stores historical rail structures and historical rail support position information corresponding to the historical rail structures.
After obtaining the current steel rail support position information corresponding to the current steel rail structure, firstly, inquiring whether a historical steel rail structure corresponding to the current steel rail structure is included in a steel rail structure information database; then, if the historical steel rail structural body corresponding to the current steel rail structural body is inquired, calling historical steel rail supporting position information corresponding to the historical steel rail structural body; and then, judging whether the current steel rail supporting position information and the historical steel rail supporting position information comprise the same supporting position information.
And after judgment, marking consistent supporting position information which is contained in the current steel rail supporting position information and is the same as the historical steel rail supporting position information as the same signal mark through a signal marking device, and marking abnormal supporting position information which is contained in the current steel rail supporting position information and is different from the historical steel rail supporting position information as an abnormal signal mark.
And the suspected rail damage information analysis part is used for carrying out the analysis of the artificial damage logic summary content and the expert experience of the rail on the abnormal support position information marked as the abnormal signal mark to obtain the suspected rail damage signal and the corresponding suspected rail support position damage information.
The suspected rail damage information analysis part comprises an artificial damage logic summary content acquisition device and a suspected damage signal analysis device.
The artificial damage logic summary content acquisition device is used for acquiring artificial damage logic summary content of the steel rail.
The summary contents of the artificial damage logic of the steel rail comprise the summary contents of the damage logic of the rail head part, the summary contents of the damage logic of the rail web part and the summary contents of the damage logic of the rail bottom part;
the summary contents of the damage logic of the rail head part comprise rail head surface damage, first-line wave generation, double waves and rail head abnormity;
the summary contents of the damage logic of the rail web part comprise rail web upper oblique fracture, rail web lower oblique fracture, rail web horizontal fracture and rail web abnormity;
the summary contents of the damage logic of the rail web part comprise rail bottom transverse cracks, rail bottom longitudinal cracks and rail bottom abnormity.
The suspected damage signal analysis device is used for carrying out artificial damage logic summary content and expert experience on the abnormal support position information marked as the abnormal signal mark to obtain suspected damage signals of the steel rail and the corresponding suspected damage information of the current steel rail support position.
The suspected injury signal analysis device comprises the following steps:
analyzing the abnormal supporting position information by the summary content of the artificial damage logic of the steel rail to obtain the wave output type of the abnormal supporting position information;
and the expert system performs logic analysis on the wave output type of the support position information to obtain a suspected rail damage signal.
The wave-emitting type of the abnormal supporting position information is processed by a relative position judgment operator, a supporting position surrounding environment judgment operator and a supporting position energy weight operator in expert experience to obtain a suspected rail damage signal.
The suspected damage signal comprises a part of abnormal signals and the rest of abnormal signals.
And part of abnormal signals are marked as suspected damage signals and output, and the rest of abnormal signals are marked as artificial damage logic summary data of the steel rail.
In addition, after a large amount of artificial damage logic summary data are obtained, the contents of the artificial damage logic summary of the steel rail can be supplemented, and the contents can also be used as reference data for improving the direction of the positioning analysis method, so that the further improvement in the future is facilitated.
As shown in fig. 5, the present embodiment provides a rail damage determination system based on a convolutional neural network, including: the system comprises a current steel rail supporting position information acquisition part, a supporting position information signal marking part, a suspected steel rail damage information analysis part and a steel rail structural body damage trend database.
The specific contents of the current steel rail supporting position information acquiring part, the supporting position information signal marking part and the steel rail suspected damage information analyzing part are the same as those of the current steel rail supporting position information acquiring part, the supporting position information signal marking part and the steel rail suspected damage information analyzing part described in fig. 4, and are not repeated.
The steel rail structure damage trend database is used for storing updated steel rail damage trend prediction information after comparing and integrating current steel rail structure information of the same damage position with a plurality of historical steel rail structure information of different periods.
After a suspected rail damage signal, the corresponding suspected rail support position damage information of the current rail support position and the current rail structural body corresponding to the suspected rail support position damage information of the current rail are obtained, whether a historical rail structural body corresponding to the current rail structural body is stored in a database of the rail structural body information is judged. After the historical steel rail structural body corresponding to the current steel rail structural body is inquired, historical steel rail supporting position suspected damage information corresponding to the historical steel rail structural body in different historical periods is called, and the suspected damage information of the historical steel rail supporting position is compared with the suspected damage information of the current steel rail supporting position.
During comparison, firstly, intelligently aligning the suspected damage information of the historical steel rail supporting position with the suspected damage information of the current steel rail supporting position; and then, integrating and displaying the suspected damage information of the historical steel rail supporting position and the base information of the suspected damage position, the suspected damage rail type and the like in the current steel rail supporting position. In the integration and display process, the method has great significance for the reexamination of a line inspection worker, mainly embodies the full application of historical data, facilitates data management, improves the inspection accuracy, and lays a foundation for the exploration of the rule of our injury trend.
As shown in fig. 6, the present embodiment provides a rail damage determination system based on a convolutional neural network, including: the system comprises a current steel rail support position information acquisition part, a support position information signal marking part, a suspected steel rail damage information analysis part, a steel rail structure damage trend database and a steel rail structure model.
The specific contents of the current steel rail supporting position information acquisition part, the supporting position information signal marking part and the suspected steel rail damage information analysis part are the same as those of the current steel rail supporting position information acquisition part, the supporting position information signal marking part, the suspected steel rail damage information analysis part and the steel rail structure damage trend database described in fig. 5, and are not repeated.
The steel rail structure model is formed by modeling according to the steel rail damage information.
After suspected damage information of the current steel rail supporting position corresponding to the current steel rail structure is obtained, multi-dimensional information integration is carried out on the suspected damage information of the current steel rail supporting position, data support is provided for the accuracy of intelligent measurement and calculation, more accurate data reference is provided for the hierarchical 'structure' confirmation process, and the accuracy of damage analysis is improved. And according to different basic information of the line, orderly modeling the line from the starting point to the end point, orderly integrating the information to generate a line model, and storing modeling data.
The foregoing is merely a preferred embodiment of the invention, which is intended to be illustrative and not limiting. It will be understood by those skilled in the art that various changes, modifications and equivalents may be made therein without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (10)
1. A rail damage determination method based on a convolutional neural network is characterized by comprising the following steps:
s1, preprocessing the original multi-dimensional time-series steel rail data, analyzing the convolutional neural network, and outputting current steel rail support position information corresponding to the current steel rail structure;
s2, comparing the current steel rail supporting position information with historical steel rail supporting position information, marking consistent supporting position information included in the current supporting position information as the same signal mark, and marking abnormal supporting position information included in the current supporting position information as an abnormal signal mark;
and S3, carrying out artificial damage logic summary content and expert experience analysis on the abnormal support position information marked as the abnormal signal mark to obtain a suspected damage signal of the steel rail and the corresponding suspected damage information of the current steel rail support position.
2. The method for determining the rail damage based on the convolutional neural network as claimed in claim 1, wherein in step S1, the method comprises the following sub-steps:
s11, training or deducing the original multi-dimensional time-series steel rail data to form structural data of a probe channel;
and S12, analyzing the structural data of the probe channel by the convolutional neural network to output current steel rail support position information corresponding to the current steel rail structural body.
3. The method for determining the rail damage based on the convolutional neural network as claimed in claim 1, wherein in step S2, the method comprises the following sub-steps:
s21, inquiring whether the steel rail structure information database comprises a historical steel rail structure corresponding to the current steel rail structure, and calling historical steel rail support position information corresponding to the historical steel rail structure if the historical steel rail structure corresponding to the current steel rail structure is inquired;
s22, judging whether the current steel rail supporting position information and the historical steel rail supporting position information comprise the same supporting position information;
and S23, marking consistent supporting position information which is contained in the current steel rail supporting position information and is the same as the historical steel rail supporting position information as the same signal mark, and marking abnormal supporting position information which is contained in the current steel rail supporting position information and is different from the historical steel rail supporting position information as an abnormal signal mark.
4. The method for determining the rail damage based on the convolutional neural network as claimed in claim 1, wherein in step S3, the method comprises the following sub-steps:
s31, acquiring logic summary content of the artificial damage of the steel rail;
and S32, carrying out artificial damage logic summary content and expert experience on the abnormal support position information marked as the abnormal signal mark to obtain a suspected damage signal of the steel rail and the corresponding suspected damage information of the current steel rail support position.
5. A rail damage determination method based on a convolutional neural network as claimed in any one of claims 1 to 4, further comprising step S4, comparing and integrating the current rail structure information of the same damage position with the historical rail structure information of a plurality of different periods, and updating the rail structure damage trend database capable of predicting rail damage trend.
6. The method for determining the rail damage based on the convolutional neural network as claimed in claim 5, further comprising a step S5 of modeling according to the rail damage information to generate a current rail structure model.
7. A determination system using the rail damage determination method based on the convolutional neural network as set forth in claim 1, comprising: the system comprises a current steel rail supporting position information acquisition part, a supporting position information signal marking part and a steel rail suspected damage information analysis part;
the current steel rail support bit information acquisition part is used for outputting current steel rail support bit information corresponding to a current steel rail structure body after preprocessing and convolutional neural network analysis are carried out on original multi-dimensional time-series steel rail data;
the support bit information signal marking part is used for comparing the current steel rail support bit information with historical steel rail support bit information, marking consistent support bit information included in the current support bit information as the same signal mark and marking abnormal support bit information included in the current support bit information as an abnormal signal mark;
and the suspected rail damage information analysis part is used for carrying out artificial damage logic summary content and expert experience analysis on the abnormal support position information marked as the abnormal signal mark to obtain a suspected rail damage signal and the corresponding suspected rail support position damage information.
8. The determination system according to claim 7, wherein the current steel rail supporting bit information acquiring portion comprises a convolutional neural network, which is used for analyzing the structured data of the probe channel to output the current steel rail supporting bit information corresponding to the current steel rail structure.
9. The determination system according to claim 8, further comprising a rail structure damage trend database for storing updated rail damage trend prediction information after comparing and integrating current rail structure information of the same damage location with historical rail structure information of a plurality of different periods.
10. The determination system of claim 9, further comprising a rail structure model that is formed from modeling rail damage information.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110150634.5A CN114862743A (en) | 2021-02-04 | 2021-02-04 | Method and system for determining rail damage based on convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110150634.5A CN114862743A (en) | 2021-02-04 | 2021-02-04 | Method and system for determining rail damage based on convolutional neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114862743A true CN114862743A (en) | 2022-08-05 |
Family
ID=82622925
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110150634.5A Pending CN114862743A (en) | 2021-02-04 | 2021-02-04 | Method and system for determining rail damage based on convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114862743A (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105548197A (en) * | 2015-12-08 | 2016-05-04 | 深圳大学 | Non-contact steel rail surface flaw inspection method and device |
JP2019143999A (en) * | 2018-02-16 | 2019-08-29 | 公益財団法人鉄道総合技術研究所 | Sleeper deterioration estimation method and sleeper deterioration estimation apparatus |
CN110263867A (en) * | 2019-06-24 | 2019-09-20 | 株洲时代电子技术有限公司 | A kind of rail defects and failures classification method |
CN110363403A (en) * | 2019-06-27 | 2019-10-22 | 中国铁道科学研究院集团有限公司 | Railway track damage forecast method and device |
CN111855810A (en) * | 2020-07-20 | 2020-10-30 | 济南大学 | Rail foot damage identification method and system based on recurrent neural network |
CN112147221A (en) * | 2020-09-22 | 2020-12-29 | 济南大学 | Steel rail screw hole crack identification method and system based on ultrasonic flaw detector data |
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
- 2021-02-04 CN CN202110150634.5A patent/CN114862743A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105548197A (en) * | 2015-12-08 | 2016-05-04 | 深圳大学 | Non-contact steel rail surface flaw inspection method and device |
JP2019143999A (en) * | 2018-02-16 | 2019-08-29 | 公益財団法人鉄道総合技術研究所 | Sleeper deterioration estimation method and sleeper deterioration estimation apparatus |
CN110263867A (en) * | 2019-06-24 | 2019-09-20 | 株洲时代电子技术有限公司 | A kind of rail defects and failures classification method |
CN110363403A (en) * | 2019-06-27 | 2019-10-22 | 中国铁道科学研究院集团有限公司 | Railway track damage forecast method and device |
CN111855810A (en) * | 2020-07-20 | 2020-10-30 | 济南大学 | Rail foot damage identification method and system based on recurrent neural network |
CN112147221A (en) * | 2020-09-22 | 2020-12-29 | 济南大学 | Steel rail screw hole crack identification method and system based on ultrasonic flaw detector data |
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 |
Non-Patent Citations (1)
Title |
---|
姚锋涛: "基于超声回波信号的钢轨伤损识别与分类研究", 优秀硕士学位论文全文库, no. 1, 15 January 2021 (2021-01-15) * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111322985A (en) | Tunnel clearance analysis method, device and system based on laser point cloud | |
CN105416097B (en) | Contact Line Detection tool car and method based on laser radar | |
CN109708615A (en) | A kind of subway tunnel limit dynamic testing method based on laser scanning | |
CN107609304A (en) | The fault diagnosis and prediction system and method based on PHM of LONG-SPAN RAILWAY bridge | |
US20220355839A1 (en) | Monitoring, predicting and maintaining the condition of railroad elements with digital twins | |
CN110789566B (en) | Track defect monitoring method and monitoring equipment based on axle box acceleration signal | |
CN109278053A (en) | A kind of railway tunnel detection method | |
CN111563530A (en) | Intelligent bridge disease inspection method and system | |
CN111272093B (en) | Roadway deformation monitoring method | |
CN114912641A (en) | Road network bridge and tunnel maintenance decision and intelligent inspection system based on BIM | |
EP1862593B2 (en) | GPS aided continuous route survey system with multi sensors | |
CN117252042B (en) | Urban underground space comprehensive bearing capacity evaluation system and method | |
CN114862743A (en) | Method and system for determining rail damage based on convolutional neural network | |
CN116090322A (en) | Intelligent decision visualization system for pipeline integrity | |
CN117789064A (en) | Highway side slope unmanned aerial vehicle patrols and examines early warning system based on multisource data | |
KR20150019401A (en) | Apparatus for inspecting underground tunnel and control method thereof | |
CN108205016A (en) | The industrial computer system of sound emission safety of tower crane detection | |
CN109766794B (en) | Automatic real-time road detection method and system thereof | |
RU2177144C1 (en) | Way to construct multiparametric digital models of construction objects, to analyze and model their states ( variants ) | |
CN114660070A (en) | Tunnel three-dimensional fault detection and identification method and system | |
CN114462647A (en) | Highway maintenance management system based on detect cloud platform | |
CN116427221B (en) | Rapid nondestructive testing method for degradation state of ballast bed | |
CN206618550U (en) | The device monitored in real time based on bridge defect plate-girder degree of disturbing | |
CN112160306B (en) | Rock-soil sampling layout system and method based on artificial intelligence analysis | |
CN113588779B (en) | Data comparison analysis method for intelligent steel rail damage judgment algorithm |
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 |