CN111091545A - Method for detecting loss fault of bolt at shaft end of rolling bearing of railway wagon - Google Patents
Method for detecting loss fault of bolt at shaft end of rolling bearing of railway wagon Download PDFInfo
- Publication number
- CN111091545A CN111091545A CN201911272493.3A CN201911272493A CN111091545A CN 111091545 A CN111091545 A CN 111091545A CN 201911272493 A CN201911272493 A CN 201911272493A CN 111091545 A CN111091545 A CN 111091545A
- Authority
- CN
- China
- Prior art keywords
- image
- detected
- bolt
- shaft end
- rolling bearing
- 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.)
- Granted
Links
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
- G06T7/0006—Industrial image inspection using a design-rule based approach
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30164—Workpiece; Machine component
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for detecting a loss fault of a bolt at the shaft end of a rolling bearing of a railway wagon, belongs to the field of image processing, and aims to solve the problems of low efficiency, low training precision and low classification detection speed of classification detection on a part easy to lose of the wagon by adopting a Faster Rcnn network model. The method comprises the following steps: firstly, acquiring a passing image of a bolt at the shaft end of a rolling bearing of a truck as an image to be detected; inputting the image to be detected in the step one into a target classification detection model for classification judgment, and outputting the image to be detected as a loss confidence coefficient and a normal confidence coefficient; step three, judging whether the confidence coefficient of the image to be detected for the lost class is greater than a set threshold value of the lost class, and if so, performing fault alarm; if not, executing the step four; step four, judging whether the confidence coefficient of the image to be detected in the normal class is greater than a normal class set threshold, if so, returning to the step one, and continuing to process the next detected image; if not, a fault alarm is carried out.
Description
Technical Field
The invention belongs to the field of image processing, and relates to a railway wagon bolt loss detection technology.
Background
With the rapid development of the railway industry in China in recent years, rail wagons are also developed in the direction of heavy load and high speed, faults occur in the area of an axle shaft increasingly in the process of high-speed running of the wagons, particularly, a rolling bearing axle end bolt plays a very critical role, the rolling bearing axle end bolt is an important component on the axle shaft of a side bogie, the rolling bearing axle end bolt is mainly used for fixing an axle end front cover on a specified position, due to the fact that the rolling bearing axle end bolt is subjected to cargo handling, foreign object impact, locking function failure of a sealing lock and the like, the axle end bolt can gradually separate from the fixed position due to vibration in the process of running of the axle shaft, the axle end bolt starts to loosen and flee out, even serious faults of bolt loss occur.
In order to avoid faults, the bolts at the shaft end of the rolling bearing can be pertinently checked in daily detection, and the manual detection mode of manual image reading is prone to various problems of untimely fault finding, low efficiency and the like caused by the problems of fatigue, responsibility and the like in manual vehicle detection. Therefore, there is currently a technology for classifying output fault types using an intelligent model, such as target classification using a fast Region Convolutional Neural Network (fast Convolutional Neural Network) model shown in fig. 1, where the fast Rcnn model first uses VGG16(VGG is a shorthand for visual geometry Group) as a base Network, extracts feature layers of an image using 13 Convolutional layers, 13 relu activation function layers, and 4 pooling layers, and the feature layers are shared for a subsequent Region suggestion Network (RPN) and a fully connected layer; and the RPN layer outputs suggested features, combines with the feature map to extract a suggested feature map, and sends the feature map to a subsequent full-connection layer to judge the target category. When a fast Rcnn network model is conventionally constructed, 13 Conv layers of a basic feature extraction network are processed layer by layer according to a sequence, and a required feature map is extracted from the final Conv layer.
Disclosure of Invention
The invention aims to solve the problems of low training efficiency, low training precision and low classification detection speed caused by low resolution of an extracted feature map when a fast Rcnn network model is adopted to classify and detect parts easy to lose of a wagon, and provides a method for detecting a fault of losing a bolt at the shaft end of a rolling bearing of a railway wagon.
The invention relates to a method for detecting a loss fault of a bolt at the shaft end of a rolling bearing of a railway wagon, which comprises the following steps of:
firstly, acquiring a passing image of a bolt at the shaft end of a rolling bearing of a truck as an image to be detected;
inputting the image to be detected in the step one into a target classification detection model for classification judgment, and outputting the image to be detected as a loss confidence coefficient and a normal confidence coefficient;
the target classification detection model is trained and constructed by adopting a Faster Rcnn network model, the Faster Rcnn network model adopts an Oxford university visual geometry group VGG16 convolutional neural network as a basic network, feature graphs of conv4 and conv5 layers in a convolutional neural network layer are respectively extracted, and feature graphs formed by weighting and fusing the feature graphs of the conv4 and the conv5 layers are used for proposing a network RPN layer and a full connection layer for a subsequent region;
step three, judging whether the confidence coefficient of the image to be detected for the lost class is greater than a set threshold value of the lost class, and if so, performing fault alarm; if not, executing the step four;
step four, judging whether the confidence coefficient of the image to be detected in the normal class is greater than a normal class set threshold, if so, returning to the step one, and continuing to process the next detected image; if not, a fault alarm is carried out.
Preferably, the process of constructing the training set required by the target classification detection model in the second step is as follows:
step two, acquiring a truck passing image of a truck side bogie through a truck operation failure dynamic image detection system (TFDS);
secondly, performing coarse positioning according to the truck wheel base information, and intercepting an image of an area where a bolt at the axle end of the rolling bearing is located as a sample image;
step two, collecting a large number of sample images of different vehicle types under various conditions and at different stations according to the step two;
fourthly, manufacturing a shaft end bolt loss fault sample in a manual simulation mode to realize data amplification, and balancing the positive sample and the negative sample;
and step two, labeling all the images after data amplification, wherein the labeling types comprise: normal class and loss class;
and step two, converting the marked image data into a data set as a training set.
Preferably, the sample images collected in the second step and the third step under various conditions are sample images influenced by shooting angles, rain water, mud stains, oil stains and black paint natural conditions or artificial conditions.
Preferably, the data amplification method for the sample image in the second step and the fourth step includes performing contrast adjustment, sample angle rotation, flipping and translation operations on the sample image under random conditions.
The invention has the beneficial effects that:
1. the method has the advantages that the shaft end bolt loss fault is detected by adopting the deep learning model, the fault detection efficiency can be improved, and the labor cost and the time cost are reduced.
2. The method has the advantages that the method for detecting the loss fault of the bolt at the shaft end of the rolling bearing by adopting the deep learning target detection algorithm is high in efficiency, can adapt to the truck image shot in the complex environment, and avoids the limitation of the traditional image detection algorithm.
3. The convolutional neural network is built by performing feature extraction on different convolutional layers and classifying after combination, so that the calculated amount is basically not increased, and meanwhile, the convolutional layers with higher resolution can be utilized, the detection precision of small targets is improved, and the training efficiency, the training precision and the detection speed of the model are improved.
Drawings
FIG. 1 is a diagram of a conventional fast Rcnn network model architecture;
FIG. 2 is a schematic structural diagram of a detection method for the loss fault of the bolt at the shaft end of the rolling bearing of the railway wagon;
FIG. 3 is a diagram of the improved fast Rcnn network model structure adopted by the present invention, in which FC is the full link layer, Score is the confidence of the detection target, and BBox is the coordinate of the detection target.
Detailed Description
The following detailed description of the embodiments of the present invention will be provided with reference to the drawings and examples, so that how to apply the technical means to solve the technical problems and achieve the technical effects can be fully understood and implemented. It should be noted that, as long as there is no conflict, the embodiments and the features of the embodiments of the present invention may be combined with each other, and the technical solutions formed are within the scope of the present invention.
First embodiment, the present embodiment is described below with reference to fig. 2 and 3, and the method for detecting a missing fault of a shaft end bolt of a rolling bearing of a railway wagon according to the present embodiment is characterized by comprising the following steps:
firstly, acquiring a passing image of a bolt at the shaft end of a rolling bearing of a truck as an image to be detected;
the network model of deep learning training is good and bad, and an important factor is the selection of samples. The diversity of sample selection and the adaptability of the samples to learning network models with different depths directly influence the accuracy of fault target detection, so that proper samples need to be selected according to actual conditions. To ensure the diversity of samples, it is necessary to collect, as much as possible, the sample images of the bolts at the shaft ends of the rolling bearings under various complex conditions when the trucks pass through the sample images in the collecting process, i.e., different passing times and different environments are considered, and the differences of the bolt parts at the shaft ends of different vehicle types are also considered, such as the interferences of passing vehicles in the morning and evening, passing vehicles in rainy and snowy days, foreign matters, and the like. After diversified samples are collected, a data amplification mode is continuously adopted, the data samples are subjected to appropriate rotation, translation and contrast adjustment, noise is added to enrich the samples, and the robustness of the target detection model is improved through enriching the samples. Because real faults of the rolling bearing shaft end bolt are rarely lost, overfitting can be caused by training the model under the condition of insufficient samples, the mode of manual simulation is adopted, the shaft end bolt lost fault samples are simulated and manufactured, positive and negative samples of the sample set are balanced as much as possible, and the model trained in this way has higher robustness and stability. The collected sample is processed and kept consistent with the shaft end bolt image used when the model is used as much as possible, so that the stability and the accuracy of model detection can be improved.
And (3) marking samples of normal and lost shaft end bolts by using a LabelImg tool, wherein the normal bolts of different vehicle types only need to be classified into one type of mark due to similar structures of shaft end bolt parts of different vehicle types. The marked sample data set can be used for training a deep learning shaft end bolt target detection model.
Inputting the image to be detected in the step one into a target classification detection model for classification judgment, and outputting the image to be detected as a loss confidence coefficient and a normal confidence coefficient;
the target classification detection model is trained and constructed by adopting a Faster Rcnn network model, the Faster Rcnn network model adopts an Oxford university visual geometry group VGG16 convolutional neural network as a basic network, feature graphs of conv4 and conv5 layers in a convolutional neural network layer are respectively extracted, and feature graphs formed by weighting and fusing the feature graphs of the conv4 and the conv5 layers are used for proposing a network RPN layer and a full connection layer for a subsequent region;
the invention has the innovation points that a VGG16 target detection and classification model of Faster Rcnn under the TensorFlow deep learning system environment is selected, roi-posing feature extraction is respectively carried out on a conv4 layer and a conv5 layer, classification is carried out after combination to build a convolutional neural network, two feature graphs are weighted and fused and then are sent to a subsequent RPN layer and a full connection layer, so that the calculated amount is not basically increased, a conv4 layer with higher resolution ratio can be utilized, the resolution ratio of the feature graphs is increased, the detection precision of small targets is improved, the training efficiency and the training precision are improved, and the detection speed of the model is improved. The improved Faster Rcnn model consists essentially of the following components:
1) creating a basic feature extraction network:
the basic feature extraction network of the fast Rcnn network target detection method is formed by connecting 13 conv convolution layers, 13 relu activation functions and 4 poiling pooling layers to obtain a feature diagram featuremap of a shaft end bolt subimage, and the feature diagram featuremap is transmitted to an RPN network and a full connection layer for use.
The network basic feature extraction layer is characterized in that a slim library which can simplify model establishment is introduced, a 64-channel 3 x 3 convolution kernel is established for 2 times for convolution, and a pooling kernel with the size of 2 x 2 is selected for pooling; creating 2 times of a 3 multiplied by 3 convolution kernel of 128 channels for convolution, and selecting a pooling kernel with the size of 2 multiplied by 2 for pooling; creating 3 times of 3 multiplied by 3 convolution kernels of 256 channels for convolution, and selecting pooling kernels with the size of 2 multiplied by 2 for pooling; creating 3 times of 512-channel 3 × 3 convolution kernels for convolution, and selecting pooling kernels with the size of 2 × 2 for pooling; and 3 times of 3 multiplied by 3 convolution kernels of 512 channels are created again for convolution, and feature information features maps extracted through the network model are stored.
2) Creating RPN networks
And creating an RPN network, judging and calculating to generate candidate regions regionproposals through a 3 x 3 convolution and a softmax function.
Respectively performing roi-posing feature extraction on the conv4 convolutional layer and the conv5 convolutional layer, and then ensuring the feature scales of different feature map layers to be consistent through canonical combination and two layers of collected image feature map features. The convolutional neural network built in this way basically does not increase the calculated amount, and simultaneously can utilize a conv4 layer with higher resolution, so that the resolution of the feature map is increased, the detection precision of small targets is improved, the training efficiency and the training precision are improved, and the detection speed of the model is also improved.
And outputting candidate region responses by using the RPN, synthesizing the weighted and fused feature map and the candidate region responses, and sending the synthesized feature map and the candidate region responses to a full-connection layer for judging and classifying the target.
3) Object detection classification
The target detection classification part calculates the specific category of each pronodal by using the obtained pronodal feature map through full connected layer fullonconnected layers and softmax functions, and outputs a probability vector; meanwhile, a regression judgment link accurate target detection frame is added.
3. Training of network models
And after the training sample set and the network model are built, preparing to start training the network model.
When the network model is trained, proper parameters need to be selected properly, proper iteration times and a network model training structure are selected according to experience knowledge and sample conditions, and the training type of the shaft end bolt module needs to be modified.
The training sample set is classified into a test set, a training set and a sample set according to the model structure, and the sample set is automatically classified according to various preset sample proportions by writing a python program.
And starting to train a rolling bearing shaft end bolt target detection model by using the constructed network model structure and the training parameters, storing model training files with different iteration times in the training process, and selecting an optimal training model as a rolling bearing shaft end bolt loss fault detection model after the training is finished.
And testing the detection accuracy and the recognition rate of the network model through the test set data, and storing the final detection result.
Step three, judging whether the confidence coefficient of the image to be detected for the lost class is greater than a set threshold value of the lost class, and if so, performing fault alarm; if not, executing the step four;
step four, judging whether the confidence coefficient of the image to be detected in the normal class is greater than a normal class set threshold, if so, returning to the step one, and continuing to process the next detected image; if not, a fault alarm is carried out.
Reading a passing image on a TFDS truck image detection system server, roughly positioning a subimage of a shaft end bolt part area, carrying out real-time fault detection on a shaft end bolt lost target detection depth learning framework by calling the subimage, and transmitting a detection result exceeding the detection result needing alarming to a TFDS fault detection platform in real time according to a loss alarming threshold set by a model; and when the missing fault is not detected on the passing sub-image, further judgment is carried out, if the normal bolt is not detected under the set normal bolt threshold value, the condition that the normal and missing bolt targets are not found on the sub-image is indicated, and in order to ensure that the real fault is not reported in a missing manner, the situation that the report is not reported due to the abnormal image is avoided, alarm information is directly output, and the condition that whether the fault is real or not is further verified manually is waited.
Although the embodiments of the present invention have been described above, the above descriptions are only for the convenience of understanding the present invention, and are not intended to limit the present invention. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (4)
1. The method for detecting the loss fault of the bolt at the shaft end of the rolling bearing of the railway wagon is characterized by comprising the following steps of:
firstly, acquiring a passing image of a bolt at the shaft end of a rolling bearing of a truck as an image to be detected;
inputting the image to be detected in the step one into a target classification detection model for classification judgment, and outputting the image to be detected as a loss confidence coefficient and a normal confidence coefficient;
the target classification detection model is trained and constructed by adopting a Faster Rcnn network model, the Faster Rcnn network model adopts a visual geometric group VGG16 convolutional neural network as a basic network, feature graphs of conv4 and conv5 layers in a convolutional neural network layer are respectively extracted, and feature graphs formed by weighting and fusing the feature graphs of the conv4 and the conv5 layers are used for proposing a network RPN layer and a full connection layer for a subsequent region;
step three, judging whether the confidence coefficient of the image to be detected for the lost class is greater than a set threshold value of the lost class, and if so, performing fault alarm; if not, executing the step four;
step four, judging whether the confidence coefficient of the image to be detected in the normal class is greater than a normal class set threshold, if so, returning to the step one, and continuing to process the next detected image; if not, a fault alarm is carried out.
2. The method for detecting the loss fault of the shaft end bolt of the rolling bearing of the railway wagon as claimed in claim 1, wherein the process of constructing the training set required by the target classification detection model in the second step is as follows:
step two, acquiring a truck passing image of a truck side bogie through a truck operation failure dynamic image detection system (TFDS);
secondly, performing coarse positioning according to the truck wheel base information, and intercepting an image of an area where a bolt at the axle end of the rolling bearing is located as a sample image;
step two, collecting a large number of sample images of different vehicle types under various conditions and at different stations according to the step two;
fourthly, manufacturing a shaft end bolt loss fault sample in a manual simulation mode to realize data amplification, and balancing the positive sample and the negative sample;
and step two, labeling all the images after data amplification, wherein the labeling types comprise: normal class and loss class;
and step two, converting the marked image data into a data set as a training set.
3. The method for detecting the loss fault of the shaft end bolt of the rolling bearing of the railway wagon as claimed in claim 2, wherein the sample images collected in the second step and the third step under various conditions refer to the sample images influenced by natural conditions or artificial conditions of shooting angles, rain, mud, oil and black paint.
4. The method for detecting the loss fault of the shaft end bolt of the rolling bearing of the railway wagon as claimed in claim 2, wherein the method for performing data amplification on the sample image in the second step and the fourth step comprises the steps of performing contrast adjustment, sample angle rotation, overturning and translation on the sample image under random conditions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911272493.3A CN111091545B (en) | 2019-12-12 | 2019-12-12 | Method for detecting loss fault of bolt at shaft end of rolling bearing of railway wagon |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911272493.3A CN111091545B (en) | 2019-12-12 | 2019-12-12 | Method for detecting loss fault of bolt at shaft end of rolling bearing of railway wagon |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111091545A true CN111091545A (en) | 2020-05-01 |
| CN111091545B CN111091545B (en) | 2021-01-22 |
Family
ID=70394885
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911272493.3A Active CN111091545B (en) | 2019-12-12 | 2019-12-12 | Method for detecting loss fault of bolt at shaft end of rolling bearing of railway wagon |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111091545B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111652212A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting loss fault of fastening bolt at end part of cross rod based on deep learning |
| CN111862029A (en) * | 2020-07-15 | 2020-10-30 | 哈尔滨市科佳通用机电股份有限公司 | Fault detection method for bolt part of vertical shock absorber of railway motor train unit |
| CN112733771A (en) * | 2021-01-18 | 2021-04-30 | 哈尔滨市科佳通用机电股份有限公司 | Railway train jumper wire foreign matter fault identification method and system |
| CN113158966A (en) * | 2021-05-08 | 2021-07-23 | 浙江浩腾电子科技股份有限公司 | Detection method for recognizing behaviors of non-motor vehicle cyclists and cyclists based on deep learning |
| CN115170923A (en) * | 2022-07-19 | 2022-10-11 | 哈尔滨市科佳通用机电股份有限公司 | Fault identification method for loss of railway wagon supporting plate nut |
| CN115170883A (en) * | 2022-07-19 | 2022-10-11 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting loss fault of brake cylinder piston push rod open pin |
| CN115331086A (en) * | 2022-08-17 | 2022-11-11 | 哈尔滨市科佳通用机电股份有限公司 | Brake shoe breaking and rivet losing fault detection method |
| CN115346068A (en) * | 2022-08-02 | 2022-11-15 | 哈尔滨市科佳通用机电股份有限公司 | Automatic generation method for bolt loss fault image of railway freight train |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106226050A (en) * | 2016-07-15 | 2016-12-14 | 北京航空航天大学 | A kind of TFDS fault automatic identifying method |
| CN106778740A (en) * | 2016-12-06 | 2017-05-31 | 北京航空航天大学 | A kind of TFDS non-faulting image detecting methods based on deep learning |
| CN108257114A (en) * | 2017-12-29 | 2018-07-06 | 天津市万贸科技有限公司 | A kind of transmission facility defect inspection method based on deep learning |
| CN109166094A (en) * | 2018-07-11 | 2019-01-08 | 华南理工大学 | A kind of insulator breakdown positioning identifying method based on deep learning |
| CN109767427A (en) * | 2018-12-25 | 2019-05-17 | 北京交通大学 | The detection method of train rail fastener defect |
| US20190243735A1 (en) * | 2018-02-05 | 2019-08-08 | Wuhan University | Deep belief network feature extraction-based analogue circuit fault diagnosis method |
| CN110310261A (en) * | 2019-06-19 | 2019-10-08 | 河南辉煌科技股份有限公司 | A kind of Contact Net's Suspension Chord defects detection model training method and defect inspection method |
| CN110827251A (en) * | 2019-10-30 | 2020-02-21 | 江苏方天电力技术有限公司 | Power transmission line locking pin defect detection method based on aerial image |
-
2019
- 2019-12-12 CN CN201911272493.3A patent/CN111091545B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106226050A (en) * | 2016-07-15 | 2016-12-14 | 北京航空航天大学 | A kind of TFDS fault automatic identifying method |
| CN106778740A (en) * | 2016-12-06 | 2017-05-31 | 北京航空航天大学 | A kind of TFDS non-faulting image detecting methods based on deep learning |
| CN108257114A (en) * | 2017-12-29 | 2018-07-06 | 天津市万贸科技有限公司 | A kind of transmission facility defect inspection method based on deep learning |
| US20190243735A1 (en) * | 2018-02-05 | 2019-08-08 | Wuhan University | Deep belief network feature extraction-based analogue circuit fault diagnosis method |
| CN109166094A (en) * | 2018-07-11 | 2019-01-08 | 华南理工大学 | A kind of insulator breakdown positioning identifying method based on deep learning |
| CN109767427A (en) * | 2018-12-25 | 2019-05-17 | 北京交通大学 | The detection method of train rail fastener defect |
| CN110310261A (en) * | 2019-06-19 | 2019-10-08 | 河南辉煌科技股份有限公司 | A kind of Contact Net's Suspension Chord defects detection model training method and defect inspection method |
| CN110827251A (en) * | 2019-10-30 | 2020-02-21 | 江苏方天电力技术有限公司 | Power transmission line locking pin defect detection method based on aerial image |
Non-Patent Citations (1)
| Title |
|---|
| 浩瀚之水: ""目标检测 Faster R-CNN深入理解&&改进方法汇总"", 《CSDN》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111652212A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting loss fault of fastening bolt at end part of cross rod based on deep learning |
| CN111862029A (en) * | 2020-07-15 | 2020-10-30 | 哈尔滨市科佳通用机电股份有限公司 | Fault detection method for bolt part of vertical shock absorber of railway motor train unit |
| CN112733771A (en) * | 2021-01-18 | 2021-04-30 | 哈尔滨市科佳通用机电股份有限公司 | Railway train jumper wire foreign matter fault identification method and system |
| CN112733771B (en) * | 2021-01-18 | 2021-10-01 | 哈尔滨市科佳通用机电股份有限公司 | Railway train jumper wire foreign matter fault identification method and system |
| CN113158966A (en) * | 2021-05-08 | 2021-07-23 | 浙江浩腾电子科技股份有限公司 | Detection method for recognizing behaviors of non-motor vehicle cyclists and cyclists based on deep learning |
| CN115170923A (en) * | 2022-07-19 | 2022-10-11 | 哈尔滨市科佳通用机电股份有限公司 | Fault identification method for loss of railway wagon supporting plate nut |
| CN115170883A (en) * | 2022-07-19 | 2022-10-11 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting loss fault of brake cylinder piston push rod open pin |
| CN115170883B (en) * | 2022-07-19 | 2023-03-14 | 哈尔滨市科佳通用机电股份有限公司 | Brake cylinder piston push rod opening pin loss fault detection method |
| CN115346068A (en) * | 2022-08-02 | 2022-11-15 | 哈尔滨市科佳通用机电股份有限公司 | Automatic generation method for bolt loss fault image of railway freight train |
| CN115331086A (en) * | 2022-08-17 | 2022-11-11 | 哈尔滨市科佳通用机电股份有限公司 | Brake shoe breaking and rivet losing fault detection method |
| CN115331086B (en) * | 2022-08-17 | 2023-08-08 | 哈尔滨市科佳通用机电股份有限公司 | Brake shoe breakage and rivet loss fault detection method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111091545B (en) | 2021-01-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111091545B (en) | Method for detecting loss fault of bolt at shaft end of rolling bearing of railway wagon | |
| CN111080603B (en) | Method for detecting breakage fault of shaft end bolt of railway wagon | |
| CN111091543B (en) | Railway wagon swing bolster spring loss fault target detection method | |
| CN111091555B (en) | Brake shoe breaking target detection method | |
| CN111080608A (en) | Method for recognizing closing fault image of automatic brake valve plug handle of railway wagon in derailment | |
| CN112330631B (en) | Railway wagon brake beam pillar rivet pin collar loss fault detection method | |
| CN111080617B (en) | Railway wagon brake beam pillar round pin loss fault identification method | |
| CN111080600A (en) | Fault identification method for split pin on spring supporting plate of railway wagon | |
| CN111080612B (en) | Truck bearing damage detection method | |
| CN111652295B (en) | Railway wagon coupler yoke key joist falling fault identification method | |
| CN111080621B (en) | Method for identifying railway wagon floor damage fault image | |
| CN114627383B (en) | Small sample defect detection method based on metric learning | |
| CN112906534A (en) | Lock catch loss fault detection method based on improved Faster R-CNN network | |
| CN111862029A (en) | Fault detection method for bolt part of vertical shock absorber of railway motor train unit | |
| CN111161224A (en) | Casting internal defect grading evaluation system and method based on deep learning | |
| CN111652211B (en) | Method for detecting foreign matter hanging fault of motor car anti-snaking shock absorber mounting seat | |
| CN111079822A (en) | Method for identifying dislocation fault image of middle rubber and upper and lower plates of axle box rubber pad | |
| CN113221839B (en) | Automatic truck image identification method and system | |
| CN112329737A (en) | Vehicle detection method based on improved Faster RCNN | |
| CN111091551A (en) | Method for detecting loss fault of brake beam strut opening pin of railway wagon | |
| CN113203464A (en) | Sensor fault detection method of dynamic vehicle-mounted weighing system | |
| CN111091552A (en) | Method for identifying closing fault image of angle cock handle of railway wagon | |
| CN112329858B (en) | Image recognition method for breakage fault of anti-loosening iron wire of railway motor car | |
| CN116206222A (en) | Power transmission line fault detection method and system based on lightweight target detection model | |
| CN108846421B (en) | Image classification method based on deep 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |