CN111079748A - Method for detecting oil throwing fault of rolling bearing of railway wagon - Google Patents
Method for detecting oil throwing fault of rolling bearing of railway wagon Download PDFInfo
- Publication number
- CN111079748A CN111079748A CN201911272538.7A CN201911272538A CN111079748A CN 111079748 A CN111079748 A CN 111079748A CN 201911272538 A CN201911272538 A CN 201911272538A CN 111079748 A CN111079748 A CN 111079748A
- Authority
- CN
- China
- Prior art keywords
- unit
- image
- group
- rolling bearing
- fault
- 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
- 238000005096 rolling process Methods 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000013528 artificial neural network Methods 0.000 claims abstract description 38
- 238000013145 classification model Methods 0.000 claims abstract description 34
- 230000006870 function Effects 0.000 claims abstract description 26
- 230000008569 process Effects 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 11
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000000605 extraction Methods 0.000 claims description 23
- 238000005070 sampling Methods 0.000 claims description 21
- 230000004927 fusion Effects 0.000 claims description 7
- 238000011176 pooling Methods 0.000 claims description 6
- 238000013135 deep learning Methods 0.000 claims description 4
- 230000003321 amplification Effects 0.000 claims description 3
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 18
- 238000006243 chemical reaction Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 238000005461 lubrication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
- G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a fault detection method for oil slinging of a rolling bearing of a railway wagon, and relates to a fault detection method for a railway wagon. The invention aims to solve the problems of poor accuracy and low efficiency of the conventional detection of the oil throwing fault of the rolling bearing of the railway wagon. The process is as follows: firstly, acquiring a linear array image; secondly, coarse positioning; thirdly, preprocessing a data set image; fourthly, classifying fault targets: and fourthly: building a neural network classification model; fourthly, two: normalizing the preprocessed data set image to be a training set and inputting the training set into a neural network classification model; fourthly, three: the neural network classification model loss function is the average value of the cross entropy loss functions of all the three preprocessed data set images; fourthly, four: updating the weight parameters of the neural network classification model; and fourthly, fifthly: obtaining a trained neural network classification model; and fifthly, judging whether the linear array image of the railway wagon to be detected is the oil throwing fault image of the rolling bearing. The invention is used for the field of oil throwing fault detection of the truck roller bearing.
Description
Technical Field
The invention relates to a fault detection method for a railway wagon.
Background
With the rapid development of Chinese economy, the demand of people on logistics is increasing day by day. The rail wagon transportation plays a vital role in logistics development, and the oil slinging fault of a rolling bearing of the rail wagon influences the safety and speed of logistics transportation. When the rolling bearing throws away oil, the effective lubrication action in the bearing can be reduced, so that the friction force and the impact force between the bearing retainer and the roller are increased, the bearing retainer is easily damaged, and the logistics transportation safety is further influenced. The rolling bearing oil shedding fault image is manually and repeatedly checked for a long time, so that the problems of dry work, easiness in distraction and low efficiency exist, and the accuracy cannot be guaranteed.
Disclosure of Invention
The invention aims to solve the problems of poor detection accuracy and low efficiency of the conventional oil slinging fault of a rolling bearing of a railway freight car, and provides an oil slinging fault detection method of the rolling bearing of the railway freight car.
The method for detecting the oil throwing fault of the rolling bearing of the railway wagon comprises the following specific processes:
firstly, acquiring a linear array image;
step two, coarse positioning;
thirdly, preprocessing a data set image;
step four, classifying fault targets; the specific process is as follows:
step four, firstly: building a neural network classification model, which comprises a feature extraction network, a multi-scale fusion unit and a classification layer;
step four and step two: normalizing the data set image preprocessed in the third step to 512 x 512, and inputting the normalized data set image serving as a training set into the neural network classification model built in the fourth step;
step four and step three: suppose that the true label of a sample is ytY of the sampletProbability of 1 being ypThen the cross entropy loss function for this sample is shown as:
log(yt|yp)＝-(yt×log(yp)+(1-yt)log(1-yp))
the neural network classification model loss function is the average value of the cross entropy loss functions of the data set images preprocessed in all the third step;
step four: performing error back propagation and gradient descent according to the cross entropy loss function, and updating the weight parameters of the neural network classification model;
step four and five: repeating the fourth step, the third step, the fourth step and the fourth step until the loss function is gradually converged and stabilized, determining that the weight parameters of the current neural network classification model are the weight parameters of the trained neural network classification model, and further obtaining the trained deep learning rolling bearing oil slinging fault neural network classification model;
and fifthly, judging whether the linear array image of the railway wagon to be detected is the oil throwing fault image of the rolling bearing.
The invention has the beneficial effects that:
the invention obtains the whole-train image of the motor train by shooting the running truck. And combining the knowledge in the fields of image processing, pattern recognition, deep learning and the like. The oil throwing fault automatic identification and alarm are realized, the alarm result is only confirmed manually, and finally the machine inspection operation is realized to replace the manual inspection operation, so that the labor cost of a unit can be saved, and the operation quality and the operation efficiency can be improved.
A linear array camera (also called a line scanning camera) is used for shooting a truck moving at a high speed and shooting images on two sides of the truck. And obtaining an interested area containing the oil slinging fault in the large online array image according to the wheel base information and the prior information of the oil slinging fault position. And carrying out data amplification on the region of interest of the oil shedding fault, constructing a multi-label neural network training set, building a dense neural network classification model, training the neural network model until the model converges, and converting the precision of the parameter weight of the classification model. In practical application, the neural network classification weight of the conversion progress is loaded, whether the shot component image is an oil slinging fault image or not is judged, and an alarm is given to an oil slinging fault area.
The automatic image identification is adopted to replace manual detection of the oil throwing fault of the rolling bearing, the identification standards of the oil throwing fault of the rolling bearing are unified, and the automatic image identification is not influenced by the occupational quality and responsibility of workers. Compared with the traditional machine vision detection method of manual standard feature extraction, the rolling bearing oil shedding fault detection method based on deep learning has high flexibility, accuracy and robustness. The adoption of a multi-label data set is less in false alarm than a single-label data set, and the adoption of a dense network is higher in accuracy than that of a common network. The model precision conversion can accelerate the recognition speed of the oil throwing of the rolling bearing and realize real-time detection.
Drawings
FIG. 1 is an overall flow chart of the present invention;
FIG. 2a is a normal schematic view of bogie K2;
FIG. 2b is a schematic drawing of the oil slinging of the bogie K2;
FIG. 2c is a normal schematic view of bogie K5;
FIG. 2d is a schematic drawing of the oil slinging of the bogie K5;
FIG. 2e is a normal schematic view of bogie K6;
FIG. 2f is a schematic drawing of the truck K6 undergoing oil slinging;
FIG. 3 is a diagram of Dense Block (Dense Block) cells;
fig. 4 is a diagram of a classification network structure.
Detailed Description
The first embodiment is as follows: the embodiment is described with reference to fig. 1, and the method for detecting the oil slinging fault of the rolling bearing of the railway wagon comprises the following specific processes:
firstly, acquiring a linear array image;
step two, coarse positioning;
thirdly, preprocessing a data set image;
step four, classifying fault targets; the specific process is as follows:
step four, firstly: because the fault area of the rolling bearing for oil throwing is larger and the proportion of the fault area occupying the area of interest is larger, the invention adopts a target classification method based on dense network DenseNet; because the requirement on the detection precision is higher in the automatic identification and detection of the truck, and the problem of the precision of the oil throwing fault of the rolling bearing is solved, the dense network DenseNet is used as a feature extraction network, the dense network process ensures that the relation between network layers is ensured on the basis of deepening the network layer number, and simultaneously, the multi-scale features are added, so that more texture information of the oil throwing of the rolling bearing at the bottom layer is provided, and the detection precision is ensured.
Building a neural network classification model, which comprises a feature extraction network, a multi-scale fusion unit and a classification layer;
step four and step two: normalizing the data set image preprocessed in the third step to 512 x 512, and inputting the normalized data set image serving as a training set into the neural network classification model built in the fourth step;
step four and step three: a loss function, wherein the loss function is binary cross entropy, and for a sample (x, y), x is a corresponding label of the sample y;
in the binary problem, the set of sample values may be {0, 1}, assuming that the true label of a sample is ytY of the sampletProbability of 1 being ypThen the cross entropy loss function for this sample is shown as:
log(yt|yp)＝-(yt×log(yp)+(1-yt)log(1-yp))
the neural network classification model loss function is the average value of the cross entropy loss functions of the data set images preprocessed in all the third step;
step four: performing error back propagation and gradient descent according to the cross entropy loss function, and updating the weight parameters of the neural network classification model;
step four and five: and repeating the fourth step, the third step, the fourth step, the fifth step, the sixth step, the seventh step, the eleventh step, the twelfth step and the eleventh step, performing multi-round training on a plurality of groups of rolling bearing oil slinging training sample pictures until the loss function is gradually converged and stabilized, determining that the weight parameters of the current neural network classification model are the weight parameters of the trained neural network classification model.
And fifthly, judging whether the linear array image of the railway wagon to be detected is the oil throwing fault image of the rolling bearing.
The second embodiment is as follows: the first embodiment is different from the first embodiment in that the line array image is obtained in the first step; the specific process is as follows:
the method comprises the steps that a linear array camera is carried by utilizing fixed equipment, a railway wagon moving at a high speed is shot, and whole-wagon images of the upper portion, two sides and the bottom of the railway wagon are shot;
aiming at a linear array camera (also called a line scanning camera), the shooting frequency of the linear array camera is calculated according to the moving speed of a measured object, continuous multiple times of shooting are carried out, and a plurality of shot strip-shaped images are combined into a complete image, so that seamless splicing can be realized, and a two-dimensional image with a large visual field and high precision is generated.
Other steps and parameters are the same as those in the first embodiment.
The third concrete implementation mode: the difference between the second embodiment and the first or second embodiment is that the second step is performed by coarse positioning; the specific process is as follows:
according to the wheel base information of hardware, the priori knowledge of the oil throwing position of a rolling bearing and the like, an interested area is cut out from the complete image information of the whole vehicle, so that the calculated amount is reduced, and the identification speed is improved.
Other steps and parameters are the same as those in the first or second embodiment.
The fourth concrete implementation mode: this embodiment differs from one of the first to third embodiments in that the data set image is preprocessed in step three; the specific process is as follows:
step three, firstly: establishing an original data set;
obtaining an interested area image of the part to be identified according to the rough positioning, and establishing an original data set;
step three: performing data amplification operation on the original data set established in the third step by adopting an image processing mode of contrast enhancement, histogram equalization and random scaling;
due to the influence of interference factors such as train speed, outdoor sunlight, rain and snow and the like, problems of different contrast, image stretching and the like of linear array images of a running truck can occur, and aiming at the specific problem of automatic identification of component images, an image enhancement method of contrast enhancement, histogram equalization and random scaling is adopted, so that not only can more training data be obtained, but also the robustness of a model can be improved.
Step three: data marking:
because the different types of the bogie and the fault forms of the oil throwing of the rolling bearing are different, if the bogie type is not classified, only normal oil throwing and oil throwing are classified, the effect of classifying the oil throwing of the rolling bearing is poor, a multi-label classification data set is adopted, a multi-label comprises the bogie type and whether the oil is thrown, and data and labels are shown in figures 2a, 2b, 2c, 2d, 2e and 2f, so that the multi-label classification can obtain a better oil throwing classification effect.
Other steps and parameters are the same as those in one of the first to third embodiments.
The fifth concrete implementation mode: the difference between the embodiment and one of the first to fourth specific embodiments is that in the first step, because the fault area of the rolling bearing in oil throwing is large and the proportion of the fault area in the area of interest is large, the target classification method based on dense network DenseNet is adopted; because the requirement on the detection precision is higher in the automatic identification and detection of the truck, and the problem of the precision of the oil throwing fault of the rolling bearing is solved, the dense network DenseNet is used as a feature extraction network, the dense network process ensures that the relation between network layers is ensured on the basis of deepening the network layer number, and simultaneously, the multi-scale features are added, so that more texture information of the oil throwing of the rolling bearing at the bottom layer is provided, and the detection precision is ensured.
Building a neural network classification model, which comprises a feature extraction network, a multi-scale fusion unit and a classification layer; the specific process is as follows:
the reference feature extraction part of the network is shown in fig. 3 and 4, a Dense network DenseNet ensures the relation among network layers on the basis of deepening the network layers, the training speed of the network is accelerated, the accuracy of a model is greatly improved, the problem of feature loss of a common serial deep feedforward network is solved on the premise of not increasing the calculation complexity, context information is fully utilized, and more rolling bearing oil shedding fault bottom layer texture features can be multiplexed by using a Dense Block (Dense Block) in 4 stages (stages) to form a high-speed path in a hierarchy.
The feature extraction network adopts 4 downsampled units of Dense Block 1(Dense Block1), Dense Block 2(DenseBlock2), Dense Block 3(Dense Block3) and Dense Block 4(Dense Block 4);
the dense block1 unit comprises six groups of convolution units, each group of convolution units comprises convolution of 1 x 1 and 3 x 3, wherein the output of the convolution unit of the first group is used as the input of the convolution unit of the second group, the third group, the fourth group, the fifth group and the sixth group, the output of the convolution unit of the second group is used as the input of the convolution unit of the third group, the fourth group, the fifth group and the sixth group, the output of the convolution unit of the third group is used as the input of the convolution unit of the fourth group, the fifth group and the sixth group, the output of the convolution unit of the fourth group is used as the input of the convolution unit of the fifth group and the sixth group, and the output of the convolution unit of the fifth group is used as the input of the convolution unit of the sixth group;
the dense block2 unit is the same as the dense block1 unit;
the dense block3 unit is the same as the dense block1 unit;
dense block4 units are the same as dense block1 units;
the multi-scale feature fusion process is as shown in fig. 3 and 4, and the multi-scale fusion unit directly splices the outputs of different levels of the first sub-scale feature extraction unit, the second sub-scale feature extraction unit and the third sub-scale feature extraction unit to the final output to form a high-speed path between levels of each stage;
the first sub-scale feature extraction unit outputs the output of the dense block1 through a down-sampling unit, the down-sampling unit consists of a maximum pooling layer, the output of the down-sampling layer is convoluted by 1 x 1, and finally the output of the down-sampling layer is output to the second sub-scale feature unit through a connection operation;
the second sub-scale feature extraction unit is characterized in that the output of the dense block2 passes through a down-sampling unit, the down-sampling unit consists of a maximum pooling layer, the output of the down-sampling layer and the first sub-scale feature unit pass through a 1 x 1 convolution and are finally output to a third sub-scale feature unit through a connection operation;
the third sub-scale feature extraction unit is characterized in that the output of the Dense Block3 passes through a down-sampling unit, the down-sampling unit consists of a maximum pooling layer, the output of the down-sampling layer and the second sub-scale feature unit pass through 1 x 1 convolution, and finally are fused with a Dense Block4 through a connection operation output and input into the last classification layer;
the classification layer adopts a sigmoid function to replace a conventional softmax activation function, because each category of multi-label classification is crossed, each category is not mutually exclusive, and the sigmoid function is shown as the following formula:
sigmoid binary function: converting the scoring result into probability for classification;
the input x belongs to any real value, and the output range is 0 to 1;
according to the classification network provided by the invention, dense block units are used in each stage to multiplex the oil throwing characteristics of the rolling bearing at the bottom layer to form an inner-level high-speed passage, the characteristics of the output of each stage after down sampling are directly spliced to the final output to form an inter-level high-speed passage, and the combination of the inner-level high-speed passage and the inter-level high-speed passage can enhance the proportion of the oil throwing characteristics of the rolling bearing at the bottom layer in the final classification layer and improve the oil throwing accuracy of the classification rolling bearing.
Other steps and parameters are the same as in one of the first to fourth embodiments.
The sixth specific implementation mode: the fifth step is to judge whether the linear array image of the railway wagon to be detected is an oil throwing fault image of the rolling bearing;
step five, first: aiming at the problems of high real-time requirement on detection and real-time problem of oil shedding fault of a rolling bearing in automatic identification and detection of a truck, an acceleration reasoning engine is used for model conversion, the weight precision of a trained classification model of a neural network is converted from 32-bit precision to 16-bit precision, the calculated amount of the neural network is reduced, and the acceleration purpose is achieved;
step five two: predicting a neural network;
obtaining a linear array image of the railway wagon to be detected, inputting the linear array image into a neural network classification model converted into 16-bit precision, and performing classification prediction;
step five and step three: according to the prediction result, if the image is an oil throwing fault image of the rolling bearing, the information of the oil throwing fault of the rolling bearing is uploaded to an alarm platform; and if the rolling bearing oil throwing fault image is not detected, detecting the next railway wagon linear array image to be detected.
Other steps and parameters are the same as those in one of the first to fifth embodiments.
The present invention is capable of other embodiments and its several details are capable of modifications in various obvious respects, all without departing from the spirit and scope of the present invention.
Claims (6)
1. The method for detecting the oil throwing fault of the rolling bearing of the railway wagon is characterized by comprising the following steps of: the method comprises the following specific processes:
firstly, acquiring a linear array image;
step two, coarse positioning;
thirdly, preprocessing a data set image;
step four, classifying fault targets; the specific process is as follows:
step four, firstly: building a neural network classification model, which comprises a feature extraction network, a multi-scale fusion unit and a classification layer;
step four and step two: normalizing the data set image preprocessed in the third step to 512 x 512, and inputting the normalized data set image serving as a training set into the neural network classification model built in the fourth step;
step four and step three: suppose that the true label of a sample is ytY of the sampletProbability of 1 being ypThen the cross entropy loss function for this sample is shown as:
log(yt|yp)＝-(yt×log(yp)+(1-yt)log(1-yp))
the neural network classification model loss function is the average value of the cross entropy loss functions of the data set images preprocessed in all the third step;
step four: performing error back propagation and gradient descent according to the cross entropy loss function, and updating the weight parameters of the neural network classification model;
step four and five: repeating the fourth step, the third step, the fourth step and the fourth step until the loss function is gradually converged and stabilized, determining that the weight parameters of the current neural network classification model are the weight parameters of the trained neural network classification model, and further obtaining the trained deep learning rolling bearing oil slinging fault neural network classification model;
and fifthly, judging whether the linear array image of the railway wagon to be detected is the oil throwing fault image of the rolling bearing.
2. The method for detecting the oil slinging fault of the rolling bearing of the railway wagon according to claim 1, wherein the method comprises the following steps: acquiring a linear array image in the first step; the specific process is as follows:
the method comprises the steps that a linear array camera is carried by utilizing fixed equipment, moving railway wagons are shot, and whole wagon images of the upper portion, two sides and the bottom of the railway wagons are shot;
and calculating the shooting frequency of the linear array camera according to the moving speed of the object to be measured, continuously shooting, and combining the shot bar images into a complete image.
3. The method for detecting the oil slinging fault of the rolling bearing of the railway wagon according to claim 1 or 2, wherein the method comprises the following steps: coarse positioning in the second step; the specific process is as follows:
and cutting out an interested area from the complete image information of the whole vehicle according to the wheel base information of the hardware and the prior knowledge of the oil throwing position of the rolling bearing.
4. The method for detecting the oil slinging fault of the rolling bearing of the railway wagon according to claim 3, wherein the method comprises the following steps: preprocessing the data set image in the third step; the specific process is as follows:
step three, firstly: establishing an original data set;
obtaining an interested area image of the part to be identified according to the rough positioning, and establishing an original data set;
step three: performing data amplification operation on the original data set established in the third step by adopting an image processing mode of contrast enhancement, histogram equalization and random scaling;
step three: data marking:
a multi-label classification dataset is employed, the multi-label including the bogie type and whether to fling oil.
5. The method for detecting the oil slinging fault of the rolling bearing of the railway wagon according to claim 4, wherein the method comprises the following steps: establishing a neural network classification model in the fourth step, wherein the neural network classification model comprises a feature extraction network, a multi-scale fusion unit and a classification layer; the specific process is as follows:
the feature extraction network adopts 4 units of down-sampled dense blocks 1, dense blocks 2, dense blocks 3 and dense blocks 4;
the dense block1 unit comprises six groups of convolution units, each group of convolution units comprises convolution of 1 x 1 and 3 x 3, wherein the output of the convolution unit of the first group is used as the input of the convolution unit of the second group, the third group, the fourth group, the fifth group and the sixth group, the output of the convolution unit of the second group is used as the input of the convolution unit of the third group, the fourth group, the fifth group and the sixth group, the output of the convolution unit of the third group is used as the input of the convolution unit of the fourth group, the fifth group and the sixth group, the output of the convolution unit of the fourth group is used as the input of the convolution unit of the fifth group and the sixth group, and the output of the convolution unit of the fifth group is used as the input of the convolution unit of the sixth group;
the dense block2 unit is the same as the dense block1 unit;
the dense block3 unit is the same as the dense block1 unit;
dense block4 units are the same as dense block1 units;
the multi-scale fusion unit is used for directly splicing the outputs of different levels of the first sub-scale feature extraction unit, the second sub-scale feature extraction unit and the third sub-scale feature extraction unit to the final output to form a high-speed path between levels of each stage;
the first sub-scale feature extraction unit outputs the output of the dense block1 through a down-sampling unit, the down-sampling unit consists of a maximum pooling layer, the output of the down-sampling layer is convoluted by 1 x 1, and finally the output of the down-sampling layer is output to the second sub-scale feature unit through a connection operation;
the second sub-scale feature extraction unit is characterized in that the output of the dense block2 passes through a down-sampling unit, the down-sampling unit consists of a maximum pooling layer, the output of the down-sampling layer and the first sub-scale feature unit pass through a 1 x 1 convolution and are finally output to a third sub-scale feature unit through a connection operation;
the third sub-scale feature extraction unit is characterized in that the output of the Dense Block3 passes through a down-sampling unit, the down-sampling unit consists of a maximum pooling layer, the output of the down-sampling layer and the second sub-scale feature unit pass through 1 x 1 convolution, and finally are fused with a Dense Block4 through a connection operation output and input into the last classification layer;
the classification layer adopts a sigmoid function, and the sigmoid function is shown as the following formula:
sigmoid function: converting the scoring result into probability for classification;
the input x is of any real value and the output ranges from 0 to 1.
6. The method for detecting the oil slinging fault of the rolling bearing of the railway wagon according to claim 5, wherein the method comprises the following steps: judging whether the linear array image of the railway wagon to be detected is an oil throwing fault image of the rolling bearing or not;
step five, first: converting the weight precision of the classification model of the trained neural network from 32 bits to 16 bits;
step five two: predicting a neural network;
obtaining a linear array image of the railway wagon to be detected, inputting the linear array image into a neural network classification model converted into 16-bit precision, and performing classification prediction;
step five and step three: according to the prediction result, if the image is an oil throwing fault image of the rolling bearing, the information of the oil throwing fault of the rolling bearing is uploaded to an alarm platform; and if the rolling bearing oil throwing fault image is not detected, detecting the next railway wagon linear array image to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911272538.7A CN111079748A (en) | 2019-12-12 | 2019-12-12 | Method for detecting oil throwing fault of rolling bearing of railway wagon |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911272538.7A CN111079748A (en) | 2019-12-12 | 2019-12-12 | Method for detecting oil throwing fault of rolling bearing of railway wagon |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111079748A true CN111079748A (en) | 2020-04-28 |
Family
ID=70314050
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911272538.7A Pending CN111079748A (en) | 2019-12-12 | 2019-12-12 | Method for detecting oil throwing fault of rolling bearing of railway wagon |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111079748A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111665819A (en) * | 2020-06-08 | 2020-09-15 | 杭州电子科技大学 | Deep learning multi-model fusion-based complex chemical process fault diagnosis method |
| CN111738367A (en) * | 2020-08-17 | 2020-10-02 | 成都中轨轨道设备有限公司 | Part classification method based on image recognition |
| CN112001123A (en) * | 2020-08-27 | 2020-11-27 | 杭州电子科技大学 | Chemical process performance evaluation method based on deep convolutional network feature fusion |
| CN112364925A (en) * | 2020-11-16 | 2021-02-12 | 哈尔滨市科佳通用机电股份有限公司 | Deep learning-based rolling bearing oil shedding fault identification method |
| CN112633132A (en) * | 2020-12-18 | 2021-04-09 | 合肥工业大学 | Bearing fault diagnosis method and system based on two-dimensional vibration image enhancement |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109272048A (en) * | 2018-09-30 | 2019-01-25 | 北京工业大学 | A kind of mode identification method based on depth convolutional neural networks |
| CN109376683A (en) * | 2018-11-09 | 2019-02-22 | 中国科学院计算技术研究所 | A kind of video classification methods and system based on dense graph |
| CN109409222A (en) * | 2018-09-20 | 2019-03-01 | 中国地质大学（武汉） | A kind of multi-angle of view facial expression recognizing method based on mobile terminal |
| CN109614869A (en) * | 2018-11-10 | 2019-04-12 | 天津大学 | A kind of pathological image classification method based on multi-scale compress rewards and punishments network |
| CN110084292A (en) * | 2019-04-18 | 2019-08-02 | 江南大学 | Object detection method based on DenseNet and multi-scale feature fusion |
| CN110231825A (en) * | 2019-06-21 | 2019-09-13 | 中国神华能源股份有限公司 | Vehicular intelligent cruising inspection system and method |
-
2019
- 2019-12-12 CN CN201911272538.7A patent/CN111079748A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109409222A (en) * | 2018-09-20 | 2019-03-01 | 中国地质大学（武汉） | A kind of multi-angle of view facial expression recognizing method based on mobile terminal |
| CN109272048A (en) * | 2018-09-30 | 2019-01-25 | 北京工业大学 | A kind of mode identification method based on depth convolutional neural networks |
| CN109376683A (en) * | 2018-11-09 | 2019-02-22 | 中国科学院计算技术研究所 | A kind of video classification methods and system based on dense graph |
| CN109614869A (en) * | 2018-11-10 | 2019-04-12 | 天津大学 | A kind of pathological image classification method based on multi-scale compress rewards and punishments network |
| CN110084292A (en) * | 2019-04-18 | 2019-08-02 | 江南大学 | Object detection method based on DenseNet and multi-scale feature fusion |
| CN110231825A (en) * | 2019-06-21 | 2019-09-13 | 中国神华能源股份有限公司 | Vehicular intelligent cruising inspection system and method |
Non-Patent Citations (4)
| Title |
|---|
| WATERSINK: "《DenseNet解读》", 《HTTPS://BLOG.CSDN.NET/QQ_14845119/ARTICLE/DETAILS/79272082》 * |
| 小小将: "《DenseNet：比ResNet更优的CNN模型》", 《HTTPS://ZHUANLAN.ZHIHU.COM/P/37189203》 * |
| 张江勇: ""基于深度学习的动车关键部位故障图像检测"", 《万方数据知识服务平台》 * |
| 打着灯笼摸黑: "《动手学深度学习》第二十三天---稠密连接网络（DenseNet）", 《HTTPS://BLOG.CSDN.NET/QQ_32166627/ARTICLE/DETAILS/82956735》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111665819A (en) * | 2020-06-08 | 2020-09-15 | 杭州电子科技大学 | Deep learning multi-model fusion-based complex chemical process fault diagnosis method |
| CN111738367A (en) * | 2020-08-17 | 2020-10-02 | 成都中轨轨道设备有限公司 | Part classification method based on image recognition |
| CN111738367B (en) * | 2020-08-17 | 2020-11-13 | 成都中轨轨道设备有限公司 | Part classification method based on image recognition |
| CN112001123A (en) * | 2020-08-27 | 2020-11-27 | 杭州电子科技大学 | Chemical process performance evaluation method based on deep convolutional network feature fusion |
| CN112364925A (en) * | 2020-11-16 | 2021-02-12 | 哈尔滨市科佳通用机电股份有限公司 | Deep learning-based rolling bearing oil shedding fault identification method |
| CN112364925B (en) * | 2020-11-16 | 2021-06-04 | 哈尔滨市科佳通用机电股份有限公司 | Deep learning-based rolling bearing oil shedding fault identification method |
| CN112633132A (en) * | 2020-12-18 | 2021-04-09 | 合肥工业大学 | Bearing fault diagnosis method and system based on two-dimensional vibration image enhancement |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111079748A (en) | Method for detecting oil throwing fault of rolling bearing of railway wagon | |
| CN111652227B (en) | Method for detecting damage fault of bottom floor of railway wagon | |
| CN111091555B (en) | Brake shoe breaking target detection method | |
| CN113436169B (en) | Industrial equipment surface crack detection method and system based on semi-supervised semantic segmentation | |
| CN103279765B (en) | Steel wire rope surface damage detection method based on images match | |
| CN111080609B (en) | Brake shoe bolt loss detection method based on deep learning | |
| CN111080617B (en) | Railway wagon brake beam pillar round pin loss fault identification method | |
| CN110047070B (en) | Method and system for identifying rail wear degree | |
| CN112330591A (en) | Steel rail surface defect detection method and device capable of achieving sample-less learning | |
| CN106290388A (en) | A kind of insulator breakdown automatic testing method | |
| CN111311567A (en) | Method for identifying fastener and steel rail diseases of track line image | |
| CN111080612B (en) | Truck bearing damage detection method | |
| CN113436157B (en) | Vehicle-mounted image identification method for pantograph fault | |
| CN102346844B (en) | Device and method for identifying fault of losing screw bolts for truck center plates | |
| CN113837007B (en) | Bridge load distribution identification method based on vehicle matching deep learning in bridge monitoring video | |
| US11495022B2 (en) | Method for recognizing an object of a mobile unit | |
| CN111080607B (en) | Rolling bearing oil slinging fault detection method based on image recognition | |
| CN116823800A (en) | Bridge concrete crack detection method based on deep learning under complex background | |
| CN112115770A (en) | Method and system for identifying autonomous inspection defects of unmanned aerial vehicle of overhead line | |
| CN111768417A (en) | Railway wagon overrun detection method based on monocular vision 3D reconstruction technology | |
| CN111079629A (en) | Railway wagon center plate out-of-position fault image identification method | |
| CN108648210B (en) | Rapid multi-target detection method and device under static complex scene | |
| CN116579992A (en) | Small target bolt defect detection method for unmanned aerial vehicle inspection | |
| CN107392313B (en) | Steel rail identification method based on deep learning | |
| CN115424128A (en) | Fault image detection method and system for lower link of freight car bogie |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200428 |
|
| RJ01 | Rejection of invention patent application after publication |