CN113945569B - Fault detection method and device for ion membrane - Google Patents
Fault detection method and device for ion membrane Download PDFInfo
- Publication number
- CN113945569B CN113945569B CN202111163342.1A CN202111163342A CN113945569B CN 113945569 B CN113945569 B CN 113945569B CN 202111163342 A CN202111163342 A CN 202111163342A CN 113945569 B CN113945569 B CN 113945569B
- Authority
- CN
- China
- Prior art keywords
- model
- ion membrane
- image
- membrane
- ion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 239000012528 membrane Substances 0.000 title claims abstract description 257
- 238000001514 detection method Methods 0.000 title claims abstract description 132
- 238000012549 training Methods 0.000 claims abstract description 89
- 238000002372 labelling Methods 0.000 claims abstract description 64
- 238000000034 method Methods 0.000 claims abstract description 38
- 239000001257 hydrogen Substances 0.000 claims abstract description 37
- 229910052739 hydrogen Inorganic materials 0.000 claims abstract description 37
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 claims abstract description 36
- 238000004519 manufacturing process Methods 0.000 claims abstract description 36
- 238000012360 testing method Methods 0.000 claims abstract description 34
- 238000012545 processing Methods 0.000 claims abstract description 21
- 238000013508 migration Methods 0.000 claims abstract description 14
- 230000005012 migration Effects 0.000 claims abstract description 14
- 238000003062 neural network model Methods 0.000 claims abstract description 10
- 238000010200 validation analysis Methods 0.000 claims abstract description 10
- 230000006870 function Effects 0.000 claims description 29
- 238000012795 verification Methods 0.000 claims description 24
- 230000004913 activation Effects 0.000 claims description 21
- 238000010606 normalization Methods 0.000 claims description 20
- 238000003860 storage Methods 0.000 claims description 19
- 238000013526 transfer learning Methods 0.000 claims description 19
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 7
- 230000002159 abnormal effect Effects 0.000 claims description 5
- 230000007797 corrosion Effects 0.000 claims description 5
- 238000005260 corrosion Methods 0.000 claims description 5
- 238000005336 cracking Methods 0.000 claims description 5
- 230000008021 deposition Effects 0.000 claims description 5
- 230000035515 penetration Effects 0.000 claims description 5
- 230000007257 malfunction Effects 0.000 claims 2
- 238000013527 convolutional neural network Methods 0.000 abstract description 15
- 150000002500 ions Chemical class 0.000 description 157
- 238000010586 diagram Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 4
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000001301 oxygen Substances 0.000 description 3
- 229910052760 oxygen Inorganic materials 0.000 description 3
- 230000002265 prevention Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000015556 catabolic process Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 150000001450 anions Chemical class 0.000 description 1
- 238000005341 cation exchange Methods 0.000 description 1
- 150000001768 cations Chemical class 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005868 electrolysis reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 150000002431 hydrogen Chemical class 0.000 description 1
- 239000003014 ion exchange membrane Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 229910001868 water Inorganic materials 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Chemical & Material Sciences (AREA)
- Pathology (AREA)
- Immunology (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Signal Processing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Separation Using Semi-Permeable Membranes (AREA)
Abstract
The invention discloses a fault detection method and device for an ion membrane. Wherein the method comprises the following steps: acquiring an ion membrane image set, wherein the ion membrane image set comprises a plurality of ion membranes with different fault categories; labeling the ion membrane image set based on the fault category to obtain a labeling data set, wherein the labeling data set comprises: training set, validation set and test set; training a target residual error network model by using the marking data set to obtain an ion membrane fault detection model, wherein the target residual error network model is a neural network model obtained by performing natural image characteristic migration; and detecting whether the ion membrane of the hydrogen production equipment fails by adopting the ion membrane failure detection model to obtain a failure detection result. The invention solves the technical problems that the existing ionic membrane fault detection method based on the convolutional neural network model is easy to generate overfitting and cannot give consideration to the detection processing speed and the detection accuracy in the prior art.
Description
Technical Field
The invention relates to the field of detection of ionic membranes, in particular to a fault detection method and device of an ionic membrane.
Background
The membrane cell electrolysis method is a method of separating an electrolytic product by dividing a cell into an anode chamber and a cathode chamber by a cation exchange membrane. The ion exchange membrane has the characteristic of selective permeation to anions and cations, and allows ions with one charge to pass through and limits ions with opposite charges to pass through, so that the aim of separating hydrogen, oxygen and water is fulfilled, but in the actual production process, the early warning is realized when the oxygen content in the hydrogen production storage exceeds 2%, and explosion occurs when the oxygen content exceeds 4%, so that the method is an important link for purposely detecting the faults of the ion membrane.
In the prior art, fault detection is carried out on an ion membrane, namely, the real-time state of the ion membrane of hydrogen production equipment is generally acquired, a video frame is extracted from a history detection image library and a shot video to form an image set, and the image set is preprocessed, however, in the prior art, the following problems exist in the use of the prior convolutional neural network-based deep learning process: 1) As shown in table 1 below, because of too many model parameters, if the training dataset is limited, overfitting can easily occur; the larger the network is, the more parameters are, the greater the calculation complexity is, and the application is difficult; the deeper the network, the more likely the gradient dispersion problem occurs (i.e. the more backward the gradient passes through, the more likely it disappears), and it is difficult to optimize the model.
TABLE 1
2) The ion membrane images of the hydrogen production equipment have various fault modes, the difference between the ion membrane images is sometimes obvious, the difference between some faults is not large, particularly the diagnostic images shot by the same image means are not completely spliced in a symmetrical mode, the image intervals after splicing are different, in addition, the acquisition of a data set is difficult, and the acquired data are required to be marked by professionals. Therefore, the data set labeling cost of the multi-label learning task is high, and the labeled data is too small in scale, so that the characteristic learning advantage cannot be fully exerted due to the fact that the fitting is easily caused by huge parameters of the characteristic learning model.
In view of the above problems, no effective solution has been proposed at present.
Disclosure of Invention
The embodiment of the invention provides a fault detection method and device for an ion membrane, which at least solve the technical problems that the existing ion membrane fault detection method based on a convolutional neural network model is easy to generate overfitting and cannot achieve the detection processing speed and the detection accuracy in the prior art.
According to an aspect of an embodiment of the present invention, there is provided a fault detection method for an ion membrane, including: acquiring an ion membrane image set, wherein the ion membrane image set comprises a plurality of ion membranes with different fault categories; labeling the ion membrane image set based on the fault category to obtain a labeling data set, wherein the labeling data set comprises: training set, validation set and test set; training a target residual error network model by using the marking data set to obtain an ion membrane fault detection model, wherein the target residual error network model is a neural network model obtained by performing natural image characteristic migration; and detecting whether the ion membrane of the hydrogen production equipment fails by adopting the ion membrane failure detection model to obtain a failure detection result.
Optionally, acquiring the ion membrane image set includes: acquiring a historical ion membrane image and a current ion membrane image, wherein the historical ion membrane image is an ion membrane image in a historical detection image library, and the current ion membrane image is an ion membrane image obtained by shooting an ion membrane in the hydrogen production equipment currently; and obtaining the ionic membrane image set based on the historical ionic membrane image and the current ionic membrane image.
Optionally, before labeling the ion membrane image set based on the fault category to obtain a labeled data set, the method further includes: scaling all the ionic membrane images in the ionic membrane image set to obtain all the adjusted ionic membrane images; adding category labels to all the adjusted ion membrane images, wherein the category labels are used for indicating the fault categories, and the fault categories comprise at least one of the following: deformation, peeling, cracking, heave, corrosion, deposition, fouling, penetration of foreign bodies.
Optionally, before labeling the ion membrane image set based on the fault category to obtain a labeled data set, the method further includes: carrying out data pretreatment on all the ionic membrane images in the ionic membrane image set to obtain pretreated ionic membrane images; the data preprocessing is used for processing missing values and abnormal values in the ionic membrane image; and extracting an image characteristic value from the preprocessed ion membrane image to obtain an image characteristic vector.
Optionally, after labeling the ionic membrane image set based on the fault category to obtain a labeled data set, the method further includes: acquiring the data processing requirement of the target residual error network model; based on the data processing requirements, the first proportion of the labeling data set is used as the training set, the second proportion of the labeling data set is used as the verification set, and the third proportion of the labeling data set is used as the test set, wherein the first proportion is larger than the second proportion and the third proportion.
Optionally, before training the target residual network model by using the labeling data set to obtain the ion membrane fault detection model, the method further includes: building an initial residual error network model; and migrating the natural image characteristics of the natural image set training model to the initial residual error network model to obtain the target residual error network model.
Optionally, the network structure of the initial residual network model is a vector convolution Conv1-batch standardization Batch Normalization algorithm-linear rectification ReLU1 activation function and a Conv2-Batch Normalization algorithm-ReLU 2 activation function; the convolution layer of the target residual network model adopts 3*3 convolution kernel, the network structure of the target residual network model is Conv1-Batch Normalization algorithm-ReLU 1 activation function, conv2-Batch Normalization algorithm-ReLU 2 activation function, conv3-Batch Normalization algorithm-ReLU 3 activation function is added at the quick connection position in the target residual network model, the output of the upper layer in the target residual network model is used as the input of the convolution layer at the quick connection position, and the first output of Conv1, the second output of Conv2 and the third output of Conv3 are used as the input of the next training stage.
Optionally, the number of convolution kernels in Conv1-10 convolution layers in the target residual network model is 64, the number of convolution kernels in Conv11-22 convolution layers is 128, the number of convolution kernels in Conv23-34 convolution layers is 256, and the number of convolution kernels in Conv35-40 convolution layers is 512.
Optionally, training the target residual error network model by using the labeling data set, and obtaining the ion membrane fault detection model includes: inputting all image data in the labeling data set into the target residual error network model for learning to obtain a transfer learning model; the training set is imported into the transfer learning model for training so as to update model parameters of the transfer learning model and obtain a training model; the verification set is imported into the training model to be trained so as to adjust model parameters of the training model, and a verification model is obtained; and introducing the test set into the verification model for testing to obtain the ion membrane fault detection model.
Optionally, detecting whether the ion membrane of the hydrogen production device has a fault by using the ion membrane fault detection model to obtain a fault detection result, including: and detecting the test set by adopting the ion membrane fault detection model to determine whether the ion membrane of the hydrogen production equipment has faults or not, so as to obtain the fault detection result.
According to another aspect of the embodiment of the present invention, there is also provided a fault detection device for an ion membrane, including: the system comprises an acquisition module, a detection module and a control module, wherein the acquisition module is used for acquiring an ion membrane image set, and the ion membrane image set contains a plurality of ion membranes with different fault categories; the labeling module is used for labeling the ion membrane image set based on the fault category to obtain a labeling data set, wherein the labeling data set comprises: training set, validation set and test set; the training module is used for training the target residual error network model by adopting the marking data set to obtain an ion membrane fault detection model, wherein the target residual error network model is a neural network model obtained by natural image feature migration; the detection module is used for detecting whether the ion membrane of the hydrogen production equipment fails or not by adopting the ion membrane failure detection model, and a failure detection result is obtained.
According to another aspect of the embodiments of the present invention, there is also provided a nonvolatile storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the fault detection method of any one of the above-mentioned ion membranes.
According to another aspect of an embodiment of the present invention, there is also provided an electronic device including a memory, in which a computer program is stored, and a processor configured to run the computer program to perform the fault detection method of an ion membrane described in any one of the above.
According to another aspect of the embodiment of the present invention, there is also provided a processor for running a program, where the program is configured to execute any one of the above fault detection methods for an ion membrane when running.
In the embodiment of the invention, an ionic membrane image set is acquired, wherein the ionic membrane image set comprises a plurality of ionic membranes with different fault categories; labeling the ion membrane image set based on the fault category to obtain a labeling data set, wherein the labeling data set comprises: training set, validation set and test set; training a target residual error network model by using the marking data set to obtain an ion membrane fault detection model, wherein the target residual error network model is a neural network model obtained by performing natural image characteristic migration; the ion membrane fault detection model is used for detecting whether the ion membrane of the hydrogen production equipment breaks down or not to obtain a fault detection result, the aim of avoiding the occurrence of the fitting phenomenon in the ion membrane fault detection process and considering the detection processing speed and the detection accuracy is fulfilled, the technical effect of ensuring the accurate prevention and control management of the hydrogen production equipment in the hydrogen production process is realized, and the technical problem that the fitting is easy to occur and the detection processing speed and the detection accuracy cannot be considered in the prior art by using the conventional ion membrane fault detection method based on the convolutional neural network model is solved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiments of the invention and together with the description serve to explain the invention and do not constitute a limitation on the invention. In the drawings:
FIG. 1 is a flow chart of a method of fault detection of an ionic membrane according to an embodiment of the present invention;
FIG. 2 is a schematic block diagram of an alternative initial residual network model according to an embodiment of the present invention;
FIG. 3 is a schematic block diagram of an alternative target residual network model according to an embodiment of the invention;
FIG. 4 is a schematic diagram of a network architecture of an alternative target residual network model according to an embodiment of the invention;
fig. 5 is a schematic structural diagram of an ion membrane fault detection device according to an embodiment of the present invention.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
First, in order to facilitate understanding of the embodiments of the present invention, some terms or nouns referred to in the present invention will be explained below:
the convolutional neural network (Convolutional Neural Network, CNN) is a feedforward neural network containing convolutional calculation and having a depth structure, is one of representative algorithms of deep learning, has characteristic learning capability, can automatically learn characteristic expression of original image pixel data in a large number of training samples, does not depend on manual design characteristics, and can obtain characteristic description better than the original data expression capability. A CNN image classification model with high classification precision and good robustness is newly built, and a data set with rich categories and large sample size is needed to train the model. Because the pre-training CNN model is fully trained on the large data set by the network weight parameters, the migration learning method can discover the unchanged characteristics and structures of the fields between two areas which are related to each other and are different from each other, and the data of the target area is directly migrated and multiplexed by using the rich label data which are different from the auxiliary area but are related to the auxiliary area.
The residual network ResNet has extremely high accuracy of computer vision detection, and has low reference quantity compared with VGGNet and GoogleNet, and has extremely outstanding effect. The ResNet structure can accelerate the training of the neural network very fast, the accuracy of the model is also improved greatly, and the ResNet model is superior to the common network structure in the aspects of convergence performance, classification performance, network correspondence and the like. The main innovation of the ResNet model is that the problem of model degradation is found, and a residual structure is introduced according to the problem, wherein the residual structure directly transmits the result of the previous layer to the next layer network, so that after the model is further deepened, the error at least keeps consistent and does not continue to increase, and the ResNet model can have more model layers and higher accuracy at the same time.
Along with the increase of the model depth, the model accuracy is gradually separated from a common network, so that the problem of model performance degradation under the depth condition is solved to a certain extent, and the performance of each task in the computer vision field is greatly improved in the later development.
Example 1
According to an embodiment of the present invention, there is provided an embodiment of a fault detection method for an ion membrane, it being noted that the steps shown in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowcharts, in some cases the steps shown or described may be performed in an order different from that herein.
Fig. 1 is a flowchart of a fault detection method of an ion membrane according to an embodiment of the present invention, as shown in fig. 1, the method includes the steps of:
step S102, an ion membrane image set is obtained, wherein the ion membrane image set comprises a plurality of ion membranes with different fault categories;
step S104, labeling the ion membrane image set based on the fault category to obtain a labeling data set, wherein the labeling data set comprises: training set, validation set and test set;
step S106, training a target residual error network model by using the marking data set to obtain an ion membrane fault detection model, wherein the target residual error network model is a neural network model obtained by performing natural image feature migration;
and S108, detecting whether the ion membrane of the hydrogen production equipment is faulty or not by adopting the ion membrane fault detection model, and obtaining a fault detection result.
In the embodiment of the invention, an ionic membrane image set is acquired, wherein the ionic membrane image set comprises a plurality of ionic membranes with different fault categories; labeling the ion membrane image set based on the fault category to obtain a labeling data set, wherein the labeling data set comprises: training set, validation set and test set; training a target residual error network model by using the marking data set to obtain an ion membrane fault detection model, wherein the target residual error network model is a neural network model obtained by performing natural image characteristic migration; the ion membrane fault detection model is used for detecting whether the ion membrane of the hydrogen production equipment breaks down or not to obtain a fault detection result, the aim of avoiding the occurrence of the fitting phenomenon in the ion membrane fault detection process and considering the detection processing speed and the detection accuracy is fulfilled, the technical effect of ensuring the accurate prevention and control management of the hydrogen production equipment in the hydrogen production process is realized, and the technical problem that the fitting is easy to occur and the detection processing speed and the detection accuracy cannot be considered in the prior art by using the conventional ion membrane fault detection method based on the convolutional neural network model is solved.
In the embodiment of the application, on the basis of the existing residual network ResNet structure, the ion membrane fault detection method of the hydrogen production equipment based on the ResNet structure with the improved staggered block structure and the migration learning method is provided, and has important significance for accurate prevention and control management in the hydrogen production process.
It should be noted that, the target residual error network model in the embodiment of the application utilizes the migration learning method to learn the general characteristics of the general field image, so that the problem of model overfitting caused by too small scale of the ion membrane image field of the hydrogen production equipment can be alleviated. For example, the embodiment of the application uses a large-scale natural image to train an initial residual network model, general features (such as color, texture, shape and the like) of the natural image are migrated to the field of the hydrogen production equipment ion membrane, a target residual network model is obtained, and the multi-label classification performance can be effectively improved by utilizing the local information of the labeled object class in the image.
Optionally, in the embodiment of the present application, an ion membrane image set may be generated based on a history detection image library and a high-definition photo taken in real time, and a label is added to the ion membrane image set according to a detection report of the ion membrane; for example, the detection report of the ion membrane is subjected to fault detection classification, and different fault categories are counted to be used as the labeling data set B.
After the image preprocessing is completed, dividing the marked data set B into a training set B1, a verification set B2 and a test set B3 according to the data processing requirement of a target residual error network model (namely, an improved ResNet CNN structure); inputting all image data in the labeling data set B into a target residual error network model for learning to obtain a target residual error network model, namely a transfer learning model M; the training set B1 is led into the transfer learning model M for training so as to update model parameters of the transfer learning model and obtain a training model M1; the verification set B2 is led into the training model to be trained so as to adjust model parameters of the training model M1, and a verification model M2 is obtained; the test set M3 is led into the verification model M2 for testing, so that the ion membrane fault detection model is obtained, and the ion membrane fault detection model M2 can be used for fault detection of the ion membrane.
Optionally, in the embodiment of the present application, the target residual network model res net uses a mode of 3*3 convolution kernels, and a plurality of small convolution kernels replace large convolution kernels, so that model parameters can be reduced, and the number of nonlinear activation functions is increased, so that the calculation amount of the model is reduced. However, small convolution kernel stacks increase model depth, model training becomes difficult, and training time increases. According to the embodiment of the application, by providing an improved ResNet structure model and combining with a feature diagram in a network structure, the number of layers of the network is reduced, redundant connection is removed, the network structure is simplified, and therefore the training time of the network is reduced; adding a convolution layer in the shortcut connection to form a new residual block type, combining residual blocks of different types, and improving the learning capacity of the whole network structure; the number of convolution kernels in part of convolution layers is increased, and the capability of extracting features of a network structure is improved.
In the embodiment of the application, the concept of transfer learning is adopted, and the problem of fitting a small sample is converted into the problem of training and adjusting parameters of the existing model by transferring a mature ResNet model, so that the speed of image recognition is greatly improved. In an alternative embodiment of the present application, the experimental results shown in the following tables 2 and 3 indicate that the improved res net model of the present application is a compact and effective network structure, and has the advantage of simple training on the premise of completing the detection task, that is, the average accuracy of fault detection is higher and the average time is less.
TABLE 2
TABLE 3 Table 3
In an alternative embodiment, acquiring an ion membrane image set includes:
step S202, acquiring a historical ionic membrane image and a current ionic membrane image;
step S204, obtaining the ion membrane image set based on the historical ion membrane image and the current ion membrane image.
In this embodiment of the present application, the historical ion membrane image is an ion membrane image in a historical detection image library, and the current ion membrane image is an ion membrane image obtained by shooting an ion membrane in the hydrogen production device. The ion membrane image set can be obtained by acquiring a historical ion membrane image in a historical detection image library and a current ion membrane image photographed in real time.
In an alternative embodiment, before labeling the ion membrane image set based on the fault class to obtain a labeled data set, the method further includes:
step S302, scaling all the ion membrane images in the ion membrane image set to obtain all the adjusted ion membrane images;
step S304, adding a category label to all the adjusted ion membrane images, wherein the category label is used for indicating the fault category, and the fault category comprises at least one of the following: deformation, peeling, cracking, heave, corrosion, deposition, fouling, penetration of foreign bodies.
In the embodiment of the application, the image size of the ion membrane image in the obtained ion membrane image set is adjusted, for example, all the ion membrane images are scaled to 1024 pixels at the shortest side in equal proportion, so that the adjusted ion membrane image is obtained. And then adding class labels to all the adjusted ion membrane images, wherein the class labels are used for indicating the fault classes, and the fault classes comprise at least one of the following: deformation, peeling, cracking, heave, corrosion, deposition, fouling, penetration of foreign bodies.
As an alternative embodiment, after adding a category label to the adjusted ion membrane image, adding the labeled ion membrane image set to the target residual network model, reading the adjusted ion membrane image and converting it into a Python voice numpy array, and feeding it and its corresponding category label to the target residual network model.
In an alternative embodiment, before labeling the ion membrane image set based on the fault class to obtain a labeled data set, the method further includes:
step S402, carrying out data preprocessing on all the ionic membrane images in the ionic membrane image set to obtain preprocessed ionic membrane images; the data preprocessing is used for processing missing values and abnormal values in the ionic membrane image;
and step S404, extracting an image characteristic value from the preprocessed ion membrane image to obtain an image characteristic vector.
Optionally, the missing value and the abnormal value of the ion membrane image can be processed by carrying out data preprocessing on all ion membrane images in the ion membrane image set, and the image characteristic value and the characteristic vector can be extracted.
In an alternative embodiment, after labeling the ion membrane image set based on the fault class to obtain a labeled data set, the method further includes:
Step S502, obtaining the data processing requirement of the target residual error network model;
step S504, based on the data processing requirement, uses the first proportion of the labeling data set as the training set, uses the second proportion of the labeling data set as the verification set, and uses the third proportion of the labeling data set as the test set, wherein the first proportion is greater than the second proportion and the third proportion.
Alternatively, in the embodiment of the present application, 70% of the labeled dataset B may be randomly used as the training set B1, 15% of the labeled dataset B may be used as the verification set B2, and the remaining 15% of the labeled dataset B may be used as the test set B3.
In an alternative embodiment, before training the target residual network model using the labeling data set to obtain the ion membrane fault detection model, the method further includes:
step S602, an initial residual error network model is built;
step S604, migrating the natural image features of the natural image set training model to the initial residual network model to obtain the target residual network model.
In the above alternative embodiment, the initial residual network model is built, and the natural image features of the natural image set training model are migrated to the initial residual network model, that is, the initial residual network model is trained by using a large-scale natural image, so as to obtain the target residual network model.
Through the embodiment of the application, general features (such as color, texture, shape and the like) of the natural image are migrated to the field of the hydrogen production equipment ion membrane to obtain the target residual network model, and the multi-label classification performance can be effectively improved by utilizing the local information of the labeled object types in the image.
In an alternative embodiment, the network structure of the initial residual network model is a vector convolution Conv1-batch normalization Batch Normalization algorithm-linear rectification ReLU1 activation function and a Conv2-Batch Normalization algorithm-ReLU 2 activation function; the convolution layer of the target residual network model adopts 3*3 convolution kernel, the network structure of the target residual network model is Conv1-Batch Normalization algorithm-ReLU 1 activation function, conv2-Batch Normalization algorithm-ReLU 2 activation function, conv3-Batch Normalization algorithm-ReLU 3 activation function is added at the quick connection position in the target residual network model, the output of the upper layer in the target residual network model is used as the input of the convolution layer at the quick connection position, and the first output of Conv1, the second output of Conv2 and the third output of Conv3 are used as the input of the next training stage.
As shown in fig. 2 and 3, the network structure of the initial residual network model (the original res net spread block model) is Conv1-Batch Normalization algorithm-ReLU 1 activation function-Conv 2-Batch Normalization algorithm-ReLU 2 activation function, the network structure of the target residual network model (the modified res net spread block model) is Conv1-Batch Normalization algorithm-ReLU 1 activation function-Conv 2-Batch Normalization algorithm-ReLU 2 activation function, and a Conv3-Batch Normalization algorithm-ReLU 3 activation function is added at the shortcut connection, the output of the upper layer is used as the input of the convolution layer at the shortcut connection, and the output of Conv1 and Conv2 and the output of Conv3 on the shortcut connection are used as the input of the next stage.
According to the embodiment of the application, the technical problems of too few samples, long model training time, low model precision and the like in the ionic membrane research of the hydrogen production equipment can be solved, and the improved ResNet CNN structure is used.
Compared with the original ResNet CNN structure, the structural design of the convolutional neural network is of great importance, and intuitively, the deeper the hierarchical structure of the convolutional neural network is, the more nodes are, the stronger the feature expression capability is. The individual detection is carried out by using network structures such as AlexNet, googleNet with fewer network layers, and the accuracy is relatively low. The ResNet model with more network layers is used for detection, and the training time is longer, so that trade-off is needed in detection accuracy and training time.
The improved model idea in the embodiment of the application is to combine the feature diagram in the network structure, reduce the layer number of the network, remove redundant connection, simplify the network structure, thereby reducing the training time of the network; adding a convolution layer in the shortcut connection to form a new residual block type, combining residual blocks of different types, and improving the learning capacity of the whole network structure; the number of convolution kernels in part of convolution layers is increased, and the capability of extracting features of a network structure is improved.
In an alternative embodiment, the number of convolution kernels in Conv1-10 convolution layers, the number of convolution kernels in Conv11-22 convolution layers, the number of convolution kernels in Conv23-34 convolution layers, and the number of convolution kernels in Conv35-40 convolution layers in the target residual network model are 64, 128, 256, and 512 respectively.
As shown in fig. 4, 7 modified res net models and 9 original res net models are used in the examples of the present application. In order to enhance the characteristic expression capability of the improved ResNet model, the number of convolution kernels in a part of convolution layers is increased. The number of convolution kernels in Conv1-10 convolution layers is 64, the number of convolution kernels in Conv11-22 convolution layers is increased to 128, the number of convolution kernels in Conv23-34 convolution layers is increased to 256, and the number of convolution kernels in Conv35-40 convolution layers is increased to 512.
Through the embodiment of the application, the improved ResNet difference block is utilized to increase the quick connection, a novel difference block structure is combined, the gradient is enlarged, the problem of gradient disappearance is avoided, the gradient is enlarged, the learning convergence speed is high, the training speed can be greatly accelerated, and the ion membrane fault detection efficiency is improved.
In an alternative embodiment, training the target residual network model using the labeling data set, and obtaining the ion membrane fault detection model includes:
step S702, inputting all image data in the marked data set into the target residual error network model for learning to obtain a transfer learning model;
step S704, the training set is imported into the transfer learning model for training so as to update model parameters of the transfer learning model and obtain a training model;
Step S706, the verification set is imported into the training model to train so as to adjust model parameters of the training model and obtain a verification model;
step S708, the test set is led into the verification model for testing, and the ion membrane fault detection model is obtained.
In the above embodiment of the present application, all image data in the labeling data set B are input into the initial residual network model for learning, so as to obtain the target residual network model, namely, the migration learning model M; the training set B1 is led into the transfer learning model M for training so as to update model parameters of the transfer learning model and obtain a training model M1; the verification set B2 is led into the training model to be trained so as to adjust model parameters of the training model M1, and a verification model M2 is obtained; the test set M3 is led into the verification model M2 for testing, so that the ion membrane fault detection model is obtained, and the ion membrane fault detection model M2 can be used for fault detection of the ion membrane.
In an alternative embodiment, detecting whether the ion membrane of the hydrogen production device has a fault by using the ion membrane fault detection model to obtain a fault detection result includes:
And step S802, detecting the test set by adopting the ion membrane fault detection model to determine whether the ion membrane of the hydrogen production equipment has faults or not, and obtaining the fault detection result.
In the embodiment of the application, the ion membrane fault detection model M2 is adopted to test the test set B3, so that the ion membrane fault detection result of the hydrogen production equipment can be obtained.
According to the embodiment of the application, the image set of the ImageNet is utilized to carry out improved ResNet model learning, a learning result is transferred to the hydrogen production equipment ionic membrane image recognition process, the problem of a small sample with difficult sampling is solved by using the concept of transfer learning, and the efficiency of the improved ResNet model is greatly improved.
Example 2
According to an embodiment of the present invention, there is further provided an apparatus embodiment for implementing the foregoing method for detecting a fault of an ion membrane, and fig. 5 is a schematic structural diagram of an apparatus for detecting a fault of an ion membrane according to an embodiment of the present invention, as shown in fig. 5, where the foregoing apparatus for detecting a fault of an ion membrane includes: an acquisition module 50, a labeling module 52, a training module 54, and a detection module 56, wherein:
an acquisition module 50, configured to acquire an ion membrane image set, where the ion membrane image set includes a plurality of ion membranes with different fault categories; the labeling module 52 is configured to label the ion membrane image set based on the fault type to obtain a labeled data set, where the labeled data set includes: training set, validation set and test set; the training module 54 is configured to train a target residual network model by using the labeling data set to obtain an ion membrane fault detection model, where the target residual network model is a neural network model obtained by performing natural image feature migration; and the detection module 56 is used for detecting whether the ion membrane of the hydrogen production equipment has faults by adopting the ion membrane fault detection model so as to obtain a fault detection result.
It should be noted that each of the above modules may be implemented by software or hardware, for example, in the latter case, it may be implemented by: the above modules may be located in the same processor; alternatively, the various modules described above may be located in different processors in any combination.
Here, the acquiring module 50, the labeling module 52, the training module 54 and the detecting module 56 correspond to steps S102 to S108 in embodiment 1, and the modules are the same as the examples and application scenarios implemented by the corresponding steps, but are not limited to those disclosed in embodiment 1. It should be noted that the above modules may be run in a computer terminal as part of the apparatus.
It should be noted that, the optional or preferred implementation manner of this embodiment may be referred to the related description in embodiment 1, and will not be repeated here.
The fault detection device of the ionic membrane may further include a processor and a memory, where the acquisition module 50, the labeling module 52, the training module 54, the detection module 56, and the like are stored as program units, and the processor executes the program units stored in the memory to implement corresponding functions.
The processor comprises a kernel, the kernel fetches corresponding program units from the memory, and one or more of the kernels can be arranged. The memory may include volatile memory, random Access Memory (RAM), and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), among other forms in computer readable media, the memory including at least one memory chip.
According to an embodiment of the present application, there is also provided an embodiment of a nonvolatile storage medium. Optionally, in this embodiment, the nonvolatile storage medium includes a stored program, where the device in which the nonvolatile storage medium is located is controlled to execute the fault detection method of any one of the ion membranes when the program runs.
Alternatively, in this embodiment, the above-mentioned nonvolatile storage medium may be located in any one of the computer terminals in the computer terminal group in the computer network or in any one of the mobile terminals in the mobile terminal group, and the above-mentioned nonvolatile storage medium includes a stored program.
Optionally, the program controls the device in which the nonvolatile storage medium is located to perform the following functions when running: acquiring an ion membrane image set, wherein the ion membrane image set comprises a plurality of ion membranes with different fault categories; labeling the ion membrane image set based on the fault category to obtain a labeling data set, wherein the labeling data set comprises: training set, validation set and test set; training a target residual error network model by using the marking data set to obtain an ion membrane fault detection model, wherein the target residual error network model is a neural network model obtained by performing natural image characteristic migration; and detecting whether the ion membrane of the hydrogen production equipment fails by adopting the ion membrane failure detection model to obtain a failure detection result.
Optionally, the program controls the device in which the nonvolatile storage medium is located to perform the following functions when running: acquiring a historical ion membrane image and a current ion membrane image, wherein the historical ion membrane image is an ion membrane image in a historical detection image library, and the current ion membrane image is an ion membrane image obtained by shooting an ion membrane in the hydrogen production equipment currently; and obtaining the ionic membrane image set based on the historical ionic membrane image and the current ionic membrane image.
Optionally, the program controls the device in which the nonvolatile storage medium is located to perform the following functions when running: scaling all the ionic membrane images in the ionic membrane image set to obtain all the adjusted ionic membrane images; adding category labels to all the adjusted ion membrane images, wherein the category labels are used for indicating the fault categories, and the fault categories comprise at least one of the following: deformation, peeling, cracking, heave, corrosion, deposition, fouling, penetration of foreign bodies.
Optionally, the program controls the device in which the nonvolatile storage medium is located to perform the following functions when running: carrying out data pretreatment on all the ionic membrane images in the ionic membrane image set to obtain pretreated ionic membrane images; the data preprocessing is used for processing missing values and abnormal values in the ionic membrane image; and extracting an image characteristic value from the preprocessed ion membrane image to obtain an image characteristic vector.
Optionally, the program controls the device in which the nonvolatile storage medium is located to perform the following functions when running: acquiring the data processing requirement of the target residual error network model; based on the data processing requirements, the first proportion of the labeling data set is used as the training set, the second proportion of the labeling data set is used as the verification set, and the third proportion of the labeling data set is used as the test set, wherein the first proportion is larger than the second proportion and the third proportion.
Optionally, the program controls the device in which the nonvolatile storage medium is located to perform the following functions when running: building an initial residual error network model; and migrating the natural image characteristics of the natural image set training model to the initial residual error network model to obtain the target residual error network model.
Optionally, the program controls the device in which the nonvolatile storage medium is located to perform the following functions when running: inputting all image data in the labeling data set into the target residual error network model for learning to obtain a transfer learning model; the training set is imported into the transfer learning model for training so as to update model parameters of the transfer learning model and obtain a training model; the verification set is imported into the training model to be trained so as to adjust model parameters of the training model, and a verification model is obtained; and introducing the test set into the verification model for testing to obtain the ion membrane fault detection model.
Optionally, the program controls the device in which the nonvolatile storage medium is located to perform the following functions when running: and detecting the test set by adopting the ion membrane fault detection model to determine whether the ion membrane of the hydrogen production equipment has faults or not, so as to obtain the fault detection result.
According to an embodiment of the present application, there is also provided an embodiment of a processor. Optionally, in this embodiment, the processor is configured to run a program, where the program executes any one of the fault detection methods of the ionic membrane during running the program.
According to an embodiment of the present application, there is also provided an embodiment of an electronic device, including a memory, in which a computer program is stored, and a processor configured to run the computer program to perform the fault detection method of any one of the above-mentioned ion membranes.
According to an embodiment of the present application, there is also provided an embodiment of a computer program product adapted to perform a program of steps of a fault detection method for an ion membrane initialized with any one of the above, when executed on a data processing device.
The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
In the foregoing embodiments of the present invention, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.
In the several embodiments provided in the present application, it should be understood that the disclosed technology content may be implemented in other manners. The above-described embodiments of the apparatus are merely exemplary, and the division of the units, for example, may be a logic function division, and may be implemented in another manner, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some interfaces, units or modules, or may be in electrical or other forms.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable non-volatile storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a non-volatile storage medium, including instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned nonvolatile storage medium includes: a U-disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a removable hard disk, a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.
Claims (11)
1. A fault detection method for an ion membrane, comprising:
acquiring an ion membrane image set, wherein the ion membrane image set comprises a plurality of ion membranes with different fault categories;
labeling the ion membrane image set based on the fault category to obtain a labeling data set, wherein the labeling data set comprises: training set, validation set and test set;
training a target residual error network model by adopting the marking data set to obtain an ion membrane fault detection model, wherein the target residual error network model is a neural network model obtained by natural image feature migration;
detecting whether the ion membrane of the hydrogen production equipment fails or not by adopting the ion membrane failure detection model to obtain a failure detection result;
before training the target residual error network model by adopting the labeling data set to obtain an ion membrane fault detection model, the method further comprises the following steps:
Building an initial residual error network model;
migrating natural image features of a natural image set training model to the initial residual error network model to obtain the target residual error network model;
the network structure of the initial residual network model is a vector convolution Conv1-batch standardized Batchnormal algorithm-linear rectification ReLU1 activation function and a Conv2-Batch Normalization algorithm-ReLU 2 activation function; the convolution layer of the target residual network model adopts 3*3 convolution kernel, the network structure of the target residual network model is Conv1-Batch Normalization algorithm-ReLU 1 activation function, conv2-Batch Normalization algorithm-ReLU 2 activation function, conv3-Batch Normalization algorithm-ReLU 3 activation function is added at the quick connection position in the target residual network model, the output of the upper layer in the target residual network model is used as the input of the convolution layer at the quick connection position, and the first output of Conv1, the second output of Conv2 and the third output of Conv3 are used as the input of the next training stage.
2. The method of claim 1, wherein acquiring an ion membrane image set comprises:
acquiring a historical ionic membrane image and a current ionic membrane image, wherein the historical ionic membrane image is an ionic membrane image in a historical detection image library, and the current ionic membrane image is an ionic membrane image obtained by shooting an ionic membrane in hydrogen production equipment currently;
And obtaining the ionic membrane image set based on the historical ionic membrane image and the current ionic membrane image.
3. The method of claim 1, wherein prior to labeling the ion membrane image set based on the fault category to obtain a labeled dataset, the method further comprises:
scaling all the ionic membrane images in the ionic membrane image set to obtain all the adjusted ionic membrane images;
adding a category label to all the adjusted ionic membrane images, wherein the category label is used for indicating the fault category, and the fault category comprises at least one of the following: deformation, peeling, cracking, heave, corrosion, deposition, fouling, penetration of foreign bodies.
4. The method of claim 1, wherein prior to labeling the ion membrane image set based on the fault category to obtain a labeled dataset, the method further comprises:
carrying out data preprocessing on all the ionic membrane images in the ionic membrane image set to obtain preprocessed ionic membrane images; the data preprocessing is used for processing missing values and abnormal values in the ionic membrane image;
And extracting an image characteristic value from the preprocessed ion membrane image to obtain an image characteristic vector.
5. The method of claim 1, wherein after labeling the ion membrane image set based on the fault category to obtain a labeled data set, the method further comprises:
acquiring the data processing requirement of the target residual error network model;
based on the data processing requirements, taking a first proportion of the marked data set as the training set, a second proportion of the marked data set as the verification set, and a third proportion of the marked data set as the test set, wherein the first proportion is greater than the second proportion and the third proportion.
6. The method of claim 1, wherein the number of convolution kernels in Conv1-10 convolution layers, the number of convolution kernels in Conv11-22 convolution layers, the number of convolution kernels in Conv23-34 convolution layers, and the number of convolution kernels in Conv35-40 convolution layers in the target residual network model are 64, 128, 256, and 512 respectively.
7. The method of claim 1, wherein training a target residual network model using the annotation data set, the obtaining an ion membrane fault detection model comprising:
Inputting all image data in the labeling data set to the target residual error network model for learning to obtain a transfer learning model;
the training set is imported into the transfer learning model for training so as to update model parameters of the transfer learning model and obtain a training model;
the verification set is imported into the training model to be trained so as to adjust model parameters of the training model, and a verification model is obtained;
and importing the test set into the verification model for testing to obtain the ion membrane fault detection model.
8. The method of claim 7, wherein detecting whether the ion membrane of the hydrogen plant is malfunctioning using the ion membrane malfunction detection model to obtain a malfunction detection result comprises:
and detecting the test set by adopting the ion membrane fault detection model to determine whether the ion membrane of the hydrogen production equipment has faults or not, so as to obtain the fault detection result.
9. An ion membrane failure detection apparatus, wherein the ion membrane failure detection apparatus performs the ion membrane failure detection method according to any one of claims 1 to 8, comprising:
The system comprises an acquisition module, a detection module and a control module, wherein the acquisition module is used for acquiring an ion membrane image set, wherein the ion membrane image set contains a plurality of ion membranes with different fault categories;
the labeling module is used for labeling the ion membrane image set based on the fault category to obtain a labeling data set, wherein the labeling data set comprises: training set, validation set and test set;
the training module is used for training the target residual error network model by adopting the marking data set to obtain an ion membrane fault detection model, wherein the target residual error network model is a neural network model obtained by natural image feature migration;
and the detection module is used for detecting whether the ion membrane of the hydrogen production equipment fails or not by adopting the ion membrane failure detection model to obtain a failure detection result.
10. A non-volatile storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the fault detection method of an ionic membrane of any one of claims 1 to 8.
11. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, the processor being arranged to run the computer program to perform the method of fault detection of an ionic membrane according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111163342.1A CN113945569B (en) | 2021-09-30 | 2021-09-30 | Fault detection method and device for ion membrane |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111163342.1A CN113945569B (en) | 2021-09-30 | 2021-09-30 | Fault detection method and device for ion membrane |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113945569A CN113945569A (en) | 2022-01-18 |
| CN113945569B true CN113945569B (en) | 2023-12-26 |
Family
ID=79329764
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111163342.1A Active CN113945569B (en) | 2021-09-30 | 2021-09-30 | Fault detection method and device for ion membrane |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113945569B (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108388890A (en) * | 2018-03-26 | 2018-08-10 | 南京邮电大学 | A kind of neonatal pain degree assessment method and system based on human facial expression recognition |
| CN108564123A (en) * | 2018-04-10 | 2018-09-21 | 复旦大学附属肿瘤医院 | A kind of Thyroid Neoplasms smear image classification method and its device |
| CN109299705A (en) * | 2018-10-24 | 2019-02-01 | 电子科技大学 | Rotary machinery fault diagnosis method based on one-dimensional depth residual error convolutional neural networks |
| CN109840593A (en) * | 2019-01-28 | 2019-06-04 | 华中科技大学鄂州工业技术研究院 | Diagnose the method and apparatus of solid oxide fuel battery system failure |
| CN110031226A (en) * | 2019-04-12 | 2019-07-19 | 佛山科学技术学院 | A kind of diagnostic method and device of bearing fault |
| CN110188720A (en) * | 2019-06-05 | 2019-08-30 | 上海云绅智能科技有限公司 | A kind of object detection method and system based on convolutional neural networks |
| CN110263692A (en) * | 2019-06-13 | 2019-09-20 | 北京数智源科技有限公司 | Container switch gate state identification method under large scene |
| CN110503154A (en) * | 2019-08-27 | 2019-11-26 | 携程计算机技术（上海）有限公司 | Method, system, electronic equipment and the storage medium of image classification |
| WO2019232830A1 (en) * | 2018-06-06 | 2019-12-12 | 平安科技（深圳）有限公司 | Method and device for detecting foreign object debris at airport, computer apparatus, and storage medium |
| CN110728654A (en) * | 2019-09-06 | 2020-01-24 | 台州学院 | Automatic pipeline detection and classification method based on deep residual error neural network |
| CN111310862A (en) * | 2020-03-27 | 2020-06-19 | 西安电子科技大学 | Deep neural network license plate positioning method based on image enhancement in complex environment |
| CN112418328A (en) * | 2020-11-25 | 2021-02-26 | 哈尔滨市科佳通用机电股份有限公司 | Deep learning-based drain outlet cover plate non-closing in-place fault detection method |
| CN113095420A (en) * | 2021-04-20 | 2021-07-09 | 池州学院 | Insulator fault detection method based on improved YOLOv3 |
-
2021
- 2021-09-30 CN CN202111163342.1A patent/CN113945569B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108388890A (en) * | 2018-03-26 | 2018-08-10 | 南京邮电大学 | A kind of neonatal pain degree assessment method and system based on human facial expression recognition |
| CN108564123A (en) * | 2018-04-10 | 2018-09-21 | 复旦大学附属肿瘤医院 | A kind of Thyroid Neoplasms smear image classification method and its device |
| WO2019232830A1 (en) * | 2018-06-06 | 2019-12-12 | 平安科技（深圳）有限公司 | Method and device for detecting foreign object debris at airport, computer apparatus, and storage medium |
| CN109299705A (en) * | 2018-10-24 | 2019-02-01 | 电子科技大学 | Rotary machinery fault diagnosis method based on one-dimensional depth residual error convolutional neural networks |
| CN109840593A (en) * | 2019-01-28 | 2019-06-04 | 华中科技大学鄂州工业技术研究院 | Diagnose the method and apparatus of solid oxide fuel battery system failure |
| CN110031226A (en) * | 2019-04-12 | 2019-07-19 | 佛山科学技术学院 | A kind of diagnostic method and device of bearing fault |
| CN110188720A (en) * | 2019-06-05 | 2019-08-30 | 上海云绅智能科技有限公司 | A kind of object detection method and system based on convolutional neural networks |
| CN110263692A (en) * | 2019-06-13 | 2019-09-20 | 北京数智源科技有限公司 | Container switch gate state identification method under large scene |
| CN110503154A (en) * | 2019-08-27 | 2019-11-26 | 携程计算机技术（上海）有限公司 | Method, system, electronic equipment and the storage medium of image classification |
| CN110728654A (en) * | 2019-09-06 | 2020-01-24 | 台州学院 | Automatic pipeline detection and classification method based on deep residual error neural network |
| CN111310862A (en) * | 2020-03-27 | 2020-06-19 | 西安电子科技大学 | Deep neural network license plate positioning method based on image enhancement in complex environment |
| CN112418328A (en) * | 2020-11-25 | 2021-02-26 | 哈尔滨市科佳通用机电股份有限公司 | Deep learning-based drain outlet cover plate non-closing in-place fault detection method |
| CN113095420A (en) * | 2021-04-20 | 2021-07-09 | 池州学院 | Insulator fault detection method based on improved YOLOv3 |
Non-Patent Citations (6)
| Title |
|---|
| 制氢工艺水碳比神经网络模型研究;张涛,胡宁,胡友谱,李为民;石油与天然气化工(06);全文 * |
| 在线优化在离子膜氯碱生产中的应用;董浩,钱积新,武培筠;化工学报(第01期);第42-43页 * |
| 基于数据融和的制氢设备故障诊断和监控系统;成回中;;实验室研究与探索(12);全文 * |
| 基于迁移学习与多标签平滑策略的图像自动标注;汪鹏;张奥帆;王利琴;董永峰;;计算机应用(11);全文 * |
| 基于迁移学习的滚动轴承在线故障诊断;毛冠通;洪流;王景霖;;航空科学技术(01);全文 * |
| 采用改进Faster R-CNN的遥感图像目标检测方法;李玉峰;顾曼璇;赵亮;;信号处理(08);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113945569A (en) | 2022-01-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110555474B (en) | Photovoltaic panel fault detection method based on semi-supervised learning | |
| CN109710636B (en) | Unsupervised industrial system anomaly detection method based on deep transfer learning | |
| CN109740617A (en) | A kind of image detecting method and device | |
| CN110569814B (en) | Video category identification method, device, computer equipment and computer storage medium | |
| CN111046550B (en) | Fuel cell voltage prediction system based on space-time characteristics | |
| CN109271957B (en) | Face gender identification method and device | |
| CN112148997B (en) | Training method and device for multi-modal countermeasure model for disaster event detection | |
| CN116760583B (en) | Enhanced graph node behavior characterization and abnormal graph node detection method | |
| CN109145743A (en) | A kind of image-recognizing method and device based on deep learning | |
| CN114330499A (en) | Method, device, equipment, storage medium and program product for training classification model | |
| CN115861210B (en) | Transformer substation equipment abnormality detection method and system based on twin network | |
| CN114612414A (en) | Image processing method, model training method, device, equipment and storage medium | |
| CN114677535A (en) | Training method of domain-adaptive image classification network, image classification method and device | |
| CN115240259A (en) | Face detection method and face detection system based on YOLO deep network in classroom environment | |
| CN109101984B (en) | Image identification method and device based on convolutional neural network | |
| CN114419406A (en) | Image change detection method, training method, device and computer equipment | |
| CN115239672A (en) | Defect detection method and device, equipment and storage medium | |
| CN113945569B (en) | Fault detection method and device for ion membrane | |
| CN117874511A (en) | Deep learning model training method, battery fault detection method, device and equipment | |
| CN113327238A (en) | Gastroscope image classification model construction method and gastroscope image classification method | |
| CN113379685A (en) | PCB defect detection method and device based on dual-channel feature comparison model | |
| CN116719060A (en) | Information fusion-based method for detecting tight combination navigation faults of deep learning network | |
| CN115797291A (en) | Circuit terminal identification method and device, computer equipment and storage medium | |
| CN115690658A (en) | Semi-supervised video abnormal behavior detection method fusing prior knowledge | |
| CN116165478A (en) | Small-current single-phase grounding fault line selection method and device under small sample condition |
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 |