CN113344919B - Method and system for detecting ceramic thermal shock damage degree based on convolutional neural network - Google Patents
Method and system for detecting ceramic thermal shock damage degree based on convolutional neural network Download PDFInfo
- Publication number
- CN113344919B CN113344919B CN202110879361.8A CN202110879361A CN113344919B CN 113344919 B CN113344919 B CN 113344919B CN 202110879361 A CN202110879361 A CN 202110879361A CN 113344919 B CN113344919 B CN 113344919B
- Authority
- CN
- China
- Prior art keywords
- training
- thermal shock
- data
- neural network
- class
- 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
- 238000000034 method Methods 0.000 title claims abstract description 58
- 230000035939 shock Effects 0.000 title claims abstract description 48
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 37
- 239000000919 ceramic Substances 0.000 title claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 65
- 229910010293 ceramic material Inorganic materials 0.000 claims abstract description 26
- 238000012795 verification Methods 0.000 claims abstract description 20
- 230000008569 process Effects 0.000 claims abstract description 17
- 230000008859 change Effects 0.000 claims abstract description 13
- 238000011478 gradient descent method Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 238000011176 pooling Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 8
- 238000013145 classification model Methods 0.000 claims description 6
- 238000005520 cutting process Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 238000004519 manufacturing process Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000003062 neural network model Methods 0.000 claims description 2
- 238000012854 evaluation process Methods 0.000 claims 1
- 238000012512 characterization method Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 10
- 238000001514 detection method Methods 0.000 description 7
- 230000006872 improvement Effects 0.000 description 7
- 239000000463 material Substances 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000011156 evaluation Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 238000001000 micrograph Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000036632 reaction speed Effects 0.000 description 1
- 230000002829 reductive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000009991 scouring Methods 0.000 description 1
- 230000035882 stress Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000008646 thermal stress Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/60—Rotation of whole images or parts thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for detecting ceramic thermal shock damage degree based on a convolutional neural network, which comprises the following steps: acquiring thermal shock experimental strength data of the ceramic material and corresponding crack pictures, and classifying and marking the crack pictures according to the change rule of the strength values; preprocessing the marked crack picture and then dividing the preprocessed crack picture into a training set and a verification set; inputting the training set and the verification set into a convolutional neural network model which is built in advance for training; training by using a random gradient descent method in a training process to obtain a classification prediction model; and inputting the crack picture to be detected into a classification prediction model, and outputting a prediction classification result. The invention provides a thermal shock damage characterization method, which optimizes the model training process, improves the classification accuracy and efficiency of the thermal shock damage degree of ceramic materials, and reduces the labor cost.
Description
Technical Field
The invention relates to the field of computer detection, in particular to a method and a system for detecting the thermal shock damage degree of ceramic based on a convolutional neural network.
Background
Ceramic materials are often used as high-temperature structural materials, and in an extremely high-temperature environment, the ceramic materials need to bear complex loads and environments such as force and stress circulation, environmental medium erosion and scouring, thermal shock and the like. Due to the inherent brittleness of the ceramic, in the process of rapid heating or thermal shock caused by high-temperature rapid cooling, the thermal stress caused by temperature difference can cause the microcracks of the material to be initiated and rapidly expanded, so that the strength of the material is suddenly attenuated, the reliability of a key thermal protection structure is seriously influenced, and the practical application of the ceramic material as a high-temperature material is restricted.
The existence and the expansion of cracks are main reasons for causing the fracture of the brittle material, so the research on the influence rule of the expansion of the microcracks on the thermal shock resistance of the ceramic material and the exploration of a method for inhibiting the expansion of the microcracks are key problems which need to be solved urgently for improving the thermal shock resistance of the ceramic.
There are many methods for characterizing crack propagation, and the analysis of the shot images of the material is most direct, including macrocrack images, scanning electron microscope images, and the like. However, the statistics of the appearance characteristics of the cracks requires a large amount of manpower, the time cost is high, the defects of difficult distinguishing by naked eyes, low recognition rate and poor generalization capability of manually extracting the characteristics are faced with complex crack images.
The method has the advantages of simple neural network identification, low classification accuracy, low reaction speed, high cost and incapability of solving the learning problem of small samples because a large amount of experimental sample data is needed in the process of training the model. Meanwhile, aiming at the model training process, the convergence rate of the random gradient decline is very low, the coverage pooling is not easy to fit, and the method is difficult to adapt to crack pictures with complex characteristics. And the shallow network structure is difficult to accurately represent the thermal shock damage degree and the crack appearance mapping relation.
Disclosure of Invention
The invention aims to provide a method and a system for detecting the degree of thermal shock damage of ceramic based on a convolutional neural network, provides a method for representing the thermal shock damage, optimizes a model training process, improves the classification accuracy and the classification efficiency of the degree of thermal shock damage of ceramic materials, and reduces the labor cost.
In order to solve the technical problem, the invention provides a method for detecting the degree of thermal shock damage of ceramic based on a convolutional neural network, which comprises the following steps:
s1: acquiring thermal shock experimental strength data of the ceramic material and corresponding crack pictures, and classifying and marking the crack pictures according to the change rule of the strength values;
s2: preprocessing the marked crack picture and then dividing the preprocessed crack picture into a training set and a verification set;
s3: inputting the training set and the verification set into a convolutional neural network model which is built in advance for training;
s4: training by using a random gradient descent method in a training process to obtain a classification prediction model;
s5: and inputting the crack picture to be detected into a classification prediction model, and outputting a prediction classification result.
As a further improvement of the present invention, in step S1, the crack pictures are classified and marked according to the intensity value change rule, specifically, according to the graph of the intensity value change rule after the ceramic material is subjected to thermal shock, the crack pictures corresponding to different intensities are divided into a plurality of intervals according to the intensity attenuation condition, data labels are respectively and correspondingly produced, and the picture file names and the label names are mapped one by one.
As a further improvement of the present invention, the step S2 specifically includes the following steps:
s21: cutting and binarizing the marked crack picture to obtain standardized image data;
s22: the data size of the image data is increased by adopting zooming, rotating and random cutting, and data enhancement is carried out;
s23: and dividing the image data after data enhancement and the corresponding label into a training set and a verification set, and manufacturing a corresponding data reader, wherein the data reader reads the picture and the corresponding classification name.
As a further improvement of the present invention, the building of the convolutional neural network model in step S3 includes the steps of: based on the AlexNet convolutional neural network, 5 layers of convolutional layers and 3 layers of pooling layers are used, and finally 3 layers of full-connection layers are used, wherein the pooling layers adopt a maximum pooling method.
As a further improvement of the method, when the convolution neural network model performs convolution processing on the input image, a ReLU activation function is used to obtain the pooling layer.
As a further improvement of the present invention, the step S4 specifically includes the following steps:
s41: placing data in a multidimensional vector by using a data reader, wherein the multidimensional vector comprises batches of read-in pictures, the number of each batch and the size parameter of the pictures;
s42: inputting the multidimensional vector into a built convolutional neural network model, and adjusting parameters of each layer of the convolutional neural network model;
s43: training a model by using a random gradient descent method, and selecting a proper learning rate and training times;
s44: according to the comparison between the verification set and the result in the model training process, obtaining a training error;
s45: and according to the training error, adjusting the learning rate and the training times after each training is finished to obtain an accurate prediction classification model.
As a further improvement of the present invention, after the classification prediction model is completed in step S4, a plurality of test sets are established and input into the prediction classification model, and the accuracy, precision and recall rate are used to evaluate the performance of the prediction classification model.
As a further improvement of the present invention, the accuracy is that all samples with correct prediction account for the proportion of all statistical samples; the accuracy rate is the proportion of the actual positive sample to all the predicted positive samples; the recall ratio is the proportion of positive samples predicted to account for actual positive samples.
Ceramic thermal shock damage degree detection system based on convolutional neural network includes:
the classification marking module is used for acquiring thermal shock experiment intensity data of the ceramic material and corresponding crack pictures and classifying and marking the crack pictures according to the intensity value change rule;
the picture processing module is used for dividing the marked crack picture into a training set and a verification set after preprocessing;
the model training module is used for building a convolutional neural network model and inputting a training set and a verification set into the model for training;
the model establishing module is used for training by using a random gradient descent method in the training process to obtain a classification prediction model;
and the detection output module is used for inputting the crack picture to be detected into the classification prediction model and outputting a prediction classification result.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the above method are implemented when the processor executes the computer program.
The invention has the beneficial effects that: the method is based on convolutional neural network model classification identification, and compared with manual classification visual identification, the method has the advantages that the identification rate is higher and the detection speed is higher in the face of complex crack images; the model is simple in overall structure and easy to realize; on the detection of the thermal shock damage degree of the ceramic, the method helps to establish a correlation model of the crack characteristics and the damage degree of the ceramic material, and provides a new idea for improving the thermal shock resistance of the ceramic material.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a schematic flow diagram of a method of practicing the present invention;
FIG. 3 is a schematic diagram of the convolutional neural network structure of the present invention;
FIG. 4 is a schematic flow chart of the training process of the present invention;
FIG. 5 is a schematic view of an original crack to be detected according to the present invention;
labeled as: 1. a first type of crack picture; 2. a second type of crack picture; 3. a third type of crack picture; 4. and a fourth type of crack picture.
Detailed Description
The present invention is further described below in conjunction with the following figures and specific examples so that those skilled in the art may better understand the present invention and practice it, but the examples are not intended to limit the present invention.
Example one
As shown in fig. 1, an embodiment of the present invention provides a method for detecting a degree of thermal shock damage of a ceramic based on a convolutional neural network, including the following steps:
s1: acquiring thermal shock experimental strength data of the ceramic material and corresponding crack pictures, and classifying and marking the crack pictures according to the change rule of the strength values;
s2: preprocessing the marked crack picture and then dividing the preprocessed crack picture into a training set and a verification set;
s3: inputting the training set and the verification set into a convolutional neural network model which is built in advance for training;
s4: training by using a random gradient descent method in a training process to obtain a classification prediction model;
s5: and inputting the crack picture to be detected into a classification prediction model, and outputting a prediction classification result.
Specifically, as shown in fig. 2, in the first step, thermal shock experimental strength data of the ceramic material and a corresponding ceramic material crack picture are obtained. The method specifically comprises the following steps: performing a thermal shock experiment, acquiring experimental data, and recording the strength value of the material;
and secondly, manufacturing data labels according to the experimental data and the pictures, and classifying the data labels into four categories according to the damage degree. The method specifically comprises the following steps: dividing the crack pictures corresponding to different intensities into four intervals according to the intensity attenuation condition according to the intensity value change rule curve graph after the ceramic material is subjected to thermal shock, wherein the four intervals correspond to labels 0, 1, 2 and 3 respectively, and the label manufacturing adopts a mode of mapping picture file names and label names one by one;
and thirdly, preprocessing the crack picture with the manufactured label, performing binarization processing on the image by preprocessing expansion, adjusting the size of the image, and adopting a data enhancement method when the experimental data is less, and finally dividing the processed picture into a training set and a verification set, wherein the training set is used for training the model, and the verification set is used for evaluating the generalization capability of the model. The method specifically comprises the following steps: obtaining standardized image data by a cutting and binarization method; the data volume is increased by adopting a data enhancement method of zooming, rotating and cutting immediately; dividing the picture data and the corresponding label into two parts, and manufacturing a corresponding data reader which can read the picture and the corresponding classification name;
and fourthly, building a convolutional neural network model. The method specifically comprises the following steps: as shown in fig. 3, based on an AlexNet convolutional neural network, 5 convolutional layers, 3 pooling layers and finally 3 full-connected layers are used, the pooling layers adopt a maximum pooling method, and when a convolutional neural network model performs convolutional processing on an input image, a ReLU activation function is used to obtain the pooling layers, wherein the ReLU is used as an activation function of CNN, so that the problem of gradient dispersion can be effectively avoided, the maximum pooling method can learn the edge and texture structure of the image, and is very suitable for processing a crack image;
and fifthly, inputting the processed training set and the verification set into a model for training. Referring to fig. 4, specifically, the method includes:
step S41, using the data reader, putting the data in a multi-dimensional vector (S) ((S))n,k,h,w) In (1),nis the batch of pictures that are read in,kis the number of each batch that is to be processed,handwis the size parameter of the picture;
step S42, inputting the vector into the built neural network, and adjusting each layer of parameters of the neural network;
step S43, training the model by using a stochastic gradient descent method, and selecting a proper learning rate and training times;
step S44, obtaining a training error according to the comparison of the result of the verification set and the result of the model training process;
step S45, according to the training error, adjusting the learning rate and the training times after each training, thereby improving the identification accuracy and obtaining an accurate model;
and sixthly, model prediction, namely inputting new crack picture data into the trained prediction model, outputting a prediction result, and evaluating the performance of the algorithm model based on classification algorithm evaluation indexes such as accuracy, precision (the probability of actually being a positive sample in all samples predicted to be positive), recall (the probability of being predicted to be a positive sample in the samples actually being positive) and the like.
Example two
The embodiment of the invention provides a method for detecting the degree of thermal shock damage of ceramic based on a convolutional neural network, which is used for evaluating an algorithm of the invention according to the following algorithm evaluation indexes on the basis of the first embodiment, and specifically comprises the following steps:
a. establishing a plurality of different test sets, inputting the thermal shock damage picture of the ceramic material into a model, and obtaining a classification prediction result;
b. for each classification of the thermal shock damage degree of the ceramic material, several data indexes are counted, for example, if one data is the first type and is predicted to be the first type, i.e. the real type, in the thermal shock damage degree (more than 75%) of the ceramic material of the first typeTP(True Positive) if a data is of the first class, but predicted to be of the other class, i.e. false negative classFN(False Negative) if a data is of another class, but is predicted to be of the first class, i.e. False positive classFP(False Positive), if an instance is other class, and is predicted to be other class, it is true negative classTN(True Negative）；
c. The accuracy rate is the proportion of all samples with correct prediction to all statistical samples, and the prediction accuracy rate is definedAccComprises the following steps:
d. accuracy rate is the probability of actually being the first class among all samples predicted to be the first class, and defines the prediction accuracy ratePreComprises the following steps:
e. recall is when actually of the first typeThe probability of being predicted as the first class in the sample of (2), defining the prediction accuracyRecallComprises the following steps:
f. the other three types except the first type adopt the same steps as the first type to count the relevant data.
Specifically, statistics is performed according to the above algorithm evaluation indexes, the number of pictures of four statistics amounts in the prediction result corresponding to each class is counted in table 1, the number of crack samples of each class in the test set is 500, and the four statistics amounts are true classes respectivelyTP(True Positive), false negative classFN(False Negative), False positive classFP(False Positive), true negative classTN(True Negative）：
TABLE 1
From Table 1, it can be seen thatTPIs greater thanFNThe number of the (c) component(s),TNis greater thanFPThe identification accuracy is high, then the performance of the algorithm model of the invention is evaluated according to a calculation formula in the specification, and as shown in table 2, the statistics of the algorithm evaluation indexes of the test set are as follows:
TABLE 2
The prediction accuracy of each type is more than 89%, the accuracy is more than 78%, which shows that the recognition rate of the model for detecting the thermal shock damage degree of the ceramic materials with different morphological characteristics is high, the recall rate of each type is more than 78%, which shows that the recognition error rate of the model is low, and the probability of misrecognition into other classifications is low. The third type of recognition result is the best, the prediction accuracy rate reaches 92.1%, the accuracy rate reaches 86.4%, and the recall rate is 82.6%. The reason is that the crack characteristics of the third type are obviously distinguished from other types, as shown in fig. 5, a first type crack picture 1, a second type crack picture 2, a third type crack picture 3 and a fourth type crack picture 4 are sequentially arranged from top to bottom, the appearance of the crack is gradually complicated, the thermal shock damage degree is gradually increased and the strength is gradually reduced according to literature and experimental test results, wherein the characteristics of the third type crack picture 3 are more obvious than those of the other types. The fourth type of crack picture 4 has a lower recognition effect than the first types of crack pictures because when the thermal shock temperature difference exceeds 500 ℃, the complexity of the crack is increased to a limited extent, the intensity change is not obvious, the feature distinction from the third type of crack picture 3 is not obvious, the number of wrong judgments for the third type is large, and the recognition rate is slightly low. Except for the fourth type, the damage degree identification accuracy of the other three types is over 90 percent, and the accuracy is over 80 percent.
EXAMPLE III
Based on the same inventive concept, the present embodiment provides a system for detecting ceramic thermal shock damage degree based on a convolutional neural network, and the principle of solving the problem is similar to the method for detecting ceramic thermal shock damage degree based on a convolutional neural network, and repeated parts are not described again.
Ceramic thermal shock damage degree detection system based on convolutional neural network includes:
the classification marking module is used for acquiring thermal shock experiment intensity data of the ceramic material and corresponding crack pictures and classifying and marking the crack pictures according to the intensity value change rule;
the picture processing module is used for dividing the marked crack picture into a training set and a verification set after preprocessing;
the model training module is used for building a convolutional neural network model and inputting a training set and a verification set into the model for training;
the model establishing module is used for training by using a random gradient descent method in the training process to obtain a classification prediction model;
and the detection output module is used for inputting the crack picture to be detected into the classification prediction model and outputting a prediction classification result.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above-mentioned embodiments are merely preferred embodiments for fully illustrating the present invention, and the scope of the present invention is not limited thereto. The equivalent substitution or change made by the technical personnel in the technical field on the basis of the invention is all within the protection scope of the invention. The protection scope of the invention is subject to the claims.
Claims (6)
1. The method for detecting the degree of ceramic thermal shock damage based on the convolutional neural network is characterized by comprising the following steps: the method comprises the following steps:
s1: acquiring thermal shock experimental intensity data of the ceramic material and corresponding crack pictures, and classifying and marking the crack pictures according to the change rule of the intensity data;
s2: preprocessing the marked crack picture and then dividing the preprocessed crack picture into a training set and a verification set;
s3: inputting the training set and the verification set into a convolutional neural network model which is built in advance for training;
s4: training by using a random gradient descent method in a training process to obtain a classification prediction model;
s5: inputting a crack picture to be detected into a classification prediction model, and outputting a prediction classification result;
the step S1 of classifying and marking the crack pictures according to the intensity data change rule specifically includes dividing the crack pictures corresponding to different intensities into four intervals according to the intensity attenuation condition according to the intensity data change rule curve graph after the ceramic material is subjected to thermal shock, and respectively and correspondingly making data labels, wherein the picture file names and the label names are mapped one by one;
after the classification prediction model is completed in step S4, establishing a plurality of test sets and inputting the test sets into the prediction classification model, and evaluating the performance of the prediction classification model by using the accuracy, precision and recall;
the accuracy is the proportion of all samples with correct prediction in all statistical samples; the accuracy rate is the proportion of the actual positive sample to all the predicted positive samples; the recall ratio is a proportion of positive samples predicted to account for actual positive samples;
the specific evaluation process comprises the following steps:
a. establishing a plurality of different test sets, inputting the thermal shock damage picture of the ceramic material into a model, and obtaining a classification prediction result;
b. for each class of ceramic material thermal shock damage, several data indicators are counted, wherein in the first class of ceramic material thermal shock damage, if a data is itself of the first class and is predicted to be of the first classThe first category, i.e. true categoryTPIf a data is of the first class, but predicted to be of another class, it is a false negative classFNIf a data is of another class, but is predicted to be of the first class, it is a false positive classFPIf a data is of other class and is predicted to be of other class, it is true negative classTN；
c. The accuracy rate is defined as the proportion of all samples with correct prediction to all statistical samplesAccComprises the following steps:
d. the accuracy rate is the probability of being actually the first class among all samples predicted to be the first class, and defines the accuracy ratePreComprises the following steps:
e. recall is the probability of being predicted as class one in a sample that is actually class one, defining the recallRecallComprises the following steps:
f. and the other three types except the first type all adopt the same step operation as the first type, and the number of the pictures of the four statistics values in the prediction result corresponding to each type is counted.
2. The method for detecting the degree of thermal shock damage of ceramic based on the convolutional neural network as claimed in claim 1, wherein: the step S2 specifically includes the following steps:
s21: cutting and binarizing the marked crack picture to obtain standardized image data;
s22: the data size of the image data is increased by adopting zooming, rotating and random cutting, and data enhancement is carried out;
s23: and dividing the image data after data enhancement and the corresponding label into a training set and a verification set, and manufacturing a corresponding data reader, wherein the data reader reads the picture and the corresponding classification name.
3. The method for detecting the degree of thermal shock damage of ceramic based on the convolutional neural network as claimed in claim 1, wherein: the building of the convolutional neural network model in the step S3 includes the steps of: based on the AlexNet convolutional neural network, 5 layers of convolutional layers and 3 layers of pooling layers are used, and finally 3 layers of full-connection layers are used, wherein the pooling layers adopt a maximum pooling method.
4. The method for detecting the degree of thermal shock damage of ceramic based on the convolutional neural network as claimed in claim 3, wherein: and when the convolution neural network model performs convolution processing on the input image, a ReLU activation function is used to obtain a pooling layer.
5. The method for detecting the degree of thermal shock damage of ceramic based on the convolutional neural network as claimed in claim 1, wherein: the step S4 specifically includes the following steps:
s41: placing data in a multidimensional vector by using a data reader, wherein the multidimensional vector comprises batches of read-in pictures, the number of each batch and the size parameter of the pictures;
s42: inputting the multidimensional vector into a built convolutional neural network model, and adjusting parameters of each layer of the convolutional neural network model;
s43: training a model by using a random gradient descent method, and selecting a proper learning rate and training times;
s44: according to the comparison between the verification set and the result in the model training process, obtaining a training error;
s45: and according to the training error, adjusting the learning rate and the training times after each training is finished to obtain an accurate prediction classification model.
6. A computer arrangement comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method according to any of claims 1-5 are implemented by the processor when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110879361.8A CN113344919B (en) | 2021-08-02 | 2021-08-02 | Method and system for detecting ceramic thermal shock damage degree based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110879361.8A CN113344919B (en) | 2021-08-02 | 2021-08-02 | Method and system for detecting ceramic thermal shock damage degree based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113344919A CN113344919A (en) | 2021-09-03 |
| CN113344919B true CN113344919B (en) | 2021-11-23 |
Family
ID=77480756
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110879361.8A Active CN113344919B (en) | 2021-08-02 | 2021-08-02 | Method and system for detecting ceramic thermal shock damage degree based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113344919B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458231A (en) * | 2019-08-12 | 2019-11-15 | 广东工业大学 | A kind of ceramic product detection method, device and equipment |
| CN112258495A (en) * | 2020-11-02 | 2021-01-22 | 郑州大学 | Building wood crack identification method based on convolutional neural network |
-
2021
- 2021-08-02 CN CN202110879361.8A patent/CN113344919B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458231A (en) * | 2019-08-12 | 2019-11-15 | 广东工业大学 | A kind of ceramic product detection method, device and equipment |
| CN112258495A (en) * | 2020-11-02 | 2021-01-22 | 郑州大学 | Building wood crack identification method based on convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| 裂纹分析特征对脆性材料热冲击后统计剩余强度的影响;孟松鹤等;《2018年全国固体体力学学术会议》;20181125;第1页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113344919A (en) | 2021-09-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111179251B (en) | Defect detection system and method based on twin neural network and by utilizing template comparison | |
| CN111814704B (en) | Full convolution examination room target detection method based on cascade attention and point supervision mechanism | |
| CN109509187B (en) | Efficient inspection algorithm for small defects in large-resolution cloth images | |
| TWI585397B (en) | Optimization of unknown defect rejection for automatic defect classification | |
| CN110610475B (en) | Visual defect detection method of deep convolutional neural network | |
| CN107967475A (en) | A kind of method for recognizing verification code based on window sliding and convolutional neural networks | |
| KR20170091716A (en) | Automatic defect classification without sampling and feature selection | |
| CN112418130A (en) | Banana maturity detection method and device based on BP neural network | |
| CN113128335B (en) | Method, system and application for detecting, classifying and finding micro-living ancient fossil image | |
| TW200818060A (en) | Clustering system, and defect kind judging device | |
| CN113222913B (en) | Circuit board defect detection positioning method, device and storage medium | |
| CN106290392A (en) | A kind of little micro-bearing surface pitting defects online test method and system thereof | |
| US6336086B1 (en) | Method and system for analyzing wafer processing order | |
| CN104978569B (en) | A kind of increment face identification method based on rarefaction representation | |
| CN114119460B (en) | Semiconductor image defect recognition method, device, computer equipment and storage medium | |
| CN114155474A (en) | Damage identification technology based on video semantic segmentation algorithm | |
| CN114399763A (en) | Single-sample and small-sample micro-body ancient biogenetic fossil image identification method and system | |
| CN113658108A (en) | Glass defect detection method based on deep learning | |
| CN118037655A (en) | SSD-based photovoltaic panel flaw detection method and device | |
| CN113344919B (en) | Method and system for detecting ceramic thermal shock damage degree based on convolutional neural network | |
| CN114065798A (en) | Visual identification method and device based on machine identification | |
| CN112085727A (en) | Intelligent identification method for scale structure on surface of hot rolled steel | |
| Bairwa et al. | Classification of Fruits Based on Shape, Color and Texture using Image Processing Techniques | |
| Rahiddin et al. | Local texture representation for timber defect recognition based on variation of LBP | |
| CN114185785A (en) | Natural language processing model test case reduction method for deep neural network |
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 |