CN110737021A - Fault recognition method and model training method and device thereof - Google Patents
Fault recognition method and model training method and device thereof Download PDFInfo
- Publication number
- CN110737021A CN110737021A CN201911078637.1A CN201911078637A CN110737021A CN 110737021 A CN110737021 A CN 110737021A CN 201911078637 A CN201911078637 A CN 201911078637A CN 110737021 A CN110737021 A CN 110737021A
- Authority
- CN
- China
- Prior art keywords
- fault
- model
- recognition
- attribute data
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 87
- 238000000034 method Methods 0.000 title claims abstract description 78
- 239000002245 particle Substances 0.000 claims abstract description 38
- 238000012706 support-vector machine Methods 0.000 claims abstract description 35
- 238000005457 optimization Methods 0.000 claims abstract description 31
- 238000004364 calculation method Methods 0.000 claims abstract description 28
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 24
- 238000012545 processing Methods 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 238000013139 quantization Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 15
- 238000003745 diagnosis Methods 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 20
- 230000008569 process Effects 0.000 description 13
- 230000015654 memory Effects 0.000 description 12
- 238000004891 communication Methods 0.000 description 7
- 238000009826 distribution Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000007621 cluster analysis Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 239000011435 rock Substances 0.000 description 3
- 238000005452 bending Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000011295 pitch Substances 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000007596 consolidation process Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000012854 evaluation process Methods 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/30—Analysis
- G01V1/307—Analysis for determining seismic attributes, e.g. amplitude, instantaneous phase or frequency, reflection strength or polarity
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/30—Analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/63—Seismic attributes, e.g. amplitude, polarity, instant phase
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/64—Geostructures, e.g. in 3D data cubes
- G01V2210/642—Faults
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
The invention provides fault recognition methods and a model training method and device thereof, and relates to the field of fault diagnosis, wherein the model training method comprises the steps of obtaining fault attribute data of a fault area, generating a sample data set according to the fault attribute data, inputting the sample data set into a preset support vector machine model for training, calculating the support vector machine model through a particle swarm optimization algorithm, obtaining a model for fault recognition when a calculation result meets a preset threshold value, outputting a fault recognition result by inputting fault attribute data to be recognized into a fault recognition model which is trained in advance, training by using improved attribute data as input data to obtain a fault recognition model, and realizing fault recognition through the support vector machine model in combination with the particle swarm optimization algorithm, so that the recognition accuracy of small faults is improved in steps, and the overall effect of fault recognition is improved.
Description
Technical Field
The invention relates to the field of fault diagnosis, in particular to a method for identifying faults and a method and a device for training a model of the method.
Background
Fault diagnosis is of great significance in geological exploration, and faults are usually diagnosed through seismic attributes in the prior art. The seismic attributes are kinematic, dynamic, geometrical and statistical characteristics obtained by performing mathematical calculation on seismic data, and response effects of the seismic attributes are different for faults with different fault distances. The variance attribute in the seismic attribute is taken as an example, the variance attribute can reveal the existence of the fault, but the difference between a large fault and a small fault is several orders of magnitude under the common condition, so that the large fault covers the small fault information, the small fault information identification effect is poor, and the overall fault identification effect is reduced.
Disclosure of Invention
In view of this, the present invention aims to provide fault identification methods and model training methods and apparatuses thereof, wherein the fault identification model is obtained by training improved attribute data as input data, and is realized by combining a support vector machine model with a particle swarm optimization algorithm, so that the identification accuracy of small faults is improved, and the overall effect of fault identification is further improved in steps.
, an embodiment of the present invention provides methods for model training for fault recognition, the method including:
acquiring fault attribute data of a fault area;
generating a sample data set according to fault attribute data;
inputting the sample data set into a preset support vector machine model for training;
and calculating the support vector machine model through a particle swarm optimization algorithm, and obtaining a model for fault identification when a calculation result meets a preset threshold value.
In , the step of generating a sample data set based on fault attribute data includes:
performing classification processing on the fault attribute data to obtain classified fault attribute data;
carrying out correlation calculation on the attribute data of the fault classified into to obtain a correlation result of the attribute data of the fault;
and filtering the fault attribute meeting the preset correlation threshold value to obtain a sample data set of the model.
In , the step of performing correlation calculation on the classified fault attribute data to obtain a correlation result of the fault attribute data includes:
carrying out correlation coefficient calculation on the attribute data of the fault subjected to the quantization of to obtain a correlation coefficient result;
and performing R-type clustering analysis on the correlation number result to obtain a correlation result of fault attribute data.
In , the step of filtering the fault attribute satisfying the preset correlation threshold to obtain a sample data set of the model includes:
judging whether the value of the fault attribute is smaller than a preset correlation threshold value or not;
if so, setting the value of the fault attribute to 0;
if not, the value of the fault attribute is set to 1 when the value of the fault attribute satisfies y '(x) × y' (x-1) <0& y '(x) <0, and the value of the fault attribute is set to 0 when the value of the fault attribute does not satisfy y' (x) × y '(x-1) <0& y' (x) <0, where x is the value of the fault attribute, y '(x) is the th derivative of the value of the fault attribute, and y' (x) is the second derivative of the value of the fault attribute.
In , the above step of calculating the support vector machine model by the particle swarm optimization algorithm, and stopping training when the calculation result satisfies a preset threshold to obtain a model for fault recognition includes:
obtaining a penalty factor and a radial basis kernel function parameter of a support vector machine model;
calculating the penalty factor and the radial basis kernel function parameter by utilizing a particle swarm optimization algorithm to obtain the calculation results of the penalty factor and the radial basis kernel function parameter;
and when the punishment factor and the calculation result of the radial basis kernel function parameter both meet the preset threshold value, stopping training to obtain a model for fault identification.
In , the preset threshold satisfies a fitness function of the particle swarm optimization algorithm, and the fitness function satisfies the recognition accuracy of the support vector machine model.
In a second aspect, an embodiment of the present invention provides an fault identification method, including:
acquiring fault attribute data to be identified;
and inputting fault attribute data to be recognized into a fault recognition model which is trained in advance, and outputting a fault recognition result, wherein the fault recognition model is obtained by training any model training method for fault recognition in the aspect of .
In a third aspect, an embodiment of the present invention provides kinds of model training apparatuses for fault recognition, where the apparatus includes:
the data acquisition module is used for acquiring fault attribute data of a fault area;
the data generation module is used for generating a sample data set according to the fault attribute data;
the training module is used for inputting the sample data set into a preset support vector machine model for training;
and the second training module is used for calculating the support vector machine model through a particle swarm optimization algorithm, and obtaining a model for fault recognition when a calculation result meets a preset threshold value.
In a fourth aspect, an embodiment of the present invention provides fault identification apparatuses, including:
the data acquisition module is used for acquiring fault attribute data to be identified;
and the recognition module is used for inputting fault attribute data to be recognized into a fault recognition model which is trained in advance and outputting a fault recognition result, and the fault recognition model is obtained by training any model training method for fault recognition in the aspect.
In a fifth aspect, an embodiment of the present invention provides electronic devices, the electronic devices including a processor and a storage device, the storage device having stored thereon a computer program, which, when executed by the processor, performs the methods as provided in the and second aspects.
The embodiment of the invention has the following beneficial effects that the embodiment of the invention provides fault recognition methods and a model training method and a device thereof, the fault recognition model training method firstly acquires fault attribute data of a fault region, generates a sample data set according to the fault attribute data, then inputs the sample data set into a preset support vector machine model for training, calculates the support vector machine model through a particle swarm optimization algorithm in the training process, obtains a model for fault recognition when a calculation result meets a preset threshold value, firstly acquires fault attribute data to be recognized in the fault recognition process of the trained fault recognition model, then inputs the fault attribute data to be recognized into a fault recognition model which is trained in advance, outputs a fault recognition result, trains by using the improved attribute data as input data to obtain a fault recognition model, realizes fault recognition by combining the support vector machine model with the particle swarm optimization algorithm, and further improves the fault recognition precision, thereby being beneficial to improving the overall fault recognition effect.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flowchart of a model training method for fault recognition according to an embodiment of the present invention;
FIG. 2 is a flowchart of step S102 in a model training method for fault recognition according to an embodiment of the present invention;
fig. 3 is a flowchart of step S202 in the model training method for fault recognition according to the embodiment of the present invention;
fig. 4 is a flowchart of step S203 in the model training method for fault recognition according to the embodiment of the present invention;
FIG. 5 is a flowchart of step S104 in the model training method for fault recognition according to the embodiment of the present invention;
FIG. 6 is a flow chart of a fault identification method provided by an embodiment of the invention;
FIG. 7 is a diagram of the effect of identifying a fault in a region identified by a vector machine model with unmodified attributes according to an embodiment of the present invention;
FIG. 8 is a diagram of the effect of identifying a fault in a location identified by a vector machine model with improved attributes according to an embodiment of the present invention;
FIG. 9 is a schematic structural diagram of a model training apparatus for fault recognition according to an embodiment of the present invention;
fig. 10 is a schematic structural diagram of a fault identification device according to an embodiment of the present invention;
fig. 11 is a schematic structural diagram of electronic devices according to an embodiment of the present invention.
Icon:
901-data acquisition module, 902-data generation module, 903- th training module, 904-second training module, 1001-data acquisition module, 1002-identification module, 101-processor, 102-memory, 103-bus and 104-communication interface.
Detailed Description
To make the objects, technical solutions and advantages of the embodiments of the present invention clearer and more complete, the technical solutions of the present invention will be described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention.
The fault is the displacement phenomenon formed by the fracture when the rock stratum or rock mass on the earth crust is under the stress and the rock strength limit is lower than the pressure value. The fault diagnosis plays a decisive role in the safety of mineral production, the conventional fault identification is realized by observing the characteristics of amplitude, phase, time difference and the like on an earthquake section, and the obtained fault identification result is influenced by subjective factors of identification personnel because the change of a small fault on a time section is very small and is difficult to observe by naked eyes. Seismic attributes are therefore commonly used in the prior art to identify faults.
The seismic attributes are kinematic, dynamic, geometric and statistical features obtained by mathematical computation of seismic data, and response effects of the seismic attributes are different for faults with different fault distances, the seismic attributes of fault interpretation are numerous, wherein the variance attribute and the curvature attribute are seismic attributes commonly used for fault identification, the variance attribute describes transverse inhomogeneity of an underlying layer through similarity of signals of adjacent seismic channels, and therefore can represent discontinuity of a stratum, the curvature attribute reflects the degree of bending of a stratum when the stratum is compressed by structural stress, the larger the absolute value of the curvature is, the larger the bending degree is, the more the variance attribute is commonly used for explaining the fault with fault fracture, and the more the curvature attribute is commonly used for explaining the fault with curvature, therefore, the prior art only can explain a certain aspect of geological phenomena by using the attribute, and has larger limitation.
For faults with different fault distances, the response degree of the seismic attribute is different, for example, the variance attribute is used, and as the fault distance of the fault increases, the corresponding variance value also continuously increases. The variance at a fault at a distance of 1 m is of the order of 10-6The variance at a fault distance of 20 m is of the order of 10-2The difference in magnitude between the two is obvious. The variance attribute can reveal the existence of faults, but the large faults easily cover the small fault information because the magnitude of the variance of the large faults and the small faults is different greatly. In practice, the number of pitch changes is very complex, often accompanied by variations in the various pitches, and therefore the values are prominent and moreoverNeglecting the condition of small value, the identification result of the small fault is not obvious, and the overall effect of fault identification is reduced.
In view of the above problems in the conventional fault recognition, the present invention aims to provide fault recognition methods and model training methods and apparatuses thereof, which can be applied to fault recognition processes and implemented by using related software or hardware, and are described below by way of embodiments.
For facilitating understanding of the present embodiment, firstly, the model training methods for fault recognition disclosed in the present embodiment are described in detail, and a flowchart of the method is shown in fig. 1, and includes:
step S101, fault attribute data of a fault area are obtained.
The fault area is preferably selected from areas with complex fault scenes, such as areas containing various types of large faults and small faults, and the fault area under various complex scenes is used as input data of the model, so that the identification precision of the model is improved. The fault area also needs to select the area only containing the large fault or the area only containing the small fault, and the accuracy when the large fault or the small fault is identified independently is guaranteed.
The fault attribute data is data for measuring the specific trend and size of the fault and can be selected from various fault attributes related to the variance attribute and the curvature attribute. The selection of fault attributes is not as great as possible, but rather, it is necessary to make a targeted selection according to actual conditions. If the fault attribute data are selected too much, the training time of the model is increased; if the fault attribute data is selected too little, the model identification precision is reduced.
Step S102, generating a sample data set according to fault attribute data.
The fault attribute data obtained in step S101 is combined with actual fault distribution to generate a sample data set. For example, the actual fault distribution in the fault region is only a large fault, and the acquired attribute data of the large fault and the large fault result are combined at the moment to be used as the forward data of model training; similarly, the acquisition of forward data for small faults is similar to that for large faults described above; if the fault region is a combination of a large fault and a small fault, the manner in which the sample data set is generated is also similar to that described above.
In order to improve the training effect of the model, steps of processing can be performed on the sample data set, for example, a threshold value of the quantity of attribute data can be set to reduce the quantity of attributes in the sample data set, a grouping processing can be performed to limit attribute values in the attribute data, the attribute values of the attribute data can be improved, and abnormal data caused by measurement errors can be removed .
Step S103, inputting the sample data set into a preset support vector machine model for training.
The support vector machine model is initialized before the input of the sample data set, and the state of the support vector machine model at the moment can be the state that the initialization process is just completed and the training is not started; or may be a state already in training.
After the sample data set is input into the support vector machine model, relevant parameters of the model are changed through relevant operation, and therefore the identification precision of the model is improved. For example, a penalty factor of the model is optimized in the training process of the sample data set, the penalty factor is a parameter for representing the tolerance of the error, and the larger the value of the penalty factor is, the more intolerable the error occurs, and the overfitting phenomenon is relatively easier to occur; conversely, the smaller the value of the penalty factor, the more the under-fitting phenomenon is relatively easy to occur.
There are also parameters that are important for training in the support vector machine model, namely, the parameters of the radial basis kernel function, which determine the distribution of data after mapping to a new feature space, and the larger the value, the less the support vector, the more overfit is likely to occur, and the smaller the value, the more the support vector, the more under-fit is likely to occur.
The process of inputting the sample data set into the preset support vector machine model for training also includes optimization of other parameters of the model, which is not described herein again.
And step S104, calculating the support vector machine model through a particle swarm optimization algorithm, and obtaining a model for fault identification when a calculation result meets a preset threshold value.
The particle swarm optimization algorithm is also called as the particle swarm optimization algorithm, and can complete training of connection weight in the artificial neural network, structural design, learning rule adjustment, feature selection, initialization of the connection weight, rule extraction and the like.
Relevant parameters in the support vector machine model, such as a penalty factor and a radial basis function parameter, can be optimized steps through a particle swarm optimization algorithm, in the particle swarm optimization algorithm, each optimization problem is to obtain a numerical value of the penalty factor or the radial basis function parameter, the parameters are called particles, and through an iteration process, the particles update themselves by tracking two extreme values, wherein extreme values are the optimal solutions found by the particles, the other extreme values are the optimal solutions found by the whole model.
The particle swarm optimization comprises the steps of initializing particles, evaluating the initialized particles, determining a global optimal solution through a fitness function in the evaluation process, updating the numerical values of the particles through sharing the optimal solution, finishing the optimization process when the particles meet a stopping condition, evaluating the particles in steps if the stopping condition is not met, and judging whether the stopping condition is met or not, wherein the stopping condition generally comprises two types, is that the number of times of particle iteration reaches a threshold value, and is that whether the global optimal solution meets a minimum limit or not.
The training effect of the support vector machine model is further improved through optimization of a particle swarm optimization algorithm, and finally, when a calculation result meets a preset threshold value, a model for fault recognition is obtained, wherein the preset threshold value is data for stopping training of the model and can be obtained through a loss function, and can also be set according to the recognition accuracy in a certain link in the training process of the model.
In the model training method for fault identification provided by the embodiment of the invention, the improved attribute data is used as input data for training to obtain a fault identification model, the attribute data can contain various types of attribute data, the limitation that only one attribute can be used for explanation in the prior art is reduced, the fusion of multiple attributes is realized, a large amount of seismic attribute information can be integrated in , the information contained in the data can be fully excavated, repeated miscellaneous information is removed, and the fault explanation precision and efficiency are improved.
As shown in fig. 2, in , in some embodiments, the step S102 includes:
in step S201, the tomographic attribute data is subjected to consolidation to obtain consolidated tomographic attribute data.
The attribute data of the fault comprise various attribute values, and the magnitude of each attribute data is different greatly, so that the fault attribute is necessary to be subjected to the treatment of classifying , the classification treatment aims to perform dimension elimination on the attribute data, all data ranges are specific numerical value intervals by classifying respective attribute values, and all the attribute data can be limited to be between-1 and +1, for example.
Step S202, correlation calculation is carried out on the classified fault attribute data, and a correlation result of the fault attribute data is obtained.
The purpose of the correlation calculation is to perform steps of limitation on fault attribute data, select attribute data with strong correlation according to the calculation result of the correlation, and reject the attribute data with the correlation, so that redundant data of input data used for model training can be reduced, and the accuracy of the model training can be improved.
For example, as shown in fig. 3, in some embodiments of , the step may include:
and step S21, calculating the correlation coefficient of the attribute data of the fault after being converted into to obtain a correlation coefficient result.
The correlation coefficient can be calculated by using a simple correlation coefficient formula, a complex correlation coefficient formula and a typical correlation coefficient formula, and the obtained result is used for the calculation of the next step in the subsequent step S22.
And step S22, performing R-type clustering analysis on the correlation number result to obtain the correlation result of the fault attribute data.
In the field of cluster analysis, the cluster analysis is generally divided into two categories according to the different classification objects, wherein categories are used for processing the classification and are called Q types, and categories are used for processing the variables and are called R types.
By this step, the correlation coefficient result in step S21 is further -step calculated by R-type cluster analysis to obtain a new correlation result.
Step S203, filtering the fault attribute meeting the preset correlation threshold value to obtain a sample data set of the model.
The filtering of fault attributes that meet a predetermined threshold of correlation is performed on the attributes in the dataset, in some embodiments , as shown in fig. 4, the method may include the following steps:
step S401, judging whether the fault attribute value is smaller than a preset correlation threshold value.
The setting of the correlation threshold is carried out according to the combination of the fault attributes and the actual situation, if the attribute with relatively more correlation needs to be selected, the correlation threshold is set to be lower , and less fault attributes are cut off, otherwise, if the attribute with relatively less correlation needs to be selected, the correlation threshold is set to be higher , and more fault attributes are cut off.
Step S402, if yes, setting the value of the fault attribute to 0, if no, setting the value of the fault attribute to 1 when the value of the fault attribute satisfies y '(x) × y' (x-1) <0& y '(x) <0, and setting the value of the fault attribute to 0 when the value of the fault attribute does not satisfy y' (x) × y '(x-1) <0& y' (x) <0, wherein x is the value of the fault attribute, y '(x) is order derivative of the value of the fault attribute, and y' (x) is the second order derivative of the value of the fault attribute.
This step can be understood as an improvement on fault attribute data, and a specific judgment condition is obtained by calculating order derivatives and second order derivatives of original attributes in the improvement process.
In the specific implementation process, the steps can be performed according to the following logic:
assuming that the original attribute value is V, then there are:
the above is a notation in computer language, a is a variable parameter by which the condition of removing abnormal data in fault attribute data can be adjusted, if the parameter setting is larger, it indicates that more original fault attribute data needs to be retained, otherwise, if the parameter setting is smaller, it indicates that less original fault attribute data needs to be retained.
From the above logic, when the original fault attribute data is less than the preset a threshold, the fault attribute data is set to 0, thus some abnormal data can be eliminated, when the relation y '(x) × y' (x-1) <0& y ″ (x) <0 is satisfied, the fault attribute data is set to 1, thus the attribute data can be limited between 0-1, if any of the above conditions is not satisfied, the fault attribute data is set to 0, through the improvement, the improved attribute value replaces the original fault attribute value, a new data set is formed, and the training of the subsequent model is facilitated.
As shown in fig. 5, in , in some embodiments, the step S104 includes:
step S501, obtaining a penalty factor and a radial basis function parameter of the support vector machine model.
The penalty factor is which is an important parameter for supporting the vector machine model and is a parameter for representing the tolerance of the error, the larger the value of the penalty factor is, the less tolerant to the error is, the more overfit is relatively easy to occur, and on the contrary, the smaller the value of the penalty factor is, the less fit is relatively easy to occur.
The parameter of the radial basis kernel function is which is an important parameter of the model of the support vector, the parameter determines the distribution of the data after being mapped to a new feature space, the larger the value of the parameter is, the less the support vector is, the overfitting is more easily generated, and the smaller the value of the parameter is, the more the support vector is, the under-fitting is more easily generated.
Step S502, calculating the penalty factor and the radial basis kernel function parameter by utilizing a particle swarm optimization algorithm to obtain the calculation result of the penalty factor and the radial basis kernel function parameter.
In the particle swarm optimization, each optimization problem is to obtain the value of a penalty factor or a radial basis kernel function parameter, the parameters are called particles, and the particles update themselves by tracking two extreme values through an iterative process, wherein extreme values are the optimal solutions found by the particles, and extreme values are the optimal solutions found by the whole model.
And S503, stopping training when the punishment factors and the calculation results of the radial basis kernel function parameters both meet a preset threshold value, and obtaining a model for fault recognition.
The preset threshold is data of stopping training of the model, can be acquired through a loss function, and can also be set according to the recognition accuracy in a certain link in the training process of the model, in embodiments, the preset threshold meets the fitness function of the particle swarm optimization algorithm, and the fitness function meets the recognition accuracy of the support vector machine model.
In the sample dataset of the model, various types of fault attributes are contained, for example. For example: in the embodiment, faults in a certain coal mine are identified, 5 attributes including variance, strike curvature, maximum amplitude, dip angle deviation and dip angle continuity of fault attribute data in a sample data set are selected as input attributes of a support vector machine according to correlation results among the attributes, and a final fault identification model is obtained through the model training method.
The embodiment of the invention provides fault identification methods, as shown in fig. 6, the method comprises the following steps:
in step S601, fault attribute data to be identified is acquired.
The fault attribute data in this step is attribute data of the fault to be identified, and the attribute data used needs to be maintained consistent with the attribute data used in model training.
Step S602, inputting fault attribute data to be recognized into a fault recognition model which is trained in advance, and outputting a fault recognition result.
The fault recognition model in the step is obtained by training through the model training method for fault recognition mentioned in the embodiment, and the fault recognition result is finally obtained through the recognition operation of the model. The result can be displayed in a picture mode, and coordinate point data can also be directly given.
For comparison, the fault recognition result and the result of manual recognition are plotted at , the fault recognition model in fig. 7 does not improve the fault data, the fault recognition model in fig. 8 improves the fault data, the black area in the figure is the fault recognized manually, and the gray area is the result of fault recognition model.
It can be seen from fig. 7 and 8 that most faults can be identified by using a fault identification model, but small faults in fig. 7 are not identified, such as F1 and F4, while the fault F1 and F4 in fig. 8 have better identification effect, and the identification result is continuously distributed and conforms to the conventional fault distribution rule, at C in the figure, fig. 7 has identification conditions of discontinuous distribution and does not coincide with an artificial identification area, but the problem does not exist in fig. 8, at D in the figure, the fault distribution does not exist, but a large number of faults are identified at D in fig. 7, errors occur, but the problem does not exist in fig. 8.
Corresponding to the above-mentioned embodiment of the model training method for fault recognition, the present embodiment also provides kinds of model training apparatuses for fault recognition, as shown in fig. 9, the apparatuses including:
the data acquisition module 901 is used for acquiring fault attribute data of a fault area;
a data generating module 902, configured to generate a sample data set according to fault attribute data;
, a training module 903, configured to input the sample data set into a preset support vector machine model for training;
and the second training module 904 is configured to calculate the support vector machine model through a particle swarm optimization algorithm, and obtain a model for fault identification when a calculation result meets a preset threshold.
The implementation principle and the generated technical effect of the model training device for fault recognition provided by the embodiment of the invention are the same as those of the embodiment of the model training method for fault recognition, and for the sake of brief description, corresponding contents in the embodiment of the method can be referred to where the embodiment is not mentioned.
Corresponding to the above-described embodiment of the fault identification method, the present embodiment also provides kinds of fault identification apparatuses, as shown in fig. 10, including:
a data acquisition module 1001 configured to acquire fault attribute data to be identified;
and the recognition module 1002 is used for inputting fault attribute data to be recognized into a fault recognition model which is trained in advance and outputting a fault recognition result, wherein the fault recognition model is obtained by training any model training method for fault recognition of item in the th aspect.
The fault recognition device provided by the embodiment of the invention has the same implementation principle and technical effect as the fault recognition method, and for brief description, the embodiment can refer to the corresponding content in the method embodiment.
The embodiment also provides electronic devices, which are shown in fig. 11, and include a processor 101 and a memory 102, where the memory 102 is used to store or more computer instructions, and or more computer instructions are executed by the processor to implement the above fault recognition model training method and fault recognition method.
The server shown in fig. 11 further includes a bus 103 and a communication interface 104, and the processor 101, the communication interface 104, and the memory 102 are connected through the bus 103.
The Memory 102 may include a high-speed Random Access Memory (RAM) and may also include a non-volatile Memory (non-volatile Memory), such as at least disk memories, the bus 103 may be an ISA bus, a PCI bus, or an EISA bus, etc. the buses may be divided into an address bus, a data bus, a control bus, etc. for ease of illustration, only double-headed arrows are shown in fig. 11, but they do not represent only buses or types of buses.
The communication interface 104 is used for connecting with at least user terminals and other network units through a network interface, and sending the packaged IPv4 message or IPv4 message to the user terminals through the network interface.
The Processor 101 may be an type integrated circuit chip, which has signal Processing capability, and in the implementation process, the steps of the above method may be implemented by instructions in the form of hardware integrated logic circuits or software in the Processor 101, the Processor 101 may be a general-purpose Processor, which includes a Central Processing Unit (CPU), a Network Processor (NP), etc., a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Array (Field-Programmable Gate Array, FPGA), or other Programmable logic devices, discrete , or transistor logic devices, discrete hardware components, and may implement or execute the methods, steps, and logic blocks disclosed in the embodiments of the present disclosure, the general-purpose Processor may be a microprocessor or any conventional Processor, etc., and the steps and logic blocks of the methods, steps, and logic blocks disclosed in the embodiments of the present disclosure may be implemented by a flash memory module, or a combination of the aforementioned methods, read-only-Programmable memory module, or a read-only-erasable memory, or a combination of the above method, the steps and steps may be implemented by a read-only-Programmable memory module, or a read-only-erasable memory module.
An embodiment of the present invention further provides computer-readable storage media, on which a computer program is stored, which, when executed by a processor, performs the steps of the method of the foregoing embodiment.
The above-described embodiments of the apparatus are merely illustrative, e.g., the division of units into logical functional divisions, and other divisions may be realized in practice, e.g., multiple units or components may be combined or integrated into another systems, or features may be omitted or not implemented.another point, the shown or discussed coupling or direct coupling or communication connection between each other may be through communication interfaces, indirect coupling or communication connection between devices or units, which may be electrical, mechanical or other forms.
The units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, that is, may be located in places, or may also be distributed on multiple network units.
In addition, functional units in the embodiments of the present invention may be integrated into processing units, or each unit may exist alone physically, or two or more units are integrated into units.
It is to be understood that the present invention may be embodied in a software product, which is stored in storage media and includes instructions for causing computer devices (which may be personal computers, servers, or network devices) to execute all or part of the steps of the methods according to the embodiments of the present invention.
Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1, A model training method for fault recognition, the method comprising:
acquiring fault attribute data of a fault area;
generating a sample data set according to the fault attribute data;
inputting the sample data set into a preset support vector machine model for training;
and calculating the support vector machine model through a particle swarm optimization algorithm, and obtaining the model for fault identification when a calculation result meets a preset threshold value.
2. The model training method according to claim 1, wherein the step of generating a sample data set from the fault attribute data comprises:
performing classification processing on the fault attribute data to obtain classified fault attribute data;
carrying out correlation calculation on the classified fault attribute data to obtain a correlation result of the fault attribute data;
and filtering the fault attributes meeting the preset correlation threshold value to obtain a sample data set of the model.
3. The model training method according to claim 2, wherein the step of performing correlation calculation on the fault attribute data subjected to the quantization of to obtain a correlation result of the fault attribute data comprises:
carrying out correlation coefficient calculation on the attribute data of the fault subjected to the quantization of to obtain a correlation coefficient result;
and performing R-type clustering analysis on the correlation coefficient result to obtain a correlation result of the fault attribute data.
4. The model training method according to claim 2, wherein the step of performing filtering processing on the fault attribute satisfying a preset correlation threshold to obtain a sample data set of the model comprises:
judging whether the value of the fault attribute is smaller than a preset correlation threshold value or not;
if so, setting the value of the fault attribute to 0;
if not, setting the value of the fault attribute to be 1 when the value of the fault attribute satisfies y '(x) × y' (x-1) <0& y '(x) <0, and setting the value of the fault attribute to be 0 when the value of the fault attribute does not satisfy y' (x) × y '(x-1) <0& y' (x) <0, wherein x is the value of the fault attribute, y '(x) is th derivative of the value of the fault attribute, and y' (x) is the second derivative of the value of the fault attribute.
5. The model training method according to claim 1, wherein the step of calculating the support vector machine model by a particle swarm optimization algorithm, stopping training when a calculation result satisfies a preset threshold, and obtaining the model for fault recognition comprises:
obtaining a penalty factor and a radial basis kernel function parameter of the support vector machine model;
calculating the penalty factors and the radial basis kernel function parameters by utilizing a particle swarm optimization algorithm to obtain the calculation results of the penalty factors and the radial basis kernel function parameters;
and when the calculation results of the penalty factors and the radial basis kernel function parameters both meet a preset threshold value, stopping training to obtain the model for fault recognition.
6. The model training method of claim 5, wherein the preset threshold satisfies a fitness function of the particle swarm optimization algorithm; and the fitness function meets the identification accuracy of the support vector machine model.
7, A method of fault identification, the method comprising:
acquiring fault attribute data to be identified;
inputting the fault attribute data to be recognized into a fault recognition model which is trained in advance, and outputting the fault recognition result, wherein the fault recognition model is obtained by training through the model training method for fault recognition in any of claims 1-6.
8, model training device for fault recognition, characterized in that, the device includes:
the data acquisition module is used for acquiring fault attribute data of a fault area;
the data generation module is used for generating a sample data set according to the fault attribute data;
an training module, configured to input the sample data set into a preset support vector machine model for training;
and the second training module is used for calculating the support vector machine model through a particle swarm optimization algorithm, and obtaining the model for fault recognition when a calculation result meets a preset threshold value.
A fault recognition apparatus of the type , said apparatus comprising:
the data acquisition module is used for acquiring fault attribute data to be identified;
and the recognition module is used for inputting the fault attribute data to be recognized into a fault recognition model which is trained in advance and outputting the fault recognition result, wherein the fault recognition model is obtained by training through the model training method for fault recognition according to any items in claims 1-6.
10, electronic device comprising a processor and a storage means, the storage means having stored thereon a computer program for performing, when executed by the processor, the method of any of claims 1 to 6, or .
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911078637.1A CN110737021A (en) | 2019-11-06 | 2019-11-06 | Fault recognition method and model training method and device thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911078637.1A CN110737021A (en) | 2019-11-06 | 2019-11-06 | Fault recognition method and model training method and device thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110737021A true CN110737021A (en) | 2020-01-31 |
Family
ID=69272391
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911078637.1A Pending CN110737021A (en) | 2019-11-06 | 2019-11-06 | Fault recognition method and model training method and device thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110737021A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111650643A (en) * | 2020-07-03 | 2020-09-11 | 中国矿业大学(北京) | Method and system for extracting training set of seismic fault machine learning model |
| CN111856572A (en) * | 2020-07-06 | 2020-10-30 | 中国石油大学(北京) | Method and device for determining width of fault fracture belt |
| CN112130200A (en) * | 2020-09-23 | 2020-12-25 | 电子科技大学 | Fault identification method based on grad-CAM attention guidance |
| CN112346126A (en) * | 2020-11-09 | 2021-02-09 | 中国石油天然气股份有限公司 | Low-order fault identification method, device, equipment and readable storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013048798A2 (en) * | 2011-09-28 | 2013-04-04 | Saudi Arabian Oil Company | Reservoir properties prediction with least square support vector machine |
| CN104914465A (en) * | 2015-06-02 | 2015-09-16 | 中国石油天然气股份有限公司 | Volcanic rock crack quantitative prediction method and device |
| CN106485015A (en) * | 2016-10-20 | 2017-03-08 | 辽宁工程技术大学 | A kind of determination method of mine tomography coverage |
| CN104570083B (en) * | 2013-10-29 | 2017-06-06 | 中国石油天然气集团公司 | Geologic body automatic identifying method based on multi-dimensional earthquake attribute |
| CN107843920A (en) * | 2017-09-29 | 2018-03-27 | 中国石油化工股份有限公司 | A kind of cluster analysis optimizes population reservoir attribute Forecasting Methodology |
| CN109492775A (en) * | 2018-11-19 | 2019-03-19 | 中国矿业大学(北京) | A kind of detection method of geologic structure interpretation, detection device and readable storage medium storing program for executing |
-
2019
- 2019-11-06 CN CN201911078637.1A patent/CN110737021A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013048798A2 (en) * | 2011-09-28 | 2013-04-04 | Saudi Arabian Oil Company | Reservoir properties prediction with least square support vector machine |
| CN104570083B (en) * | 2013-10-29 | 2017-06-06 | 中国石油天然气集团公司 | Geologic body automatic identifying method based on multi-dimensional earthquake attribute |
| CN104914465A (en) * | 2015-06-02 | 2015-09-16 | 中国石油天然气股份有限公司 | Volcanic rock crack quantitative prediction method and device |
| CN106485015A (en) * | 2016-10-20 | 2017-03-08 | 辽宁工程技术大学 | A kind of determination method of mine tomography coverage |
| CN107843920A (en) * | 2017-09-29 | 2018-03-27 | 中国石油化工股份有限公司 | A kind of cluster analysis optimizes population reservoir attribute Forecasting Methodology |
| CN109492775A (en) * | 2018-11-19 | 2019-03-19 | 中国矿业大学(北京) | A kind of detection method of geologic structure interpretation, detection device and readable storage medium storing program for executing |
Non-Patent Citations (1)
| Title |
|---|
| 孙振宇等: "基于SVM算法的地震小断层自动识别", 《煤炭学报》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111650643A (en) * | 2020-07-03 | 2020-09-11 | 中国矿业大学(北京) | Method and system for extracting training set of seismic fault machine learning model |
| CN111650643B (en) * | 2020-07-03 | 2021-03-12 | 中国矿业大学(北京) | Method and system for extracting training set of seismic fault machine learning model |
| CN111856572A (en) * | 2020-07-06 | 2020-10-30 | 中国石油大学(北京) | Method and device for determining width of fault fracture belt |
| CN111856572B (en) * | 2020-07-06 | 2021-07-20 | 中国石油大学(北京) | Method and device for determining width of fault fracture belt |
| CN112130200A (en) * | 2020-09-23 | 2020-12-25 | 电子科技大学 | Fault identification method based on grad-CAM attention guidance |
| CN112130200B (en) * | 2020-09-23 | 2021-07-20 | 电子科技大学 | Fault identification method based on grad-CAM attention guidance |
| CN112346126A (en) * | 2020-11-09 | 2021-02-09 | 中国石油天然气股份有限公司 | Low-order fault identification method, device, equipment and readable storage medium |
| CN112346126B (en) * | 2020-11-09 | 2024-02-02 | 中国石油天然气股份有限公司 | Method, device, equipment and readable storage medium for identifying low-order faults |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110737021A (en) | Fault recognition method and model training method and device thereof | |
| CN113792453B (en) | Digital twinning-based partial discharge monitoring system, method and device | |
| Hengl et al. | Predictive soil mapping with R | |
| CN109685152B (en) | Image target detection method based on DC-SPP-YOLO | |
| CN110598845B (en) | Data processing method, data processing device, computer equipment and storage medium | |
| CN108959794B (en) | Structural frequency response dynamic model correction method based on deep learning | |
| CN111144548A (en) | Method and device for identifying working condition of pumping well | |
| CN111401472B (en) | Infrared target classification method and device based on deep convolutional neural network | |
| JP7298825B2 (en) | Learning support device, learning device, learning support method, and learning support program | |
| CN105718848A (en) | Quality evaluation method and apparatus of fingerprint images | |
| JP2014006813A (en) | Performance prediction device, performance prediction method, and program | |
| CN114492555A (en) | Method and device for identifying reasons of abnormal operation of conveying pipeline | |
| CN112182056A (en) | Data detection method, device, equipment and storage medium | |
| CN115587543A (en) | Federal learning and LSTM-based tool residual life prediction method and system | |
| CN117972314B (en) | Cloud platform monitoring method and system based on digital twinning | |
| CN111078500A (en) | Method and device for adjusting operation configuration parameters, computer equipment and storage medium | |
| CN112506907A (en) | Engineering machinery marketing strategy pushing method, system and device based on big data | |
| CN116486146A (en) | Fault detection method, system, device and medium for rotary mechanical equipment | |
| CN112346126B (en) | Method, device, equipment and readable storage medium for identifying low-order faults | |
| US20220156416A1 (en) | Techniques for comparing geometric styles of 3d cad objects | |
| CN114218462A (en) | Data classification method, device, equipment and storage medium based on LSTM | |
| CN113111948A (en) | Time sequence data classification method and system based on feature recalibration mechanism | |
| Nasser et al. | Improvement of geoid accuracy using convolution neural networks and empirical rules | |
| CN109658489B (en) | Three-dimensional grid data processing method and system based on neural network | |
| JP6831307B2 (en) | Solution calculation device, solution calculation method and solution calculation program |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200131 |