CN117892080A - Method and device for establishing and diagnosing fault diagnosis model of valve body of oilfield fracturing pump - Google Patents
Method and device for establishing and diagnosing fault diagnosis model of valve body of oilfield fracturing pump Download PDFInfo
- Publication number
- CN117892080A CN117892080A CN202311756427.XA CN202311756427A CN117892080A CN 117892080 A CN117892080 A CN 117892080A CN 202311756427 A CN202311756427 A CN 202311756427A CN 117892080 A CN117892080 A CN 117892080A
- Authority
- CN
- China
- Prior art keywords
- features
- feature
- fault
- fracturing pump
- samples
- 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
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000003745 diagnosis Methods 0.000 title claims abstract description 45
- 230000007246 mechanism Effects 0.000 claims abstract description 47
- 238000007637 random forest analysis Methods 0.000 claims abstract description 31
- 238000004458 analytical method Methods 0.000 claims abstract description 30
- 238000004364 calculation method Methods 0.000 claims description 18
- 238000001228 spectrum Methods 0.000 claims description 18
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 11
- 238000001514 detection method Methods 0.000 claims description 10
- 230000000694 effects Effects 0.000 claims description 9
- 238000012795 verification Methods 0.000 claims description 8
- 238000003066 decision tree Methods 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 230000003595 spectral effect Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 description 9
- 239000000284 extract Substances 0.000 description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000013145 classification model Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 210000003097 mucus Anatomy 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
- F04B—POSITIVE-DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS
- F04B51/00—Testing machines, pumps, or pumping installations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Mechanical Engineering (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The embodiment of the disclosure provides a method and a device for establishing and diagnosing a fault diagnosis model of an oilfield fracturing pump valve body. Comprising the following steps: collecting a plurality of groups of continuous operation data from the normal history to the fault of the oilfield fracturing pump, and dividing the plurality of groups of continuous operation data into a plurality of original samples by using a sliding window method; extracting time domain features, frequency domain features and other mechanism features in each original sample by a mechanism analysis means, and adding the time domain features, the frequency domain features and the other mechanism features into the original sample to obtain a data set; selecting an optimal feature subset from the dataset using a random forest based feature importance analysis method; and modeling the optimal feature subset by using a multi-layer perceptron to obtain the fault diagnosis model of the oilfield fracturing pump. Important features can be mined from a large number of mechanism features, the dimension of input features in the data driving model can be effectively reduced, and the fault of the oilfield fracturing pump can be more effectively identified.
Description
Technical Field
The embodiment of the disclosure belongs to the technical field of fault diagnosis of valve bodies of oilfield fracturing pumps, and particularly relates to a method and a device for establishing and diagnosing a fault diagnosis model of a valve body of an oilfield fracturing pump.
Background
The fracturing pump is core equipment of oil and gas field development fracturing equipment, and the fracturing pump pumps mucus into the stratum with a displacement pump which is greatly higher than the stratum absorption capacity, so that cracks are formed in the stratum, the diversion capacity of an oil and gas layer is improved, and the purpose of increasing yield is achieved. Meanwhile, due to the factors of high working pressure of the fracturing pump, sand and stone contained in working medium and the like, the valve body of the fracturing pump is easy to damage, and the safe operation of equipment is influenced.
In actual operation, the main faults of the fracturing pump occur on the two inlet and outlet valve bodies of the hydraulic cylinder due to the fact that the pressure is too high (up to 100 Mpa). If the fault is not detected in time, the oil field output is reduced on one hand, and if the valve body is damaged for too long, the valve seat is damaged, and the valve body is further damaged. At present, the manual inspection pump is a necessary means for the safe operation of the equipment, and the traditional manual diagnosis technology comprises the following steps: ear hearing, hand touch, measuring the magnitude of vibration value of the cylinder body, checking the pressure real-time curve (most direct and difficult to distinguish), and the like. However, the effective rate of the manual detection means is not more than 60%, so that the method has important significance for fault diagnosis of the check valve of the fracturing pump.
Because the working environment of the fracturing pump is complex and varied, and vibration signal excitation sources are numerous, the traditional fault diagnosis is difficult, and aiming at the situation, the invention adopts a means of combining data driving and mechanical fault diagnosis, extracts the time domain information and the frequency domain information of a sensor signal by using a mechanical fault diagnosis method, and models by using a data driving means, thereby realizing the fault diagnosis of the one-way valve of the fracturing pump. Compared with the traditional manual detection means, the method based on mechanism analysis and data driving can better excavate key characterization parameters of the fault of the fracturing pump, so that the potential fault of the fracturing pump can be identified more accurately.
Disclosure of Invention
The embodiment of the disclosure aims at solving at least one of the technical problems existing in the prior art and provides a method and a device for establishing and diagnosing a fault diagnosis model of an oilfield fracturing pump valve body.
In a first aspect, embodiments of the present disclosure provide a method for establishing a fault diagnosis model of an oilfield fracturing pump valve body, the method comprising:
collecting a plurality of groups of continuous operation data from the normal history to the fault of the oilfield fracturing pump, and dividing the plurality of groups of continuous operation data into a plurality of original samples by using a sliding window method;
extracting time domain features, frequency domain features and other mechanism features in each original sample by a mechanism analysis means, and adding the time domain features, the frequency domain features and the other mechanism features into the original sample to obtain a data set;
selecting an optimal feature subset from the dataset using a random forest based feature importance analysis method;
and modeling the optimal feature subset by using a multi-layer perceptron to obtain the fault diagnosis model of the oilfield fracturing pump.
In some embodiments, the collecting a plurality of sets of continuous operational data of historical normal to failure of an oilfield fracturing pump and dividing the plurality of sets of continuous operational data into a plurality of raw samples using a sliding window method comprises:
collecting a plurality of groups of continuous operation data p= { p from history of normal to failure of the oilfield fracturing pump 1 ,p 2 ,...,p n Dividing the fracturing pump operation data into a plurality of original samples by using a sliding window method; wherein,
setting the sliding window size as L, and setting the sliding step length as s1 and s2, s1 during normal and fault respectively>s2, input x of the ith sample i The method comprises the following steps:
x i ＝[p is+1 ,p is+2 ,...,p is+L ]
wherein p is i The ith value in the fracturing pump outlet pressure curve is represented, and s represents the sliding step length; if the input x of the ith sample i Including fault data, the output y of the ith sample i Sliding step s=s2; otherwise y i =0, sliding step s=s1.
In some embodiments, the means for analyzing by a mechanism extracts a time domain feature, a frequency domain feature, and other mechanism features in each of the original samples, and adds the time domain feature, the frequency domain feature, and the other mechanism features to the original samples, including:
in the time domain feature calculation step, the time domain features include a mean value u and a variance sigma 2 The range R, kurtosis K, skewness S, waveform factor Rf and peak factor Cf are calculated as follows:
R＝p max -p min
wherein p is i Representing the value of the ith in sample x, p max And p min Respectively representing the maximum value and the minimum value of the corresponding characteristics of the sample x, wherein L represents the length of a sliding window;
and/or the number of the groups of groups,
in the frequency domain feature calculation step, performing fourier transform on the discrete sample x to obtain a frequency spectrum f corresponding to each input feature and a power spectrum density P corresponding to each input feature; the frequency domain features comprise a spectrum mean value up, a spectrum kurtosis Sp and a spectrum skewness Kp, and the calculation formula is as follows:
wherein P is i Representing f in the ith frequency i A corresponding power spectral density, k representing the length of the spectrum f;
and/or the number of the groups of groups,
in the calculating step of other mechanism characteristics, main characteristic signals of the oilfield split-pressure pump are pressure signals, and the mechanism characteristics comprise differential pressure of an inlet and an outlet of the split-pressure pump and periods of different pressure signals.
In some embodiments, the selecting an optimal feature subset from the dataset using a random forest based feature importance analysis method comprises:
step 1, dividing the data set into a training set and a verification set;
step 2, modeling all the features on the training set by using a random forest algorithm to obtain importance scores score of each feature;
and 3, evaluating the classification effect of the fault category of the random forest model on the verification set by using a classification index F1_score, wherein the calculation formula is as follows:
wherein, for the fault samples, TP represents the correct number of the fault sample classification results, FP represents the number of the normal samples classified as the fault samples, FN represents the number of the normal samples classified as the fault samples; the larger the F1_score is, the better the classification effect of the model on the fault sample is;
step 4, eliminating the features with the lowest feature importance, and modeling a new feature subset by using a random forest algorithm to obtain an importance score of the remaining features;
step 5, repeating the step 3 and the step 4 until the number of the features in the feature subset is 1;
and 6, selecting the feature subset with the largest fault sample classification index F1_score as the optimal feature subset best_feature.
In some embodiments, modeling all features on the training set using a random forest algorithm, resulting in an importance score for each feature, comprises:
calculating importance score of each feature by using a random forest algorithm, and assuming that the random forest comprises M base learners, wherein the number of input features is F; the calculation mode of the coefficient of the basis is as follows:
wherein C represents the number of categories, p, in the dataset c Representing the proportion of class c samples in the node; if the node uses the ith input feature f i Then the change in the coefficient of the kunit around node q is:
wherein,representing the ith feature f i The coefficient of the Kernel score, gini, at node q q1 And Gini q2 Two branch nodes respectively representing a node q;
if the feature f i The set of nodes appearing in decision tree m is Q, feature f i The feature importance in decision tree m is:
feature f i Feature importance score VIM i The calculation formula of (2) is as follows:
the importance score set for all features is score= { VIM i },i＝1,2,3,...,F。
In some embodiments, modeling the optimal feature subset using the multi-layer perceptron to obtain an oilfield fracturing pump fault diagnosis model includes:
modeling the optimal feature subset by using a multi-layer perceptron MLP to obtain an oilfield split pump fault detection Model:
y＝Model(best_feature)
the fault detection Model consists of a plurality of fully connected layers, the dimension of the last fully connected layer is 2, and the activation function of each fully connected layer is a Softmax function, which is defined as follows:
wherein x is i Representing the output of the ith neuron node, c representing the number of neurons of the next layer;
wherein, the loss function of the fault detection Model uses F1_score, which is defined as follows:
wherein TP represents the correct number of classification results of the fault samples, FP represents the number of classification of the normal samples as the fault samples, FN represents the number of classification of the fault samples as the normal samples; the larger f1_score indicates the better the classification of the fault samples by the model.
In a second aspect, embodiments of the present disclosure provide a method of diagnosing a fault in an oilfield fracturing pump valve body, the method comprising:
collecting actual operation data of an oilfield fracturing pump;
inputting actual operation data of the oilfield fracturing pump into an oilfield fracturing pump fault diagnosis model to obtain a diagnosis result of the oilfield fracturing pump; the oilfield fracturing pump fault diagnosis model is obtained by adopting the establishment method.
In a third aspect, embodiments of the present disclosure provide an apparatus for establishing a fault diagnosis model of an oilfield fracturing pump valve body, where the apparatus includes:
the acquisition module is used for acquiring a plurality of groups of continuous operation data from the normal history to the failure of the oilfield fracturing pump and dividing the plurality of groups of continuous operation data into a plurality of original samples by using a sliding window method;
the extraction module is used for extracting time domain features, frequency domain features and other mechanism features in each original sample through a mechanism analysis means, and adding the time domain features, the frequency domain features and the other mechanism features into the original sample to obtain a data set;
an analysis module for selecting an optimal feature subset from the dataset using a random forest based feature importance analysis method;
the building module is used for modeling the optimal feature subset by using the multi-layer perceptron to obtain an oilfield fracturing pump fault diagnosis model.
In a fourth aspect, embodiments of the present disclosure provide an electronic device, comprising:
one or more processors;
and a storage unit configured to store one or more programs that, when executed by the one or more processors, enable the one or more processors to implement the method according to the foregoing description.
In a fifth aspect, embodiments of the present disclosure provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, is capable of implementing a method according to the preceding description.
The method and the device for establishing and diagnosing the fault diagnosis model of the oilfield fracturing pump valve body have the following beneficial technical effects compared with the prior art:
1. the embodiment uses a mechanism analysis means, and extracts mechanism information such as time domain information, frequency domain information and the like of the sensor signals by using a mechanical fault diagnosis method;
2. according to the embodiment, the feature importance analysis method based on the random forest is used for selecting the mechanism features, so that important features can be mined from a large number of mechanism features, and the dimension of input features in the data driving model is effectively reduced;
3. based on the whole set of processes of mechanism feature construction, feature selection and fault classification model construction, the embodiment combines the means of mechanism analysis and data analysis, and can more effectively identify faults of the oilfield fracturing pump.
Drawings
FIG. 1 is a flow chart of a method of establishing a fault diagnosis model for an oilfield fracturing pump valve body in an embodiment of the disclosure;
FIG. 2 is a schematic diagram of a fracturing pump outlet pressure curve under different conditions provided by an embodiment of the present disclosure;
FIG. 3 is a flowchart of an optimal feature subset selection method based on importance analysis provided by an embodiment of the present disclosure;
FIG. 4 is a schematic diagram of a multi-layer sensor network according to an embodiment of the present disclosure;
fig. 5 is a schematic structural diagram of an apparatus for establishing a fault diagnosis model of an oilfield fracturing pump valve body according to an embodiment of the disclosure.
Detailed Description
In order to enable those skilled in the art to better understand the technical solutions of the embodiments of the present disclosure, the embodiments of the present disclosure are described in further detail below with reference to the accompanying drawings and detailed description.
As shown in fig. 1, an embodiment of the present disclosure provides a method for establishing a fault diagnosis model of an oilfield fracturing pump valve body, the method comprising the following specific steps:
and step S101, collecting a plurality of groups of continuous operation data from the normal history to the failure of the oilfield fracturing pump, and dividing the plurality of groups of continuous operation data into a plurality of original samples by using a sliding window method.
Specifically, in this step, multiple sets of continuous operation data p= { p of the oil field fracturing pump from normal to failure are collected 1 ,p 2 ,...,p n Dividing the fracturing pump operation data into a plurality of original samples by using a sliding window method; wherein,
setting the sliding window size as L, and setting the sliding step length as s1 and s2, s1 during normal and fault respectively>s2, input x of the ith sample i The method comprises the following steps:
x i ＝[p is+1 ,p is+2 ,...,p is+L ]
wherein p is i The ith value in the fracturing pump outlet pressure curve is represented, and s represents the sliding step length; if the input of the ith sampleIn x i Including fault data, the output y of the ith sample i Sliding step s=s2; otherwise y i =0, sliding step s=s1.
As a specific example, the embodiment uses a real-time pressure curve of a hydraulic end (water outlet) of the fracturing pump as a judging basis for fault detection of the fracturing pump, and collects a pressure curve p of the hydraulic end of the fracturing pump from normal operation to fault occurrence through a fracturing pump test bed. Setting the size of a sliding window to be L=30, setting the window interval s=5, decomposing a water outlet pressure curve into a plurality of samples, and inputting x of the ith sample i The following is shown:
x i ＝[p is+1 ,p is+2 ,...,p is+L ]。
step S102, extracting time domain features, frequency domain features and other mechanism features in each original sample through a mechanism analysis method, and adding the time domain features, the frequency domain features and the other mechanism features into the original samples to obtain a data set.
Specifically, in this step, since the difference in the water outlet pressure curves of the normal sample and the failure sample is small, it is difficult to distinguish the normal and failure samples with the naked eye, so it is difficult to directly use the original water outlet pressure curve for failure diagnosis, as shown in fig. 2. According to the traditional fault diagnosis experience, the embodiment extracts corresponding time domain features and frequency domain features from the water outlet pressure curve.
Specifically, in this step, in the time domain feature calculation step, the time domain features include a mean value u and a variance σ 2 The range R, kurtosis K, skewness S, waveform factor Rf and peak factor Cf are calculated as follows:
R＝p max -p min
wherein p is i Representing the value of the ith in sample x, p max And p min The maximum value and the minimum value of the corresponding feature of the sample x are respectively represented, L represents the length of the sliding window, and this embodiment sets it to 30.
In the frequency domain feature calculation step, performing fourier transform on the discrete sample x to obtain a frequency spectrum f corresponding to each input feature and a power spectrum density P corresponding to each input feature; the frequency domain features comprise a spectrum mean value up, a spectrum kurtosis Sp and a spectrum skewness Kp, and the calculation formula is as follows:
wherein P is i Representing f in the ith frequency i The corresponding power spectral density, k, represents the length of the spectrum f.
In the calculating step of other mechanism characteristics, main characteristic signals of the oilfield split-pressure pump are pressure signals, and the mechanism characteristics comprise differential pressure of an inlet and an outlet of the split-pressure pump, periods of different pressure signals and the like.
Step S103, selecting an optimal feature subset from the data set by using a feature importance analysis method based on random forests.
Specifically, in this step, for the fracturing pump failure data set obtained in step S102, feature selection is performed by using a random forest-based importance analysis method, so as to obtain a feature subset with the greatest importance degree. The steps of this feature selection method are shown in fig. 3, and the specific implementation steps in this embodiment are as follows:
step 1, dividing the data set obtained in step 102 into a training set and a verification set, wherein the ratio of the training set to the verification set is 9:1, namely training is carried out by using 90% of data, and verification is carried out by 10% of data.
Step 2, modeling all the features on the training set by using a random forest algorithm to obtain importance scores score of each feature;
and 3, evaluating the classification effect of the fault category of the random forest model on the verification set by using a classification index F1_score, wherein the calculation formula is as follows:
wherein, for the fault samples, TP represents the correct number of the fault sample classification results, FP represents the number of the normal samples classified as the fault samples, FN represents the number of the normal samples classified as the fault samples; the larger the F1_score is, the better the classification effect of the model on the fault sample is;
step 4, eliminating the features with the lowest feature importance, and modeling a new feature subset by using a random forest algorithm to obtain an importance score of the remaining features;
step 5, repeating the step 3 and the step 4 until the number of the features in the feature subset is 1;
and 6, selecting the feature subset with the largest fault sample classification index F1_score as the optimal feature subset best_feature.
In some embodiments, modeling all features on the training set using a random forest algorithm, resulting in an importance score for each feature, comprises:
calculating importance score of each feature by using a random forest algorithm, and assuming that the random forest comprises M base learners, wherein the number of input features is F; the calculation mode of the coefficient of the basis is as follows:
wherein C represents the number of categories, p, in the dataset c Representing the proportion of class c samples in the node; if the node uses the ith input feature f i Then the change in the coefficient of the kunit around node q is:
wherein,representing the ith feature f i The coefficient of the Kernel score, gini, at node q q1 And Gini q2 Two branch nodes respectively representing a node q;
if the feature f i The set of nodes appearing in decision tree m is Q, feature f i The feature importance in decision tree m is:
feature f i Feature importance score VIM i The calculation formula of (2) is as follows:
the importance score set for all features is score= { VIM i },i＝1,2,3,...,F。
According to the above steps, the optimal feature subset obtained by the implementation selection is as follows: variance sigma 2 The kurtosis K, the skewness S and the waveform factor Rf are four, and the classification accuracy index F1_score=0.98 of the random forest model is obtained.
And step S104, modeling the optimal feature subset by using a multi-layer perceptron to obtain an oilfield fracturing pump fault diagnosis model.
Specifically, in this step, the optimal feature subset is modeled using the multi-layer perceptron MLP, where the network layers are respectively set to 8,16,8,2, and the activation function is set to softmax function and the loss function is set to f1_score function as shown in fig. 4. Wherein, the formula of the softmax function is as follows:
wherein x is i Representing the output of the ith neuron node, c represents the number of neurons of the next layer, c=2 in this embodiment;
wherein, the loss function of the fault detection Model uses F1_score, which is defined as follows:
wherein TP represents the correct number of classification results of the fault samples, FP represents the number of classification of the normal samples as the fault samples, FN represents the number of classification of the fault samples as the normal samples; the larger f1_score indicates the better the classification of the fault samples by the model.
Finally, in order to verify the effectiveness of the model, the present embodiment compares the classification effects of different classification algorithms and different feature processing modes, and the evaluation index is f1_score, as shown in the following table 1:
TABLE 1 error in temperature field estimation at different mass flows
Based on the technical scheme, compared with the prior art, the oilfield fracturing pump valve body fault diagnosis model of the embodiment has the following beneficial technical effects:
1. the embodiment uses a mechanism analysis means, and extracts mechanism information such as time domain information, frequency domain information and the like of the sensor signals by using a mechanical fault diagnosis method;
2. according to the embodiment, the feature importance analysis method based on the random forest is used for selecting the mechanism features, so that important features can be mined from a large number of mechanism features, and the dimension of input features in the data driving model is effectively reduced;
3. based on the whole set of processes of mechanism feature construction, feature selection and fault classification model construction, the embodiment combines the means of mechanism analysis and data analysis, and can more effectively identify faults of the oilfield fracturing pump.
Based on the same inventive concept, the embodiment of the disclosure further provides a fault diagnosis method for the valve body of the oilfield fracturing pump, which comprises the following steps:
collecting actual operation data of an oilfield fracturing pump;
inputting actual operation data of the oilfield fracturing pump into an oilfield fracturing pump fault diagnosis model to obtain a diagnosis result of the oilfield fracturing pump; the oilfield fracturing pump fault diagnosis model is obtained by adopting the establishment method described in the previous description, and the related description can be referred to in detail, and details are omitted here.
Based on the same inventive concept, as shown in fig. 5, an embodiment of the present disclosure further provides an apparatus for establishing a fault diagnosis model of an oilfield fracturing pump valve body, where the apparatus is applicable to the foregoing establishing method, and the foregoing may be referred to in detail, and details are not described herein. The device comprises an acquisition module 201, an extraction module 202, an analysis module 203 and a building module 204.
The collection module 201 is configured to collect a plurality of sets of continuous operation data from a normal to a fault of the oil field fracturing pump, and divide the plurality of sets of continuous operation data into a plurality of original samples by using a sliding window method. The extraction module 202 is configured to extract the time domain feature, the frequency domain feature and other mechanism features in each original sample by means of mechanism analysis, and add the time domain feature, the frequency domain feature and other mechanism features to the original sample, so as to obtain a data set. The analysis module 203 is configured to select an optimal feature subset from the dataset using a random forest based feature importance analysis method. The establishing module 204 is configured to use the multi-layer perceptron to model the optimal feature subset, so as to obtain an oilfield fracturing pump fault diagnosis model.
Based on the same inventive concept, embodiments of the present disclosure further provide an electronic device including:
one or more processors;
and a storage unit configured to store one or more programs that, when executed by the one or more processors, enable the one or more processors to implement the method according to the foregoing description.
Based on the same inventive concept, embodiments of the present disclosure also provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, is capable of implementing the method according to the preceding description.
The computer readable medium may be any apparatus, device, or system of the present invention or may exist alone.
Wherein the computer readable storage medium may be any tangible medium that can contain, or store a program that can be an electronic, magnetic, optical, electromagnetic, infrared, semiconductor system, apparatus, device, more specific examples of which include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, an optical fiber, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.
The computer-readable storage medium may also include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein, specific examples of which include, but are not limited to, electromagnetic signals, optical signals, or any suitable combination thereof.
It is to be understood that the above implementations are merely exemplary implementations employed to illustrate the principles of the disclosed embodiments, which are not limited thereto. Various modifications and improvements may be made by those skilled in the art without departing from the spirit and substance of the embodiments of the disclosure, and these modifications and improvements are also considered to be within the scope of the embodiments of the disclosure.
Claims (10)
1. The method for establishing the fault diagnosis model of the valve body of the oilfield fracturing pump is characterized by comprising the following steps of:
collecting a plurality of groups of continuous operation data from the normal history to the fault of the oilfield fracturing pump, and dividing the plurality of groups of continuous operation data into a plurality of original samples by using a sliding window method;
extracting time domain features, frequency domain features and other mechanism features in each original sample by a mechanism analysis means, and adding the time domain features, the frequency domain features and the other mechanism features into the original sample to obtain a data set;
selecting an optimal feature subset from the dataset using a random forest based feature importance analysis method;
and modeling the optimal feature subset by using a multi-layer perceptron to obtain the fault diagnosis model of the oilfield fracturing pump.
2. The method of claim 1, wherein the collecting the historical normal to failure sets of continuous operational data for the oilfield fracturing pump and dividing the sets of continuous operational data into a plurality of raw samples using a sliding window method comprises:
collecting a plurality of groups of continuous operation data p= { p from history of normal to failure of the oilfield fracturing pump 1 ,p 2 ,...,p n Dividing the fracturing pump operation data into a plurality of original samples by using a sliding window method; wherein,
setting the sliding window size as L, and setting the sliding step length as s1 and s2, s1 during normal and fault respectively>s2, input x of the ith sample i The method comprises the following steps:
x i ＝[p is+1 ,p is+2 ,...,p is+L ]
wherein p is i The ith value in the fracturing pump outlet pressure curve is represented, and s represents the sliding step length; if the input x of the ith sample i Including fault data, the output y of the ith sample i Sliding step s=s2; otherwise y i =0, sliding step s=s1.
3. The method according to claim 1 or 2, wherein the means for extracting the time domain features, frequency domain features and other mechanism features in each of the original samples by means of mechanism analysis and adding the time domain features, frequency domain features and other mechanism features to the original samples comprises:
in the time domain feature calculation step, the time domain features include a mean value u and a variance sigma 2 The range R, kurtosis K, skewness S, waveform factor Rf and peak factor Cf are calculated as follows:
R＝p max -p min
wherein p is i Representing the value of the ith in sample x, p max And p min Respectively representing the maximum value and the minimum value of the corresponding characteristics of the sample x, wherein L represents the length of a sliding window;
and/or the number of the groups of groups,
in the frequency domain feature calculation step, performing fourier transform on the discrete sample x to obtain a frequency spectrum f corresponding to each input feature and a power spectrum density P corresponding to each input feature; the frequency domain features comprise a spectrum mean value up, a spectrum kurtosis Sp and a spectrum skewness Kp, and the calculation formula is as follows:
wherein P is i Representing f in the ith frequency i A corresponding power spectral density, k representing the length of the spectrum f;
and/or the number of the groups of groups,
in the calculating step of other mechanism characteristics, main characteristic signals of the oilfield split-pressure pump are pressure signals, and the mechanism characteristics comprise differential pressure of an inlet and an outlet of the split-pressure pump and periods of different pressure signals.
4. The method according to claim 1 or 2, wherein the selecting an optimal feature subset from the dataset using a random forest based feature importance analysis method comprises:
step 1, dividing the data set into a training set and a verification set;
step 2, modeling all the features on the training set by using a random forest algorithm to obtain importance scores score of each feature;
and 3, evaluating the classification effect of the fault category of the random forest model on the verification set by using a classification index F1_score, wherein the calculation formula is as follows:
wherein, for the fault samples, TP represents the correct number of the fault sample classification results, FP represents the number of the normal samples classified as the fault samples, FN represents the number of the normal samples classified as the fault samples; the larger the F1_score is, the better the classification effect of the model on the fault sample is;
step 4, eliminating the features with the lowest feature importance, and modeling a new feature subset by using a random forest algorithm to obtain an importance score of the remaining features;
step 5, repeating the step 3 and the step 4 until the number of the features in the feature subset is 1;
and 6, selecting the feature subset with the largest fault sample classification index F1_score as the optimal feature subset best_feature.
5. The method of claim 4, wherein modeling all features on the training set using a random forest algorithm results in an importance score for each feature, comprising:
calculating importance score of each feature by using a random forest algorithm, and assuming that the random forest comprises M base learners, wherein the number of input features is F; the calculation mode of the coefficient of the basis is as follows:
wherein C represents the number of categories, p, in the dataset c Representing the proportion of class c samples in the node; if the node uses the ith input feature f i Then the change in the coefficient of the kunit around node q is:
wherein,representing the ith feature f i The coefficient of the Kernel score, gini, at node q q1 And Gini q2 Two branch nodes respectively representing a node q;
if the feature f i The set of nodes appearing in decision tree m is Q, feature f i The feature importance in decision tree m is:
feature fi i Feature importance score VIM i The calculation formula of (2) is as follows:
the importance score set for all features is score= { VIM i },i＝1,2,3,...,F。
6. The method according to claim 1 or 2, wherein modeling the optimal feature subset using a multi-layer perceptron to obtain an oilfield fracturing pump fault diagnosis model comprises:
modeling the optimal feature subset by using a multi-layer perceptron MLP to obtain an oilfield split pump fault detection Model:
y＝Model(best_feature)
the fault detection Model consists of a plurality of fully connected layers, the dimension of the last fully connected layer is 2, and the activation function of each fully connected layer is a Softmax function, which is defined as follows:
wherein x is i Representing the output of the ith neuron node, c representing the number of neurons of the next layer;
wherein, the loss function of the fault detection Model uses F1_score, which is defined as follows:
wherein TP represents the correct number of classification results of the fault samples, FP represents the number of classification of the normal samples as the fault samples, FN represents the number of classification of the fault samples as the normal samples; the larger f1_score indicates the better the classification of the fault samples by the model.
7. A method for diagnosing faults of a valve body of an oilfield fracturing pump, which is characterized by comprising the following steps:
collecting actual operation data of an oilfield fracturing pump;
inputting actual operation data of the oilfield fracturing pump into an oilfield fracturing pump fault diagnosis model to obtain a diagnosis result of the oilfield fracturing pump; wherein the oilfield fracturing pump fault diagnosis model is obtained by adopting the building method of any one of claims 1 to 6.
8. An apparatus for establishing a fault diagnosis model of an oilfield fracturing pump valve body, which is characterized in that the apparatus comprises:
the acquisition module is used for acquiring a plurality of groups of continuous operation data from the normal history to the failure of the oilfield fracturing pump and dividing the plurality of groups of continuous operation data into a plurality of original samples by using a sliding window method;
the extraction module is used for extracting time domain features, frequency domain features and other mechanism features in each original sample through a mechanism analysis means, and adding the time domain features, the frequency domain features and the other mechanism features into the original sample to obtain a data set;
an analysis module for selecting an optimal feature subset from the dataset using a random forest based feature importance analysis method;
the building module is used for modeling the optimal feature subset by using the multi-layer perceptron to obtain an oilfield fracturing pump fault diagnosis model.
9. An electronic device, comprising:
one or more processors;
a storage unit for storing one or more programs, which when executed by the one or more processors, enable the one or more processors to implement the method of any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that,
the computer program being capable of implementing the method according to any one of claims 1 to 7 when executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311756427.XA CN117892080A (en) | 2023-12-19 | 2023-12-19 | Method and device for establishing and diagnosing fault diagnosis model of valve body of oilfield fracturing pump |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311756427.XA CN117892080A (en) | 2023-12-19 | 2023-12-19 | Method and device for establishing and diagnosing fault diagnosis model of valve body of oilfield fracturing pump |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117892080A true CN117892080A (en) | 2024-04-16 |
Family
ID=90646010
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311756427.XA Pending CN117892080A (en) | 2023-12-19 | 2023-12-19 | Method and device for establishing and diagnosing fault diagnosis model of valve body of oilfield fracturing pump |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117892080A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113807606A (en) * | 2021-10-09 | 2021-12-17 | 上海交通大学 | Intermittent process quality online prediction method capable of explaining ensemble learning |
| CN116358871A (en) * | 2023-03-29 | 2023-06-30 | 哈尔滨理工大学 | Rolling bearing weak signal composite fault diagnosis method based on graph rolling network |
| CN116577254A (en) * | 2022-09-16 | 2023-08-11 | 南京大学 | Adsorption mechanism analysis and adsorption condition positioning method and device based on machine learning |
| CN116821804A (en) * | 2023-06-02 | 2023-09-29 | 西南石油大学 | Fracturing pump check valve fault on-line monitoring method based on multi-feature fusion |
-
2023
- 2023-12-19 CN CN202311756427.XA patent/CN117892080A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113807606A (en) * | 2021-10-09 | 2021-12-17 | 上海交通大学 | Intermittent process quality online prediction method capable of explaining ensemble learning |
| CN116577254A (en) * | 2022-09-16 | 2023-08-11 | 南京大学 | Adsorption mechanism analysis and adsorption condition positioning method and device based on machine learning |
| CN116358871A (en) * | 2023-03-29 | 2023-06-30 | 哈尔滨理工大学 | Rolling bearing weak signal composite fault diagnosis method based on graph rolling network |
| CN116821804A (en) * | 2023-06-02 | 2023-09-29 | 西南石油大学 | Fracturing pump check valve fault on-line monitoring method based on multi-feature fusion |
Non-Patent Citations (1)
| Title |
|---|
| 苏楠: ""基于数据驱动的热连轧板形分析与优化"", 《工程科技Ⅰ辑》, 15 March 2022 (2022-03-15), pages 1 - 5 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111853555B (en) | Water supply pipe network blind leakage identification method based on dynamic process | |
| CN112036042B (en) | Power equipment abnormality detection method and system based on variational modal decomposition | |
| CN112097126B (en) | Water supply network pipe burst pipeline accurate identification method based on deep neural network | |
| CN105546352A (en) | Natural gas pipeline tiny leakage detection method based on sound signals | |
| CN109886433A (en) | The method of intelligent recognition city gas pipeline defect | |
| CN112668526A (en) | Bolt group loosening positioning monitoring method based on deep learning and piezoelectric active sensing | |
| CN116821804A (en) | Fracturing pump check valve fault on-line monitoring method based on multi-feature fusion | |
| CN117609836A (en) | Electromagnetic sensitivity prediction and health management method for integrated module | |
| CN114021620B (en) | BP neural network feature extraction-based electric submersible pump fault diagnosis method | |
| CN111898744A (en) | TDLAS trace gas concentration detection method based on pooled LSTM | |
| Huang et al. | Interpretable real-time monitoring of pipeline weld crack leakage based on wavelet multi-kernel network | |
| CN117633690A (en) | Rotary machine health state monitoring method and equipment based on data driving | |
| CN117037841A (en) | Acoustic signal hierarchical cavitation intensity identification method based on hierarchical transition network | |
| CN113239618A (en) | Gas pipeline detection and identification method based on acoustic signal characteristic analysis | |
| CN117892080A (en) | Method and device for establishing and diagnosing fault diagnosis model of valve body of oilfield fracturing pump | |
| CN116772122A (en) | Natural gas pipeline leakage fault diagnosis method, system, equipment and medium | |
| CN114548555B (en) | Axial flow compressor stall surge prediction method based on deep autoregressive network | |
| CN116026588A (en) | Bearing fault diagnosis and early warning method based on ensemble learning | |
| Guo et al. | A novel approach to bearing prognostics based on impulse-driven measures, improved morphological filter and practical health indicator construction | |
| CN112733381A (en) | Noise simulation method based on physical mechanism | |
| CN113051809A (en) | Virtual health factor construction method based on improved restricted Boltzmann machine | |
| CN118313052B (en) | Basement reverse construction major structure subsides monitoring system | |
| Kats et al. | Optimization of the features extraction method in cyber physical systems of monitoring energy infrastructure facilities | |
| CN118468091B (en) | Transformer fault diagnosis method and device and computer readable storage medium | |
| KR102699113B1 (en) | Device and method for assessing the target company's ESG performance to predicting the target company's ESG performance data using estimation model |
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 |