Establishment of fault diagnosis model for oilfield fracturing pump valve body and diagnosis method and device

Abstract

The disclosed embodiment provides a method and device for establishing and diagnosing a fault diagnosis model for a valve body of an oilfield fracturing pump. The method includes: collecting multiple groups of continuous operation data from normal to faulty history of an oilfield fracturing pump, and using a sliding window method to divide the multiple groups of continuous operation data into multiple original samples; extracting time domain features, frequency domain features and other mechanism features from 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 to obtain a data set; using a feature importance analysis method based on random forests to select the optimal feature subset from the data set; using a multi-layer perceptron to model the optimal feature subset to obtain an oilfield fracturing pump fault diagnosis model. Important features can be mined from a large number of mechanism features, effectively reducing the dimensionality of input features in a data-driven model, and more effectively identifying oilfield fracturing pump faults.

Description

Technical Field

[0001] The disclosed embodiments belong to the technical field of oilfield fracturing pump valve body fault diagnosis, and specifically relate to the establishment of an oilfield fracturing pump valve body fault diagnosis model and a diagnosis method and device. Background Art

[0002] The fracturing pump is the core equipment of fracturing equipment for oil and gas field development. The fracturing pump pumps the mucus into the formation at a displacement that greatly exceeds the absorption capacity of the formation, forming cracks in the formation, improving the conductivity of the oil and gas layer, and thus achieving the purpose of increasing production. At the same time, due to the high working pressure of the fracturing pump and the presence of sand and gravel in the working medium, the valve body of the fracturing pump is easily damaged, affecting the safe operation of the equipment.

[0003] In actual work, due to excessive pressure (up to 100Mpa), the main fault of the fracturing pump occurs on the two inlet and outlet valve bodies of the hydraulic cylinder. If the fault is not detected in time, on the one hand, it will cause a decrease in oil field production, and on the other hand, if the valve body is damaged for too long, it will cause damage to the valve seat and further damage to the valve body. At present, manual pump inspection is a necessary means for the safe operation of such equipment. Traditional manual diagnosis techniques include: listening, touching, measuring the vibration value of the cylinder body, and checking the real-time pressure curve (the most direct and difficult to distinguish). However, the efficiency of manual detection methods does not exceed 60%, so fault diagnosis of the check valve of the fracturing pump is of great significance.

[0004] Due to the complex and changeable working environment of the fracturing pump and the numerous vibration signal excitation sources, traditional fault diagnosis is difficult. In view of this situation, the present invention adopts a combination of data-driven and mechanical fault diagnosis methods, using the mechanical fault diagnosis method to extract the time domain information and frequency domain information of the sensor signal, and then uses data-driven methods to model, so as to achieve fault diagnosis of the check valve of the fracturing pump. Compared with traditional manual detection methods, the method based on mechanism analysis and data-driven can better explore the key characterization parameters of the fracturing pump fault, thereby more accurately identifying the potential faults of the fracturing pump. Summary of the invention

[0005] The embodiments of the present disclosure aim to solve at least one of the technical problems existing in the prior art, and provide a method and device for establishing a fault diagnosis model for an oilfield fracturing pump valve body and a diagnosis method.

[0006] In a first aspect, an embodiment of the present disclosure provides a method for establishing a fault diagnosis model for a fracturing pump valve body in an oilfield, the method comprising:

[0007] Collect multiple groups of historical continuous operation data of oilfield fracturing pumps from normal to faulty, and divide the multiple groups of continuous operation data into multiple original samples using a sliding window method;

[0008] By means of mechanism analysis, time domain features, frequency domain features and other mechanism features are extracted from each of the original samples, and the time domain features, frequency domain features and other mechanism features are added to the original samples to obtain a data set;

[0009] Using a feature importance analysis method based on random forest, selecting the optimal feature subset from the data set;

[0010] The optimal feature subset is modeled using a multi-layer perceptron to obtain an oilfield fracturing pump fault diagnosis model.

[0011] In some embodiments, the collecting of multiple sets of historical normal to fault continuous operation data of the oilfield fracturing pump and dividing the multiple sets of continuous operation data into multiple original samples using a sliding window method includes:

[0012] Collect multiple groups of continuous operation data p={p1, p2, ..., p n}, and using a sliding window method to divide the fracturing pump operation data into a plurality of said original samples; wherein,

[0013] Set the sliding window size to L, set the sliding step size to s1 and s2 for normal and fault conditions, s1>s2, and the input x of the i-th sample i for:

[0014] x i =[p is+1 ,p is+2 ,...,p is+L ]

[0015] Among them, p i represents the i-th value in the fracturing pump outlet pressure curve, s represents the sliding step length; if the input x of the i-th sample i contains fault data, then the output y of the i-th sample i =1, sliding step s = s2; otherwise y i =0, sliding step s=s1.

[0016] In some embodiments, the extracting 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, includes:

[0017] In the time domain feature calculation step, the time domain features include mean u, variance σ 2 , range R, kurtosis K, skewness S, waveform factor Rf and peak factor Cf are calculated as follows:

[0018]

[0019]

[0020] R=p max -p min

[0021]

[0022]

[0023]

[0024]

[0025] Among them, p i represents the value of the i-th sample x, p max and p min They represent the maximum and minimum values ​​of the features corresponding to sample x, respectively, and L represents the length of the sliding window;

[0026] and / or,

[0027] In the frequency domain feature calculation step, the discrete sample x is subjected to Fourier transform to obtain the spectrum f corresponding to each input feature and its corresponding power spectrum density P; the frequency domain features include the spectrum mean up, the spectrum kurtosis Sp and the spectrum skewness Kp, and the calculation formula is as follows:

[0028]

[0029]

[0030]

[0031] Among them, P i Indicates the frequency f in the ith frequency i The corresponding power spectral density, k represents the length of the spectrum f;

[0032] and / or,

[0033] In the calculation step of other mechanism characteristics, the main characteristic signals of the oilfield fracturing pump are all pressure signals, and the mechanism characteristics include the inlet and outlet pressure difference of the fracturing pump and the periods of different pressure signals.

[0034] In some embodiments, the selecting the optimal feature subset from the data set using a feature importance analysis method based on random forests comprises:

[0035] Step 1: Divide the data set into a training set and a validation set;

[0036] Step 2: Use the random forest algorithm to model all the features in the training set and obtain the importance score of each feature;

[0037] Step 3: Use the classification indicator F1_score to evaluate the classification effect of the random forest model on the fault category on the validation set. The calculation formula is as follows:

[0038]

[0039]

[0040]

[0041] Among them, for fault samples, TP represents the number of correct fault sample classification results, FP represents the number of normal samples classified as fault samples, and FN represents the number of fault samples classified as normal samples; the larger the F1_score, the better the classification effect of the model on fault samples;

[0042] Step 4: Remove the features with the lowest feature importance, and use the random forest algorithm to model the new feature subset to obtain the importance score of the remaining features;

[0043] Step 5: Repeat steps 3 and 4 until the number of features in the feature subset is 1;

[0044] Step 6: Select the feature subset with the largest fault sample classification index F1_score as the optimal feature subset best_feature.

[0045] In some embodiments, the random forest algorithm is used to model all features in the training set to obtain an importance score of each feature, including:

[0046] The importance score of each feature is calculated using the random forest algorithm. Assume that the random forest contains M base learners and the number of input features is F. The Gini coefficient of the qth node in the mth base learner is calculated as follows:

[0047]

[0048] Among them, C represents the number of categories in the data set, p c Indicates the proportion of samples of category c in the node; if the node uses the i-th input feature f i , then the change in the Gini coefficient before and after node q is:

[0049]

[0050] in, represents the i-th feature f i The Gini coefficient score at node q, Gini q1 and Gini q2 Respectively represent the two branch nodes of node q;

[0051] If the feature f i The node set that appears in the decision tree m is Q, then the feature f i The feature importance of m in the decision tree is:

[0052]

[0053] Feature i Feature Importance Score VIM i The calculation formula is as follows:

[0054]

[0055] The importance score set of all features is score = {VIM i},i=1,2,3,...,F.

[0056] In some embodiments, the method of using a multi-layer perceptron to model the optimal feature subset to obtain an oilfield fracturing pump fault diagnosis model includes:

[0057] The optimal feature subset is modeled using a multi-layer perceptron MLP to obtain the oilfield fracturing pump fault detection model Model:

[0058] y = Model (best_feature)

[0059] Among them, the fault detection model Model consists of multiple fully connected layers, and the dimension of the last fully connected layer is 2. The activation function of each fully connected layer is the Softmax function, which is defined as follows:

[0060]

[0061] Among them, x i represents the output of the i-th neuron node, and c represents the number of neurons in the next layer;

[0062] Among them, the loss function of the fault detection model Model uses F1_score, which is defined as follows:

[0063]

[0064]

[0065]

[0066] Among them, TP represents the number of correct classification results of fault samples, FP represents the number of normal samples classified as fault samples, and FN represents the number of fault samples classified as normal samples. The larger the F1_score, the better the classification effect of the model on fault samples.

[0067] In a second aspect, an embodiment of the present disclosure provides a method for diagnosing a fault of a valve body of an oilfield fracturing pump, the method comprising:

[0068] Collect actual operation data of oilfield fracturing pumps;

[0069] The actual operation data of the oilfield fracturing pump is input into the oilfield fracturing pump fault diagnosis model to obtain the diagnosis result of the oilfield fracturing pump; wherein, the oilfield fracturing pump fault diagnosis model is obtained by the establishment method described above.

[0070] In a third aspect, an embodiment of the present disclosure provides a device for establishing a fault diagnosis model for a fracturing pump valve body in an oilfield, the device comprising:

[0071] A collection module, used to collect multiple groups of historical normal to fault continuous operation data of the oilfield fracturing pump, and divide the multiple groups of continuous operation data into multiple original samples using a sliding window method;

[0072] An extraction module, used to extract time domain features, frequency domain features and other mechanism features from each of the original samples by means of mechanism analysis, and add the time domain features, frequency domain features and other mechanism features to the original samples to obtain a data set;

[0073] An analysis module, configured to select an optimal feature subset from the data set using a feature importance analysis method based on random forest;

[0074] A module is established to model the optimal feature subset using a multi-layer perceptron to obtain an oilfield fracturing pump fault diagnosis model.

[0075] In a fourth aspect, an embodiment of the present disclosure provides an electronic device, including:

[0076] one or more processors;

[0077] A storage unit is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors can implement the method described above.

[0078] In a fifth aspect, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method described above can be implemented.

[0079] The establishment of the oilfield fracturing pump valve body fault diagnosis model and the diagnosis method and device of the disclosed embodiment have the following beneficial technical effects compared with the prior art:

[0080] 1. This embodiment uses the means of mechanism analysis and the method of mechanical fault diagnosis to extract mechanism information such as time domain information and frequency domain information of sensor signals;

[0081] 2. This embodiment uses a feature importance analysis method based on random forest to select mechanism features, which can mine important features from a large number of mechanism features and effectively reduce the dimension of input features in the data-driven model;

[0082] 3. Based on the whole process of mechanism feature construction, feature selection, and fault classification model building, this embodiment combines the means of mechanism analysis with data analysis, which can more effectively identify oilfield fracturing pump faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 It is a flow chart of a method for establishing a fault diagnosis model of an oilfield fracturing pump valve body according to an embodiment of the present disclosure;

[0084] Figure 2 A schematic diagram of a water outlet pressure curve of a fracturing pump under different working conditions provided by an embodiment of the present disclosure;

[0085] Figure 3 A flow chart of an optimal feature subset selection method based on importance analysis provided in an embodiment of the present disclosure;

[0086] Figure 4 A schematic diagram of a multi-layer perceptron network structure provided in an embodiment of the present disclosure;

[0087] Figure 5 It is a structural schematic diagram of a device for establishing an oilfield fracturing pump valve body fault diagnosis model according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0088] 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 further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0089] like Figure 1 As shown, an embodiment of the present disclosure provides a method for establishing an oilfield fracturing pump valve body fault diagnosis model, the method comprising the following specific steps:

[0090] Step S101: collecting multiple groups of historical continuous operation data of oilfield fracturing pumps from normal to faulty, and dividing the multiple groups of continuous operation data into multiple original samples using a sliding window method.

[0091] Specifically, in this step, multiple groups of continuous operation data p={p1, p2, ..., p n}, and using a sliding window method to divide the fracturing pump operation data into a plurality of said original samples; wherein,

[0092] Set the sliding window size to L, set the sliding step size to s1 and s2 for normal and fault conditions, s1>s2, and the input x of the i-th sample i for:

[0093] x i =[p is+1 ,p is+2 ,...,p is+L ]

[0094] Among them, p i represents the i-th value in the fracturing pump outlet pressure curve, s represents the sliding step length; if the input x of the i-th sample i contains fault data, then the output y of the i-th sample i =1, sliding step s = s2; otherwise y i =0, sliding step s=s1.

[0095] As a specific example, this embodiment uses the real-time pressure curve of the hydraulic end (water outlet) of the fracturing pump as the judgment basis for fracturing pump fault detection, and collects the fracturing pump outlet pressure curve p from normal operation to failure through the fracturing pump test bench. Set the sliding window size to L = 30, the window interval s = 5, and decompose the outlet pressure curve into several samples. The input x of the i-th sample i As shown below:

[0096] x i =[p is+1 ,p is+2 ,...,p is+L ].

[0097] Step S102: extracting the time domain features, frequency domain features and other mechanism features from 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 to obtain a data set.

[0098] Specifically, in this step, since the difference between the outlet water pressure curves of normal samples and faulty samples is small, it is difficult to distinguish between normal and faulty samples with the naked eye, so it is difficult to directly use the original outlet water pressure curve for fault diagnosis, such as Figure 2 According to the traditional fault diagnosis experience, this embodiment extracts the corresponding time domain features and frequency domain features from the outlet water pressure curve.

[0099] Specifically, in this step, in the time domain feature calculation step, the time domain features include mean u, variance σ 2 , range R, kurtosis K, skewness S, waveform factor Rf and peak factor Cf are calculated as follows:

[0100]

[0101]

[0102] R=p max -p min

[0103]

[0104]

[0105]

[0106]

[0107] Among them, p i represents the value of the i-th sample x, p max and p min They represent the maximum and minimum values ​​of the features corresponding to sample x respectively, and L represents the length of the sliding window, which is set to 30 in this embodiment.

[0108] In the frequency domain feature calculation step, the discrete sample x is subjected to Fourier transform to obtain the spectrum f corresponding to each input feature and its corresponding power spectrum density P; the frequency domain features include the spectrum mean up, the spectrum kurtosis Sp and the spectrum skewness Kp, and the calculation formula is as follows:

[0109]

[0110]

[0111]

[0112] Among them, P i Indicates the frequency f in the ith frequency i The corresponding power spectral density, k represents the length of the spectrum f.

[0113] In the calculation step of other mechanism characteristics, the main characteristic signals of the oilfield fracturing pump are all pressure signals, and the mechanism characteristics include the inlet and outlet pressure difference of the fracturing pump and the period of different pressure signals.

[0114] Step S103: using a feature importance analysis method based on random forest to select an optimal feature subset from the data set.

[0115] Specifically, in this step, for the fracturing pump fault data set obtained in step S102, a random forest-based importance analysis method is used to perform feature selection to obtain a feature subset with the greatest importance. The steps of this feature selection method are as follows: Figure 3 As shown, the specific implementation steps in this embodiment are as follows:

[0116] Step 1: Divide the data set obtained in step S102 into a training set and a validation set, wherein the ratio of the training set to the validation set is 9:1, that is, 90% of the data is used for training and 10% of the data is used for validation.

[0117] Step 2: Use the random forest algorithm to model all the features in the training set and obtain the importance score of each feature;

[0118] Step 3: Use the classification indicator F1_score to evaluate the classification effect of the random forest model on the fault category on the validation set. The calculation formula is as follows:

[0119]

[0120]

[0121]

[0122] Among them, for fault samples, TP represents the number of correct fault sample classification results, FP represents the number of normal samples classified as fault samples, and FN represents the number of fault samples classified as normal samples; the larger the F1_score, the better the classification effect of the model on fault samples;

[0123] Step 4: Remove the features with the lowest feature importance, and use the random forest algorithm to model the new feature subset to obtain the importance score of the remaining features;

[0124] Step 5: Repeat steps 3 and 4 until the number of features in the feature subset is 1;

[0125] Step 6: Select the feature subset with the largest fault sample classification index F1_score as the optimal feature subset best_feature.

[0126] In some embodiments, the random forest algorithm is used to model all features in the training set to obtain an importance score of each feature, including:

[0127] The importance score of each feature is calculated using the random forest algorithm. Assume that the random forest contains M base learners and the number of input features is F. The Gini coefficient of the qth node in the mth base learner is calculated as follows:

[0128]

[0129] Among them, C represents the number of categories in the data set, p c Indicates the proportion of samples of category c in the node; if the node uses the i-th input feature f i , then the change in the Gini coefficient before and after node q is:

[0130]

[0131] in, represents the i-th feature f i The Gini coefficient score at node q, Gini q1 and Gini q2 Respectively represent the two branch nodes of node q;

[0132] If the feature f i The node set that appears in the decision tree m is Q, then the feature f i The feature importance of m in the decision tree is:

[0133]

[0134] Feature i Feature Importance Score VIM i The calculation formula is as follows:

[0135]

[0136] The importance score set of all features is score = {VIM i},i=1,2,3,...,F.

[0137] According to the above steps, the optimal feature subset selected in this implementation is: variance σ 2 , kurtosis K, skewness S and waveform factor Rf, a total of four, the classification accuracy index F1_score = 0.98 of the random forest model.

[0138] Step S104: Use a multi-layer perceptron to model the optimal feature subset to obtain an oilfield fracturing pump fault diagnosis model.

[0139] Specifically, in this step, a multi-layer perceptron MLP is used to model the optimal feature subset, where the number of network layers is set to 8, 16, 8, and 2, respectively, such as Figure 4As shown in the figure, the activation function is set to the softmax function and the loss function is the F1_score function. The formula of the softmax function is as follows:

[0140]

[0141] Among them, x i represents the output of the i-th neuron node, c represents the number of neurons in the next layer, and in this embodiment, c=2;

[0142] Among them, the loss function of the fault detection model Model uses F1_score, which is defined as follows:

[0143]

[0144]

[0145]

[0146] Among them, TP represents the number of correct classification results of fault samples, FP represents the number of normal samples classified as fault samples, and FN represents the number of fault samples classified as normal samples. The larger the F1_score, the better the classification effect of the model on fault samples.

[0147] Finally, in order to verify the effectiveness of the model, this embodiment compares the classification effects under different classification algorithms and different feature processing methods, and the evaluation index is F1_score, as shown in Table 1 below:

[0148] Table 1 Errors in temperature field estimation at different mass flow rates

[0149]

[0150] Based on the above technical solution, compared with the prior art, the oilfield fracturing pump valve body fault diagnosis model of this embodiment has the following beneficial technical effects:

[0151] 1. This embodiment uses the means of mechanism analysis and the method of mechanical fault diagnosis to extract mechanism information such as time domain information and frequency domain information of sensor signals;

[0152] 2. This embodiment uses a feature importance analysis method based on random forest to select mechanism features, which can mine important features from a large number of mechanism features and effectively reduce the dimension of input features in the data-driven model;

[0153] 3. Based on the whole process of mechanism feature construction, feature selection, and fault classification model building, this embodiment combines the means of mechanism analysis with data analysis, which can more effectively identify oilfield fracturing pump faults.

[0154] Based on the same inventive concept, an embodiment of the present disclosure further provides a method for diagnosing a fault of a valve body of an oilfield fracturing pump, the method comprising:

[0155] Collect actual operation data of oilfield fracturing pumps;

[0156] The actual operation data of the oilfield fracturing pump is input into the oilfield fracturing pump fault diagnosis model to obtain the diagnosis result of the oilfield fracturing pump; wherein, the oilfield fracturing pump fault diagnosis model is obtained by the establishment method described in the above text, and the details can be referred to the relevant records in the above text, which will not be repeated here.

[0157] Based on the same inventive concept, Figure 5 As shown, the embodiment of the present disclosure also provides a device for establishing a fault diagnosis model of a fracturing pump valve body in an oilfield, which is applicable to the establishment method described above. For details, please refer to the relevant records in the above text, and no further description is given here. The device includes a collection module 201, an extraction module 202, an analysis module 203 and an establishment module 204.

[0158] The acquisition module 201 is used to collect multiple groups of continuous operation data from normal to faulty history of oilfield fracturing pumps, and divide the multiple groups of continuous operation data into multiple original samples using a sliding window method. The extraction module 202 is used to extract the time domain features, frequency domain features and other mechanism features in each of the original samples by means of mechanism analysis, and add the time domain features, frequency domain features and other mechanism features to the original samples to obtain a data set. The analysis module 203 is used to select the optimal feature subset from the data set using a feature importance analysis method based on random forests. The establishment module 204 is used to model the optimal feature subset using a multi-layer perceptron to obtain an oilfield fracturing pump fault diagnosis model.

[0159] Based on the same inventive concept, an embodiment of the present disclosure further provides an electronic device, including:

[0160] one or more processors;

[0161] A storage unit is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors can implement the method described above.

[0162] Based on the same inventive concept, an embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method described above can be implemented.

[0163] The computer-readable medium may be included in the apparatus, device, or system of the present invention, or may exist independently.

[0164] Among them, computer-readable storage media can be any tangible media that contains or stores a program, which can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment. More specific examples include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0165] The computer-readable storage medium may also include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code, specific examples of which include but are not limited to electromagnetic signals, optical signals, or any suitable combination thereof.

[0166] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the embodiments of the present disclosure, but the embodiments of the present disclosure are not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the embodiments of the present disclosure, and these modifications and improvements are also considered to be within the protection scope of the embodiments of the present disclosure.

Claims

1. A method for establishing a fault diagnosis model for an oilfield fracturing pump valve body, characterized in that: The method comprises: Collect multiple groups of historical continuous operation data of oilfield fracturing pumps from normal to faulty, and divide the multiple groups of continuous operation data into multiple original samples using a sliding window method; By means of mechanism analysis, time domain features, frequency domain features and other mechanism features are extracted from each of the original samples, and the time domain features, frequency domain features and other mechanism features are added to the original samples to obtain a data set; Using a feature importance analysis method based on random forest, selecting the optimal feature subset from the data set; The optimal feature subset is modeled using a multi-layer perceptron to obtain an oilfield fracturing pump fault diagnosis model; The extracting of time domain features, frequency domain features and other mechanism features from 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, includes: In the time domain feature calculation step, the time domain features include mean u, variance σ 2 , range R, kurtosis K, skewness S, waveform factor Rf and peak factor Cf are calculated as follows: R=p max -p min Among them, p i represents the value of the i-th sample x, p max and p min They represent the maximum and minimum values ​​of the features corresponding to sample x, respectively, and L represents the length of the sliding window; and / or, In the frequency domain feature calculation step, the discrete sample x is subjected to Fourier transform to obtain the spectrum f corresponding to each input feature and its corresponding power spectrum density P; the frequency domain features include the spectrum mean up, the spectrum kurtosis Sp and the spectrum skewness Kp, and the calculation formula is as follows: Among them, P i Indicates the frequency f in the ith frequency i The corresponding power spectral density, k represents the length of the spectrum f; and / or, In the calculation step of the other mechanism characteristics, the main characteristic signals of the oilfield fracturing pump are all pressure signals, and the other mechanism characteristics include the inlet and outlet pressure difference of the fracturing pump and the periods of different pressure signals.

2. The method according to claim 1, characterized in that The method collects multiple groups of continuous operation data of the oilfield fracturing pump from normal to faulty, and divides the multiple groups of continuous operation data into multiple original samples using a sliding window method, including: Collect multiple groups of continuous operation data p={p1, p2, ..., p n }, and using a sliding window method to divide the fracturing pump operation data into a plurality of said original samples; wherein, Set the sliding window size to L, set the sliding step size to s1 and s2 for normal and fault conditions, s1>s2, and the input x of the i-th sample i for: x i =[p is+1 ,p is+2 ,...,p is+L ] Among them, p i represents the i-th value in the fracturing pump outlet pressure curve, s represents the sliding step length; if the input x of the i-th sample i contains fault data, then the output y of the i-th sample i =1, sliding step s = s2; otherwise y i =0, sliding step s=s1.

3. The method according to claim 1 or 2, characterized in that: The method of using a feature importance analysis method based on random forest to select an optimal feature subset from the data set includes: Step 1: Divide the data set into a training set and a validation set; Step 2: Use the random forest algorithm to model all the features in the training set and obtain the importance score of each feature; Step 3: Use the classification indicator F1_score to evaluate the classification effect of the random forest model on the fault category on the validation set. The calculation formula is as follows: Among them, for fault samples, TP represents the number of correct fault sample classification results, FP represents the number of normal samples classified as fault samples, and FN represents the number of fault samples classified as normal samples; the larger the F1_score, the better the classification effect of the model on fault samples; Step 4: Remove the features with the lowest feature importance, and use the random forest algorithm to model the new feature subset to obtain the importance score of the remaining features; Step 5: Repeat steps 3 and 4 until the number of features in the feature subset is 1; Step 6: Select the feature subset with the largest fault sample classification index F1_score as the optimal feature subset best_feature.

4. The method according to claim 3, characterized in that The random forest algorithm is used to model all features in the training set to obtain the importance score of each feature, including: The importance score of each feature is calculated using the random forest algorithm. Assume that the random forest contains M base learners and the number of input features is F. The Gini coefficient of the qth node in the mth base learner is calculated as follows: Among them, C represents the number of categories in the data set, p c Indicates the proportion of samples of category c in the node; if the node uses the i-th input feature f i , then the change in the Gini coefficient before and after node q is: in, represents the i-th feature f i The Gini coefficient score at node q, Gini q1 and Gini q2 Respectively represent the two branch nodes of node q; If the feature f i The node set that appears in the decision tree m is Q, then the feature f i The feature importance of m in the decision tree is: Feature i Feature Importance Score VIM i The calculation formula is as follows: The importance score set of all features is score = {VIM i },i=1,2,3,...,F.

5. The method according to claim 1 or 2, characterized in that: The method uses a multi-layer perceptron to model the optimal feature subset to obtain an oilfield fracturing pump fault diagnosis model, including: The optimal feature subset is modeled using a multi-layer perceptron MLP to obtain the oilfield fracturing pump fault detection model Model: y = Model (best_feature) Among them, the fault detection model Model consists of multiple fully connected layers, and the dimension of the last fully connected layer is 2. The activation function of each fully connected layer is the Softmax function, which is defined as follows: Among them, x i represents the output of the i-th neuron node, and c represents the number of neurons in the next layer; Among them, the loss function of the fault detection model Model uses F1_score, which is defined as follows: Among them, TP represents the number of correct classification results of fault samples, FP represents the number of normal samples classified as fault samples, and FN represents the number of fault samples classified as normal samples. The larger the F1_score, the better the classification effect of the model on fault samples.

6. A method for diagnosing oilfield fracturing pump valve body faults, characterized in that: The method comprises: Collect actual operation data of oilfield fracturing pumps; The actual operation data of the oilfield fracturing pump is input into the oilfield fracturing pump fault diagnosis model to obtain the diagnosis result of the oilfield fracturing pump; wherein the oilfield fracturing pump fault diagnosis model is obtained by the establishment method described in any one of claims 1 to 5.

7. A device for establishing a fault diagnosis model for a fracturing pump valve in an oil field, characterized in that: The device comprises: A collection module, used to collect multiple groups of historical normal to fault continuous operation data of the oilfield fracturing pump, and divide the multiple groups of continuous operation data into multiple original samples using a sliding window method; An extraction module, used to extract time domain features, frequency domain features and other mechanism features from each of the original samples by means of mechanism analysis, and add the time domain features, frequency domain features and other mechanism features to the original samples to obtain a data set; An analysis module, configured to select an optimal feature subset from the data set using a feature importance analysis method based on random forest; Establish a module for modeling the optimal feature subset using a multi-layer perceptron to obtain a fault diagnosis model for oilfield fracturing pumps; The extraction module is specifically used for: In the time domain feature calculation, the time domain features include mean u, variance σ 2 , range R, kurtosis K, skewness S, waveform factor Rf and peak factor Cf are calculated as follows: R=p max -p min Among them, p i represents the value of the i-th sample x, p max and p min They represent the maximum and minimum values ​​of the features corresponding to sample x, respectively, and L represents the length of the sliding window; and / or, In the frequency domain feature calculation, the discrete sample x is Fourier transformed to obtain the spectrum f corresponding to each input feature and its corresponding power spectrum density P; the frequency domain features include the spectrum mean up, the spectrum kurtosis Sp and the spectrum skewness Kp, and the calculation formula is as follows: Among them, P i Indicates the frequency f in the ith frequency i The corresponding power spectral density, k represents the length of the spectrum f; and / or, In the calculation of the other mechanism characteristics, the main characteristic signals of the oilfield fracturing pump are all pressure signals, and the other mechanism characteristics include the inlet and outlet pressure difference of the fracturing pump and the periods of different pressure signals.

8. An electronic device, characterized in that: include: one or more processors; A storage unit, used to store one or more programs, which, when executed by the one or more processors, enable the one or more processors to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 can be implemented.

Images