Fault detection method and device, program product and storage medium

Abstract

The embodiment of the invention discloses a fault detection method and device, a program product and a storage medium, and the method comprises the steps: obtaining the operation data of an oil delivery pump, carrying out the time-frequency feature extraction of the operation data according to a predetermined mechanism model, and obtaining the first feature of the operation data, the first feature comprises a time domain feature, a frequency domain feature and a time-frequency domain feature of the operation data; performing data feature extraction on the operation data based on a pre-trained neural network model to obtain a second feature of the operation data; and obtaining a fault detection result of the operation data based on the first feature, the second feature and a predetermined confidence decision mode, and displaying the fault detection result to a user. According to the method, the features extracted by the mechanism model and the features extracted by the neural network model are fused, the advantages of the two models can be fully utilized, the robustness of fault detection is enhanced, and misjudgment caused by defects of a single model is reduced.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing, and in particular, to a fault detection method, device, program product, and storage medium. Background Art

[0002] During the production operation of long-distance oil and gas pipelines, oil transfer pumps are the core equipment to ensure the stable and efficient transportation of oil and gas. Therefore, it is crucial to monitor the oil transfer pumps in real time and diagnose faults. During the operation of each key component of the oil transfer pump, damage may be caused for various reasons. For example, improper assembly, poor lubrication, intrusion of moisture and foreign objects, corrosion, and overload may all lead to premature damage of the bearings.

[0003] Most of the current oil transfer pump fault detection methods are based on the historical operation data of the oil transfer pump to train a detection model for user fault detection based on deep learning or machine learning, and use the trained detection model to detect faults in the oil transfer pump. However, this method has a complex calculation method and poor interpretability of the fault diagnosis results, reducing the fault detection efficiency of the oil pipeline. Summary of the Invention

[0004] The embodiments of the present invention provide a fault detection method, device, program product, and storage medium, which can reduce the complexity of the calculation process while ensuring the accuracy of the fault detection results of the oil pipeline, and at the same time make the fault diagnosis have a certain interpretability, improving the fault detection efficiency of the oil pipeline.

[0005] In a first aspect, the embodiments of the present invention provide a fault detection method, including:

[0006] Obtain the operation data of the oil transfer pump,

[0007] Extract time-frequency features from the operation data according to a pre-determined mechanism model to obtain the first feature of the operation data, where the first feature includes the time-domain feature, frequency-domain feature, and time-frequency domain feature of the operation data;

[0008] Extract data features from the operation data based on a pre-trained neural network model to obtain the second feature of the operation data;

[0009] Obtain the fault detection result of the operation data based on the first feature, the second feature, and a pre-determined confidence decision method, and display the fault detection result to the user.

[0010] In a second aspect, the embodiments of the present invention provide a fault detection device, and the device includes:

[0011] A data acquisition module, configured to obtain the operation data of the oil transfer pump,

[0012] The first processing module is configured to extract time-frequency features from the operation data according to a pre-determined mechanism model, so as to obtain first features of the operation data, where the first features include time-domain features, frequency-domain features, and time-frequency domain features of the operation data;

[0013] The second processing module is configured to extract data features from the operation data based on a pre-trained neural network model, so as to obtain second features of the operation data;

[0014] The result determination module is configured to obtain a fault detection result of the operation data based on the first features, the second features, and a pre-determined confidence decision method, and display the fault detection result to the user.

[0015] In a third aspect, an embodiment of the present invention further provides an electronic device, where the electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the program, it implements the fault detection method as described in any one of the embodiments of the present invention.

[0016] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the fault detection method as described in any one of the embodiments of the present invention.

[0017] In a fifth aspect, an embodiment of the present invention provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the fault detection method as described in any one of the embodiments of the present invention.

[0018] In the embodiment of the present invention, the operation data of the oil transfer pump is obtained, time-frequency features are extracted from the operation data according to a pre-determined mechanism model, so as to obtain first features of the operation data, where the first features include time-domain features, frequency-domain features, and time-frequency domain features of the operation data; data features are extracted from the operation data based on a pre-trained neural network model, so as to obtain second features of the operation data; a fault detection result of the operation data is obtained based on the first features, the second features, and a pre-determined confidence decision method, and the fault detection result is displayed to the user. The method of the embodiment of the present invention can obtain the time-domain, frequency-domain, and time-frequency domain features of the oil transfer pump during operation through a mechanism model. The process of feature extraction by the mechanism model is simple and has good interpretability, which helps users understand the basis of the fault detection result. By automatically extracting complex features in the operation data through a neural network model, the accuracy of fault detection can be improved. Fusing the features extracted by the mechanism model and the features extracted by the neural network model can make full use of the advantages of the two models, reduce the computational complexity while enhancing the robustness of fault detection, and also avoid misjudgment caused by the defects of a single model. Description of the Drawings

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for the embodiments will be briefly introduced below. It should be understood that the following accompanying drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related accompanying drawings can also be obtained based on these drawings.

[0020] Figure 1 The first flowchart of a fault detection method provided by an embodiment of the present invention;

[0021] Figure 2 The schematic diagram of the time-domain characteristics of the operation data provided by an embodiment of the present invention;

[0022] Figure 3 The schematic diagram of performing EMD decomposition on the operation data provided by an embodiment of the present invention;

[0023] Figure 4 The second flowchart of a fault detection method provided by an embodiment of the present invention;

[0024] Figure 5 The structural schematic diagram of performing fault detection on an oil transfer pump provided by an embodiment of the present invention;

[0025] Figure 6 The structural schematic diagram of a fault detection device provided by an embodiment of the present invention;

[0026] Figure 7 The structural schematic diagram of an electronic device provided by an embodiment of the present invention. Specific Embodiments

[0027] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Additionally, it should be noted that for the sake of convenience of description, only parts related to the present invention rather than all structures are shown in the accompanying drawings.

[0028] Figure 1 The first flowchart of a fault detection method provided by an embodiment of the present invention. The method of the embodiment of the present invention ensures the accuracy of the fault detection result of the oil pipeline while reducing the complexity of the calculation process, making the fault diagnosis have a certain interpretability, and improving the fault detection efficiency of the oil pipeline. This method can be executed by a fault detection device provided by an embodiment of the present invention, and the device can be implemented in a software and / or hardware manner. The following embodiments will be described by taking the integration of the device in an electronic device as an example. The electronic device can be a server or a computer device, etc. Referring to Figure 1 , the method can specifically include the following steps:

[0029] Step 101: Obtain the operation data of the oil transfer pump.

[0030] Among them, the oil transfer pump is a mechanical device used to transport oil or petroleum products (such as crude oil, gasoline, diesel, and lubricating oil, etc.). The operation data of the oil transfer pump is the data generated when the oil transfer pump is in the working state. When a fault detection of the oil transfer pump is required, the server can obtain the operation data of the oil transfer pump in real time or at a preset time interval. The operation data of the oil transfer pump at least includes vibration signals, and may also include the inlet pressure, outlet pressure, operating temperature, operating state of the motor, and the flow rate of the oil transfer pump, etc.

[0031] In an optional implementation manner, various sensors are included on the oil transfer pump, such as vibration sensors, pressure sensors, temperature sensors, current sensors, voltage sensors, and flow sensors. When a fault detection of the oil transfer pump is required, the vibration signal of the oil transfer pump is collected by the vibration sensor, and the inlet and outlet pressures of the oil transfer pump are detected by the pressure sensor. The operating temperature of the device is collected by the temperature sensor. The operating state of the monitored motor is collected by the current sensor and the voltage sensor. The flow rate of the oil transfer pump is collected by the flow sensor.

[0032] Step 102: Extract the time-frequency characteristics of the operation data according to a pre-determined mechanism model to obtain the first characteristics of the operation data.

[0033] Among them, the first characteristics include the time-domain characteristics, frequency-domain characteristics, and time-frequency domain characteristics of the operation data. The mechanism model is a pre-established mathematical model based on physical, chemical, or engineering principles, used to analyze the operation data of the oil transfer pump, so as to obtain the time-domain characteristics, frequency-domain characteristics, time-frequency domain characteristics, etc. of the operation data. The time-domain characteristics are statistical characteristics extracted from time-series data, used to describe the amplitude change, fluctuation, etc. of the vibration signal. Exemplarily, the time-domain characteristics of the operation data of the oil transfer pump are shown in Table 1 below:

[0034]

[0035]

[0036] Table 1

[0037] Exemplarily, Figure 2 is the schematic diagram of the time-domain characteristics of the operation data provided by the embodiment of the present invention. Figure 2 The horizontal axis in represents the time domain, and the vertical axis represents the specific value. The frequency-domain characteristics are obtained by performing a Fourier transform on the vibration signal, used to describe the frequency distribution and energy distribution of the vibration signal. Through the frequency-domain characteristics, the spectrum information of the operation data can be obtained, and the spectrum information can reflect whether a certain type of fault has occurred in the oil transfer pump. Exemplarily, the characteristic frequency data of common faults of the oil transfer pump are shown in Table 2 below:

[0038] Fault Name Frequency Feature Vibration Stability Poor Rotor Alignment 2 times the fundamental frequency, often accompanied by 1 and 3 times the fundamental frequency and higher harmonics Stable Rotor Unbalance Concentrated 1 times the fundamental frequency Temperature Rotor Bending 1 times the fundamental frequency, often accompanied by 2 times the fundamental frequency Stable Bearing Oil Film Whirl 0.42 - 0.48 times the fundamental frequency, often accompanied by 1 times the fundamental frequency Unstable Bearing Oil Film Oscillation 0.42 - 0.48 times the fundamental frequency Unstable Foundation Distortion 1 and 2 times the fundamental frequency accompanied by 3 times the fundamental frequency and higher harmonics Stable Assembly Looseness 0.3 - 0.5 times the fundamental frequency accompanied by 1 times the fundamental frequency and higher harmonics Unstable

[0039] Table 2

[0040] The time-frequency domain feature is a feature that combines the information of time and frequency. In this solution, the operation data can be decomposed by Empirical Mode Decomposition (EMD) to obtain Intrinsic Mode Functions (IMFs). EMD is an adaptive signal processing method used to decompose a complex data set into a series of simple, locally characteristic intrinsic mode functions. Exemplarily, Figure 3 is a schematic diagram of performing EMD decomposition on the operation data provided by an embodiment of the present invention. As Figure 3 shown, each sub-graph (IMF1 to IMF5) represents an intrinsic mode function. The horizontal axis of each sub-graph represents the time series, that is, the acquisition time points of the operation data. The vertical axis represents the amplitude or intensity of the signal.

[0041] Specifically, after obtaining the operation data of the oil transfer pump, the server can perform data preprocessing on the operation data, such as deleting abnormal data (such as outliers, missing values, and error values). Remove the noise in the operation data through wavelet transform or Fast Fourier Transform (FFT), and perform normalization processing on the operation data after removing the noise to obtain operation data with a unified scale. After preprocessing the operation data, input the preprocessed operation data into the mechanism model. The mechanism model can analyze the time domain features of the operation data according to the time series relationship of the operation data. Convert the vibration signal from the time domain to the frequency domain through Fourier transform to obtain the frequency domain features of the operation data. Further, the mechanism model can extract the time-frequency domain features of the operation data through a pre-set time-frequency domain feature extraction algorithm. For example, use the short-time Fourier transform to divide the signal into multiple short time periods and perform Fourier transform respectively to obtain the time-frequency domain features. Use wavelet transform to perform multi-scale analysis on the signal through wavelet basis functions to obtain the time-frequency domain features. Use wavelet packet decomposition to extract the energy features of the signal in different frequency ranges to obtain the time-frequency domain features.

[0042] Step 103: Extract data features from the operation data based on a pre-trained neural network model to obtain the second feature of the operation data.

[0043] Among them, the neural network model is a pre-trained model for rich feature extraction of operation data. The neural network model includes convolutional neural network, recurrent neural network, attention mechanism, and graph neural network, etc. The neural network model in this solution can be a convolutional neural network. In an optional implementation manner, after obtaining the operation data, the operation data is input into the neural network model, and local feature extraction is performed on the operation data through the input layer and convolutional layer of the neural network model to obtain the second candidate feature of the operation data; the pooling layer of the neural network model is used to perform feature dimensionality reduction on the second candidate feature, and the fully connected layer and output layer of the neural network model are used to process the second candidate feature after feature dimensionality reduction to obtain the second feature of the operation data.

[0044] The neural network model in this solution can also be a recurrent neural network. In an optional implementation manner, after obtaining the operation data, the data of each time step in the operation data can be input into the pre-trained recurrent neural network, and the time step data is updated through the hidden layer of the recurrent neural network. The updated data is input into the output layer, and the output layer generates output information according to the hidden state of the hidden layer, and the second feature is obtained according to the output information.

[0045] Step 104: Obtain the fault detection result of the operation data based on the first feature, the second feature, and the pre-determined confidence decision method, and display the fault detection result to the user.

[0046] Among them, the first feature is the feature of the operation data output by the mechanism model, and the second feature is the feature of the operation data output by the neural network model. The confidence decision method is pre-determined by the server and is used to instruct the server to determine the final fault detection result according to the first feature and the second feature. The fault detection result includes whether a fault occurs, the fault type, and the fault data, etc.

[0047] In an optional implementation manner, after obtaining the first feature and the second feature, the first feature and the second feature can be weighted and fused to form a comprehensive feature. For example, through a feature pyramid network combined with a convolutional neural network and a long short-term memory network layer, the first feature and the second feature are processed simultaneously to obtain the comprehensive feature. After obtaining the comprehensive feature, the comprehensive feature can be input into a classifier (such as a fully connected layer or a support vector machine) for fault classification to obtain the confidence of various types of faults corresponding to the operation data, and the final fault detection result is determined according to the confidence.

[0048] In an alternative embodiment, after obtaining the first feature and the second feature, the first feature and the second feature may be subjected to feature fusion to obtain a fusion feature set; for each type of fusion feature in the fusion feature set, the current type of fusion feature is respectively input into a mechanism model and a neural network model to obtain first diagnostic data output by the mechanism model and second diagnostic data output by the neural network model; the first diagnostic data includes a first diagnostic result, a first similarity, and a first confidence level, and the second diagnostic data includes a second diagnostic result, a second similarity, and a second confidence level; when the first diagnostic result is the same as the second diagnostic result, the first diagnostic result or the second diagnostic result is determined as the fault detection result of the current type of fusion feature; when the first diagnostic result and the second diagnostic result are different, the fault detection result of the current type of fusion feature is determined based on the first confidence level, the first similarity, the second confidence level, and the second similarity. The fault detection result of the operation data is determined based on the fault detection results of all types of fusion features.

[0049] After obtaining the fault detection result, a fault detection report of the oil transfer pump may be generated according to the fault detection result. The fault detection report includes whether the oil transfer pump has a fault, the type of fault that occurred, the location where the fault occurred, the severity of the fault, and a recommended solution, etc. After obtaining the fault detection report, the fault detection report may be sent to the user, and the user may process the oil transfer pump in a timely manner when the oil transfer pump has a fault according to the fault detection report.

[0050] The technical solution of this embodiment is to obtain the operation data of the oil transfer pump; perform time-frequency feature extraction on the operation data according to a pre-determined mechanism model to obtain the first feature of the operation data, where the first feature includes the time-domain feature, the frequency-domain feature, and the time-frequency domain feature of the operation data; perform data feature extraction on the operation data based on a pre-trained neural network model to obtain the second feature of the operation data; obtain the fault detection result of the operation data based on the first feature, the second feature, and a pre-determined confidence decision method, and display the fault detection result to the user. The technical solution of this embodiment can obtain the time-domain, frequency-domain, and time-frequency domain features during the operation of the oil transfer pump through the mechanism model. The process of feature extraction by the mechanism model is simple and has good interpretability, which helps users understand the basis of the fault detection result. Automatically extract complex features in the operation data through the neural network model, which can improve the accuracy of fault detection. Fusing the features extracted by the mechanism model and the features extracted by the neural network model can make full use of the advantages of the two models, reduce the computational complexity while enhancing the robustness of fault detection, and can also avoid misjudgment caused by the defects of a single model.

[0051] Figure 4 This is the second flowchart of a fault detection method provided by an embodiment of the present invention. This embodiment is a refinement based on the above embodiment. The specific method may be asFigure 4 As shown, the method may include the following steps:

[0052] Step 401, obtain the operation data of the oil transfer pump.

[0053] Step 402, perform time-frequency feature extraction on the operation data according to a pre-determined mechanism model to obtain the first feature of the operation data.

[0054] Among them, the first feature includes the time-domain feature, frequency-domain feature, and time-frequency domain feature of the operation data.

[0055] Step 403, input the operation data into the neural network model, and perform local feature extraction on the operation data through the input layer and convolutional layer of the neural network model to obtain the second candidate feature of the operation data.

[0056] Among them, the neural network model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.

[0057] Specifically, after obtaining the operation data, preprocess the operation data, thereby converting the operation data into one-dimensional time series data, and set the number of nodes in the input layer of the convolutional neural network to the dimension of the one-dimensional time series data. Process the one-dimensional time series data through the input layer, and input the processed data into the convolutional layer. The convolutional layer can perform a convolution operation on the processed data through a sliding window to extract the local features of the data and obtain a local feature map, that is, the second candidate feature of the operation data.

[0058] Step 404, perform feature dimensionality reduction on the second candidate feature through the pooling layer of the neural network model, and use the fully connected layer and output layer of the neural network model to process the second candidate feature after feature dimensionality reduction to obtain the second feature of the operation data.

[0059] Among them, the pooling layer is used to reduce the dimension of the feature map while retaining important feature information, thereby improving the computational efficiency and generalization ability of the model. In this solution, a non-linear activation function is added between the convolutional layer and the pooling layer to introduce non-linear factors, thereby enhancing the expression ability and fitting ability of the model. After obtaining the second candidate feature, activate the second candidate feature according to the non-linear activation function, and input the activated second candidate feature into the pooling layer of the convolutional neural network. Perform a pooling operation (such as max pooling and average pooling) on the second candidate feature through the pooling layer. Flatten the feature map obtained after the pooling operation and connect it to the fully connected layer (the number of nodes in the fully connected layer can be the dimension of the one-dimensional time series data), map the feature map to the output space through the fully connected layer to obtain an output feature vector. Input the output feature vector into the output layer, and output the second feature through the output layer.

[0060] Step 405, perform feature fusion on the first feature and the second feature to obtain a fusion feature set.

[0061] Feature fusion is the integration of features extracted from different sources or different models to form a more comprehensive and representative feature representation. The fused feature set includes all fused features obtained by fusion of the first feature and the second feature. After obtaining the first feature and the second feature, the first feature and the second feature are fused to obtain a fused feature set. In this solution, optionally, the first feature and the second feature are fused to obtain a fused feature set, including the following steps A1-A2:

[0062] Step A1: Perform weighted feature fusion on the first feature and the second feature according to each pre-set weight combination, and obtain feature fusion results of each weight combination.

[0063] Among them, each weight combination is a weight combination predetermined by the server based on domain big data and historical operation data of the oil pump. The feature fusion result of each weight combination includes the fusion feature obtained after the server fuses the first feature and the second feature according to each weight combination. Specifically, after obtaining the first feature and the second feature, the first feature and the second feature are weightedly fused using each pre-set weight combination to obtain the fusion feature corresponding to each weight combination. Exemplarily, a series of pre-set weight combinations are α (α = 0, 0.1, 0.2, ..., 1.0). For each weight α, the fusion feature F is: F fusion = α × F1 + (1-α) × F2, where F1 represents the first feature and F2 represents the second feature.

[0064] Step A2: Evaluate the fusion results of each feature based on the pre-set performance indicators, and determine the target weight combination according to the evaluation results; perform weighted feature fusion on the first feature and the second feature based on the target weight combination to obtain a fused feature set.

[0065] Among them, the performance metrics are used to evaluate the rationality and accuracy of each fused feature on the mechanism model or neural network model. The performance metrics include accuracy, recall, and the area under the receiver operating characteristic curve. The target weight combination is the weight combination with the best evaluation result among all weight combinations. After obtaining the fused feature results of each feature, evaluate the fused features generated by each weight combination to obtain the evaluation results of the fused features corresponding to each weight combination on each performance metric. Exemplarily, for the weight combination α = 0.7, substitute α into F_fusion = α × F1 + (1 - α) × F2 to obtain the fused feature. After obtaining the fused feature, evaluate its accuracy and recall in the convolutional neural network or mechanism model to obtain the evaluation result of this weight combination. After obtaining the evaluation results of all weight combinations, select the weight combination with the best performance as the target weight combination. Of course, before performing fault detection on the oil transfer pump, the target weight combination can also be determined in advance according to the historical operation data of the oil transfer pump.

[0066] Further, after obtaining the target weight combination, use the target weight combination to perform weighted feature fusion on the first feature and the second feature to obtain a set of fused features. The set of fused features includes all fused features, and the dimension of each fused feature can be 10 dimensions. By optimizing the feature representation through feature fusion, the recognition ability of the model can be improved. By selecting the optimal weight combination to enhance the model performance, the accuracy and reliability of fault detection can be ensured.

[0067] Step 406: For each type of fused feature in the set of fused features, input the current type of fused feature into the mechanism model and the neural network model respectively to obtain the first diagnostic data output by the mechanism model and the second diagnostic data output by the neural network model.

[0068] Among them, the first diagnostic data is the data obtained by diagnosing the fusion features output by the mechanism model. The first diagnostic data includes the first diagnostic result, the first similarity, and the first confidence level. The second diagnostic data is the data obtained by diagnosing the fusion features output by the neural network model. The second diagnostic data includes the second diagnostic result, the second similarity, and the second confidence level. In this solution, optionally, the fusion features of the current type are respectively input into the mechanism model and the neural network model to obtain the first diagnostic data output by the mechanism model and the second diagnostic data output by the neural network model, including: inputting the fusion features of the current type into the mechanism model respectively to obtain the first diagnostic result of the fusion features of the current type, and calculating the first similarity and the first confidence level between the fusion features of the current type and the standard features of the fusion features of the current type determined in advance through the mechanism model; inputting the fusion features of the current type into the neural network model respectively to obtain the second diagnostic result of the fusion features of the current type, and calculating the second similarity and the second confidence level between the fusion features of the current type and the standard features of the fusion features of the current type determined in advance through the neural network model.

[0069] Among them, the standard features are used to help the mechanism model and the neural network model diagnose the fusion features. After obtaining the first feature and the second feature, a feature set T corresponding to the operation data is constructed according to the first feature and the second feature. It is assumed that the feature set T contains seven types of fault data, and each type of data has multiple samples. For each type of data, calculate the central coordinate (mean vector) of its feature vector and determine it as the standard feature T of this type. s (The seven types of fault data correspond to seven Ts s ), assuming that the fusion feature is 10-dimensional, then T s = [t1, t2, t3, t4, t5, t6, t7, t8, t9, t 10 . After determining the standard features, for each sample (each fusion feature) in each type of data, calculate its cosine similarity with each standard feature: Among them, T i represents the i-th fusion feature, and S i represents the cosine similarity. After obtaining the cosine similarity, sort the cosine similarity S i in descending order, and update the first confidence level according to the sorting result of the similarity. For example: the confidence level C i corresponding to the highest cosine similarity S 0i is 1.0; the confidence level corresponding to the 100th highest cosine similarity decreases to 0.9. The confidence level corresponding to the 200th highest cosine similarity decreases to 0.8, and so on. According to this method, further, input each fusion feature into the mechanism model to obtain the first diagnostic result R1 of each fusion feature output by the mechanism model (the mechanism model determines the first diagnostic result R1 according to the similarity S i and the confidence level Ci The output fault diagnosis result), the first confidence level C1, and the first similarity S1. Similarly, each fused feature is input into the neural network model to obtain the second diagnosis result R2, the second confidence level C2, and the second similarity S2 of each fused feature output by the neural network model.

[0070] Step 407: Determine the fault detection result of the fused feature of the current type based on the first diagnostic data and the second diagnostic data; determine the fault detection result of the operation data based on the fault detection results of the fused features of all types.

[0071] Specifically, after obtaining the first diagnosis result and the second diagnosis result of the fused features of each type, the fault detection result of the fused feature of each type can be determined according to the first confidence level, the first similarity, the second confidence level, and the second similarity. In this solution, optionally, determining the fault detection result of the fused feature of the current type based on the first diagnostic data and the second diagnostic data includes the following steps B1 - step B2:

[0072] Step B1: When the first diagnosis result is the same as the second diagnosis result, determine the first diagnosis result or the second diagnosis result as the fault detection result of the fused feature of the current type.

[0073] Specifically, when the first diagnosis result is the same as the second diagnosis result, it means that the fault detection results of the mechanism model and the neural network model for the fused feature are the same, and then directly determine the same diagnosis result as the fault detection result of the fused feature.

[0074] Step B2: When the first diagnosis result and the second diagnosis result are different, if the second confidence level is greater than the first confidence level and the second similarity is greater than the first similarity, determine the second diagnosis result as the fault detection result of the fused feature of the current type; if the first confidence level is greater than the second confidence level and the first similarity is greater than the second similarity, determine the first diagnosis result as the fault detection result of the fused feature of the current type; if the first confidence level is equal to the second confidence level, determine the diagnosis result corresponding to the similarity with the larger similarity value among the first similarity and the second similarity as the fault detection result of the fused feature of the current type.

[0075] Specifically, the confidence level reflects the reliability assessment of the model for the diagnostic result. The higher the confidence level, the more accurate the result output by the model is considered. The similarity reflects the matching degree between the fused feature and the known fault mode. The higher the similarity, the closer the fused feature is to the fault type. Therefore, when the confidence levels are the same, the similarity can be used as an auxiliary index to determine the fault detection result of the fused feature. When the first diagnostic result of the mechanism model is different from the second diagnostic result of the neural network model, the first confidence level and the second confidence level, as well as the first similarity and the second similarity, are compared. If the first confidence level is greater than the second confidence level and the first similarity is greater than the second similarity, the first diagnostic result output by the mechanism model is determined as the fault detection result of the fused feature. Of course, if the first confidence level is less than the second confidence level and the first similarity is less than the first confidence level, the second diagnostic result output by the neural network model is determined as the fault detection result of the fused feature. If the first confidence level is equal to the second confidence level and the first confidence level is greater than the second confidence level, the first diagnostic result output by the mechanism model is determined as the fault detection result of the fused feature. If the first confidence level is equal to the second confidence level and the first confidence level is less than the second confidence level, the second diagnostic result output by the neural network model is determined as the fault detection result of the fused feature.

[0076] By comprehensively considering the consistency, confidence level and similarity of the first diagnostic result and the second diagnostic result, the reliability and uniqueness of the final fault detection result can be ensured. The above steps utilize the advantages of the mechanism model and the neural network model to improve the accuracy and robustness of fault diagnosis.

[0077] In this solution, before obtaining the operation data of the oil transfer pump, the historical operation data of the oil transfer pump can be used to train the overall model of this solution (including the neural network model, the mechanism model and the feature fusion part) to obtain an overall model that can accurately determine whether the oil transfer pump has a fault. Exemplarily, Figure 5 is a schematic structural diagram of fault detection for the oil transfer pump provided by an embodiment of the present invention. As Figure 5 shown, the overall model includes a signal feature layer, a model layer, a decision layer and an output layer. The signal feature layer is used for data acquisition and signal extraction, including obtaining video features, sampling frequency, temperature, rotation speed and bearing type, etc., and also includes deep features extracted by a Convolutional Neural Network (CNN). The model layer includes a convolutional layer, a rectified linear unit activation layer, a pooling layer and a fully connected layer, etc. The decision layer includes a mechanism model, a big data model (which can be the neural network model in this solution) and a confidence level judgment module.

[0078] Exemplarily, the historical operation data of the oil transfer pump: the fault types include normal, bearing fault, evacuation and cavitation fault, and looseness fault. 100 samples of normal, bearing fault, evacuation and cavitation fault, and looseness fault are respectively extracted for verification, and the fault determination accuracy rate is 91%, as shown in Table 3 below:

[0079]

[0080] Table 3

[0081] Exemplarily, 100 groups of on-site rolling bearing fault data sent by third-party users are received. Among them, there are 59 groups of data of centrifugal pumps, 18 groups of data of centrifugal fans, and 23 groups of data of motors. Each group of data has 3 sample data. Taking 59 groups of centrifugal pump data (a total of 59 * 3 = 177 samples) as test data, the accuracy rate of the overall model is 98.8%, as shown in Table 4 below:

[0082]

[0083] Table 4

[0084] Step 408: Display the fault detection result to the user.

[0085] In the technical solution of this embodiment, the operation data of the oil transfer pump is acquired. Time-frequency feature extraction is performed on the operation data according to a pre-determined mechanism model to obtain the first features of the operation data, where the first features include the time-domain features, frequency-domain features, and time-frequency domain features of the operation data. The operation data is input into a neural network model, and local feature extraction is performed on the operation data through the input layer and convolutional layer of the neural network model to obtain the second candidate features of the operation data. Feature dimension reduction is performed on the second candidate features through the pooling layer of the neural network model, and the second candidate features after feature dimension reduction are processed through the fully connected layer and output layer of the neural network model to obtain the second features of the operation data. Feature fusion is performed on the first features and the second features to obtain a fusion feature set. For each type of fusion feature in the fusion feature set, the current type of fusion feature is respectively input into the mechanism model and the neural network model to obtain the first diagnostic data output by the mechanism model and the second diagnostic data output by the neural network model. The first diagnostic data includes a first diagnostic result, a first similarity, and a first confidence level, and the second diagnostic data includes a second diagnostic result, a second similarity, and a second confidence level; the fault detection result of the current type of fusion feature is determined based on the first diagnostic data and the second diagnostic data; the fault detection result of the operation data is determined based on the fault detection results of all types of fusion features. The fault detection result is presented to the user. The technical solution of this embodiment fully utilizes the advantages of the two models by fusing the time-frequency features extracted by the mechanism model and the local features extracted by the neural network model, improving the accuracy of fault detection and the interpretability of the model. By diagnosing through two models respectively and comprehensively determining the final fault detection result based on the results of both, the risk of misjudgment by a single model is reduced. The technical solution of this embodiment realizes accurate, efficient, and interpretable fault detection for the operation data of the oil transfer pump, which helps to improve the operation reliability of the equipment, reduce the maintenance cost, and enhance the operation safety.

[0086] Figure 6 FIG. is a schematic structural diagram of a fault detection device provided by an embodiment of the present invention, and this device is applicable to execute the fault detection method provided by an embodiment of the present invention. As Figure 6 shown, this device may specifically include:

[0087] A data acquisition module 601, configured to acquire the operation data of the oil transfer pump,

[0088] A first processing module 602, configured to perform time-frequency feature extraction on the operation data according to a pre-determined mechanism model to obtain the first features of the operation data, where the first features include the time-domain features, frequency-domain features, and time-frequency domain features of the operation data;

[0089] The second processing module 603 is configured to extract data features from the operation data based on a pre-trained neural network model, and obtain second features of the operation data;

[0090] The result determination module 604 is configured to obtain a fault detection result of the operation data based on the first feature, the second feature, and a pre-determined confidence decision method, and display the fault detection result to the user.

[0091] Optionally, the second processing module 602 is specifically configured to: input the operation data into the neural network model, and perform local feature extraction on the operation data through the input layer and the convolutional layer of the neural network model to obtain second candidate features of the operation data;

[0092] Perform feature dimensionality reduction on the second candidate features through the pooling layer of the neural network model, and process the second candidate features after feature dimensionality reduction by using the fully connected layer and the output layer of the neural network model to obtain second features of the operation data.

[0093] Optionally, the result determination module 604 is specifically configured to: perform feature fusion on the first feature and the second feature to obtain a fused feature set;

[0094] For each type of fused feature in the fused feature set, input the current type of fused feature into the mechanism model and the neural network model respectively to obtain first diagnostic data output by the mechanism model and second diagnostic data output by the neural network model; determine a fault detection result of the current type of fused feature based on the first diagnostic data and the second diagnostic data;

[0095] Determine a fault detection result of the operation data based on the fault detection results of all types of fused features.

[0096] Optionally, the result determination module 604 is further configured to: perform weighted feature fusion on the first feature and the second feature according to each pre-set weight combination, and obtain a feature fusion result of each weight combination;

[0097] Evaluate each feature fusion result based on each pre-set performance index, and determine a target weight combination according to the evaluation result;

[0098] Perform weighted feature fusion on the first feature and the second feature based on the target weight combination to obtain the fused feature set.

[0099] Optionally, the first diagnostic data includes a first diagnostic result, a first similarity, and a first confidence level, and the second diagnostic data includes a second diagnostic result, a second similarity, and a second confidence level; the result determination module 604 is further configured to: input the fusion features of the current type into the mechanism model respectively to obtain the first diagnostic result of the fusion features of the current type, and calculate the first similarity and the first confidence level between the fusion features of the current type and the standard features of the fusion features of the current type determined in advance through the mechanism model;

[0100] Input the fusion features of the current type into the neural network model respectively to obtain the second diagnostic result of the fusion features of the current type, and calculate the second similarity and the second confidence level between the fusion features of the current type and the standard features of the fusion features of the current type determined in advance through the neural network model.

[0101] Optionally, the result determination module 604 is further configured to: when the first diagnostic result is the same as the second diagnostic result, determine the first diagnostic result or the second diagnostic result as the fault detection result of the fusion features of the current type;

[0102] When the first diagnostic result is different from the second diagnostic result, determine the fault detection result of the fusion features of the current type based on the first confidence level, the first similarity, the second confidence level, and the second similarity.

[0103] The fault detection device provided by the embodiments of the present invention can execute the fault detection method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. The content not described in detail in this embodiment can be referred to the description in any method embodiment of the present invention.

[0104] The embodiments of the present invention also provide a computer program product.

[0105] The various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer program products, the one or more computer program products can include one or more computer programs, the one or more computer programs can be executed and / or interpreted on a programmable system including at least one programmable processor, the programmable processor can be a dedicated or general-purpose programmable processor, can receive data and instructions from a storage system, at least one input device, and at least one output device, and can transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0106] Figure 7 Schematic diagram of a structure of an electronic device provided for an embodiment of the present invention, refer to Figure 7 , Figure 7 The electronic device 12 shown is merely an example and should not impose any limitation on the functions and scope of use of the embodiments of this application. As Figure 7 shown, the electronic device 12 is presented in the form of a general-purpose computing device. The components of the electronic device 12 can include but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0107] The bus 18 represents one or more of several types of bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. By way of example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0108] The electronic device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the electronic device 12, including volatile and non-volatile media, removable and non-removable media.

[0109] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The electronic device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 can be used for reading and writing on non-removable, non-volatile magnetic media ( Figure 7 not shown, typically referred to as a "hard disk drive"). Although Figure 7 not shown in the figure, a disk drive for reading and writing on removable non-volatile disks (such as "floppy disks") and an optical disk drive for reading and writing on removable non-volatile optical disks (such as CD-ROM, DVD-ROM or other optical media) can be provided. In these cases, each drive can be connected to the bus 18 through one or more data media interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present application.

[0110] A program / utility 40 having a set (at least one) of program modules 46 can be stored, for example, in the memory 28. Such program modules 46 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. The program modules 46 generally execute the functions and / or methods in the embodiments described in the present application.

[0111] The electronic device 12 can also communicate with one or more external devices 14 (such as a keyboard, a pointing device, a display 24, etc.), and can also communicate with one or more devices that enable a user to interact with the electronic device 12, and / or communicate with any device that enables the electronic device 12 to communicate with one or more other computing devices (such as a network card, a modem, etc.). Such communication can be carried out through the input / output (I / O) interface 22. Also, the electronic device 12 can communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 20. As shown in the figure, the network adapter 20 communicates with other modules of the electronic device 12 through the bus 18. It should be understood that although Figure 7 not shown in the figure, other hardware and / or software modules can be used in combination with the electronic device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0112] The processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28, for example, implementing a fault detection method provided by an embodiment of the present invention: obtaining the operation data of the oil transfer pump; performing time-frequency feature extraction on the operation data according to a pre-determined mechanism model to obtain a first feature of the operation data, where the first feature includes the time-domain feature, frequency-domain feature, and time-frequency domain feature of the operation data; performing data feature extraction on the operation data based on a pre-trained neural network model to obtain a second feature of the operation data; obtaining a fault detection result of the operation data based on the first feature, the second feature, and a pre-determined confidence decision method, and presenting the fault detection result to the user.

[0113] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements a fault detection method provided by all embodiments of the present invention: obtaining the operation data of the oil transfer pump; performing time-frequency feature extraction on the operation data according to a pre-determined mechanism model to obtain a first feature of the operation data, where the first feature includes the time-domain feature, frequency-domain feature, and time-frequency domain feature of the operation data; performing data feature extraction on the operation data based on a pre-trained neural network model to obtain a second feature of the operation data; obtaining a fault detection result of the operation data based on the first feature, the second feature, and a pre-determined confidence decision method, and presenting the fault detection result to the user. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electronic device, apparatus, or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or combined with an instruction execution electronic device, apparatus, or device.

[0114] A computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution electronic device, apparatus, or device.

[0115] The program code contained on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wire, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0116] The computer program code for performing the operations of the present invention may be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0117] Note that the above is only the preferred embodiment of the present invention and the applied technical principles. Those skilled in the art will understand that the present invention is not limited to the specific embodiments herein, and various obvious changes, re-adjustments, and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments. Without departing from the concept of the present invention, it may also include more other equivalent embodiments, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A fault detection method, characterized in that: The method comprises: Obtain the operating data of the oil pump; Extracting time-frequency features of the operation data according to a predetermined mechanism model to obtain a first feature of the operation data, wherein the first feature includes a time domain feature, a frequency domain feature, and a time-frequency domain feature of the operation data; Extracting data features from the operating data based on a pre-trained neural network model to obtain a second feature of the operating data; A fault detection result of the operating data is obtained based on the first feature, the second feature and a predetermined confidence decision method, and the fault detection result is displayed to a user.

2. The method according to claim 1, characterized in that Extracting data features from the operating data based on a pre-trained neural network model to obtain a second feature of the operating data includes: Inputting the operation data into the neural network model, performing local feature extraction on the operation data through the input layer and convolution layer of the neural network model, and obtaining second candidate features of the operation data; The second candidate feature is subjected to feature dimensionality reduction through the pooling layer of the neural network model, and the second candidate feature after feature dimensionality reduction is processed using the fully connected layer and the output layer of the neural network model to obtain the second feature of the operating data.

3. The method according to claim 1, characterized in that Obtaining a fault detection result of the operating data based on the first feature, the second feature and a predetermined confidence decision method includes: Performing feature fusion on the first feature and the second feature to obtain a fused feature set; For each type of fusion feature in the fusion feature set, the fusion feature of the current type is input into the mechanism model and the neural network model respectively to obtain the first diagnostic data output by the mechanism model and the second diagnostic data output by the neural network model; based on the first diagnostic data and the second diagnostic data, a fault detection result of the fusion feature of the current type is determined; A fault detection result of the operation data is determined based on the fault detection results of all types of fused features.

4. The method according to claim 3, characterized in that The first feature and the second feature are subjected to feature fusion to obtain a fused feature set, including: Performing weighted feature fusion on the first feature and the second feature according to each preset weight combination, and obtaining a feature fusion result of each weight combination; Evaluate the feature fusion results based on the pre-set performance indicators, and determine the target weight combination according to the evaluation results; The first feature and the second feature are weightedly fused based on the target weight combination to obtain the fused feature set.

5. The method according to claim 3, characterized in that: The first diagnostic data includes a first diagnostic result, a first similarity and a first confidence, and the second diagnostic data includes a second diagnostic result, a second similarity and a second confidence; the current type of fusion features are respectively input into the mechanism model and the neural network model to obtain the first diagnostic data output by the mechanism model and the second diagnostic data output by the neural network model, including: Inputting the fusion features of the current type into the mechanism model respectively to obtain a first diagnosis result of the fusion features of the current type, and calculating a first similarity and a first confidence between the fusion features of the current type and a predetermined standard feature of the fusion features of the current type through the mechanism model; The current type of fusion features are respectively input into the neural network model to obtain a second diagnosis result of the current type of fusion features, and the second similarity and second confidence of the current type of fusion features and the predetermined standard features of the current type of fusion features are calculated through the neural network model.

6. The method according to claim 5, characterized in that Determining a fault detection result of the current type of fusion feature based on the first diagnostic data and the second diagnostic data includes: When the first diagnosis result is the same as the second diagnosis result, determining that the first diagnosis result or the second diagnosis result is a fault detection result of the current type of fusion feature; When the first diagnosis result and the second diagnosis result are different, a fault detection result of the current type of fused features is determined based on the first confidence, the first similarity, the second confidence, and the second similarity.

7. The method according to claim 6, characterized in that When the first diagnosis result and the second diagnosis result are different, determining a fault detection result of the current type of fusion feature based on the first confidence, the first similarity, the second confidence, and the second similarity includes: When the first diagnosis result and the second diagnosis result are different, if the second confidence is greater than the first confidence and the second similarity is greater than the first similarity, determining that the second diagnosis result is a fault detection result of the current type of fusion feature; If the first confidence is greater than the second confidence and the first similarity is greater than the second similarity, determining that the first diagnosis result is a fault detection result of the current type of fusion feature; If the first confidence is equal to the second confidence, the diagnosis result corresponding to the similarity with a larger similarity value between the first similarity and the second similarity is determined as the fault detection result of the current type of fusion feature.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the computer program implements a fault detection method according to any one of claims 1 to 7.

9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the fault detection method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the fault detection method according to any one of claims 1 to 7 is implemented.

Images