Abnormality detection method and system based on oil field equipment

Abstract

The invention relates to an anomaly detection method and system based on oil field equipment, and belongs to the technical field of industrial equipment fault diagnosis, and the anomaly detection method comprises the steps: receiving a transmission shaft vibration signal collected by a vibration sensor installed on a drilling machine transmission shaft of the oil field equipment; acquiring environmental data and lubricant state data in real time; performing data preprocessing on the transmission shaft vibration signal, the environment data and the lubricant state data to obtain multi-dimensional time sequence data; vibration signal features are extracted, and compensation is carried out based on environment data and lubricant state data; generating a dynamic feature threshold range based on a dynamic threshold model, and comparing the dynamic feature threshold range with feature values in the compensated vibration signal feature set to obtain abnormal vibration signal data; and extracting frequency characteristics of different time periods, performing abnormal mode matching, obtaining a corresponding abnormal mode matching result, and sending corresponding maintenance prompt information to the management terminal. According to the invention, the accuracy and reliability of abnormity identification can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of industrial equipment fault diagnosis, and in particular to an abnormality detection method and system based on oil field field equipment. Background Art

[0002] With the acceleration of industrialization and the continuous growth of energy demand, the scale and complexity of oilfield equipment have increased significantly. As an important base for energy production, oilfield equipment needs to operate efficiently and for a long time in a complex environment. As one of the core components of oilfield equipment, the main function of the drilling rig drive shaft is to transfer power from the drive device to other components to drive the drill bit to complete the drilling operation. The stability and performance of the drive shaft directly affect the operating efficiency of the equipment and the overall production capacity of the oilfield. However, oilfield operating equipment has long been facing harsh environmental conditions such as high temperature, high humidity, high pressure, vibration and corrosion. In addition, continuous high-intensity operation can easily cause wear, fatigue damage and even failure of the drive shaft and its related components (such as bearings, lubrication systems, etc.).

[0003] At present, due to the complexity of the operating environment of oilfield equipment, the sources of abnormal signals may be very diverse. Common abnormalities include bearing wear, poor lubrication, mechanical looseness, etc. These abnormal signals are usually manifested in the form of vibration, but because of their complex characteristics and rapid changes, it is difficult to accurately identify abnormal patterns or quickly locate the cause of the fault, which in turn affects the accuracy of abnormal detection. Summary of the invention

[0004] In order to improve the accuracy of anomaly identification, the present application provides an anomaly detection method and system based on oilfield field equipment.

[0005] In the first aspect, the present application provides an abnormality detection method based on oil field field equipment, which adopts the following technical solution: An abnormality detection method based on oil field field equipment, the abnormality detection method comprising: Receiving a transmission shaft vibration signal collected by a vibration sensor; wherein the vibration sensor is installed on a drilling rig transmission shaft of an oil field field equipment; Acquire environmental data and lubricant status data of the oil field field equipment in real time; Preprocessing the transmission shaft vibration signal, environmental data and lubricant state data to obtain multi-dimensional time series data; Extracting vibration signal features from the multidimensional time series data, and compensating the vibration signal features based on the environmental data and the lubricant state data to obtain a compensated vibration signal feature set; Generate a dynamic feature threshold range based on a pre-trained dynamic threshold model, and compare it with the feature value in the compensated vibration signal feature set to obtain abnormal vibration signal data; Decomposing the abnormal vibration signal data in time series to extract frequency features of different time periods; Perform abnormal pattern matching on the frequency characteristics of the different time periods to obtain corresponding abnormal pattern matching results; According to the abnormal pattern matching result, the corresponding maintenance prompt information is sent to the management terminal.

[0006] By adopting the above technical solutions, through the collection of vibration signals, fusion of environmental and lubricant data, dynamic threshold detection and abnormal pattern matching, real-time monitoring and fault diagnosis of the operating status of oilfield equipment are realized, and collaborative analysis is performed using multi-dimensional data to significantly improve the accuracy and reliability of anomaly detection; through the introduction of dynamic threshold models and pattern matching methods, the system can adapt to complex environmental changes, which not only improves the operating efficiency and safety of oilfield equipment, but also reduces production interruptions and maintenance costs caused by equipment failures.

[0007] Optionally, the step of extracting vibration signal features from the multidimensional time series data and compensating the vibration signal features based on the environmental data and the lubricant state data to obtain a compensated vibration signal feature set includes: Performing spectrum decomposition and time domain analysis on the transmission shaft vibration signal in the multidimensional time series data to extract vibration signal features; Calculating a first compensation value of the environmental data to the vibration signal feature based on a pre-constructed environmental data compensation model; Calculating a second compensation value of the lubricant state data to the vibration signal feature based on a pre-constructed lubricant state compensation model; The vibration signal feature is compensated according to the first compensation value and the second compensation value to obtain a compensated vibration signal feature set.

[0008] By adopting the above technical solution, based on dual compensation of environment and lubricant, the vibration signal feature set is made more accurate, which can accurately reflect the actual working status of the equipment and provide reliable data support for subsequent fault diagnosis and life prediction.

[0009] Optionally, the method further includes a training step of the dynamic threshold model, wherein the training step includes: Acquire historical sample data; the historical sample data includes historical operation status data of the equipment, vibration signal sample characteristics, and pre-marked abnormal vibration signal labels; Preprocessing the historical sample data and dividing it into a training set and a test set; Inputting the equipment historical operating status data and vibration signal sample features in the training set into a pre-built random forest model to obtain a training label result; Comparing the training label result with the pre-labeled abnormal vibration signal label in the training set to obtain a training comparison result; The random forest model is iterated according to the training comparison result to obtain the trained dynamic threshold model.

[0010] By adopting the above technical solutions, the model can gradually reduce the cases of misjudgment and missed judgment, improve the sensitivity and accuracy of abnormal vibration signals, and the final dynamic threshold model can automatically adjust the judgment criteria according to the real-time working conditions of the equipment and the characteristics of the vibration signal, providing accurate support for subsequent real-time monitoring. In addition, the dynamic adjustment capability of the dynamic threshold model enables it to adapt to different working conditions, avoid the error of fixed thresholds, and provide accurate fault warnings.

[0011] Optionally, after the step of obtaining the trained dynamic threshold model, the step further includes: Inputting the historical operation status data of the equipment and the vibration signal sample features in the test set into the trained dynamic threshold model to obtain a test label result; Based on the pre-labeled abnormal vibration signal labels in the test set, the test label results are compared to obtain a test comparison result; According to the test comparison results, the performance indicators of the trained dynamic threshold model are evaluated and iterated until a preset performance index is met or a preset number of iterations is reached, thereby obtaining the trained dynamic threshold model.

[0012] By adopting the above technical solutions, after multiple rounds of training, testing and optimization iterations, it is ensured that the constructed dynamic threshold model can accurately and stably judge the abnormal vibration signals of the equipment, thereby providing reliable support for equipment fault warning and maintenance. The high efficiency of the model ensures that the health monitoring of the equipment can respond quickly, reduce the equipment failure rate, and improve production efficiency.

[0013] Optionally, the step of performing abnormal pattern matching on the frequency characteristics of the different time periods to obtain corresponding abnormal pattern matching results includes: Based on a preset abnormal pattern library, the frequency characteristics of the different time periods are matched using a characteristic pattern matching algorithm; Determine whether there is an abnormal pattern corresponding to each frequency feature in different time periods in the preset abnormal pattern library; if so, obtain a matching result of the known abnormal pattern; If not, the frequency features are classified based on a density clustering algorithm to obtain matching results of potential unknown abnormal patterns.

[0014] By adopting the above technical solution, we can achieve the transition from rapid matching of known abnormal patterns to exploration and classification of unknown abnormal patterns. In actual operation, the feature pattern matching algorithm effectively identifies the known patterns in the preset abnormal pattern library; and for the unmatched frequency features, the density clustering algorithm further mines its potential patterns and classifies them, thereby expanding the capabilities of the abnormal pattern library. Ultimately, the technical solution of this application has the dual capabilities of rapidly identifying known anomalies and flexibly responding to unknown anomalies, providing comprehensive technical support for equipment anomaly detection and fault diagnosis. This hierarchical matching method ensures the efficiency and scalability of fault pattern analysis.

[0015] Optionally, after obtaining the matching result of the potential unknown abnormal pattern, the following steps are further included: Standardizing the matching results of the potential unknown abnormal patterns to obtain visual result data; Sending the visualization result data to the management terminal; Receiving classification and annotation of abnormal patterns fed back by the management terminal; Performing integrity check on the abnormal pattern classification labeling; The abnormal pattern classification label is added to the preset abnormal pattern library to obtain an expanded preset abnormal pattern library.

[0016] By adopting the above technical solution, the matching results of potential unknown abnormal patterns are sent to the management terminal, combined with the classification and annotation of the management personnel, and the abnormal pattern library is dynamically expanded. Through the feedback mechanism and pattern library update, the system can continuously learn new abnormal patterns, thereby realizing the continuous optimization and intelligent upgrade of the anomaly detection system.

[0017] In the second aspect, the present application provides an abnormality detection system based on oil field field equipment, which adopts the following technical solution: An abnormality detection system based on oil field field equipment, the abnormality detection system comprising: A receiving module, used for receiving a transmission shaft vibration signal collected by a vibration sensor; wherein the vibration sensor is installed on a drilling rig transmission shaft of an oil field field equipment; A data acquisition module, used for acquiring environmental data and lubricant status data of the oil field field equipment in real time; A data processing module, used for preprocessing the transmission shaft vibration signal, environmental data and lubricant state data to obtain multi-dimensional time series data; A signal feature compensation module, used for extracting vibration signal features from the multi-dimensional time series data, and compensating the vibration signal features based on the environmental data and the lubricant state data to obtain a compensated vibration signal feature set; An abnormality comparison module, used for generating a dynamic feature threshold range based on a pre-trained dynamic threshold model, and comparing it with the feature value in the compensated vibration signal feature set to obtain abnormal vibration signal data; A frequency feature extraction module, used to perform time series decomposition on the abnormal vibration signal data and extract frequency features of different time periods; An abnormal pattern matching module is used to perform abnormal pattern matching on the frequency characteristics of the different time periods to obtain corresponding abnormal pattern matching results; The maintenance prompt module is used to send corresponding maintenance prompt information to the management terminal according to the abnormal pattern matching result.

[0018] Optionally, the signal feature compensation module includes: A vibration signal extraction unit, used to perform spectrum decomposition and time domain analysis on the transmission shaft vibration signal in the multi-dimensional time series data, and extract vibration signal features; A first compensation calculation unit, configured to calculate a first compensation value of the environmental data for the vibration signal feature based on a pre-constructed environmental data compensation model; A second compensation calculation unit, configured to calculate a second compensation value of the lubricant state data to the vibration signal feature based on a pre-constructed lubricant state compensation model; A compensation unit is used to compensate the vibration signal feature according to the first compensation value and the second compensation value to obtain a compensated vibration signal feature set.

[0019] In a third aspect, the present application provides a computer device, which adopts the following technical solution: A computer device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method as described in the first aspect.

[0020] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium stores a computer program that can be loaded by a processor and execute any one of the methods in the first aspect.

[0021] To summarize, the present application includes at least one of the following beneficial technical effects: the present application optimizes equipment operation management, reduces downtime losses caused by faults, improves the safety and production efficiency of oilfield equipment, and at the same time realizes intelligent and precise equipment status monitoring and fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1This is a first flow chart of an abnormality detection method based on oilfield field equipment according to one of the embodiments of the present application.

[0023] Figure 2 This is a second flow chart of an abnormality detection method based on oilfield field equipment according to one of the embodiments of the present application.

[0024] Figure 3 This is a third flow chart of an abnormality detection method based on oilfield field equipment in one of the embodiments of the present application.

[0025] Figure 4 This is a fourth flow chart of an abnormality detection method based on oilfield field equipment according to one of the embodiments of the present application.

[0026] Figure 5 This is a fifth flow chart of an abnormality detection method based on oilfield field equipment in one of the embodiments of the present application.

[0027] Figure 6 This is a sixth flow chart of an abnormality detection method based on oilfield field equipment in one of the embodiments of the present application. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of this application more clear, the following Figure 1-6 It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0029] The embodiment of the present application discloses an abnormality detection method based on oil field field equipment.

[0030] Reference Figure 1 , an abnormality detection method based on oil field field equipment, the abnormality detection method includes: Step S101, receiving a transmission shaft vibration signal collected by a vibration sensor; wherein the vibration sensor is installed on a drilling rig transmission shaft of an oil field field equipment; Specifically, vibration signals are an important reflection of the operating status of equipment. As a key component of oilfield field equipment, the vibration characteristics of the drilling rig drive shaft are directly related to the health status and potential failure risks of the equipment. In one embodiment of the present application, by installing a high-sensitivity triaxial acceleration sensor at a key position of the drive shaft, the vibration signal of the equipment during operation can be collected in real time, including displacement, velocity and acceleration information.

[0031] It should be noted that the sensor installation location should be selected in the bearing area of ​​the transmission shaft to capture signals of bearing wear or abnormal vibration. Vibration signals are usually collected at a high sampling rate (such as above 1kHz) to ensure signal integrity and capture of details.

[0032] Step S102, real-time acquisition of environmental data and lubricant status data of oil field equipment; Among them, the operating environment of oilfield equipment is complex and changeable. Environmental factors such as temperature and humidity may affect the operating status of the equipment; the lubricant status (such as viscosity, pH value, and metal particle content) directly reflects the changes in lubrication conditions. These factors will interfere with the characteristics of the vibration signal. By collecting environmental and lubricant status data in real time, the vibration signal can be compensated and optimized.

[0033] Specifically, temperature and humidity sensors and pressure sensors can be used to monitor the environmental conditions around the equipment in real time. At the same time, lubricant analysis sensors can be used to detect the physical (viscosity) and chemical (pH value, metal particle concentration) states of the lubricant.

[0034] Step S103, preprocessing the transmission shaft vibration signal, environmental data and lubricant state data to obtain multi-dimensional time series data; Among them, vibration signals, environmental data, and lubricant status data need to be preprocessed to remove noise and outliers, and synchronize the time axis to ensure data accuracy and time consistency. This step is the basis for subsequent feature extraction and analysis.

[0035] Specifically, for the preprocessing of vibration signals, high-frequency interference can be removed through wavelet transform denoising and low-pass filtering, while low-frequency key signals can be retained. For the processing of environmental data and lubricant data, standardization processing (such as normalization) can be used to unify the numerical range of the data, remove outliers and perform interpolation. The above three types of data are aligned by timestamps to form a structured multidimensional time series containing vibration signals, environmental parameters and lubricant status.

[0036] Step S104, extracting vibration signal features from the multidimensional time series data, and compensating the vibration signal features based on the environmental data and the lubricant state data to obtain a compensated vibration signal feature set; Among them, the vibration signal contains a lot of information, but the direct analysis of the original signal is inefficient. By extracting key features (such as frequency, amplitude, and energy), the equipment status can be more intuitively reflected. At the same time, the interference of environmental data and lubricant status needs to be corrected through the compensation model. The compensation model calculates the correction amount of the vibration signal given the environmental conditions and lubricant status, thereby obtaining the compensated vibration signal characteristics.

[0037] Step S105, generating a dynamic feature threshold range based on a pre-trained dynamic threshold model, and comparing it with the feature value in the compensated vibration signal feature set to obtain abnormal vibration signal data; Among them, the dynamic threshold model adjusts the threshold range of characteristic variables in real time based on equipment operating conditions (such as load and operating hours), and accurately judges abnormal vibration signals by comparing with real-time characteristic data.

[0038] In some embodiments, the dynamic threshold model can use a machine learning algorithm (such as random forest) to establish a multidimensional threshold model of characteristic variables based on historical equipment data. The dynamic threshold model outputs the dynamic threshold range of each characteristic variable, and then compares the compensated vibration signal characteristics with the dynamic threshold range, outputs whether there is an abnormal vibration signal, and records the abnormal time point.

[0039] Step S106, performing time series decomposition on the abnormal vibration signal data to extract frequency features of different time periods; Among them, abnormal vibration signals often contain multiple features. Through time series decomposition (such as wavelet decomposition or short-time Fourier transform), the frequency characteristics of different time periods can be extracted to analyze the specific patterns of the abnormalities.

[0040] Step S107, performing abnormal pattern matching on the frequency characteristics of different time periods to obtain corresponding abnormal pattern matching results; Among them, by matching the frequency characteristics of abnormal signals with the pre-built abnormal pattern library, the specific pattern of the abnormality can be quickly located, and the matching results are used to generate targeted maintenance prompts.

[0041] Specifically, a feature template matching algorithm (such as cosine similarity calculation) can be used to match the abnormal signal with the abnormal pattern library, and the best matching result and credibility can be output.

[0042] Step S108: sending corresponding maintenance prompt information to the management terminal according to the abnormal pattern matching result.

[0043] Among them, maintenance suggestions are generated based on the abnormal pattern matching results, such as bearing replacement, lubricant replenishment, etc., and pushed to the management terminal. By quickly identifying the cause of the abnormality, the fault location time is shortened and the pertinence and efficiency of equipment maintenance are improved.

[0044] In the above implementation, through the collection of vibration signals, fusion of environmental and lubricant data, dynamic threshold detection and abnormal pattern matching, real-time monitoring and fault diagnosis of the operating status of oilfield field equipment are achieved, and collaborative analysis using multi-dimensional data significantly improves the accuracy and reliability of abnormality detection; through the introduction of dynamic threshold models and pattern matching methods, the system can adapt to complex environmental changes, which not only improves the operating efficiency and safety of oilfield equipment, but also reduces production interruptions and maintenance costs caused by equipment failures.

[0045] Reference Figure 2 As an implementation method of step S104, the steps of extracting vibration signal features from multidimensional time series data and compensating the vibration signal features based on environmental data and lubricant state data to obtain a compensated vibration signal feature set include: Step S201, performing spectrum decomposition and time domain analysis on the transmission shaft vibration signal in the multi-dimensional time series data to extract vibration signal features; Among them, the vibration signal characteristics include frequency domain characteristics (such as frequency, energy distribution) and time domain characteristics (such as amplitude, peak value).

[0046] Specifically, the fast Fourier transform can be used to perform spectrum analysis on the transmission shaft vibration signal, converting the time domain signal into a frequency domain signal, thereby extracting the distribution of frequency components and amplitude changes in the vibration signal, which helps to identify the frequency characteristics of the transmission shaft (such as fault frequency or normal working frequency). At the same time, time domain analysis of the vibration signal usually includes calculating features such as amplitude, peak value, and root mean square (RMS). Time domain analysis can capture the changing trend of the signal, such as the fluctuation of the vibration amplitude, which is critical for early fault diagnosis.

[0047] Step S202, calculating a first compensation value of the environmental data to the vibration signal feature based on a pre-built environmental data compensation model; Among them, based on historical data and experience, a relationship model between environmental factors (such as temperature and humidity) and vibration signal characteristics is established. Through regression analysis, interpolation method or machine learning method (such as support vector regression SVR), according to real-time environmental data input, the first compensation value of the vibration signal characteristics is calculated to adjust the environmental impact part of the vibration signal characteristics.

[0048] It is understandable that temperature may cause the expansion of the drive shaft material, affecting the frequency of the vibration signal, while humidity or pressure changes may affect the friction and noise of the transmission system. By establishing an environmental data compensation model, it is possible to predict the specific impact of these environmental changes on the vibration signal. By correcting specific environmental interference in the vibration signal through environmental compensation, the vibration signal characteristics can more truly reflect the equipment status, thereby improving signal accuracy and helping to more accurately judge the equipment status.

[0049] Step S203, calculating a second compensation value of the lubricant state data to the vibration signal characteristics based on a pre-built lubricant state compensation model; Among them, the relationship model between lubricant state and vibration signal characteristics is established by using the law of lubricant state data (such as lubricant oil viscosity, temperature, etc.) affecting equipment wear and vibration signals. For example, changes in lubricant oil viscosity may cause changes in the friction coefficient of the transmission shaft, thereby affecting the amplitude and frequency of the vibration signal. By establishing a relationship model between lubricant oil state and vibration characteristics, the impact of lubricant state on vibration signals can be predicted.

[0050] Furthermore, based on real-time lubricant status data (such as oil temperature, oil viscosity, etc.), regression models, Kalman filtering and other techniques can be used to calculate a corresponding second compensation value, which is used to adjust the lubricant influence part of the vibration signal characteristics to achieve dynamic compensation of the lubricant status to the vibration signal.

[0051] It can be understood that by compensating for the effect of lubricant condition on the vibration signal, the accuracy of the vibration signal is improved. The effect of lubricant condition is a more complex factor, so this compensation allows the signal to more accurately reflect the health of the equipment and avoid misjudgment caused by changes in lubrication conditions.

[0052] Step S204: Compensate the vibration signal feature according to the first compensation value and the second compensation value to obtain a compensated vibration signal feature set.

[0053] Specifically, the first compensation value and the second compensation value are applied to the frequency domain and time domain characteristics of the vibration signal respectively to adjust the environmental and lubricant effects in the vibration signal. Specifically, the first compensation value may affect the frequency distribution and energy level of the vibration signal, and the second compensation value may affect the amplitude and energy of the signal. Therefore, the compensation process also requires weighted adjustment of each dimension of the signal (frequency domain and time domain), and finally generates a compensated vibration signal feature set, including vibration signal feature data after correction of the environment and lubricant state.

[0054] In the above implementation, the dual compensation based on the environment and the lubricant makes the vibration signal feature set more accurate, which can accurately reflect the actual working state of the equipment and provide reliable data support for subsequent fault diagnosis and life prediction.

[0055] Reference Figure 3 As a further implementation of the anomaly detection method, a training step of a dynamic threshold model is also included, and the training step includes: Step S301, obtaining historical sample data; the historical sample data includes historical operation status data of the equipment, vibration signal sample characteristics, and pre-marked abnormal vibration signal labels; Among them, the historical operating status data of the equipment includes load, operating time, temperature, pressure, etc. The vibration signal sample characteristics include multi-dimensional vibration signal characteristics collected during the operation of the equipment, such as frequency, amplitude, energy, peak value, etc.; the pre-marked abnormal vibration signal labels are derived from historical equipment failure data, marking which signal data are in normal state (labeled as "0") or abnormal state (labeled as "1").

[0056] Step S302, preprocessing the historical sample data and dividing it into a training set and a test set; The data preprocessing step includes: cleaning and normalizing the equipment historical operating status data and vibration signal sample characteristics to ensure the consistency and stability of the input data. For example, outliers are removed, missing data is filled, and the amplitude of vibration signal characteristics is standardized.

[0057] In some embodiments, the preprocessed data set is divided into a training set and a test set, for example, 80% for training and 20% for testing, to ensure that the model can be effectively verified on unknown data.

[0058] Step S303, inputting the equipment historical operation status data and vibration signal sample features in the training set into a pre-built random forest model to obtain a training label result; Among them, the integrated learning characteristics of random forests ensure that the model can automatically adjust the recognition of abnormal signals according to the input feature information, thereby improving the prediction accuracy; in addition, the independence and voting mechanism of each decision tree reduces the risk of overfitting and improves the stability of the model.

[0059] Specifically, the historical operating status data of the equipment and the sample features of the vibration signal (including frequency domain and time domain features) in the training set are used as input data. The random forest model predicts whether the signal is abnormal based on the input feature data through multiple decision trees. In each tree, different features are randomly selected for decision making, and finally a "voting mechanism" is used to determine whether it is an abnormal signal. The training label result is the prediction result made by the model based on the features of the training set, marked as "normal" or "abnormal".

[0060] Step S304, comparing the training label result with the pre-labeled abnormal vibration signal label in the training set to obtain a training comparison result; The training labels generated by the model are compared with the real labels one by one to calculate the matching degree between the two. If the model output is abnormal and the label is abnormal, it is judged as a "correct" prediction. If the model output is abnormal but the label is normal, or the model output is normal but the label is abnormal, it is a "wrong" prediction.

[0061] Step S305, iterating the random forest model according to the training comparison result to obtain a trained dynamic threshold model.

[0062] Among them, the random forest model is iteratively optimized according to the training comparison results. The model can be continuously optimized by adjusting model parameters (such as the number of trees, the maximum depth of the tree, the minimum number of sample splits, etc.), adjusting feature selection (selecting the most effective features for abnormality identification and removing redundant features), and adjusting sample weights (for misclassified samples, a larger weight can be given so that the model can better learn the features of these samples). After multiple iterations, the performance of the model on the training set will gradually improve, and finally a trained dynamic threshold model that can judge abnormal vibration signals of equipment will be obtained.

[0063] In the above implementation, the model can gradually reduce the cases of misjudgment and missed judgment, improve the sensitivity and accuracy of abnormal vibration signals, and the final dynamic threshold model can automatically adjust the judgment criteria according to the real-time working conditions of the equipment and the characteristics of the vibration signal, providing accurate support for subsequent real-time monitoring. In addition, the dynamic adjustment capability of the dynamic threshold model enables it to adapt to different working conditions, avoid the error of fixed thresholds, and provide accurate fault warnings.

[0064] Reference Figure 4 As a further implementation of training the dynamic threshold model, after the step of obtaining the trained dynamic threshold model, the method further includes: Step S401, inputting the historical operation status data of the equipment and the vibration signal sample features in the test set into the trained dynamic threshold model to obtain the test label result; The historical operating status data and vibration signal sample features of the equipment in the test set are input into the trained dynamic threshold model. The trained model uses the previously learned features and thresholds to infer whether these signals are abnormal based on the input data. Each input sample corresponds to an output label, indicating whether the equipment has an abnormal vibration signal.

[0065] It is understandable that by inputting the test set data into the trained dynamic threshold model, the performance of the model on new, unseen data can be verified. The prediction results of the model provide basic data for subsequent performance evaluation and optimization.

[0066] Step S402, comparing the test label results based on the pre-labeled abnormal vibration signal labels in the test set to obtain a test comparison result; The test label predicted by the model is compared with the pre-labeled true label. If the predicted label is consistent with the true label (for example, an abnormal vibration signal is correctly predicted as abnormal), it is a correct prediction. If the predicted label is inconsistent with the true label (for example, a normal signal is misjudged as abnormal), it is an incorrect prediction.

[0067] It is understandable that by comparing the model prediction results in the test set with the true labels, the accuracy of the model can be quantitatively analyzed. The comparison results provide a specific basis for subsequent model adjustment and optimization, ensuring that the trained model can better adapt to the actual data.

[0068] Step S403, based on the test comparison result, the performance index of the trained dynamic threshold model is evaluated and iterated until a preset performance index is met or a preset number of iterations is reached, thereby obtaining a trained dynamic threshold model.

[0069] Among them, the relevant performance indicators (such as accuracy, precision, recall, etc.) are calculated based on the test comparison results. These indicators can help quantify the performance of the model in practical applications. If the performance indicators meet the preset standards (for example, the accuracy exceeds 95%) or the maximum number of iterations is reached, the model is considered to have been trained and entered the final stage. In addition, if the performance indicators do not meet the requirements, the model can be further optimized by cross-validation and hyperparameter adjustment, and re-trained based on the optimized model.

[0070] It is understandable that through repeated iterative optimization, the model can continuously improve its performance in practical applications and reduce misjudgments and missed judgments. This process ensures that the trained model can adapt to changes in the actual environment and improve the adaptability and stability of the model.

[0071] In the above implementation, after multiple rounds of training, testing and optimization iterations, it is ensured that the constructed dynamic threshold model can accurately and stably judge the abnormal vibration signal of the equipment, thereby providing reliable support for equipment fault warning and maintenance. The high efficiency of the model ensures that the health monitoring of the equipment can respond quickly, reduce the equipment failure rate, and improve production efficiency.

[0072] Reference Figure 5 As an implementation of step S107, the step of performing abnormal pattern matching on the frequency characteristics of different time periods to obtain corresponding abnormal pattern matching results includes: Step S501, based on a preset abnormal pattern library, using a characteristic pattern matching algorithm to match frequency characteristics of different time periods; Among them, the frequency feature sets of different time periods include the main frequency, energy distribution, spectrum diagram, etc. obtained by decomposing and extracting the vibration signal. The preset abnormal pattern library is a set of feature templates containing known abnormal patterns, and each template corresponds to a specific abnormality (such as bearing wear, looseness, insufficient lubrication, etc.).

[0073] Specifically, the frequency features (such as frequency values, spectral energy, etc.) of different time periods are aligned with the templates in the pattern library. The alignment process can ensure that the input features and the template are in the same scale or frequency domain through interpolation or normalization.

[0074] In some embodiments, a feature similarity matching algorithm (such as cosine similarity, Euclidean distance, dynamic time warping DTW) can be used to calculate the similarity between the input frequency feature and the abnormal pattern template, and the frequency feature with a similarity higher than a set threshold can be matched to the corresponding abnormal pattern. The threshold can be determined based on historical data experience, and is usually set to a similarity of more than 80%.

[0075] Step S502, determining whether there is an abnormal pattern corresponding to each frequency feature in different time periods in the preset abnormal pattern library; if yes, jump to step S503; if no, jump to step S504; Step S503, obtaining a matching result of a known abnormal pattern; Here, each frequency feature is judged one by one whether it has matched a pattern template in the preset abnormal pattern library. If the feature has matched a template, the matching result is added to the known abnormal pattern matching list.

[0076] Specifically, the matching results include matching results of known abnormal patterns (corresponding to matched frequency features and their abnormal pattern identifiers) and a list of unknown frequency features (including all unmatched frequency features).

[0077] Step S504: classify the frequency features based on the density clustering algorithm to obtain matching results of potential unknown abnormal patterns.

[0078] Among them, density clustering algorithms (such as DBSCAN) can be used to classify unknown frequency features. The advantage of DBSCAN is that it does not need to preset the number of clusters. It can automatically discover clusters based on the density of data points and mark abnormal data points as noise. By performing density clustering on unknown frequency features, multiple feature categories are formed, each of which represents a potential abnormal pattern. In addition, isolated points (i.e., noise points) that cannot be clustered are marked as "unclassified features."

[0079] Specifically, two key parameters of the clustering algorithm are set: the minimum number of samples (minPts) and the neighborhood radius (ε). The parameter selection can be adjusted based on historical data. For example, for frequency features, ε can be set to a frequency value difference of less than 5Hz, and minPts can be set to contain at least 5 similar data points.

[0080] It can be understood that the use of density clustering algorithm can effectively discover the potential rules of unknown patterns and automatically classify feature points into multiple categories. The matching results of potential unknown abnormal patterns include all categories formed by clustering, and each category corresponds to a potential unknown abnormal pattern.

[0081] In the above implementation, the rapid matching of known abnormal patterns to the exploration and classification of unknown abnormal patterns is achieved. In actual operation, the feature pattern matching algorithm effectively identifies the known patterns in the preset abnormal pattern library; and for the unmatched frequency features, the density clustering algorithm further mines its potential patterns and classifies them, thereby expanding the capabilities of the abnormal pattern library. Ultimately, the technical solution of the present application has the dual capabilities of quickly identifying known anomalies and flexibly responding to unknown anomalies, providing comprehensive technical support for equipment anomaly detection and fault diagnosis. This hierarchical matching method ensures the efficiency and scalability of fault pattern analysis.

[0082] Reference Figure 6 As a further implementation of the anomaly detection method, after obtaining the matching result of the potential unknown anomaly pattern in step S504, the method further includes: Step S601, normalizing the matching results of the potential unknown abnormal patterns to obtain visualization result data; The standardization process includes extracting the core features of each category (such as main frequency, spectral energy, modal distribution) and their corresponding timestamp information, and generating visual result data (such as cluster distribution diagrams, spectrum diagrams, etc.) to facilitate quick analysis by management end users.

[0083] Step S602, sending the visualization result data to the management terminal; In some embodiments, the visualization result data can be sent to a management terminal through a communication interface (such as an API, a message queue), so that the management personnel can intuitively understand the key features of the abnormal pattern.

[0084] Step S603, receiving the abnormal pattern classification labeling fed back by the management terminal; Among them, the management personnel classify and label the potential unknown abnormal patterns according to the feature descriptions. The labeled data usually includes the abnormal pattern category (such as bearing failure, insufficient lubrication, etc.), the corresponding feature description (such as main frequency range, spectrum energy, etc.) and the management personnel's remarks (such as possible equipment status, recommended measures, etc.).

[0085] Step S604, performing integrity check on the abnormal pattern classification label; Among them, the integrity check is used to ensure that each annotation item contains the necessary information (such as category, description, timestamp, etc.). If the annotation information is missing or abnormal, the system should return a prompt message so that the management terminal user can make corrections.

[0086] It can be understood that by receiving and verifying the annotation data fed back by the management terminal, the system has established a human-machine collaborative pattern classification mechanism to ensure that the annotation information of the abnormal pattern is complete and accurate, providing a basis for the subsequent expansion of the abnormal pattern library.

[0087] Step S605: Add the abnormal pattern classification label to the preset abnormal pattern library to obtain an expanded preset abnormal pattern library.

[0088] Specifically, check whether the preset abnormal pattern library already contains the abnormal pattern marked by the management terminal. If the marked abnormal pattern is a new pattern, add the pattern to the abnormal pattern library and update the index and feature template of the pattern library. If the marked abnormal pattern already exists, update the existing pattern according to the new marking information (such as supplementing feature information, optimizing feature templates, etc.).

[0089] It is understandable that by dynamically updating the abnormal pattern library, the system has the ability to continuously learn and self-optimize, and newly discovered abnormal patterns can be quickly added to the detection system, thereby improving the system's ability to identify unknown abnormalities.

[0090] In the above implementation, the matching results of potential unknown abnormal patterns are sent to the management terminal, combined with the classification and annotation of the management personnel, and the abnormal pattern library is dynamically expanded. Through the feedback mechanism and pattern library update, the system can continuously learn new abnormal patterns, thereby realizing the continuous optimization and intelligent upgrading of the anomaly detection system.

[0091] The embodiment of the present application also discloses an abnormality detection system based on oil field field equipment.

[0092] An abnormality detection system based on oil field field equipment, the abnormality detection system comprising: A receiving module is used to receive a transmission shaft vibration signal collected by a vibration sensor; wherein the vibration sensor is installed on a drilling rig transmission shaft of an oil field field equipment; Data acquisition module, used to obtain environmental data and lubricant status data of oil field equipment in real time; A data processing module is used to pre-process the transmission shaft vibration signal, environmental data and lubricant status data to obtain multi-dimensional time series data; A signal feature compensation module is used to extract vibration signal features from multi-dimensional time series data, and compensate the vibration signal features based on environmental data and lubricant state data to obtain a compensated vibration signal feature set; An abnormal comparison module is used to generate a dynamic feature threshold range based on a pre-trained dynamic threshold model, and compare it with the feature value in the compensated vibration signal feature set to obtain abnormal vibration signal data; The frequency feature extraction module is used to decompose the abnormal vibration signal data into time series and extract the frequency features of different time periods; The abnormal pattern matching module is used to perform abnormal pattern matching on the frequency characteristics of different time periods to obtain corresponding abnormal pattern matching results; The maintenance prompt module is used to send corresponding maintenance prompt information to the management terminal according to the abnormal pattern matching results.

[0093] In the above implementation, the system can not only accurately identify abnormal vibration signals of the equipment, but also send maintenance reminder information in real time according to the matching results, forming a rapid response mechanism. By introducing environmental and lubricant state compensation and dynamic threshold adjustment technology, the detection accuracy and robustness under complex working conditions are significantly improved, and false alarms and missed alarms are reduced.

[0094] As an implementation of the signal feature compensation module, the signal feature compensation module includes: A vibration signal extraction unit is used to perform spectrum decomposition and time domain analysis on the transmission shaft vibration signal in the multi-dimensional time series data to extract the vibration signal characteristics; A first compensation calculation unit, used to calculate a first compensation value of the environmental data for the vibration signal feature based on a pre-built environmental data compensation model; A second compensation calculation unit, used to calculate a second compensation value of the lubricant state data to the vibration signal characteristics based on a pre-constructed lubricant state compensation model; The compensation unit is used to compensate the vibration signal feature according to the first compensation value and the second compensation value to obtain a compensated vibration signal feature set.

[0095] An abnormality detection system based on oilfield field equipment in an embodiment of the present application can implement any of the above-mentioned abnormality detection methods, and the specific working process of each module in the abnormality detection system can refer to the corresponding process in the above-mentioned method embodiment.

[0096] In the several embodiments provided in this application, it should be understood that the provided methods and systems can be implemented in other ways. For example, the system embodiments described above are only illustrative; for example, the division of a certain module is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0097] The embodiment of the present application also discloses a computer device.

[0098] The computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned abnormality detection method based on oilfield field equipment is implemented.

[0099] The embodiment of the present application also discloses a computer-readable storage medium.

[0100] A computer-readable storage medium stores a computer program that can be loaded by a processor and execute any one of the above-mentioned abnormality detection methods based on oilfield field equipment.

[0101] Among them, computer-readable storage media can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, apparatus or device; the program code contained on the computer-readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0102] It should be noted that in the above embodiments, the description of each embodiment has different emphases, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0103] The above are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Any feature disclosed in this specification (including the abstract and drawings), unless otherwise stated, can be replaced by other equivalent or alternative features with similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

Claims

1. An abnormality detection method based on oil field field equipment, characterized in that: The anomaly detection method comprises: Receiving a transmission shaft vibration signal collected by a vibration sensor; wherein the vibration sensor is installed on a drilling rig transmission shaft of an oil field field equipment; Acquire environmental data and lubricant status data of the oil field field equipment in real time; Preprocessing the transmission shaft vibration signal, environmental data and lubricant state data to obtain multi-dimensional time series data; Extracting vibration signal features from the multidimensional time series data, and compensating the vibration signal features based on the environmental data and the lubricant state data to obtain a compensated vibration signal feature set; Generate a dynamic feature threshold range based on a pre-trained dynamic threshold model, and compare it with the feature value in the compensated vibration signal feature set to obtain abnormal vibration signal data; Decomposing the abnormal vibration signal data in time series to extract frequency features of different time periods; Perform abnormal pattern matching on the frequency characteristics of the different time periods to obtain corresponding abnormal pattern matching results; According to the abnormal pattern matching result, the corresponding maintenance prompt information is sent to the management terminal.

2. The abnormality detection method based on oil field field equipment according to claim 1 is characterized in that: The steps of extracting vibration signal features from the multidimensional time series data and compensating the vibration signal features based on the environmental data and the lubricant state data to obtain a compensated vibration signal feature set include: Performing spectrum decomposition and time domain analysis on the transmission shaft vibration signal in the multidimensional time series data to extract vibration signal features; Calculating a first compensation value of the environmental data to the vibration signal feature based on a pre-constructed environmental data compensation model; Calculating a second compensation value of the lubricant state data to the vibration signal feature based on a pre-constructed lubricant state compensation model; The vibration signal feature is compensated according to the first compensation value and the second compensation value to obtain a compensated vibration signal feature set.

3. The abnormality detection method based on oil field field equipment according to claim 1 is characterized in that: The method further includes a training step of the dynamic threshold model, wherein the training step includes: Acquire historical sample data; the historical sample data includes historical operation status data of the equipment, vibration signal sample characteristics, and pre-marked abnormal vibration signal labels; Preprocessing the historical sample data and dividing it into a training set and a test set; Inputting the equipment historical operating status data and vibration signal sample features in the training set into a pre-built random forest model to obtain a training label result; Comparing the training label result with the pre-labeled abnormal vibration signal label in the training set to obtain a training comparison result; The random forest model is iterated according to the training comparison result to obtain the trained dynamic threshold model.

4. The abnormality detection method based on oil field field equipment according to claim 3 is characterized in that: After the step of obtaining the trained dynamic threshold model, the step further includes: Inputting the historical operation status data of the equipment and the vibration signal sample features in the test set into the trained dynamic threshold model to obtain a test label result; Based on the pre-labeled abnormal vibration signal labels in the test set, the test label results are compared to obtain a test comparison result; According to the test comparison results, the performance indicators of the trained dynamic threshold model are evaluated and iterated until a preset performance index is met or a preset number of iterations is reached, thereby obtaining the trained dynamic threshold model.

5. The abnormality detection method based on oil field field equipment according to any one of claims 1 to 4, characterized in that: The step of performing abnormal pattern matching on the frequency characteristics of the different time periods to obtain corresponding abnormal pattern matching results comprises: Based on a preset abnormal pattern library, the frequency characteristics of the different time periods are matched using a characteristic pattern matching algorithm; Determine whether there is an abnormal pattern corresponding to each frequency feature in different time periods in the preset abnormal pattern library; if so, obtain a matching result of the known abnormal pattern; If not, the frequency features are classified based on a density clustering algorithm to obtain matching results of potential unknown abnormal patterns.

6. The abnormality detection method based on oil field field equipment according to claim 5 is characterized in that: After the step of obtaining the matching result of the potential unknown abnormal pattern, the following steps are also included: Standardizing the matching results of the potential unknown abnormal patterns to obtain visual result data; Sending the visualization result data to the management terminal; Receiving classification and annotation of abnormal patterns fed back by the management terminal; Performing integrity check on the abnormal pattern classification labeling; The abnormal pattern classification label is added to the preset abnormal pattern library to obtain an expanded preset abnormal pattern library.

7. An abnormality detection system based on oil field field equipment, characterized in that: The anomaly detection system comprises: A receiving module, used for receiving a transmission shaft vibration signal collected by a vibration sensor; wherein the vibration sensor is installed on a drilling rig transmission shaft of an oil field field equipment; A data acquisition module, used for acquiring environmental data and lubricant status data of the oil field field equipment in real time; A data processing module, used for preprocessing the transmission shaft vibration signal, environmental data and lubricant state data to obtain multi-dimensional time series data; A signal feature compensation module, used for extracting vibration signal features from the multi-dimensional time series data, and compensating the vibration signal features based on the environmental data and the lubricant state data to obtain a compensated vibration signal feature set; An abnormality comparison module, used for generating a dynamic feature threshold range based on a pre-trained dynamic threshold model, and comparing it with the feature value in the compensated vibration signal feature set to obtain abnormal vibration signal data; A frequency feature extraction module, used to perform time series decomposition on the abnormal vibration signal data and extract frequency features of different time periods; An abnormal pattern matching module is used to perform abnormal pattern matching on the frequency characteristics of the different time periods to obtain corresponding abnormal pattern matching results; The maintenance prompt module is used to send corresponding maintenance prompt information to the management terminal according to the abnormal pattern matching result.

8. The abnormality detection system based on oil field field equipment according to claim 7 is characterized in that: The signal characteristic compensation module comprises: A vibration signal extraction unit, used to perform spectrum decomposition and time domain analysis on the transmission shaft vibration signal in the multi-dimensional time series data, and extract vibration signal features; A first compensation calculation unit, configured to calculate a first compensation value of the environmental data for the vibration signal feature based on a pre-constructed environmental data compensation model; A second compensation calculation unit, configured to calculate a second compensation value of the lubricant state data to the vibration signal feature based on a pre-constructed lubricant state compensation model; A compensation unit is used to compensate the vibration signal feature according to the first compensation value and the second compensation value to obtain a compensated vibration signal feature set.

9. A computer device, characterized in that: The method comprises 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 method according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the method according to any one of claims 1 to 6.

Images