Bearing anomaly detection method and system

Abstract

The invention discloses a bearing anomaly detection method and system, and the method comprises the following steps: obtaining the historical operation data of a current motion sequence of a bearing, and constructing a first parameter trend model based on the historical operation data of the current motion sequence of the bearing; acquiring historical operation data corresponding to the bearing attitude, and constructing a second parameter trend model based on the historical operation data corresponding to the bearing attitude; acquiring first parameter prediction data according to the current time point, and acquiring second parameter prediction data according to the current bearing attitude; obtaining first difference data corresponding to the first parameter prediction data and second difference data corresponding to the second parameter prediction data; and judging whether the bearing is abnormal according to the first difference data and the second difference data. The method has the beneficial effects that the influence of space and time on the bearing operation condition can be shown through the operation data corresponding to the bearing posture change and the operation data corresponding to the bearing operation time change, and the situation that the bearing does not operate in the expected state is judged in time.

Description

Technical Field

[0001] The present application relates to the field of bearing detection technology, and in particular to a bearing abnormality detection method and system. Background Art

[0002] As a key rotating component in mechanical equipment, the operating status of bearings directly affects the stability and service life of the equipment. Bearing detection can timely detect bearing defects and failures, thereby timely avoiding equipment failures caused by bearings and improving the overall reliability and stability of the equipment. There are many ways to detect bearings in the existing technology, including temperature detection, oil detection, acoustic emission, vibration analysis, geometric dimension measurement, surface quality inspection, clearance measurement, and material and heat treatment detection.

[0003] However, although there are multi-dimensional detection forms for bearing detection in the existing technology, it still relies on the accuracy of historical operating data and the accuracy of model training. It not only has a large amount of calculation but also has the problem of poor adaptability.

[0004] The Chinese patent "Rolling Bearing Abnormality Detection Device and Rolling Bearing Abnormality Detection Method", publication number: CN118435034A, publication date: August 2, 2024, specifically discloses the spectrum of the vibration data of the vibration generated in the rolling bearing, and according to the obtained spectrum, the peak frequency is detected within a specified frequency range including the theoretical frequency that brings a peak on the spectrum when an abnormality occurs, and based on the time-dependent frequency variation of the difference between the peak frequency and the pre-set reference frequency and the detected peak frequency, it is judged whether the rolling bearing has an abnormality. In this scheme, the abnormality of the rolling bearing is detected by the change in the vibration frequency over time, which is limited to the application field of the rolling bearing, and the adaptability of the bearing in other application scenarios, such as high-temperature operation, high-pressure operation, etc., is poor.

[0005] The patent "A multi-dimensional bearing stability monitoring system", publication number: CN118518358A, publication date: August 20, 2024, specifically discloses: bearing operation management module, bearing operation data detection module, bearing operation status evaluation module, bearing lubrication status evaluation module, fault diagnosis and prediction module and bearing stability analysis module. By grouping the bearings in the same working area into the same spatial group for monitoring, the bearing operation status of each working area can be more accurately grasped. Although this solution uses multiple data sources for comprehensive analysis, it relies on the accuracy of each data in the historical data. Once there is error data in the historical data, it is easy to ignore the abnormal situation of the bearing. Summary of the invention

[0006] In order to solve the problem in the prior art that bearing detection depends on the accuracy of historical operating data and the accuracy of model training, the present application provides a bearing anomaly detection method and system, which can demonstrate the influence of space and time on the operation of the bearing through the operating data corresponding to the bearing posture change and the operating data corresponding to the bearing operation time change. When the predicted data is significantly different from the real-time operating data of the bearing, it can be judged in time that the bearing is not operating in the expected state. The dual models verify each other. If a certain data is abnormal, the abnormality can still be judged by the other model, thereby ensuring accuracy while reducing the amount of calculation for model construction and improving efficiency.

[0007] To achieve the above technical purpose, a technical solution provided by the present application is a bearing abnormality detection method, comprising the following steps: obtaining historical operation data of the current action sequence of the bearing, and constructing a first parameter trend model based on the historical operation data of the current action sequence of the bearing; obtaining historical operation data corresponding to the bearing posture, and constructing a second parameter trend model based on the historical operation data corresponding to the bearing posture; obtaining first parameter prediction data according to the current time point and the first parameter trend model, and obtaining second parameter prediction data according to the current bearing posture and the second parameter trend model; obtaining real-time operation data of the bearing, obtaining first difference data between the real-time operation data of the bearing and the first parameter prediction data, and second difference data between the real-time operation data of the bearing and the second parameter prediction data; judging whether the bearing is abnormal by the first difference data and the second difference data, and if so, outputting a bearing abnormality signal. By constructing the first parameter trend model and the second parameter trend model, the bearing operation state is predicted from time and space respectively. The first parameter trend model is based on the recent operation data of the bearing, reflecting the recent overall performance change trend of the bearing, while the second parameter trend model is based on the operation data under the same posture, reflecting the performance of the bearing under specific working conditions. The two parameter prediction data can be mutually verified, reducing the possible deviation or misjudgment of a single model, and improving the sensitivity and robustness of detection.

[0008] Furthermore, the operating data at least includes vibration data, temperature data and pressure data.

[0009] Furthermore, the acquisition of historical operation data of the current motion sequence of the bearing includes: acquiring operation data of the bearing between the current time point and the time point of the last change in the motion state of the bearing as the historical operation data of the current motion sequence of the bearing.

[0010] Furthermore, the constructing of the second parameter trend model based on the historical operating data corresponding to the bearing posture includes: constructing a second parameter trend model corresponding to each bearing posture based on each bearing posture and the historical operating data corresponding to each bearing posture; or, constructing a second parameter trend model including all bearing postures based on all bearing postures and the historical operating data corresponding to all bearing postures.

[0011] Furthermore, it also includes: the determination of whether the bearing is abnormal based on the first difference data and the second difference data includes: if the first difference data is greater than the preset parameter difference threshold, the bearing is considered to be abnormal; or / and, if the second difference data is greater than the preset parameter difference threshold, the bearing is considered to be abnormal.

[0012] Furthermore, it also includes: obtaining historical abnormality detection results, and outputting an abnormality judgment feature set based on the historical abnormality detection results; judging whether the bearing is abnormal by using the first difference data, the second difference data and the abnormality judgment feature set, and if so, outputting a bearing abnormality signal.

[0013] Furthermore, outputting anomaly judgment feature sets based on historical anomaly detection results includes: dividing historical anomaly detection results based on normal results and abnormal results, extracting features from the divided historical anomaly detection results, and obtaining a normal feature set and an abnormal feature set.

[0014] Furthermore, the method of judging whether the bearing is abnormal by using the first difference data, the second difference data and the abnormal judgment feature set also includes: if one of the first difference data and the second difference data is greater than a preset parameter difference threshold, extracting features of the real-time operation data of the bearing, matching the features of the real-time operation data of the bearing with features of the abnormal judgment feature set, and judging whether the bearing is abnormal by the matching result; if both the first difference data and the second difference data are less than or equal to the preset parameter difference threshold, judging that the bearing is normal; if both the first difference data and the second difference data are greater than the preset parameter difference threshold, judging that the bearing is abnormal.

[0015] Furthermore, it also includes: comparing the first difference data and the second difference data to obtain the third difference data; when the third difference data is greater than the preset model difference threshold, obtaining the historical first difference data and the historical second difference data, obtaining the correction factor based on the historical fluctuation of the difference data, and correcting the parameter trend model based on the correction factor.

[0016] A technical solution provided by the present application is a bearing abnormality detection system for implementing the method as described above, comprising: a sensor unit for acquiring operating data of the bearing; a model building unit for building a first parameter trend model and a second parameter trend model based on the operating data output by the sensor unit; and a judgment and analysis unit for outputting a bearing abnormality judgment result based on the real-time operating data output by the sensor unit and the first parameter trend model and the second parameter model.

[0017] The beneficial effects of the present application are as follows: 1. The operating data corresponding to the changes in the bearing posture and the operating data corresponding to the changes in the bearing operating time can show the influence of space and time on the bearing operation condition, and when the predicted data is significantly different from the real-time operating data of the bearing, it can be judged in time that the bearing is not operating in the expected state.

[0018] 2. By focusing on the historical operating data of the current motion sequence of the bearing, targeted analysis of the bearing in a specific motion state can be achieved, eliminating the interference of irrelevant factors. In a continuous motion sequence, the operating data of the bearing may have subtle but important changes. Obtaining historical data during this period of time will help capture these subtle changes and will not be ignored due to excessive data volume. At the same time, it reduces the amount of data processed, improves the accuracy of fault diagnosis, and improves the efficiency of prediction output.

[0019] 3. When either the first difference data or the second difference data is greater than the preset parameter difference threshold, the characteristics of the real-time operation data of the bearing are matched with the characteristics of the abnormal judgment feature set, and further judgment is performed to further determine whether the bearing operation is abnormal in the form of a secondary judgment, thereby improving the accuracy of the judgment and avoiding false alarms.

[0020] 4. The parameters in the historical operating data corresponding to the difference data whose historical fluctuation values ​​exceed the preset fluctuation threshold are taken as the values ​​to be corrected. A correction function is constructed according to the performance value of the correction factor in the historical operating data corresponding to the difference data whose historical fluctuation values ​​meet the preset fluctuation threshold. The value to be corrected is compensated with the changing trend of each parameter, thereby correcting the error parameter value in the parameter trend model, and then reconstructing the entire parameter trend model to realize the correction and optimization of the training data, avoid the influence of erroneous data on the output value of the parameter trend model, and further improve the accuracy of bearing abnormality judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of the bearing abnormality detection method of this application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application is further described in detail below in conjunction with the drawings and examples. It should be understood that the specific implementation method described here is only an optimal embodiment of the present application, which is only used to explain the present application and does not limit the scope of protection of the present application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0023] like Figure 1 As shown, as the first embodiment of the present application, the bearing abnormality detection method includes the following steps: Acquire historical operation data of the current motion sequence of the bearing, and build a first parameter trend model based on the historical operation data of the current motion sequence of the bearing; Acquire historical operating data corresponding to the bearing posture, and construct a second parameter trend model based on the historical operating data corresponding to the bearing posture; Acquire first parameter prediction data according to the current time point and the first parameter trend model, and acquire second parameter prediction data according to the current bearing posture and the second parameter trend model; Acquire real-time operation data of the bearing, acquire first difference data between the real-time operation data of the bearing and the first parameter prediction data, and acquire second difference data between the real-time operation data of the bearing and the second parameter prediction data; The first difference data and the second difference data are used to determine whether the bearing is abnormal. If so, a bearing abnormality signal is output.

[0024] By constructing the first parameter trend model and the second parameter trend model, the bearing operating status is predicted from time and space respectively. The first parameter trend model is based on the recent operating data of the bearing, reflecting the recent overall performance change trend of the bearing, while the second parameter trend model is based on the operating data in the same posture, reflecting the performance of the bearing under specific working conditions. The two parameter prediction data can verify each other, reducing the possible deviation or misjudgment of a single model, and improving the sensitivity and robustness of the detection.

[0025] The operation data at least includes vibration data, temperature data and pressure data. The operation data of the bearing is obtained by installing various types of sensors on the bearing, such as vibration sensors, temperature sensors and pressure sensors. The historical operation data of the current action sequence of the bearing is obtained by: The operation data of the bearing between the current time point and the time point when the motion state of the bearing changes last time are obtained as the historical operation data of the current motion sequence of the bearing.

[0026] By focusing on the historical operating data of the current motion sequence of the bearing, targeted analysis of the bearing in a specific motion state can be achieved, eliminating interference from irrelevant factors. In a continuous motion sequence, there may be subtle but important changes in the operating data of the bearing. Obtaining historical data during this period helps capture these subtle changes and prevents them from being ignored due to excessive data volume. At the same time, it reduces the amount of data to be processed, improves the accuracy of fault diagnosis, and improves the efficiency of prediction output.

[0027] The historical operation data corresponding to the bearing posture are obtained including: All historical operation data including the bearing posture are obtained, and the historical operation data are screened based on the same bearing posture to obtain the historical operation data corresponding to the bearing posture.

[0028] At this time, the second parameter trend model is constructed based on the historical operation data corresponding to the bearing posture, including: A second parameter trend model corresponding to each bearing posture is constructed based on each bearing posture and the historical operating data corresponding to each bearing posture; or, a second parameter trend model including all bearing postures is constructed based on all bearing postures and the historical operating data corresponding to all bearing postures.

[0029] In the first case, the second trend model corresponding to the bearing posture is constructed based on the bearing posture. Each posture corresponds to a targeted second trend model, which can more accurately reflect the operating characteristics and degradation laws of the bearing under each posture, and can more accurately capture the subtle changes of the bearing under each posture, thereby improving the accuracy of fault diagnosis. At the same time, the model under each posture is relatively independent, which is convenient for model training, verification and updating. When the data under a certain posture changes, only the corresponding model needs to be adjusted without affecting the models of other postures, thus reducing the amount of data calculation at the same time, ensuring accuracy while improving calculation efficiency.

[0030] In the second case, a second parameter trend model containing all bearing postures is constructed based on all postures. The model covers the operating data under all postures and can fully reflect the performance changes and degradation laws of the bearing under different postures. Since the model is constructed based on data from multiple postures, it has good generalization ability and can adapt to the operating requirements of the bearing under different working conditions. When the bearing has a new posture or working condition, the model can still give a relatively accurate prediction and judgment, which helps to discover the correlation and regularity between different postures and has a wide range of applications.

[0031] The two model construction methods can be selected according to actual needs. For example, in application scenarios where higher efficiency is required but the possibility of new bearing working conditions is small, the first construction method can be selected, and in application scenarios where new bearing working conditions are likely to occur, the second construction method can be selected.

[0032] In this embodiment, judging whether the bearing is abnormal by using the first difference data and the second difference data includes: If the first difference data is greater than the preset parameter difference threshold, it is considered that the bearing is abnormal; or / and, If the second difference data is greater than the preset parameter difference threshold, the bearing is considered abnormal.

[0033] When the first difference data and the second difference data are arbitrarily greater than the preset parameter difference threshold, the bearing is considered abnormal, thereby avoiding the neglect of abnormalities caused by a certain model prediction deviation and improving the accuracy of fault detection.

[0034] In other embodiments, a first preset parameter difference threshold corresponding to the first difference data and a second preset parameter difference threshold corresponding to the second difference data may also be set. When the first difference data is greater than the first preset parameter difference threshold or / and the second difference data is greater than the second preset parameter difference threshold, the bearing is considered abnormal. By setting preset parameter difference thresholds corresponding to different models, the judgment result is more accurate, avoiding the mismatch of the difference thresholds caused by model differences, and further improving the accuracy of bearing abnormality detection.

[0035] The bearing abnormality detection method also includes: Obtain historical anomaly detection results, and output anomaly judgment feature set based on the historical anomaly detection results; The first difference data, the second difference data and the abnormality judgment feature set are used to judge whether the bearing is abnormal. If so, a bearing abnormality signal is output.

[0036] The historical anomaly detection results are the results of manual bearing anomaly detection, that is, manual experience. The abnormality judgment feature set output based on the historical anomaly detection results includes: The historical anomaly detection results are divided into normal results and abnormal results, and features are extracted from the divided historical anomaly detection results to obtain a normal feature set and an abnormal feature set.

[0037] The alarm anomalies are detected manually, and the historical anomaly detection results include the cases that are manually determined to be misjudgments and the cases that are manually determined to be abnormal. The historical anomaly detection results are divided into the cases that are manually determined to be misjudgments and the cases that are manually determined to be abnormal based on the normal results and the abnormal results. The normal features of the corresponding operating data are obtained based on the cases where the process is judged to be abnormal but manually judged to be normal. The abnormal features of the corresponding operating data are obtained based on the cases where the process is judged to be abnormal and manually judged to be abnormal, thereby constructing a normal feature set and an abnormal feature set.

[0038] At this time, judging whether the bearing is abnormal by using the first difference data, the second difference data and the abnormality judgment feature set includes: when the first difference data and the second difference data are arbitrarily greater than a preset parameter difference threshold, extracting features of the real-time operation data of the bearing; Match the features of the real-time operation data of the bearing with the features of the abnormality judgment feature set, and use the matching results to determine whether the bearing is abnormal.

[0039] When either the first difference data or the second difference data is greater than the preset parameter difference threshold, the characteristics of the real-time operation data of the bearing are matched with the characteristics of the abnormal judgment feature set, and further judgment is performed to further determine whether the bearing operation is abnormal in the form of a secondary judgment, thereby improving the accuracy of the judgment and avoiding false alarms.

[0040] In some other embodiments, judging whether the bearing is abnormal by using the first difference data, the second difference data and the abnormality judgment feature set further includes: If one of the first difference data and the second difference data is greater than a preset parameter difference threshold, extract the features of the real-time operation data of the bearing, match the features of the real-time operation data of the bearing with the features of the abnormality judgment feature set, and judge whether the bearing is abnormal based on the matching result; If the first difference data and the second difference data are both less than or equal to the preset parameter difference threshold, it is determined that the bearing has no abnormality; If both the first difference data and the second difference data are greater than the preset parameter difference threshold, it is determined that the bearing is abnormal.

[0041] In this embodiment, the matching of the abnormality judgment feature set is performed only when one of the first difference data and the second difference data is greater than the preset parameter difference threshold and the other is less than or equal to the preset parameter difference threshold, thereby reducing redundant judgments and improving abnormality judgment efficiency.

[0042] Extract the features of the real-time operation data of the bearing, match the features of the real-time operation data of the bearing with the features of the abnormality judgment feature set, and use the matching results to determine whether the bearing is abnormal, including: If the characteristics of the real-time operation data of the bearing match the normal feature set in the abnormal judgment feature set, the bearing is judged to be normal; if the characteristics of the real-time operation data of the bearing match the abnormal feature set in the abnormal judgment feature set, the bearing is judged to be abnormal.

[0043] Whether the bearing is operating abnormally is determined based on whether the characteristics of the real-time operating data of the bearing conform to the normal feature set or the abnormal feature set. In some embodiments, the operating data that is determined to be normal by the process but is found to be abnormal by human screening is also included in the historical abnormality detection results. At this time, the abnormal feature set also includes abnormal features that are determined to be normal by the process but are found to be abnormal by human screening, which can effectively avoid some special abnormalities from being ignored and improve the accuracy of abnormal screening.

[0044] As a second embodiment of the present application, the bearing abnormality detection method further includes: Comparing the first difference data with the second difference data to obtain third difference data; When the third difference data is greater than a preset model difference threshold, the historical first difference data and the historical second difference data are obtained, a correction factor is obtained based on the historical fluctuation of the difference data, and the parameter trend model is corrected based on the correction factor.

[0045] When the difference in the predicted values ​​output by the two parameter trend models is greater than the preset model difference threshold, the corresponding model to be corrected and the correction factor are obtained according to the historical fluctuations of the difference data, so that when a large error occurs in any parameter trend model, correction can be made in time, ensuring in real time that the two models are in a relatively accurate state at any time of abnormal judgment, thereby ensuring the accuracy of mutual verification of the dual models.

[0046] In this embodiment, obtaining the historical first difference data and the historical second difference data, and obtaining the correction factor based on the historical fluctuation matching of the difference data includes: Obtaining a historical fluctuation value of the first difference data according to the historical first difference data, and obtaining a historical fluctuation value of the second difference data according to the historical second difference data; The historical operation data corresponding to the difference data whose historical fluctuation value exceeds the preset fluctuation threshold is selected, and the features in the historical operation data are extracted as the correction factor.

[0047] The parameter trend model based on the correction factor includes: The historical operating data corresponding to the correction factor is compensated based on the interpolation method, and the parameter trend model is reconstructed according to the compensated correction factor.

[0048] Compensation of historical operating data corresponding to the correction factor based on interpolation method includes: Obtain historical operation data corresponding to difference data whose historical fluctuation value meets a preset fluctuation threshold, and extract correction factor parameters in the historical operation data based on the correction factor; Based on the interpolation method, a correction function is constructed with correction factor parameters to compensate for the historical operating data corresponding to the difference data whose historical fluctuation values ​​exceed the preset fluctuation threshold.

[0049] In this embodiment, the parameters in the historical operating data corresponding to the difference data whose historical fluctuation values ​​exceed the preset fluctuation threshold are taken as the values ​​to be corrected, and a correction function is constructed according to the performance value of the correction factor in the historical operating data corresponding to the difference data whose historical fluctuation values ​​meet the preset fluctuation threshold. The value to be corrected is compensated with the changing trend of each parameter, thereby correcting the error parameter value in the parameter trend model, and then reconstructing the entire parameter trend model, realizing the correction and optimization of the training data, avoiding the influence of erroneous data on the output value of the parameter trend model, and further improving the accuracy of bearing abnormality judgment.

[0050] As a third embodiment of the present application, a bearing abnormality detection system includes: A sensor unit, used to obtain operating data of the bearing; A model building unit, used to build a first parameter trend model and a second parameter trend model according to the operation data output by the sensor unit; The judgment and analysis unit is used to output the bearing abnormality judgment result according to the real-time operation data output by the sensor unit and the first parameter trend model and the second parameter model.

[0051] In this embodiment, the sensor unit includes at least a posture sensor, a temperature sensor, a vibration sensor and a pressure sensor. The bearing posture is obtained through the posture sensor, and then the corresponding historical operating data is obtained based on the bearing posture, and a second parameter trend model is constructed. The operating data corresponding to the change of the bearing posture and the operating data corresponding to the change of the bearing operation time can show the influence of space and time on the operation of the bearing, and then when the predicted data is significantly different from the real-time operation data of the bearing, it can be judged in time that the bearing is not operating in the expected state. It can be understood that in this application, the abnormal condition of the bearing comes from reasons other than the end of life. Therefore, according to the normal operation, the data output by the first parameter trend model and the second parameter trend model are both parameter states of the normal operation of the bearing. If the real-time operation data is significantly different from the predicted operation data at this time, it proves that the bearing is in an abnormal operation state.

[0052] The specific implementation described above is a preferred implementation of the bearing abnormality detection method and system of the present application, and is not intended to limit the specific implementation scope of the present application. The scope of the present application includes but is not limited to the specific implementation embodiment. All equivalent changes made in accordance with the shape and structure of the present application are within the protection scope of the present application.

Claims

1. A bearing abnormality detection method, characterized in that: The steps include: Acquire historical operation data of the current motion sequence of the bearing, and build a first parameter trend model based on the historical operation data of the current motion sequence of the bearing; Acquire historical operating data corresponding to the bearing posture, and construct a second parameter trend model based on the historical operating data corresponding to the bearing posture; Acquire first parameter prediction data according to the current time point and the first parameter trend model, and acquire second parameter prediction data according to the current bearing posture and the second parameter trend model; Acquire real-time operation data of the bearing, acquire first difference data between the real-time operation data of the bearing and the first parameter prediction data, and acquire second difference data between the real-time operation data of the bearing and the second parameter prediction data; The first difference data and the second difference data are used to determine whether the bearing is abnormal. If so, a bearing abnormality signal is output.

2. The bearing abnormality detection method according to claim 1, characterized in that: The operating data at least includes vibration data, temperature data and pressure data.

3. The bearing abnormality detection method according to claim 1, characterized in that: The acquisition of historical operation data of the current motion sequence of the bearing includes: The operation data of the bearing between the current time point and the time point when the motion state of the bearing changes last time are obtained as the historical operation data of the current motion sequence of the bearing.

4. The bearing abnormality detection method according to claim 1, characterized in that: The constructing of the second parameter trend model based on the historical operation data corresponding to the bearing posture includes: Based on each bearing posture and the historical operation data corresponding to each bearing posture, a second parameter trend model corresponding to each bearing posture is constructed respectively; or, A second parameter trend model including all bearing postures is constructed based on all bearing postures and historical operation data corresponding to all bearing postures.

5. The bearing abnormality detection method according to claim 1, characterized in that: Also includes: The determining whether the bearing is abnormal by using the first difference data and the second difference data includes: If the first difference data is greater than the preset parameter difference threshold, it is considered that the bearing is abnormal; or / and, If the second difference data is greater than the preset parameter difference threshold, the bearing is considered abnormal.

6. The bearing abnormality detection method according to claim 1, characterized in that: Also includes: Obtain historical anomaly detection results, and output anomaly judgment feature set based on the historical anomaly detection results; The first difference data, the second difference data and the abnormality judgment feature set are used to judge whether the bearing is abnormal. If so, a bearing abnormality signal is output.

7. The bearing abnormality detection method according to claim 6, characterized in that: The output of the abnormality judgment feature set based on the historical abnormality detection results includes: The historical anomaly detection results are divided into normal results and abnormal results, and features are extracted from the divided historical anomaly detection results to obtain a normal feature set and an abnormal feature set.

8. The bearing abnormality detection method according to claim 7, characterized in that: The determining whether the bearing is abnormal by using the first difference data, the second difference data and the abnormality determination feature set also includes: If one of the first difference data and the second difference data is greater than a preset parameter difference threshold, extract the features of the real-time operation data of the bearing, match the features of the real-time operation data of the bearing with the features of the abnormality judgment feature set, and judge whether the bearing is abnormal based on the matching result; If the first difference data and the second difference data are both less than or equal to the preset parameter difference threshold, it is determined that the bearing has no abnormality; If both the first difference data and the second difference data are greater than the preset parameter difference threshold, it is determined that the bearing is abnormal.

9. The bearing abnormality detection method according to claim 7, characterized in that: Also includes: Comparing the first difference data with the second difference data to obtain third difference data; When the third difference data is greater than a preset model difference threshold, the historical first difference data and the historical second difference data are obtained, a correction factor is obtained based on the historical fluctuation of the difference data, and the parameter trend model is corrected based on the correction factor.

10. A bearing abnormality detection system, used to implement the method according to any one of claims 1 to 9, characterized in that: include: A sensor unit, used to obtain operating data of the bearing; A model building unit, used to build a first parameter trend model and a second parameter trend model according to the operation data output by the sensor unit; The judgment and analysis unit is used to output the bearing abnormality judgment result according to the real-time operation data output by the sensor unit and the first parameter trend model and the second parameter model.

Images