Compressor abnormality detection method, device and computer-readable storage medium

Abstract

The embodiment of the present application discloses a compressor abnormality detection method, device, and computer-readable storage medium. The method includes: obtaining first-site operating data of the equipment to be detected and identifying the operating condition to which the first-site operating data belongs; if an abnormality detection model corresponding to the operating condition to which the first-site operating data belongs exists, inputting the first-site operating data into the abnormality detection model corresponding to the operating condition to which the data belongs for abnormality detection, and outputting the detection result; if the detection result meets the preset conditions, adding the first-site operating data to the training data set and updating the abnormality detection model corresponding to the operating condition to which the first-site operating data belongs; the preset conditions include: the abnormal value x in the detection result is greater than or equal to the abnormal alarm threshold t and less than or equal to N times t. Through this embodiment, it is achieved that the diagnostic model has high accuracy under all operating conditions without adding additional sensors or requiring additional fault experiments, and the quality of training data can be improved and the amount of calculation can be reduced.

Description

Technical Field

[0001] The embodiments of the present application relate to electrical equipment abnormality detection technology, and in particular to a compressor abnormality detection method, device, and computer-readable storage medium. Background Art

[0002] Magnetic levitation compressors are core components of chillers. Promptly detecting abnormalities in magnetic levitation compressors can effectively improve equipment reliability. Common anomaly detection methods currently fall into three categories: using vibration signals, temperature, and pressure, based on the data source; and based on signal processing, physical models, and data-driven approaches. The use of vibration sensors undoubtedly increases the cost of fault warning, and the development and application of sufficiently accurate physical models are also challenging. Summary of the Invention

[0003] The embodiments of the present application provide a compressor anomaly detection method, device, and computer-readable storage medium, which can detect anomalies in a compressor without adding additional sensors or requiring additional fault experiments, so that the diagnostic model has high accuracy under all operating conditions, improves the quality of training data, and reduces the amount of computation.

[0004] An embodiment of the present application provides a method for detecting compressor abnormality, which may include:

[0005] Acquiring first on-site operating data during the operation of the device to be detected, and identifying the operating condition to which the first on-site operating data belongs;

[0006] If an anomaly detection model corresponding to the operating condition to which the first on-site operating data belongs already exists, inputting the first on-site operating data into the anomaly detection model corresponding to the operating condition to perform anomaly detection, and outputting a detection result;

[0007] When the detection result meets the set preset conditions, the first field operation data is added to the training data set, and the anomaly detection model corresponding to the working condition to which the input first field operation data belongs can be updated, so that the updated anomaly detection model can be used in subsequent anomaly detection to perform anomaly detection; the preset conditions include: the anomaly value x contained in the detection result is greater than or equal to the anomaly alarm threshold t, and less than or equal to N times t; N is a positive integer greater than 1; wherein N is a positive integer greater than 1.

[0008] In an exemplary embodiment of the present application, the automatically updated input first field operation data corresponding to the corresponding working condition of the anomaly detection model may include:

[0009] The anomaly detection model corresponding to the corresponding working condition is retrained using the training data after adding the first on-site operation data, thereby automatically updating the anomaly detection model, and the retrained anomaly detection model is used as the updated anomaly detection model corresponding to the working condition.

[0010] In an exemplary embodiment of the present application, the method may further include: outputting the abnormality alarm threshold t and the multiple N after each abnormality detection by the abnormality detection model; and the detection result includes the abnormality alarm threshold t and the multiple N.

[0011] In an exemplary embodiment of the present application, inputting the first field operation data into an anomaly detection model corresponding to the corresponding operating condition for anomaly detection may include:

[0012] After feature extraction and normalization are performed on the first field operation data in the anomaly detection model, a preset principal component analysis algorithm is used to perform data analysis on the first field operation data, thereby achieving anomaly detection.

[0013] In an exemplary embodiment of the present application, the principal component analysis algorithm may be used to detect whether one or more of the first field operation data input into the anomaly detection model is abnormal data;

[0014] In an exemplary embodiment of the present application, the method may further include: obtaining an anomaly detection model corresponding to the operating condition to which the first on-site operating data belongs in the following manner:

[0015] Retrieve a pre-created and trained anomaly detection model corresponding to the working condition to which the first on-site operation data belongs; or,

[0016] Directly create and train an anomaly detection model corresponding to the working condition to which the first on-site operation data belongs.

[0017] In an exemplary embodiment of the present application, creating and training anomaly detection models under different operating conditions may include:

[0018] Establish anomaly detection neural network models corresponding to different working conditions;

[0019] Acquire second on-site operating data of the device to be detected during operation, identify the operating condition to which the second on-site operating data belongs, and use the acquired second on-site operating data as training data corresponding to an anomaly detection neural network model corresponding to the identified operating condition;

[0020] For the anomaly detection neural network model under each working condition, the following training steps are performed respectively: the training data corresponding to the working condition is input into the anomaly detection neural network model corresponding to the working condition, the anomaly detection neural model is trained, and the trained anomaly detection neural model is used as the anomaly detection model corresponding to the working condition.

[0021] In an exemplary embodiment of the present application, before inputting the training data corresponding to the operating condition into the anomaly detection neural network model under the corresponding operating condition for training, the method may further include:

[0022] Each time new training data is added, detecting whether the amount of the training data under the corresponding working condition has reached a preset amount threshold;

[0023] When the amount of the training data under the corresponding working condition has reached the quantity threshold, the training data under the corresponding working condition can be input into the anomaly detection neural network model under the corresponding working condition for training;

[0024] When the amount of the training data under the corresponding working condition has not reached the quantity threshold, it is possible to continue waiting for the addition of new training data under the working condition.

[0025] In an exemplary embodiment of the present application, the equipment to be detected may include a magnetic levitation compressor; the first on-site operating data and the second on-site operating data are operating data acquired by a sensor provided by the magnetic levitation compressor during normal operation; the first on-site operating data and the second on-site operating data may include, but are not limited to, any one or more of the following:

[0026] Guide vane opening feedback, secondary exhaust pressure, inverter feedback frequency, suction pressure, AZ current, FY current, FX current, RY current, RX current, FY displacement, FX displacement, RY displacement, RX displacement and AZ displacement.

[0027] In an exemplary embodiment of the present application, the method may further include:

[0028] When there is no anomaly detection model corresponding to the operating condition to which the first on-site operation data belongs among all the anomaly detection models, an anomaly detection model corresponding to the operating condition to which the first on-site operation data belongs may be directly created and trained.

[0029] An embodiment of the present application also provides a compressor abnormality detection device, which may include but is not limited to a processor and a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, and when the instructions are executed by the processor, the compressor abnormality detection method can be implemented.

[0030] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the compressor abnormality detection method is implemented.

[0031] Compared with related technologies, the embodiments of the present application may include: obtaining first-site operating data during the operation of the equipment to be detected, and identifying the operating condition to which the first-site operating data belongs; if an anomaly detection model corresponding to the operating condition to which the first-site operating data belongs already exists, the first-site operating data can be input into the anomaly detection model corresponding to the operating condition to perform anomaly detection, and output a detection result; when the detection result meets the preset conditions, the first-site operating data is added to the training data set, and the anomaly detection model under the operating condition to which the input first-site operating data belongs can be automatically updated, so that the updated anomaly detection model can be used for anomaly detection in subsequent anomaly detection; the preset conditions include: the anomaly value x contained in the detection result is greater than or equal to the anomaly alarm threshold t, and less than or equal to N times t; where N is a positive integer greater than 1. Through this embodiment, without adding additional sensors or requiring additional fault testing, the diagnostic model has high accuracy under all operating conditions, and the training data quality can be improved, reducing the amount of computation.

[0032] Other features and advantages of the present application will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present application. Other advantages of the present application can be realized and obtained by the solutions described in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings are used to provide an understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0034] Figure 1 This is a flow chart of a compressor abnormality detection method according to an embodiment of the present application;

[0035] Figure 2 This is a block diagram of the compressor abnormality detection device according to an embodiment of the present application. DETAILED DESCRIPTION

[0036] This application describes multiple embodiments, but this description is exemplary rather than restrictive, and it will be apparent to those skilled in the art that there may be more embodiments and implementations within the scope of the embodiments described herein. Although many possible feature combinations are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are also possible. Unless specifically limited, any feature or element of any embodiment may be used in combination with any other feature or element in any other embodiment, or may replace any other feature or element in any other embodiment.

[0037] This application includes and contemplates combinations of features and elements known to those of ordinary skill in the art. The embodiments, features, and elements disclosed in this application may also be combined with any conventional features or elements to form a unique inventive solution defined by the claims. Any features or elements of any embodiment may also be combined with features or elements from other inventive solutions to form another unique inventive solution defined by the claims. Therefore, it should be understood that any feature shown and / or discussed in this application may be implemented individually or in any appropriate combination. Therefore, except for the limitations made according to the appended claims and their equivalents, the embodiments are not subject to other limitations. In addition, various modifications and changes may be made within the scope of protection of the appended claims.

[0038] In addition, when describing representative embodiments, the specification may have presented the method and / or process as a specific sequence of steps. However, to the extent that the method or process does not rely on the specific order of the steps described herein, the method or process should not be limited to the steps in the specific order described. As will be understood by those skilled in the art, other orders of steps are also possible. Therefore, the specific order of the steps set forth in the specification should not be interpreted as a limitation to the claims. In addition, the claims for the method and / or process should not be limited to performing their steps in the order written, and those skilled in the art can readily understand that these orders can be changed and still remain within the spirit and scope of the embodiments of the present application.

[0039] The embodiment of the present application provides a method for detecting abnormality of a compressor, such as Figure 1 As shown, the method may include steps S101-S103:

[0040] S101, obtaining first on-site operating data during the operation of the device to be detected, and identifying the operating condition to which the first on-site operating data belongs;

[0041] S102: If an anomaly detection model corresponding to the operating condition to which the first on-site operating data belongs already exists, input the first on-site operating data into the anomaly detection model corresponding to the operating condition to perform anomaly detection, and output a detection result;

[0042] S103. When the detection result meets the preset conditions, the first field operation data is added to the training data set, and the anomaly detection model corresponding to the working condition to which the input first field operation data belongs can be automatically updated, so that the updated anomaly detection model can be used in subsequent anomaly detection to perform anomaly detection; the preset conditions may include: the anomaly value x contained in the detection result is greater than or equal to the anomaly alarm threshold t, and less than or equal to N times t; N is a positive integer greater than 1; wherein N is a positive integer greater than 1.

[0043] In an exemplary embodiment of the present application, the device to be detected may include but is not limited to a magnetic levitation compressor. The following takes a magnetic levitation compressor as an example to illustrate the embodiment of the present application.

[0044] In the exemplary embodiment of the present application, through analysis of the current abnormality detection method of the magnetic levitation compressor, it can be seen that in the current common abnormality detection method, the use of vibration signals for fault detection undoubtedly requires additional vibration sensors and other devices, which will inevitably increase the cost of magnetic levitation compressor fault warning. If additional sensors and other devices are not used, a detection model can be used. However, the development cycle of a sufficiently accurate physical model is long and requires a large amount of training data. This abnormality detection solution also has great difficulties.

[0045] In an exemplary embodiment of the present application, in order to solve the above problems, the embodiment of the present application is based on the various sensors provided by the current magnetic levitation compressor, uses the field operation data of the magnetic levitation compressor to train the anomaly detection model, and realizes automatic updating of the anomaly detection model, thereby obtaining a highly accurate anomaly detection model.

[0046] In an exemplary embodiment of the present application, for example, in order not to add additional sensors, the data collected by various sensors in the magnetic levitation compressor unit can be used to realize the abnormal state identification of the magnetic levitation compressor; in order not to require additional fault experiments, the abnormality detection model training can be realized only through the normal operation data of the magnetic levitation compressor; and by setting the abnormality detection model under different working conditions, and automatically updating the abnormality detection model under different working conditions, it can be achieved that the detection accuracy is higher under different working conditions, and the automatic update of the abnormality detection model is realized based on real-time operating data during the use of the abnormality detection model, thereby reducing the use of experimental training data.

[0047] In an exemplary embodiment of the present application, in order to implement the solution of the embodiment of the present application, an anomaly detection model under different working conditions may be first obtained.

[0048] In an exemplary embodiment of the present application, the method may further include: obtaining an anomaly detection model corresponding to the operating condition to which the first on-site operating data belongs in the following manner:

[0049] Retrieve a previously created and trained anomaly detection model corresponding to the working condition to which the first on-site operation data belongs; or,

[0050] Directly create and train an anomaly detection model corresponding to the working condition to which the first on-site operation data belongs.

[0051] In an exemplary embodiment of the present application, the anomaly detection model can be created in advance and initially trained, and the anomaly detection model obtained after the initial training is stored so that the stored anomaly detection model can be directly called when executing the embodiment of the present application.

[0052] In an exemplary embodiment of the present application, in order to achieve comprehensive fault detection of the magnetic levitation compressor, an abnormality detection model under different working conditions may be pre-established.

[0053] In the exemplary embodiments of the present application, in other embodiments, when it is found that there is no pre-stored anomaly detection model for a certain working condition when executing the embodiment of the present application, the anomaly detection model for the working condition can be directly created and trained.

[0054] In an exemplary embodiment of the present application, creating and training anomaly detection models corresponding to different operating conditions may include:

[0055] Establish anomaly detection neural network models corresponding to different working conditions;

[0056] Acquiring second on-site operating data of the device to be detected during operation, identifying the operating condition to which the second on-site operating data belongs, and using the acquired second on-site operating data as training data corresponding to an anomaly detection neural network model under the identified operating condition;

[0057] For the anomaly detection neural network model under each working condition, the following training steps can be performed respectively: the training data corresponding to the working condition is input into the anomaly detection neural network model under the corresponding working condition and the anomaly detection neural model is trained, so that the trained anomaly detection neural model can be used as the anomaly detection model corresponding to the working condition.

[0058] In an exemplary embodiment of the present application, when creating an anomaly detection model, an anomaly detection neural network model capable of implementing the principal component analysis algorithm can be first established based on a preset principal component analysis algorithm (i.e., PCA algorithm), that is, the anomaly detection neural network model can include a mathematical model of the principal component analysis algorithm.

[0059] The PCA (Principal Component Analysis) algorithm is a commonly used data analysis method. PCA transforms raw data into a set of linearly independent representations through linear transformation. It can be used to extract the main characteristic components of the data and is often used for dimensionality reduction of high-dimensional data. Applying PCA to anomaly detection of magnetic levitation compressors can decompose the data space composed of magnetic levitation compressor field operation data into two subspaces: the principal component subspace and the residual subspace. The mathematical model of the PCA algorithm can be expressed as:

[0060] A=A1+A2;

[0061] A1=PP T A;

[0062] A2=(I-PP T )A;

[0063] Wherein, A is the collected field operation data of the magnetic levitation compressor (for example, the first field operation data and the second field operation data), A1 is the projection of X in the principal component subspace, and A2 is the projection of A in the residual subspace; wherein, PP T and I-PP T They are the projection matrix, I is the unit matrix, and P is the load matrix, which are obtained from a series of field operation data of the magnetic levitation compressor under normal working conditions. The principal component subspace mainly contains the normal values ​​of the measurement data, while the residual subspace contains faults and measurement noise.

[0064] In an exemplary embodiment of the present application, the load matrix P in the abnormality detection neural network model (such as the mathematical model of the PCA algorithm) corresponding to different working conditions is different.

[0065] In an exemplary embodiment of the present application, the principal component analysis algorithm may be used to detect whether one or more field operation data input into the anomaly detection model are abnormal data.

[0066] In an exemplary embodiment of the present application, before executing the principal component analysis algorithm, operations may include but are not limited to: feature screening (ie, feature extraction) and normalization processing.

[0067] In the exemplary embodiment of the present application, since the abnormality detection model to be created is an abnormality detection model under different working conditions, the abnormality detection neural network model can be created as different abnormality detection neural network models according to different working conditions, or an average load matrix P1 can be calculated based on the average value of a series of field operation data of the magnetic levitation compressor under normal working conditions, thereby creating a unified abnormality detection neural network model based on the mathematical model of the above-mentioned PCA algorithm and the average load matrix P1, and the unified abnormality detection neural network model is trained by training data under different working conditions to obtain abnormality detection models under different working conditions.

[0068] In an exemplary embodiment of the present application, after the abnormality detection neural network model is created, training data may be acquired to train the abnormality detection neural network model to obtain an abnormality detection model for the magnetic levitation compressor.

[0069] In an exemplary embodiment of the present application, second-site operating data of the magnetic levitation compressor during operation can be obtained and used as training data. To obtain an anomaly detection model under different operating conditions, the obtained second-site operating data can be identified using a pre-set recognition algorithm to identify the operating condition corresponding to the currently obtained second-site operating data. This data can then be used as training data for that operating condition to train the established anomaly detection neural network model.

[0070] In an exemplary embodiment of the present application, when the device to be detected is a magnetic levitation compressor, the second on-site operating data collected is operating data acquired by a sensor provided by the magnetic levitation compressor during normal operation; the second on-site operating data may include, but is not limited to, any one or more of the following:

[0071] Guide vane opening feedback, secondary exhaust pressure, inverter feedback frequency, suction pressure, AZ current, FY current, FX current, RY current, RX current, FY displacement, FX displacement, RY displacement, RX displacement and AZ displacement.

[0072] In an exemplary embodiment of the present application, before inputting the training data into the anomaly detection neural network model under the corresponding working conditions for training, the method may further include:

[0073] Whenever new training data is added, check whether the number of training data under the corresponding working condition has reached a preset number threshold;

[0074] When the amount of the training data under the corresponding working condition has reached the quantity threshold, the training data under the corresponding working condition can be input into the anomaly detection neural network model under the corresponding working condition for training;

[0075] When the amount of the training data under the corresponding working condition has not reached the quantity threshold, it is possible to continue waiting for the addition of new training data under the working condition.

[0076] In an exemplary embodiment of the present application, in order to ensure the training effect, the training data needs to reach a certain amount. In addition, in order to save data, shorten the collection time of training data, and improve training efficiency, the amount of training data cannot be too large. Therefore, a quantity threshold can be set in advance. When the second field operation data collected under a certain working condition reaches the quantity threshold, model training can be started.

[0077] In the exemplary embodiments of the present application, the data threshold can be defined according to different working conditions and different application scenarios, and the detailed numerical value of the data threshold is not limited here.

[0078] In an exemplary embodiment of the present application, when it is detected that the training data in a training data set under a certain working condition reaches a preset quantity threshold, data processing can be performed on these training data, for example, including but not limited to: data labeling, marking abnormal data in the training data; and using the processed training data for model training.

[0079] In an exemplary embodiment of the present application, if it is detected that the training data in the training data set under a certain working condition does not reach a preset quantity threshold, it is in a waiting state, waiting for new second field operation data under the working condition to be added to the training data set until the amount of training data reaches the said data threshold.

[0080] In an exemplary embodiment of the present application, after the training data is input into the anomaly detection neural network model, a pre-set principal component analysis algorithm can be used in the anomaly detection neural network model to perform data analysis on the training data, thereby realizing the detection of abnormal data and realizing the training of the anomaly detection neural network model to obtain an anomaly detection model.

[0081] In an exemplary embodiment of the present application, the trained anomaly detection model can also be verified using pre-prepared verification data to verify the detection performance of the anomaly detection model. If the verification passes, the anomaly detection model can be regarded as a qualified anomaly detection model and can be put into daily anomaly detection applications. If the verification fails, the anomaly detection model can continue to be trained until it becomes a qualified anomaly detection model and can be put into daily anomaly detection applications.

[0082] In an exemplary embodiment of the present application, an initial abnormality detection model is obtained through the above scheme. Based on the initial abnormality detection model, real-time abnormality detection can be performed on the magnetic levitation compressor, and the initial abnormality detection model can be updated in real time according to the detection results.

[0083] In an exemplary embodiment of the present application, the initial anomaly detection model is an anomaly detection model under multiple different working conditions.

[0084] In an exemplary embodiment of the present application, the method may further include:

[0085] When there is no anomaly detection model corresponding to the operating condition to which the first on-site operation data belongs among all the anomaly detection models, an anomaly detection model corresponding to the operating condition to which the first on-site operation data belongs may be directly created and trained.

[0086] In the exemplary embodiment of the present application, since the working conditions that may be considered when establishing the initial anomaly detection model are not comprehensive enough, in the actual detection process, if the first field operation data under a certain working condition is obtained but the anomaly detection model under the working condition is not found, the anomaly detection model under the working condition can be directly created to continuously improve the working condition type of the anomaly detection model in practice and enrich the anomaly detection model library.

[0087] In an exemplary embodiment of the present application, during daily abnormality detection, the real-time operating data of the magnetic levitation compressor (i.e., the first on-site operating data mentioned above) can be obtained, and the operating conditions corresponding to the first on-site operating data can be identified. After determining the operating conditions, the first on-site operating data is input into the abnormality detection model corresponding to the operating conditions, and data analysis is performed through the abnormality detection model corresponding to the operating conditions to realize abnormality detection, and the detection results are output.

[0088] In an exemplary embodiment of the present application, when the device to be detected is a magnetic levitation compressor, the first on-site operating data collected in real time is operating data acquired by a sensor provided by the magnetic levitation compressor during normal operation; the first on-site operating data may include, but is not limited to, any one or more of the following:

[0089] Guide vane opening feedback, secondary exhaust pressure, inverter feedback frequency, suction pressure, AZ current, FY current, FX current, RY current, RX current, FY displacement, FX displacement, RY displacement, RX displacement and AZ displacement.

[0090] In an exemplary embodiment of the present application, inputting the first field operation data into an anomaly detection model under the corresponding working condition to perform anomaly detection may include:

[0091] After performing feature extraction and normalization on the first on-site operation data in the anomaly detection model, a pre-set principal component analysis algorithm is used to perform data analysis on the first on-site operation data, thereby achieving anomaly detection.

[0092] In an exemplary embodiment of the present application, the principal component analysis algorithm can be used to detect whether the first field operation data input into the abnormality detection model contains abnormal data.

[0093] In an exemplary embodiment of the present application, a preset operation can be performed on all the input first-site operation data in the principal component analysis algorithm, and an operation value can be obtained. The operation value can be used as an abnormal value (which can be called an SPE value) x to characterize whether there is abnormal data in all the first-site operation data. The abnormal value x can be included in the detection result and output, so as to determine whether there is abnormal data at present based on the output abnormal value.

[0094] In an exemplary embodiment of the present application, x=||A2||2=A2T(I-PPT)A2≤δ, where δ is the confidence limit of x and can be obtained through statistical analysis.

[0095] In an exemplary embodiment of the present application, in order to solve the problem that traditional data-driven methods are heavily dependent on training data, the anomaly detection model can be automatically updated using daily operating data, so that the anomaly detection model has high accuracy under all working conditions. The embodiment of the present application proposes an automatic update strategy for the anomaly detection model.

[0096] In an exemplary embodiment of the present application, the automatic update timing of the anomaly detection model can be determined based on the detection result, for example, based on the magnitude of the anomaly value x in the anomaly detection result.

[0097] In the exemplary embodiment of the present application, the anomaly detection model needs to be continuously updated because the anomaly detection model is usually unstable in the initial stage and is easily affected by factors such as temperature changes and external interference, resulting in performance fluctuations and affecting the detection results. Therefore, when determining the timing for updating the anomaly detection model, if it is detected that the anomaly detection model has caused detection performance fluctuations due to interference from adverse factors, it is determined that the update time has arrived and the update of the anomaly detection model can be executed.

[0098] In an exemplary embodiment of the present application, when the outlier value is small enough, it can be said that the input data of the anomaly detection model has no anomaly. When the outlier value is large enough, it can be said that the input data of the anomaly detection model must have an anomaly. When the outlier value is between certain values, it means that the input data has been detected to be abnormal, but the detection result may be caused by interference from adverse factors. At this time, the anomaly detection model can be updated.

[0099] In an exemplary embodiment of the present application, the size of the outlier value x can be determined by the outlier alarm threshold t. For example, when x<t, it can be determined that the outlier value x is small enough, and at this time it can be determined that the input data of the anomaly detection model has no anomaly (that is, the existing operating data is basically similar to the data in the training data set, and the difference is very small). When t≤x≤Nt, it can be determined that the anomaly detection model needs to be updated. When x>Nt, it can be determined that the outlier value x is large enough (that is, the existing operating data is very different from the data in the training data set). At this time, it can be determined that the input data of the anomaly detection model has an anomaly.

[0100] In an exemplary embodiment of the present application, whether to add the first operating data to the training data set is determined by judging the size of the outlier x, so as to achieve the purpose of accurately screening the training data set, which can improve the quality of the training data. Compared with adding any operating data to the training data set, using a large amount of invalid operating data for training also reduces the amount of calculation and workload.

[0101] In an exemplary embodiment of the present application, the method may further include: outputting the abnormality alarm threshold t and the multiple N after each abnormality detection by the abnormality detection model; and the detection result includes the abnormality alarm threshold t and the multiple N.

[0102] In an exemplary embodiment of the present application, the abnormal alarm threshold t and the multiple N are not fixed values ​​and can be continuously optimized as the abnormality detection model is updated.

[0103] In an exemplary embodiment of the present application, the automatically updated input first field operation data corresponding to the operating condition of the abnormality detection model may include:

[0104] The anomaly detection model is retrained using the training data after adding the first field operation data, thereby automatically updating the anomaly detection model, and the retrained anomaly detection model is used as the updated anomaly detection model corresponding to the corresponding working condition.

[0105] In the exemplary embodiments of the present application, the embodiments of the present application include at least the following advantages:

[0106] 1. Based on mathematical statistics, the data collected by the unit's own temperature, pressure, current, displacement and other sensors are used, eliminating the need for additional sensors and reducing the cost of abnormality detection;

[0107] 2. Only the normal operating data of the magnetic levitation compressor unit needs to be input, and no additional fault tests are required; this saves manpower and material resources.

[0108] 3. A model self-update strategy is proposed to solve the problem that traditional data-driven methods are heavily dependent on training data. This allows the anomaly detection model to be automatically updated using daily operating data, ensuring that the anomaly detection model has high accuracy under all working conditions.

[0109] The embodiment of the present application also provides a compressor abnormality detection device 1, such as Figure 2 As shown, it may include but is not limited to a processor 11 and a computer-readable storage medium 12, wherein the computer-readable storage medium 12 stores instructions. When the instructions are executed by the processor 11, the compressor abnormality detection method can be implemented.

[0110] In the exemplary embodiments of the present application, any embodiment of the aforementioned compressor abnormality detection method is applicable to the device embodiment and will not be described in detail here.

[0111] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the compressor abnormality detection method is implemented.

[0112] In the exemplary embodiments of the present application, any embodiment of the aforementioned compressor abnormality detection method is applicable to the embodiments of the computer-readable storage medium, and will not be described one by one here.

[0113] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As is well known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable, and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those skilled in the art that communication media generally embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

Claims

1. A compressor abnormality detection method, characterized in that: The method comprises: Acquire first on-site operating data of the device to be detected, and identify the operating condition to which the first on-site operating data belongs; If an anomaly detection model corresponding to the working condition to which the first on-site operation data belongs exists, inputting the first on-site operation data into the anomaly detection model corresponding to the working condition to perform anomaly detection, and outputting a detection result; Comparing the abnormal value in the detection result with the abnormal alarm threshold t to determine whether it is necessary to update the abnormality detection model corresponding to the working condition to which the first field operation data belongs; When the detection result meets the preset conditions, it is determined that the anomaly detection model corresponding to the working condition to which the first field operation data belongs needs to be updated, the first field operation data is added to the training data set, and it is detected whether the amount of training data after the first field operation data is added under the corresponding working condition has reached a preset quantity threshold. When the quantity threshold has been reached, the anomaly detection model corresponding to the working condition is retrained using the training data after the first field operation data is added, and the retrained anomaly detection model is used as the updated anomaly detection model corresponding to the working condition; the preset conditions include: the anomaly value x contained in the detection result is greater than or equal to the anomaly alarm threshold t, and less than or equal to N times t; N is a positive integer greater than 1.

2. The compressor abnormality detection method according to claim 1, characterized in that: The detection result includes the abnormal alarm threshold t and the multiple N.

3. The compressor abnormality detection method according to claim 1, characterized in that: The inputting the first field operation data into an anomaly detection model corresponding to the corresponding working condition to perform anomaly detection includes: After feature extraction and normalization are performed on the first field operation data in the anomaly detection model, a preset principal component analysis algorithm is used to perform data analysis on the first field operation data to achieve anomaly detection.

4. The compressor abnormality detection method according to claim 1, wherein: The method further includes: obtaining an anomaly detection model corresponding to the working condition to which the first field operation data belongs in the following manner: Retrieve a pre-created and trained anomaly detection model corresponding to the working condition to which the first on-site operation data belongs, or create and train an anomaly detection model corresponding to the working condition to which the first on-site operation data belongs.

5. The compressor abnormality detection method according to claim 4, characterized in that: Create and train anomaly detection models for different operating conditions, including: Establish anomaly detection neural network models under different working conditions; Acquire second on-site operating data of the device to be tested, identify the operating condition to which the second on-site operating data belongs, and use the second on-site operating data as training data corresponding to the identified operating condition; For the anomaly detection neural network model under each working condition, the following training steps are performed respectively: the training data corresponding to the working condition is input into the anomaly detection neural network model corresponding to the working condition, the anomaly detection neural model is trained, and the trained anomaly detection neural model is used as the anomaly detection model corresponding to the working condition.

6. The compressor abnormality detection method according to claim 5, characterized in that: Before inputting the training data corresponding to the operating condition into the anomaly detection neural network model corresponding to the operating condition, the method further includes: Whenever new training data is added, check whether the number of training data under the corresponding working condition reaches the preset number threshold; When the amount of training data under the corresponding working condition reaches the amount threshold, the training data under the corresponding working condition is input into the anomaly detection neural network model corresponding to the corresponding working condition; When the amount of training data under the corresponding working condition does not reach the quantity threshold, wait for new training data under the working condition to be added.

7. The compressor abnormality detection method according to claim 5, characterized in that: The device to be detected includes a magnetic levitation compressor; the first on-site operating data and the second on-site operating data are operating data obtained by a sensor provided with the magnetic levitation compressor when the compressor is operating normally; The first on-site operation data and the second on-site operation data each include any one or more of the following: Secondary exhaust pressure, suction pressure, guide vane opening feedback, inverter feedback frequency, AZ current, FX current, FY current, RX current, RY current, FX displacement, FY displacement, RX displacement, RY displacement and AZ displacement.

8. A compressor abnormality detection device, characterized in that: The method comprises a processor and a computer-readable storage medium, wherein the computer-readable storage medium stores instructions. When the instructions are executed by the processor, the compressor abnormality detection method according to any one of claims 1 to 7 is implemented.

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

Images