A method and system for monitoring and managing permanent magnetic fluid seal

Abstract

The application provides a permanent magnet fluid seal monitoring management method and system, relates to the technical field of permanent magnet fluid seal, and the method comprises the following steps: collecting a pressure time sequence signal in a sealing cavity and a coil working current signal of an excitation coil; collecting a coil impedance current signal at both ends of the excitation coil by injecting an impedance test signal into a driving circuit according to impedance analysis of a front end; constructing a health benchmark mapping model; obtaining a residual feature vector set and extracting a high-dimensional residual feature vector set; obtaining a seal failure mode through pattern recognition according to the high-dimensional residual feature vector set; and sending the seal failure mode to a seal monitoring management terminal to generate a reminder information. The application solves the technical problem that the monitoring means is relatively single in the running process of the permanent magnet fluid seal in the prior art, it is difficult to accurately monitor the running state of the seal, and the abnormal state of the seal under complex working conditions is difficult to identify in time.

Description

Technical Field

[0001] This invention relates to the field of permanent magnet fluid sealing technology, specifically to a permanent magnet fluid sealing monitoring and management method and system. Background Technology

[0002] Permanent magnet fluid sealing technology is a non-contact sealing technology that uses a magnetic field to confine a magnetic fluid to form a sealing barrier. It boasts advantages such as low leakage rate, minimal wear, and suitability for high-speed rotating environments, and is widely used in vacuum equipment, chemical equipment, and high-reliability rotating machinery. Traditional permanent magnet fluid sealing structures typically employ a fixed permanent magnet to generate a constant magnetic field, causing the magnetic fluid to form a stable sealing layer in the sealing gap region, thereby achieving media isolation.

[0003] During long-term operation, pressure fluctuations inside the sealed cavity, differences in the sealing contact structure, and changes in the local distribution of magnetofluid can easily lead to a gradual decline in sealing performance. When local anomalies occur in the sealing structure, they are usually accompanied by changes in the magnetofluid constraint state, changes in the electromagnetic circuit load, and abnormal pressure response in the sealed cavity. If these abnormalities are not identified in time, they can easily develop into local leakage or sealing failure.

[0004] Existing permanent magnet fluid seal monitoring methods typically rely on single pressure detection, leakage detection, or simple current monitoring to determine the seal status. These methods provide limited information dimensionality and are insufficient to accurately reflect the true operating state within the seal structure. Especially for different types of seal contact surfaces, the variations in their corresponding magnetofluid distribution characteristics and failure mechanisms make it difficult for traditional monitoring methods to effectively distinguish early failure characteristics under different structures, resulting in low accuracy in identifying seal anomalies. Summary of the Invention

[0005] This application provides a permanent magnet fluid seal monitoring and management method and system, which aims to solve the technical problem that the existing permanent magnet fluid seal has relatively simple monitoring methods during operation, making it difficult to accurately monitor the seal operation status and making it difficult to identify abnormal seal conditions in a timely manner under complex working conditions.

[0006] The first aspect disclosed in this application provides a method for monitoring and managing permanent magnet fluid seals. The method includes: acquiring a pressure time-series signal and an excitation coil operating current signal within a sealed cavity; the sealed cavity is sealed by controlling a drive circuit of the excitation coil to act on the permanent magnet fluid; the drive circuit of the excitation coil includes an impedance analysis front-end; an impedance test signal is injected into the drive circuit based on the impedance analysis front-end; and coil impedance current signals are acquired at both ends of the excitation coil. A health benchmark mapping model is constructed based on pressure time-series signal samples, coil operating current signal samples, and coil impedance current signal samples. The pressure time-series signal, coil operating current signal, and coil impedance current signal are input into the health benchmark mapping model to obtain a residual feature vector set; a high-dimensional residual feature vector set is extracted from the residual feature vector set; pattern recognition is performed based on the high-dimensional residual feature vector set to obtain a seal failure mode; and the seal failure mode is sent to a seal monitoring and management terminal to generate an alert message.

[0007] The second aspect of this application discloses a permanent magnet fluid seal monitoring and management system. The system is used in the aforementioned permanent magnet fluid seal monitoring and management method. The system includes: a first signal acquisition module for acquiring pressure time-series signals and excitation coil operating current signals within a sealed cavity, wherein the sealed cavity is sealed by controlling the driving circuit of the excitation coil to act on the permanent magnet fluid; a second signal acquisition module, wherein the driving circuit of the excitation coil includes an impedance analysis front-end, injects an impedance test signal into the driving circuit based on the impedance analysis front-end, and acquires coil impedance current signals at both ends of the excitation coil; a model construction module for constructing a health benchmark mapping model based on pressure time-series signal samples, coil operating current signal samples, and coil impedance current signal samples; a feature vector extraction module for inputting the pressure time-series signals, coil operating current signals, and coil impedance current signals into the health benchmark mapping model to obtain a residual feature vector set, and extracting a high-dimensional residual feature vector set from the residual feature vector set; and a reminder information generation module for performing pattern recognition based on the high-dimensional residual feature vector set to obtain a seal failure mode, and sending the seal failure mode to a seal monitoring and management terminal to generate a reminder information.

[0008] One or more technical solutions provided in this application have at least the following beneficial effects:

[0009] By synchronously acquiring the pressure time-series signal within the sealed cavity and the coil operating current signal of the excitation coil, the operating status and magnetic field drive status of the permanent magnet fluid seal are monitored simultaneously. This establishes a correspondence between the pressure status and the excitation drive status, providing fundamental data support for subsequent seal health status analysis and enabling dynamic sensing of the permanent magnet fluid sealing capability. By setting an impedance analysis front-end and injecting impedance test signals into the excitation coil drive circuit, online detection of the excitation coil impedance response characteristics is achieved. This allows the coil impedance changes to reflect the magnetic circuit changes in the permanent magnet fluid sealing area, further improving the detection capability for latent failure states such as local leakage, changes in sealing gaps, and abnormal magnetic fields. Finally, a health benchmark is constructed based on pressure time-series signal samples, coil operating current signal samples, and coil impedance current signal samples. The model is used to model the coupling relationship between pressure and electromagnetic response under healthy sealing conditions, thus forming a health benchmark for the normal operation of the sealing system and providing a reference for subsequent abnormal state identification. By inputting real-time acquired signals into the health benchmark mapping model to obtain residual feature vectors, and further extracting high-dimensional residual feature vector sets, the deviation between the current operating state and the healthy state can be quantitatively analyzed. At the same time, the ability to express local abnormal features is enhanced by combining structural types, thereby improving the distinguishability between different sealing failure states. By performing pattern recognition on the high-dimensional residual feature vector set, the sealing failure mode corresponding to the current permanent magnet fluid sealing system can be accurately identified, and the identification results can be sent to the sealing monitoring and management terminal to generate reminder information, thereby realizing online health monitoring, fault early warning, and failure state management of the permanent magnet fluid sealing system.

[0010] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of a permanent magnet fluid seal monitoring and management method provided in an embodiment of this application.

[0012] Figure 2 This is a schematic diagram of a permanent magnet fluid seal monitoring and management system provided in an embodiment of this application.

[0013] Explanation of reference numerals in the attached figures: First signal acquisition module 10, Second signal acquisition module 20, Model construction module 30, Feature vector extraction module 40, Reminder information generation module 50. Detailed Implementation

[0014] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0015] Example 1, as Figure 1 As shown in the figure, this application provides a method for monitoring and managing permanent magnet fluid seals, the method comprising: The pressure time series signal and the coil operating current signal of the excitation coil in the sealed cavity are collected. The sealed cavity is sealed by controlling the drive circuit of the excitation coil to act on the permanent magnet fluid.

[0016] The internal pressure of the permanent magnet fluid sealing area is collected in real time by a pressure sensor installed inside the sealed cavity, acquiring a pressure time-series signal that varies over time to characterize the pressure fluctuation state of the sealed cavity during operation. The operating current in the excitation coil drive circuit is simultaneously acquired to obtain the coil operating current signal. Since the excitation coil controls the magnetic field strength through the drive circuit, thereby acting on the permanent magnet fluid to form a seal, the coil operating current signal can reflect the magnetic field output state required for maintaining the seal. By synchronously correlating the pressure time-series signal and the coil operating current signal, a correspondence between the sealing pressure state and the magnetic field drive state is established, providing basic data for subsequent seal health status analysis.

[0017] The drive circuit of the excitation coil includes an impedance analysis front end. An impedance test signal is injected into the drive circuit according to the impedance analysis front end, and the coil impedance current signal at both ends of the excitation coil is collected.

[0018] An impedance analysis front-end is installed in the drive circuit of the excitation coil. An impedance test signal is injected into the drive circuit through the impedance analysis front-end to detect the electromagnetic response characteristics of the excitation coil under operating conditions. The impedance test signal can be any one of a swept frequency signal, a pulse signal, or a micro-amplitude AC disturbance signal, used to excite the excitation coil to generate a corresponding impedance response. The coil impedance current signal at both ends of the excitation coil is acquired, and the impedance amplitude change characteristics and phase change characteristics of the coil are obtained based on the coil impedance current signal. Since changes in the permanent magnet fluid sealing state will cause changes in the magnetic circuit structure, thus affecting the equivalent impedance of the excitation coil, the coil impedance current signal can characterize the changes in the health status of the sealed area.

[0019] A health benchmark mapping model is constructed based on pressure time series signal samples, coil operating current signal samples, and coil impedance current signal samples.

[0020] Pressure time-series signal samples, coil operating current signal samples, and coil impedance current signal samples under normal sealing conditions are acquired. These samples are then synchronized and aligned in time sequence, and segmented according to a preset time window. Pressure fluctuation characteristics, operating current fluctuation characteristics, operating current harmonic characteristics, impedance amplitude-frequency characteristics, and phase-frequency characteristics are extracted, and corresponding feature vector sample sets are constructed. A health benchmark mapping model is then built based on this model. This model describes the normal electromagnetic coupling characteristics of the sealing system under healthy conditions, thus forming a benchmark for healthy sealing operation.

[0021] Preferably, the health benchmark mapping model is constructed using a long short-term memory (LSTM) neural network model. Since pressure changes, operating current changes, and impedance response changes all exhibit continuous temporal correlation characteristics, the LSTM neural network model can learn the dynamic temporal coupling relationships during the operation of the sealed system. Specifically, the model input is a pressure fluctuation feature vector, and the outputs are operating current fluctuation feature vectors, operating current harmonic feature vectors, impedance amplitude-frequency feature vectors, and phase-frequency feature vectors, used for supervised training of health state samples. During training, the model parameters are iteratively updated by minimizing the mean square error between the model's predicted output and the actual acquired output until the model output error meets a preset convergence condition.

[0022] After training, the health benchmark mapping model can output the predicted coil operating current characteristics and predicted coil impedance response characteristics corresponding to the healthy operating state based on the input pressure state characteristics, thereby establishing a normal mapping relationship between the pressure state and electromagnetic response state of the sealing system in a healthy state.

[0023] The pressure time series signal, coil operating current signal, and coil impedance current signal are input into the health benchmark mapping model to obtain the residual feature vector set, and the high-dimensional residual feature vector set of the residual feature vector set is extracted.

[0024] The real-time acquired pressure time-series signal, coil operating current signal, and coil impedance current signal are input into the health benchmark mapping model. The model then outputs the corresponding predicted coil operating current signal and predicted coil impedance current signal based on the current pressure state. Subsequently, the differences between the predicted and actual coil operating current signals, and the differences between the predicted and actual coil impedance current signals, are calculated to form a residual feature vector set. Further, considering the structural type of the sealing contact surface, a corresponding sealing failure mode library is loaded, and the residual feature vector set undergoes high-dimensional mapping processing to extract directional high-dimensional residual features related to the current structural type, thus forming a high-dimensional residual feature vector set to enhance the distinguishing ability between different sealing failure modes.

[0025] The sealing failure mode is obtained by pattern recognition based on the high-dimensional residual feature vector set, and the sealing failure mode is sent to the sealing monitoring and management terminal to generate a reminder message.

[0026] The high-dimensional residual feature vector set is matched with each candidate sealing failure mode in the sealing failure mode library to calculate the corresponding feature matching degree. Based on the feature matching degree result, pattern recognition is performed to determine the sealing failure mode corresponding to the current sealing system. The sealing failure mode includes at least one of leakage trend failure, local magnetic field attenuation failure, abnormal sealing gap failure, or structural wear failure. After identifying the corresponding sealing failure mode, the failure mode information is sent to the sealing monitoring and management terminal, which generates corresponding reminder information to achieve online health monitoring and fault early warning of the permanent magnet fluid sealing system.

[0027] Furthermore, a health benchmark mapping model is constructed based on pressure time series signal samples, coil operating current signal samples, and coil impedance current signal samples. The methods include: Obtain the pressure fluctuation time-series characteristics of the pressure time-series signal sample; obtain the operating current fluctuation time-series characteristics and operating current harmonic time-series characteristics of the coil operating current signal sample; obtain the impedance amplitude-frequency characteristics and phase-frequency characteristics of the coil impedance current signal sample; establish a health benchmark mapping model with the pressure fluctuation time-series characteristics as input and the operating current fluctuation time-series characteristics, operating current harmonic time-series characteristics, impedance amplitude-frequency characteristics, and phase-frequency characteristics as outputs.

[0028] Time series analysis was performed on the collected pressure time series signal samples to extract the fluctuation characteristics of the internal pressure of the sealed cavity over time. These pressure fluctuation time series characteristics include the pressure mean, pressure change rate, pressure pulsation period, pressure peak distribution, and pressure fluctuation stability parameters, used to characterize the dynamic pressure changes of the sealing system during operation. Since changes in the permanent magnet fluid sealing state can cause abnormal pressure fluctuations inside the sealed cavity, these pressure fluctuation time series characteristics can reflect the basic operating state of the sealing system.

[0029] Time-domain and frequency-domain analyses were performed on the operating current signal samples of the excitation coil to obtain the timing characteristics of operating current fluctuations and harmonics. The timing characteristics of operating current fluctuations include the mean current, current fluctuation amplitude, current change rate, and periodic stability parameters, used to characterize the dynamic driving characteristics of the excitation coil while maintaining a sealed state. The timing characteristics of operating current harmonics include the energy distribution of each harmonic, harmonic frequency components, and harmonic amplitude variation characteristics, used to characterize the electromagnetic response state of the excitation coil during magnetic field changes. Since abnormal sealing conditions will cause changes in the load on the excitation coil, the corresponding operating current characteristics will also change.

[0030] Impedance frequency response analysis is performed on the coil impedance current signal samples to obtain the impedance variation characteristics of the excitation coil at different test frequencies. The impedance amplitude-frequency characteristic is used to characterize the response law of the coil impedance amplitude changing with frequency, and the phase frequency characteristic is used to characterize the response law of the coil impedance phase changing with frequency. Since changes in the permanent magnet fluid sealing state will cause changes in the magnetic circuit structure, thereby affecting the equivalent inductance and impedance characteristics of the excitation coil, the impedance amplitude-frequency characteristic and phase frequency characteristic can reflect the changes in the electromagnetic coupling state of the sealed region.

[0031] The pressure fluctuation time-series characteristics are used as model input, and the operating current fluctuation time-series characteristics, operating current harmonic time-series characteristics, impedance amplitude-frequency characteristics, and phase-frequency characteristics are used as model output. The correlation between these characteristics is trained and analyzed to establish a health benchmark mapping model. This health benchmark mapping model is used to characterize the normal coupling relationship between pressure changes and the excitation coil operating current response and impedance response under healthy operating conditions of the sealing system, thereby forming a health state benchmark for the sealing system and providing a reference for subsequent seal failure identification.

[0032] Furthermore, the method for inputting the pressure time series signal, coil operating current signal, and coil impedance current signal into a health benchmark mapping model to obtain a residual feature vector set includes: The health benchmark mapping model obtains the predicted coil operating current signal and the predicted coil impedance current signal based on the input pressure time series signal; obtains the operating current residual feature vector of the predicted coil operating current signal and the coil operating current signal, and obtains the impedance current residual feature vector of the predicted coil impedance current signal and the coil impedance current signal; and outputs the operating current residual feature vector and the impedance current residual feature vector as a residual feature vector set.

[0033] The real-time acquired pressure time-series signal is input into the health reference mapping model, which outputs a corresponding predicted coil operating current signal and a predicted coil impedance current signal based on the current pressure state. The predicted coil operating current signal characterizes the theoretical excitation current response required to maintain the current pressure state under healthy sealing conditions; the predicted coil impedance current signal characterizes the corresponding theoretical impedance response state under healthy sealing conditions. The health reference mapping model establishes a predictive relationship between the pressure state and the electromagnetic response state, thereby forming a healthy reference response under the current operating state.

[0034] The difference between the predicted coil operating current signal and the real-time acquired coil operating current signal is calculated to obtain the operating current residual feature vector; the difference between the predicted coil impedance current signal and the real-time acquired coil impedance current signal is calculated to obtain the impedance current residual feature vector. The operating current residual feature vector is used to characterize the degree of deviation of the current actual excitation drive state from the healthy state, and the impedance current residual feature vector is used to characterize the degree of deviation of the current magnetic circuit structure and sealing state from the healthy state. Since sealing failure will cause changes in magnetic field distribution and load state, the corresponding residual features will show abnormal changes.

[0035] The residual feature vector of the operating current is combined and associated with the residual feature vector of the impedance current to form a residual feature vector set output. The residual feature vector set is used to comprehensively characterize the overall deviation of the current sealing system from the healthy operating state, and serves as the basic input data for subsequent high-dimensional mapping analysis and sealing failure mode identification.

[0036] Furthermore, the method for extracting the high-dimensional residual feature vector set of the residual feature vector set includes: The structural type of the sealing contact surface is analyzed, wherein the sealing contact surface is the contact surface between the permanent magnet fluid and the sealing cavity; a sealing failure mode library based on the structural type is loaded, and the sealing failure mode library is analyzed to extract the directional high-dimensional residual feature vector corresponding to the structural type; a high-dimensional mapping space is constructed based on the directional high-dimensional residual feature vector; the set of residual feature vectors is mapped in a high-dimensional way in the high-dimensional mapping space to obtain a high-dimensional residual feature vector set.

[0037] Structural analysis is performed on the sealing contact surface between the permanent magnet fluid and the sealed cavity to identify the microstructure type of the sealing contact surface. This microstructure type includes stepped, keyway, threaded, or other structural types with different surface features. Since different structural types correspond to different fluid adhesion characteristics, magnetic field distribution characteristics, and local stress distribution characteristics, they will exhibit different electromagnetic response patterns during seal failure. By analyzing the structural type of the sealing contact surface, the physical characteristics of the microstructure are converted into structural activation logic in subsequent high-dimensional mapping analysis to establish the correlation between structural type and local failure characteristics.

[0038] Based on the structure type, a corresponding sealing failure mode library is loaded. Various failure samples in the library are analyzed to extract directional high-dimensional residual feature vectors that are highly correlated with the current structure type. These directional high-dimensional residual feature vectors characterize local failure features that are prone to occur under specific structure types. For example, stepped structures are more prone to high-frequency harmonic disturbances, keyway structures are more prone to periodic impacts, and threaded structures are more prone to low-frequency modulation. By extracting these directional high-dimensional residual feature vectors, the structure type provides a directional constraint on the direction of failure feature analysis, thereby improving the ability to distinguish between different failure modes.

[0039] The oriented high-dimensional residual feature vectors are used as basis vectors in the high-dimensional mapping space. Linear orthogonalization is performed on each basis vector to establish a corresponding feature subspace. A high-dimensional mapping space is then constructed based on these feature subspaces, allowing local failure features corresponding to different structural types to form independently distributed regions within the high-dimensional mapping space. This high-dimensional mapping space enhances the structural correlation and failure mode differences among residual features, thereby improving feature separation capabilities in subsequent pattern recognition processes.

[0040] The residual feature vector set is input into the high-dimensional mapping space, and high-dimensional feature mapping processing is performed according to the mapping rules corresponding to the current structure type to obtain a high-dimensional residual feature vector set. Through the high-dimensional mapping processing, the local failure features related to the current structure type in the original residual features are enhanced, while the interference of irrelevant features is suppressed, making the distribution differences of different sealing failure modes in the high-dimensional feature space more obvious, thereby providing highly discriminative feature input for subsequent sealing failure mode identification.

[0041] Furthermore, the structural types of the sealing contact surface include stepped type, keyway type, or threaded type.

[0042] The structural type of the sealing contact surface is used to characterize the surface structural features of the contact area between the permanent magnetofluid and the sealing cavity. Specifically, the stepped structure corresponds to a contact surface structure with abrupt local height changes, which easily leads to local fluid accumulation and high-frequency disturbances under the influence of a magnetic field; the keyway structure corresponds to a contact surface structure with a periodic groove distribution, which easily leads to periodic fluid impacts and synchronous fluctuations; and the threaded structure corresponds to a contact surface structure with continuous helical patterns, which easily leads to low-frequency modulated flow characteristics propagating along the thread direction. Different structural types correspond to different magnetofluid flow characteristics and magnetic field distribution characteristics, thus forming structurally specific electromagnetic response characteristics during the seal failure process.

[0043] Furthermore, the method for analyzing the sealing failure mode library and extracting the directional high-dimensional residual feature vector corresponding to the structure type includes: For each candidate sealing failure mode in the sealing failure mode library, read the pressure time series signal failure samples, coil operating current signal failure samples, and coil impedance current signal failure samples; establish a residual feature vector failure sample set of the pressure time series signal failure samples, coil operating current signal failure samples, and coil impedance current signal failure samples with the health benchmark mapping model; extract at least one hidden intermediate vector from the residual feature vector failure sample set according to the preset structure-signal association rule and output it as a directional high-dimensional residual feature vector.

[0044] For each candidate sealing failure mode in the sealing failure mode library, corresponding pressure time-series signal failure samples, coil operating current signal failure samples, and coil impedance current signal failure samples are read. The candidate sealing failure modes include at least one of the following: local leakage failure, abnormal sealing gap failure, magnetic field attenuation failure, structural wear failure, or abnormal fluid distribution failure. By reading multi-source failure sample data under different failure modes, electromagnetic response change characteristics corresponding to various failure states are established for subsequent structural correlation analysis.

[0045] The failure samples from the pressure time series signal are input into the health benchmark mapping model to obtain the corresponding predicted coil operating current signal and predicted coil impedance current signal. The differences between these signals and the actual acquired failure samples from the coil operating current signal and coil impedance current signal are calculated to establish corresponding operating current residual features and impedance current residual features. Subsequently, these residual features are combined to form a residual feature vector failure sample set corresponding to different candidate sealing failure modes, used to characterize the deviation of the failure state from the healthy state.

[0046] Based on preset structure-signal association rules, a structure-oriented analysis is performed on the residual feature vector failure sample set to extract at least one hidden intermediate vector related to the current structure type, which is then output as a directional high-dimensional residual feature vector. This hidden intermediate vector characterizes the local failure features most likely to occur under a specific structure type. Specifically, for stepped structures, the focus is on extracting high-frequency harmonic energy features from the operating current residual; for keyway structures, the focus is on extracting frequency synchronization impact features from the impedance current residual; and for threaded structures, the focus is on extracting low-frequency modulation amplitude features from the operating current residual. Through the structure-signal association rules, a directional coupling between the structural characteristics of the sealing contact surface and the failure response features is achieved, thereby improving the targeting and accuracy of subsequent failure mode identification.

[0047] Furthermore, the preset structure-signal association rules include step type association rules, keyway type association rules, and thread type association rules; The step type association rule is used to extract the high-frequency harmonic energy of the working current residual in the failure sample set of the residual feature vector as a hidden intermediate vector; the keyway type association rule is used to extract the frequency synchronous impact feature of the impedance current residual in the failure sample set of the residual feature vector as a hidden intermediate vector; the thread type association rule is used to extract the low-frequency modulation component amplitude feature of the working current residual in the failure sample set of the residual feature vector as a hidden intermediate vector.

[0048] Based on the differences in electromagnetic response of different sealing contact surface structures during permanent magnet fluid flow, association rules between structure types and signal response characteristics are pre-established. These pre-defined structure-signal association rules include step type association rules, keyway type association rules, and thread type association rules, used to selectively extract local failure features corresponding to the current structure type from the residual feature vector failure sample set. Different association rules correspond to different feature activation methods to achieve a mapping association between sealing structure characteristics and electromagnetic anomaly responses.

[0049] The step type association rule is used to extract the high-frequency harmonic energy of the operating current residual in the failure sample set of the residual feature vector as a hidden intermediate vector. Because step-type structures are prone to local magnetic field abrupt changes and fluid boundary disturbances during permanent magnet fluid flow, high-frequency harmonic fluctuations are easily generated in the excitation coil operating current. By analyzing the distribution of high-frequency harmonic energy in the operating current residual, the hidden intermediate vector related to the local disturbance of the step structure is obtained and output as the corresponding directional high-dimensional residual feature vector.

[0050] The keyway type association rule is used to extract the frequency-synchronous impact characteristics of the impedance current residuals in the failure sample set of the residual feature vector as a hidden intermediate vector. Due to the periodic slot distribution characteristics of the keyway structure, periodic impact disturbances are easily generated under the action of permanent magnet fluid flow and magnetic field coupling, resulting in synchronous impact characteristics in the excitation coil impedance response corresponding to the structural period. By performing frequency-synchronous analysis on the impedance current residuals, the corresponding periodic impact characteristics are extracted as hidden intermediate vectors to characterize the local sealing anomalies related to the keyway structure.

[0051] The thread type association rule is used to extract the amplitude characteristics of the low-frequency modulation component of the working current residual in the failure sample set of the residual feature vector as a hidden intermediate vector. Because the thread type structure has continuous helical flow guiding characteristics, a slowly changing fluid modulation effect is easily formed during the flow of permanent magnet fluid along the thread direction, resulting in a low-frequency modulation component in the excitation coil's working current. By extracting the low-frequency modulation amplitude variation in the working current residual, a hidden intermediate vector related to the thread structure is obtained and output as the corresponding directional high-dimensional residual feature vector to characterize the local sealing failure trend under the thread structure.

[0052] Furthermore, the method for constructing a high-dimensional mapping space based on the directional high-dimensional residual feature vector includes: The directional high-dimensional residual feature vectors are orthogonalized as basis vectors to extract the linear space of basis vectors; the linear space of basis vectors is then used as the high-dimensional mapping space.

[0053] The directional high-dimensional residual feature vectors are used as basis vectors in the high-dimensional mapping space. Linear orthogonalization is performed on each directional high-dimensional residual feature vector to eliminate the correlation between failure features of different structures. This orthogonalization process ensures that each basis vector corresponds to the direction of different types of local sealing failure features, thus forming mutually independent feature representation dimensions. A corresponding linear basis vector space is established based on the orthogonalized basis vectors, enabling the failure features corresponding to different structural types to be expressed separately in different feature dimensions, thereby improving the distinguishability between different sealing failure modes.

[0054] The linear space of the basis vectors is used as a high-dimensional mapping space to perform high-dimensional feature mapping on the residual feature vector set. This high-dimensional mapping space enhances local anomaly features related to the current sealing contact surface structure type while suppressing interference features unrelated to the current structure type. This makes the distribution differences of different sealing failure modes in the high-dimensional feature space more obvious, thereby improving the feature separability in subsequent pattern recognition processes.

[0055] Furthermore, the method for obtaining sealing failure modes through pattern recognition based on the high-dimensional residual feature vector set includes: The high-dimensional residual feature vector set is matched with the high-dimensional residual feature vector candidate set of each candidate sealing failure mode in the sealing failure mode library to obtain the feature matching degree; the sealing failure mode is obtained by pattern recognition according to the feature matching degree.

[0056] The high-dimensional residual feature vector set is matched with the candidate high-dimensional residual feature vector set corresponding to each candidate sealing failure mode in the sealing failure mode library, and the corresponding feature matching degree is calculated. The feature matching degree is used to characterize the similarity between the current high-dimensional residual feature and each candidate sealing failure mode. Specifically, the correlation between the current sealing state and different candidate sealing failure modes is obtained by comparing the distribution distance, feature direction consistency, and feature energy distribution among the high-dimensional feature vectors.

[0057] Pattern recognition is performed based on the feature matching results corresponding to each candidate sealing failure mode to determine the sealing failure mode corresponding to the current sealing system. When the feature matching degree corresponding to a candidate sealing failure mode meets the preset matching condition, the candidate sealing failure mode is determined as the target sealing failure mode of the current sealing system. The sealing failure mode includes at least one of the following: local leakage failure, abnormal sealing gap failure, magnetic field attenuation failure, structural wear failure, or abnormal fluid distribution failure, thereby realizing online fault identification and condition diagnosis of the permanent magnet fluid sealing system.

[0058] Example 2, based on the same inventive concept as the permanent magnet fluid seal monitoring and management method in the foregoing examples, such as... Figure 2 As shown in the figure, this application provides a permanent magnet fluid seal monitoring and management system, the system comprising: The first signal acquisition module 10 is used to acquire the pressure time series signal and the coil operating current signal of the excitation coil within the sealed cavity. The sealed cavity is sealed by controlling the drive circuit of the excitation coil to act on the permanent magnet fluid. The second signal acquisition module 20 includes an impedance analysis front-end in the drive circuit of the excitation coil. An impedance test signal is injected into the drive circuit based on the impedance analysis front-end, and the coil impedance current signal at both ends of the excitation coil is acquired. The model construction module 30 is used to construct a health benchmark mapping model based on the pressure time series signal sample, the coil operating current signal sample, and the coil impedance current signal sample. The feature vector extraction module 40 is used to input the pressure time series signal, the coil operating current signal, and the coil impedance current signal into the health benchmark mapping model to obtain a residual feature vector set, and extract a high-dimensional residual feature vector set from the residual feature vector set. The reminder information generation module 50 is used to perform pattern recognition based on the high-dimensional residual feature vector set to obtain the sealing failure mode, and send the sealing failure mode to the sealing monitoring and management terminal to generate a reminder information.

[0059] Furthermore, the model building module 30 is used to perform the following operation steps: Obtain the pressure fluctuation time-series characteristics of the pressure time-series signal sample; obtain the operating current fluctuation time-series characteristics and operating current harmonic time-series characteristics of the coil operating current signal sample; obtain the impedance amplitude-frequency characteristics and phase-frequency characteristics of the coil impedance current signal sample; establish a health benchmark mapping model with the pressure fluctuation time-series characteristics as input and the operating current fluctuation time-series characteristics, operating current harmonic time-series characteristics, impedance amplitude-frequency characteristics, and phase-frequency characteristics as outputs.

[0060] Furthermore, the feature vector extraction module 40 is used to perform the following operation steps: The health benchmark mapping model obtains the predicted coil operating current signal and the predicted coil impedance current signal based on the input pressure time series signal; obtains the operating current residual feature vector of the predicted coil operating current signal and the coil operating current signal, and obtains the impedance current residual feature vector of the predicted coil impedance current signal and the coil impedance current signal; and outputs the operating current residual feature vector and the impedance current residual feature vector as a residual feature vector set.

[0061] Furthermore, the feature vector extraction module 40 is used to perform the following operation steps: The structural type of the sealing contact surface is analyzed, wherein the sealing contact surface is the contact surface between the permanent magnet fluid and the sealing cavity; a sealing failure mode library based on the structural type is loaded, and the sealing failure mode library is analyzed to extract the directional high-dimensional residual feature vector corresponding to the structural type; a high-dimensional mapping space is constructed based on the directional high-dimensional residual feature vector; the set of residual feature vectors is mapped in a high-dimensional way in the high-dimensional mapping space to obtain a high-dimensional residual feature vector set.

[0062] Furthermore, the structural types of the sealing contact surface include stepped type, keyway type, or threaded type.

[0063] Furthermore, the feature vector extraction module 40 is used to perform the following operation steps: For each candidate sealing failure mode in the sealing failure mode library, read the pressure time series signal failure samples, coil operating current signal failure samples, and coil impedance current signal failure samples; establish a residual feature vector failure sample set of the pressure time series signal failure samples, coil operating current signal failure samples, and coil impedance current signal failure samples with the health benchmark mapping model; extract at least one hidden intermediate vector from the residual feature vector failure sample set according to the preset structure-signal association rule and output it as a directional high-dimensional residual feature vector.

[0064] Furthermore, the preset structure-signal association rules include step type association rules, keyway type association rules, and thread type association rules; the step type association rule is used to extract the high-frequency harmonic energy of the working current residual in the residual feature vector failure sample set as a hidden intermediate vector; the keyway type association rule is used to extract the frequency synchronization impact characteristics of the impedance current residual in the residual feature vector failure sample set as a hidden intermediate vector; and the thread type association rule is used to extract the low-frequency modulation component amplitude characteristics of the working current residual in the residual feature vector failure sample set as a hidden intermediate vector.

[0065] Furthermore, the feature vector extraction module 40 is used to perform the following operation steps: The directional high-dimensional residual feature vectors are orthogonalized as basis vectors to extract the linear space of basis vectors; the linear space of basis vectors is then used as the high-dimensional mapping space.

[0066] Furthermore, the reminder information generation module 50 is used to perform the following operation steps: The high-dimensional residual feature vector set is matched with the high-dimensional residual feature vector candidate set of each candidate sealing failure mode in the sealing failure mode library to obtain the feature matching degree; the sealing failure mode is obtained by pattern recognition according to the feature matching degree.

[0067] Through the foregoing detailed description of a permanent magnet fluid seal monitoring and management method, those skilled in the art can clearly understand the permanent magnet fluid seal monitoring and management system in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to the method section.

[0068] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for monitoring and managing permanent magnet fluid seals, characterized in that, The method includes: The pressure time series signal and the coil operating current signal of the excitation coil in the sealed cavity are collected. The sealed cavity is sealed by controlling the drive circuit of the excitation coil to act on the permanent magnet fluid. The drive circuit of the excitation coil includes an impedance analysis front end. An impedance test signal is injected into the drive circuit according to the impedance analysis front end, and the coil impedance current signal at both ends of the excitation coil is collected. Construct a health benchmark mapping model based on pressure time series signal samples, coil operating current signal samples, and coil impedance current signal samples; The pressure time series signal, coil operating current signal, and coil impedance current signal are input into the health benchmark mapping model to obtain a residual feature vector set, and a high-dimensional residual feature vector set is extracted from the residual feature vector set. The sealing failure mode is obtained by pattern recognition based on the high-dimensional residual feature vector set, and the sealing failure mode is sent to the sealing monitoring and management terminal to generate a reminder message.

2. The permanent magnet fluid seal monitoring and management method as described in claim 1, characterized in that, The method for constructing a health benchmark mapping model based on pressure time series signal samples, coil operating current signal samples, and coil impedance current signal samples includes: Obtain the temporal characteristics of pressure fluctuations from the pressure time series signal samples; Obtain the operating current fluctuation timing characteristics and operating current harmonic timing characteristics of the coil operating current signal sample; Obtain the impedance amplitude-frequency characteristics and phase-frequency characteristics of the coil impedance current signal sample; A health benchmark mapping model is established with the pressure fluctuation timing characteristics as input and the operating current fluctuation timing characteristics, operating current harmonic timing characteristics, impedance amplitude-frequency characteristics, and phase-frequency characteristics as outputs.

3. The permanent magnet fluid seal monitoring and management method as described in claim 2, characterized in that, The method involves inputting the pressure time series signal, coil operating current signal, and coil impedance current signal into a health benchmark mapping model to obtain a residual feature vector set, including: The health benchmark mapping model obtains the prediction coil operating current signal and the prediction coil impedance current signal based on the input pressure time series signal. Obtain the predicted coil operating current signal and the operating current residual feature vector of the coil operating current signal, and obtain the predicted coil impedance current signal and the impedance current residual feature vector of the coil impedance current signal; The operating current residual feature vector and the impedance current residual feature vector are output as a residual feature vector set.

4. The permanent magnet fluid seal monitoring and management method as described in claim 1, characterized in that, The method for extracting the high-dimensional residual feature vector set of the residual feature vector set includes: The structural type of the sealing contact surface is analyzed, and the sealing contact surface is the contact surface between the permanent magnet fluid and the sealing cavity; Load the sealing failure mode library based on the structure type, and analyze the sealing failure mode library to extract the directional high-dimensional residual feature vector corresponding to the structure type; Construct a high-dimensional mapping space based on the directional high-dimensional residual feature vector; The residual feature vector set is mapped in a high-dimensional space to obtain a high-dimensional residual feature vector set.

5. The permanent magnet fluid seal monitoring and management method as described in claim 4, characterized in that, The structural types of the sealing contact surface include stepped type, keyway type, or threaded type.

6. The permanent magnet fluid seal monitoring and management method as described in claim 4, characterized in that, The method for analyzing the sealing failure mode library and extracting the directional high-dimensional residual feature vector corresponding to the structure type includes: For each candidate seal failure mode in the seal failure mode library, read the pressure time series signal failure sample, the coil operating current signal failure sample, and the coil impedance current signal failure sample. Establish a set of residual feature vector failure samples of the pressure time series signal failure samples, coil operating current signal failure samples, and coil impedance current signal failure samples and the health benchmark mapping model; According to the preset structure-signal association rules, at least one hidden intermediate vector is extracted from the set of failed residual feature vector samples and output as a directional high-dimensional residual feature vector.

7. The permanent magnet fluid seal monitoring and management method as described in claim 6, characterized in that, The preset structure-signal association rules include step type association rules, keyway type association rules, and thread type association rules; The step type association rule is used to extract the high-frequency harmonic energy of the working current residual in the failure sample set of the residual feature vector as a hidden intermediate vector. The keyway type association rule is used to extract the frequency synchronous impact feature of the impedance current residual in the failure sample set of the residual feature vector as a hidden intermediate vector. The thread type association rule is used to extract the low-frequency modulation component amplitude features of the working current residual in the failure sample set of the residual feature vector as a hidden intermediate vector.

8. The permanent magnet fluid seal monitoring and management method as described in claim 4, characterized in that, The method for constructing a high-dimensional mapping space based on the directional high-dimensional residual feature vector includes: The directional high-dimensional residual feature vectors are orthogonalized as basis vectors to extract the linear space of basis vectors. The linear space of the basis vectors is used as the high-dimensional mapping space.

9. The permanent magnet fluid seal monitoring and management method as described in claim 6, characterized in that, The method for obtaining sealing failure modes by pattern recognition based on the high-dimensional residual feature vector set includes: The feature matching degree is obtained by matching the high-dimensional residual feature vector set with the high-dimensional residual feature vector candidate set of each candidate sealing failure mode in the sealing failure mode library. Seal failure modes are obtained by pattern recognition based on feature matching degree.

10. A permanent magnet fluid seal monitoring and management system, characterized in that, For implementing the permanent magnet fluid seal monitoring and management method according to any one of claims 1-9, the system comprises: The first signal acquisition module is used to acquire the pressure time sequence signal and the coil operating current signal of the excitation coil in the sealed cavity. The sealed cavity is sealed by controlling the drive circuit of the excitation coil to act on the permanent magnet fluid. The second signal acquisition module includes an impedance analysis front-end in the drive circuit of the excitation coil. An impedance test signal is injected into the drive circuit according to the impedance analysis front-end, and the coil impedance current signal at both ends of the excitation coil is acquired. The model building module is used to build a health benchmark mapping model based on pressure time series signal samples, coil operating current signal samples, and coil impedance current signal samples. The feature vector extraction module is used to input the pressure time series signal, coil operating current signal and coil impedance current signal into the health benchmark mapping model to obtain the residual feature vector set, and extract the high-dimensional residual feature vector set of the residual feature vector set; The reminder information generation module is used to obtain the sealing failure mode by performing pattern recognition based on the high-dimensional residual feature vector set, and send the sealing failure mode to the sealing monitoring and management terminal to generate reminder information.

Images