A vibration monitoring signal filtering and fault monitoring method and system

Abstract

The application relates to the technical field of vibration fault monitoring, and discloses a vibration monitoring signal filtering and fault monitoring method and system, which comprises the following steps: constructing a filtering analysis graph and matching a target analysis line according to an initially acquired vibration monitoring signal; extracting a vibration feature vector corresponding to each vibration sensor and constructing a sensor network; screening a target sensor in the sensor network according to the target analysis line, analyzing a vibration fault transmission path, and screening out a fault sensor and a corresponding fault vibration signal; converting a time-domain vibration signal into an angular-domain order ratio signal based on a real-time rotating speed signal, filtering and reconstructing the fault vibration signal, and obtaining an updated vibration signal; screening out a first sensor associated with the fault sensor in the vibration fault transmission path, analyzing the propagation and inhibition effect of the first sensor on the fault, and obtaining a fault analysis result; and the application can enhance the accuracy of fault positioning, quantify the propagation range and influence degree of the fault, and improve the fault monitoring performance.

Description

Technical Field

[0001] This application relates to the field of vibration fault monitoring technology, and more specifically to a vibration monitoring signal filtering and fault monitoring method and system. Background Technology

[0002] Currently, in the field of mechanical equipment condition monitoring, vibration analysis is the core means of diagnosing rotating component faults. Traditional methods typically use fixed filtering parameters and thresholds to analyze time-domain or frequency-domain signals under static or single operating conditions. However, in industrial settings, equipment speed and load are often dynamically changing, causing vibration signal characteristics to drift. Fixed-parameter filtering methods struggle to accurately extract fault features. Furthermore, relying on data from a single sensor and lacking analysis of the correlation between multiple sensors and fault propagation paths leads to inaccurate fault location and poor monitoring and early warning capabilities, failing to meet the demands of modern industry for high-precision predictive fault maintenance.

[0003] The existing technology has the following problems: the filtering parameters are fixed and cannot be dynamically adjusted according to real-time operating conditions such as speed and load, resulting in poor filtering effect and inaccurate fault feature extraction under real-time changing operating conditions; it adopts a single data analysis method, processing the data of each sensor independently, ignoring the spatiotemporal correlation of vibration characteristics between sensors, resulting in low accuracy of fault analysis results; it only performs simple fault identification and lacks in-depth analysis of fault origin and transmission path, resulting in slow fault response; to solve at least one of the above problems, this application proposes a vibration monitoring signal filtering and fault monitoring method and system. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this application is to provide a vibration monitoring signal filtering and fault monitoring method and system, which can effectively solve the problems in the background technology. The specific technical solution of this application is as follows:

[0005] A vibration monitoring signal filtering and fault monitoring method, comprising:

[0006] Based on the pre-acquired initial vibration monitoring signal, analyze the filter analysis lines corresponding to different speeds and load conditions of the historical vibration monitoring signals, construct a filter analysis diagram, and match the target analysis line in the filter analysis diagram according to the real-time operating conditions;

[0007] The vibration characteristics of the vibration monitoring signal corresponding to each vibration sensor are extracted, a vibration feature vector is constructed, the correlation of vibration feature vectors between different vibration sensors is analyzed, and a sensor network is constructed.

[0008] According to the target analysis line, target sensors are screened in the sensor network, the vibration fault transmission path between target sensors is analyzed, and faulty sensors and corresponding fault vibration signals are screened out.

[0009] Based on the real-time rotational speed signal, the time-domain vibration signal is converted into an angular domain order ratio signal. Combined with the filtering analysis process in the target analysis line, the fault vibration signal is filtered and reconstructed to obtain an updated vibration signal.

[0010] Based on the updated vibration signal, a first sensor associated with the fault sensor is selected in the vibration fault transmission path through a preset fault analysis model. The effect of the first sensor on the propagation and suppression of the fault is analyzed to obtain the fault analysis results for fault monitoring.

[0011] Specifically, the step of analyzing the filter analysis lines corresponding to different speeds and load conditions of historical vibration monitoring signals based on the pre-acquired initial vibration monitoring signals, constructing a filter analysis diagram, and matching the target analysis line in the filter analysis diagram according to the real-time operating conditions includes:

[0012] Based on the pre-acquired initial vibration monitoring signal, modal decomposition is performed on the historical vibration monitoring signal to extract the operating condition characteristics corresponding to different speed and load conditions.

[0013] Generate corresponding filter analysis lines according to the operating condition characteristics, establish the correlation between corresponding filter analysis lines by combining the correlation between different operating conditions, and construct a filter analysis diagram.

[0014] The real-time operating conditions are mapped onto the filter analysis graph, and the first analysis line with a path distance less than a preset distance threshold is selected. The first analysis line is then merged and dynamically interpolated and corrected based on the changes in operating conditions to obtain the target analysis line.

[0015] Specifically, corresponding filter analysis lines are generated according to the described operating condition characteristics. The correlation between these lines is established based on the relationships between different operating conditions, and a filter analysis graph is constructed, including:

[0016] Calculate the energy proportion of the working condition feature under the corresponding working condition, and filter out the first working condition feature whose energy proportion is greater than the preset energy threshold.

[0017] Based on the frequency distribution range of the first working condition characteristics, corresponding filtering parameters are generated, a first filtering analysis line is constructed, and the filtering process of the fault data by the first filtering analysis line is simulated by a preset filtering optimization model to optimize the parameters and obtain a second filtering analysis line.

[0018] Using the second filter analysis line corresponding to each working condition as a node, the correlation between different working conditions is analyzed, and node connections are established between the corresponding second filter analysis lines to construct a filter analysis graph.

[0019] Specifically, the process of extracting vibration features from the vibration monitoring signal corresponding to each vibration sensor, constructing a vibration feature vector, analyzing the correlation of vibration feature vectors between different vibration sensors, and constructing a sensor network includes:

[0020] Extract the vibration features of the vibration monitoring signal corresponding to each vibration sensor and construct a vibration feature vector;

[0021] Each vibration sensor is treated as a node. By calculating the correlation between vibration feature vectors of different vibration sensors, node associations are established between vibration sensors whose correlation is greater than a preset correlation threshold, thus obtaining the first sensor network.

[0022] Based on the physical location between vibration sensors, the connections of nodes in the first sensor network whose physical distance is greater than a preset propagation distance are pruned, and connections are established between vibration sensors whose physical distance is less than or equal to the preset propagation distance and which have not yet been established, thus obtaining the sensor network.

[0023] Specifically, target sensors are screened in the sensor network according to the target analysis line, the vibration fault transmission path between target sensors is analyzed, and faulty sensors and their corresponding fault vibration signals are screened out, including:

[0024] Analyze the filter frequency band range in the target analysis line, calculate the energy ratio of each sensor vibration signal in the corresponding filter frequency band range, and screen out target sensors whose energy ratio is greater than a preset ratio threshold.

[0025] Based on the sensor network structure, the vibration fault transmission path between target sensors is analyzed, and faulty sensors and their corresponding fault vibration signals are screened out.

[0026] Specifically, the step of analyzing the vibration fault transmission path between target sensors based on the sensor network structure and filtering out faulty sensors and their corresponding faulty vibration signals includes:

[0027] Based on the sensor network structure, the centrality feature of each target sensor is calculated, and the candidate fault sensor with the highest centrality feature value is selected. According to the connection relationship in the sensor network, the first fault transmission path connected to the candidate fault sensor is selected.

[0028] Calculate the vibration characteristic correlation between adjacent nodes in the first fault transmission path, remove path segments whose vibration characteristic correlation is less than a preset vibration correlation threshold, and obtain the second fault transmission path.

[0029] In the second fault transmission path, the characteristic change trend between adjacent nodes is analyzed to screen out fault sensors and corresponding fault vibration signals.

[0030] Specifically, the process of converting the time-domain vibration signal into an angular-domain order signal based on the real-time rotational speed signal, and then filtering and reconstructing the fault vibration signal in conjunction with the filtering analysis process in the target analysis line to obtain an updated vibration signal, includes:

[0031] Based on the real-time rotational speed signal, the time-domain vibration signal is converted into an angular domain order signal to obtain the fault angular domain signal corresponding to the fault vibration signal;

[0032] Based on the filtering analysis process in the target analysis line, the fault angular domain signal is filtered and reconstructed to obtain the updated vibration signal.

[0033] Specifically, the step of filtering and reconstructing the fault angle domain signal according to the filtering analysis process in the target analysis line to obtain the updated vibration signal includes:

[0034] The fault angle domain signal is filtered according to the frequency parameters in the target analysis line to obtain the first filtered signal. The envelope signal is then extracted by Hilbert transform and the order ratio spectrum is analyzed to obtain the envelope order ratio spectrum.

[0035] Analyze the energy distribution of the corresponding fault features in the envelope order ratio spectrum, select the resonant frequency bands with energy distributions greater than the preset energy distribution threshold, and filter and reconstruct the first filtered signal according to the resonant frequency bands to obtain the updated vibration signal.

[0036] Specifically, based on the updated vibration signal, a first sensor associated with the fault sensor is selected from the vibration fault transmission path using a preset fault analysis model. The effect of the first sensor on fault propagation and suppression is analyzed to obtain fault analysis results, including:

[0037] Based on the updated vibration signal, the first sensor associated with the fault sensor is selected in the vibration fault transmission path through a preset fault analysis model, and the first sensor set is obtained.

[0038] Analyze the impact response characteristics of each sensor signal in the first sensor set, analyze the propagation and suppression effects of the first sensor on the fault, calculate the transmission delay and energy change of the fault impact between the sensors, and construct the fault propagation feature vector.

[0039] Based on the fault propagation feature vector, the fault situation is analyzed to obtain the fault analysis results.

[0040] A vibration monitoring signal filtering and fault monitoring system, used to implement the aforementioned vibration monitoring signal filtering and fault monitoring method, includes:

[0041] The target analysis line construction module analyzes the filter analysis lines corresponding to different speeds and load conditions of historical vibration monitoring signals based on the pre-acquired initial vibration monitoring signals, constructs a filter analysis diagram, and matches the target analysis line in the filter analysis diagram according to the real-time operating conditions.

[0042] The sensor network construction module extracts the vibration characteristics of the vibration monitoring signal corresponding to each vibration sensor, constructs a vibration feature vector, analyzes the correlation of vibration feature vectors between different vibration sensors, and constructs a sensor network.

[0043] The fault analysis module filters target sensors in the sensor network according to the target analysis line, analyzes the vibration fault transmission path between target sensors, and filters out faulty sensors and corresponding faulty vibration signals.

[0044] The signal filtering module converts the time-domain vibration signal into an angular-domain order ratio signal based on the real-time rotational speed signal. Combined with the filtering analysis process in the target analysis line, the fault vibration signal is filtered and reconstructed to obtain an updated vibration signal.

[0045] The fault monitoring module, based on the updated vibration signal, uses a preset fault analysis model to select the first sensor associated with the fault sensor in the vibration fault transmission path, analyzes the first sensor's effect on fault propagation and suppression, and obtains fault analysis results for fault monitoring.

[0046] The beneficial effects of this application are as follows: Historical operating conditions are correlated with optimal filtering parameters, and corresponding target analysis lines are adaptively generated through real-time matching and interpolation correction; a sensor network is constructed by establishing connections between sensors, providing a topology for analyzing the fault propagation process; combined with angular domain order analysis and fault propagation feature vectors, the vibration fault transmission path is analyzed, which not only locates the fault source but also quantifies the fault propagation and suppression process. Operating condition-adaptive filtering effectively highlights fault components, improving the signal-to-noise ratio and feature extraction accuracy; sensor network and transmission path analysis enhance the accuracy of fault location; and analyzing the vibration fault transmission path not only detects faults but also quantifies the fault's propagation range and impact, achieving in-depth fault diagnosis and improving fault monitoring performance. Attached Figure Description

[0047] Figure 1 This is a flowchart illustrating the process of a vibration monitoring signal filtering and fault monitoring method according to an embodiment of this application.

[0048] Figure 2 This is a schematic diagram of the filter analysis diagram in the embodiments of this application;

[0049] Figure 3 This is a schematic diagram of the first sensor network in an embodiment of this application;

[0050] Figure 4 This is a schematic diagram of a vibration monitoring signal filtering and fault monitoring system according to an embodiment of this application. Detailed Implementation

[0051] The present application will be further described in detail below with reference to the accompanying drawings and embodiments.

[0052] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0053] Hereinafter, the terms "first," "second," and other generic terms are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0054] refer to Figure 1 The image shows a specific implementation of a vibration monitoring signal filtering and fault monitoring method according to this application, including:

[0055] S101. Based on the pre-acquired initial vibration monitoring signal, analyze the filter analysis lines corresponding to different speeds and load conditions of the historical vibration monitoring signal, construct the filter analysis diagram, and match the target analysis line in the filter analysis diagram according to the real-time operating conditions.

[0056] S102. Extract the vibration characteristics of the vibration monitoring signal corresponding to each vibration sensor, construct a vibration feature vector, analyze the correlation of vibration feature vectors between different vibration sensors, and construct a sensor network.

[0057] S103. According to the target analysis line, filter the target sensors in the sensor network, analyze the vibration fault transmission path between the target sensors, and filter out the faulty sensors and the corresponding faulty vibration signals.

[0058] S104. Based on the real-time rotational speed signal, the time-domain vibration signal is converted into an angular domain order ratio signal. Combined with the filtering analysis process in the target analysis line, the fault vibration signal is filtered and reconstructed to obtain an updated vibration signal.

[0059] S105. Based on the updated vibration signal, a first sensor associated with the fault sensor is selected in the vibration fault transmission path through a preset fault analysis model. The effect of the first sensor on the propagation and suppression of the fault is analyzed to obtain the fault analysis result for fault monitoring.

[0060] Traditional methods suffer from poor fault monitoring performance due to fixed filter parameters, neglect of operating conditions, and insufficient correlation among multiple sensors. This solution achieves condition-adaptive filtering and multi-sensor collaborative fault diagnosis by constructing a dynamic filter analysis graph, sensor network, and fault propagation path analysis. It can accurately identify fault sources and analyze fault propagation behavior, providing accurate data support for equipment condition assessment and predictive maintenance, and meeting the needs of industrial sites for high-precision fault monitoring.

[0061] In this embodiment, modal decomposition is performed on the pre-acquired initial vibration monitoring signal to decompose the non-stationary signal into corresponding intrinsic mode functions. Operating condition characteristics representing different speed and load combinations are extracted. Filter analysis lines corresponding to different speed and load conditions in historical vibration monitoring signals are analyzed, and a filter analysis graph is constructed. The target analysis line is then matched to the filter analysis graph according to the real-time operating conditions. By constructing the filter analysis graph and performing real-time matching, the corresponding filter parameters can be adaptively tracked and accurately matched according to the real-time operating conditions of the equipment. The filter frequency band can be dynamically adjusted according to changes in speed and load. In complex and dynamically changing operating conditions, it can accurately cover the frequency band where fault characteristics are located, effectively avoiding the problems of missed fault characteristics and background noise introduction caused by changes in operating conditions. This improves the quality and reliability of the vibration signal and provides accurate data support for the fault monitoring process.

[0062] Specifically, vibration characteristics of the vibration monitoring signal corresponding to each vibration sensor are extracted, including but not limited to root mean square value and kurtosis index in the time domain, and centroid frequency and envelope demodulation spectrum characteristics in the frequency domain. Vibration feature vectors are constructed, and the correlation between vibration feature vectors of different vibration sensors is analyzed to build a sensor network. By constructing a sensor network, correlation analysis and monitoring of sensors can be performed. Combining the mutual influence and propagation relationship of vibration behavior between measuring points, a topological structure is provided for accurately locating fault sources and identifying fault transmission paths, thereby enhancing the accuracy and correlation of the fault monitoring process.

[0063] Specifically, based on the filtering frequency band determined by the target analysis line, target sensors are screened in the sensor network. Target sensors have a stronger ability to acquire resonance or impact responses caused by faults. The vibration fault transmission path between target sensors is analyzed to screen out faulty sensors and their corresponding fault vibration signals. By combining adaptive filtering and network analysis, the accuracy and efficiency of identifying faulty sensors and fault transmission paths can be improved. Based on the fault transmission path, the fault source and fault propagation process can be quickly located, improving the depth and accuracy of fault monitoring. It can effectively distinguish between root cause faults and transmission effects, avoid misjudging affected normal components as fault sources, and improve the accuracy of fault analysis results.

[0064] Furthermore, based on the real-time rotational speed signal, the time-domain vibration signal of the fault sensor is resampled at equal angles and converted into an angular domain order ratio signal. This ensures that the fault characteristics, which are strictly synchronized with the rotational speed, are accurately reflected in the angular domain spectrum, avoiding spectral ambiguity caused by rotational speed changes in time-frequency analysis. Combined with the filtering analysis process in the target analysis line, the fault vibration signal is filtered and reconstructed. According to the frequency parameters in the target analysis line, the angular domain signal is bandpass filtered to extract the first filtered signal, which includes potential fault components. The first filtered signal is then subjected to Hilbert transform to extract the envelope signal and undergo order ratio analysis to obtain the envelope order ratio spectrum. The resonant frequency band where energy is concentrated at the corresponding fault characteristic order ratio is identified. Using the resonant frequency band as a filtering parameter, the first filtered signal is subjected to secondary filtering and reconstruction to obtain a new vibration signal with a high signal-to-noise ratio and significantly enhanced fault impact components.

[0065] It should be noted that the secondary filtering reconstruction, which combines angular domain order ratio conversion with envelope order ratio spectrum analysis, can enhance the characteristics of periodic impact faults. It can effectively filter background noise and interference that are independent of rotational speed and amplify the weak fault impact components that are synchronized with rotational speed. This can amplify fault characteristics in the context of early faults or strong noise, thereby improving the accuracy and precision of fault monitoring.

[0066] Specifically, based on the updated vibration signal, a pre-defined fault analysis model is used to select the first sensor associated with the faulty sensor in the vibration fault transmission path. The propagation and suppression effects of the first sensor on the fault are analyzed. The time delay, attenuation or amplification ratio of the impact energy when the fault impact is transmitted from the fault source to each first sensor are calculated. The severity and propagation direction of the fault are analyzed to obtain fault analysis results for fault monitoring. By deeply analyzing the fault propagation process in the sensor network, the propagation trend of the fault and its potential risks to related equipment can be analyzed, thereby formulating more predictive and targeted maintenance strategies and improving the safety and economy of equipment operation.

[0067] Preferably, the fault analysis model integrates multiple analysis strategies, including inferring the main direction and velocity of fault propagation based on the order of time delays; assessing structural damping characteristics based on the gradient of energy attenuation; if the energy attenuation in a certain direction is abnormally small, structural resonance or loosening may exist; and performing similarity matching between the current feature vector and feature vectors in the historical fault case library to determine the fault type and severity level. The construction process of the historical fault case library includes: injecting known faults into the test bench or actual equipment and extracting fault propagation feature vectors; labeling each case with fault type and severity level to obtain a standard template library. During matching, the Euclidean distance or cosine similarity between the current feature vector and each template vector is calculated, and the fault type and severity level corresponding to the most similar template are taken as a reference. The fault analysis model outputs fault analysis results, including the fault source location, fault type, severity, propagation path, impact range, and the suppression or amplification effect of each associated point.

[0068] This application correlates historical operating conditions with optimal filtering parameters, adaptively generating corresponding target analysis lines through real-time matching and interpolation correction. It establishes connections between sensors to construct a sensor network, providing a topology for analyzing fault propagation. Combining angular domain order analysis and fault propagation feature vectors, it analyzes vibration fault transmission paths, enabling not only fault source localization but also quantification of fault propagation and suppression processes. Operating condition-adaptive filtering effectively highlights fault components, improving signal-to-noise ratio and feature extraction accuracy. Sensor network and transmission path analysis enhance fault localization accuracy. Analyzing vibration fault transmission paths not only detects faults but also quantifies their propagation range and impact, achieving in-depth fault diagnosis and improving fault monitoring performance.

[0069] Furthermore, based on the pre-acquired initial vibration monitoring signal, the filter analysis lines corresponding to different speeds and load conditions of historical vibration monitoring signals are analyzed, a filter analysis diagram is constructed, and the target analysis line is matched in the filter analysis diagram according to the real-time operating conditions, including:

[0070] S201. Based on the pre-acquired initial vibration monitoring signal, perform modal decomposition on the historical vibration monitoring signal to extract the operating condition characteristics corresponding to different speed and load conditions.

[0071] S202. Generate corresponding filter analysis lines according to the operating condition characteristics, establish the correlation between corresponding filter analysis lines by combining the correlation between different operating conditions, and construct a filter analysis diagram.

[0072] S203. Map the real-time operating conditions onto the filter analysis graph, select the first analysis line whose path distance is less than the preset distance threshold, merge the first analysis line and perform dynamic interpolation correction on the first analysis line in combination with the changes in operating conditions to obtain the target analysis line.

[0073] In this embodiment, based on the pre-acquired initial vibration monitoring signal, the initial vibration monitoring signal is acquired in real time by sensors at various measuring points. Under different speed and load conditions, the intrinsic modal components and their energy distribution excited in the vibration monitoring signal of the mechanical equipment will change significantly. The variational mode decomposition method is used to perform mode decomposition on a large number of pre-acquired historical vibration monitoring signals covering different speed and load combinations. The variational mode decomposition method is an adaptive signal decomposition method. By constructing and solving a variational problem, the input vibration signal is decomposed into multiple eigenmode functions with specific center frequencies and finite bandwidths, which can effectively avoid mode aliasing. After decomposition, for each historical operating point, the energy of all its modal components is calculated to obtain the operating condition characteristics corresponding to different speed and load conditions. By modal decomposition and extracting modal features, feature components related to the equipment operating conditions and representing dynamic responses can be quickly extracted from the complex original vibration signal, providing accurate data support for building a filtering knowledge base.

[0074] Specifically, corresponding filter analysis lines are generated according to the characteristics of the working conditions. The correlation between the corresponding filter analysis lines is established by combining the correlation between different working conditions, and a filter analysis graph is constructed. By constructing a filter analysis graph by associating the filter strategies of each working condition, the similarity and reachability between filter strategies of different working conditions can be analyzed through the node connection relationship. This provides a topology structure for quickly and accurately matching and generating target filter strategies, thereby improving the accuracy and efficiency of formulating target analysis lines.

[0075] Specifically, the real-time operating point determined by the real-time monitored speed and load values ​​is mapped onto the filter analysis graph. In the filter analysis graph, the Euclidean distance from the real-time operating point to all nodes is calculated as the path distance. The first analysis line with a path distance less than a preset distance threshold is selected. The distance threshold can be set according to the accuracy requirements of the fault monitoring process. The selected first analysis lines are weighted and averaged according to the distance between their corresponding historical operating points and the real-time operating points. The fused analysis lines are then dynamically interpolated and corrected according to the changes in operating conditions. The dynamic difference correction is achieved by analyzing the rate of change of speed, calculating the mean between adjacent parameters among the fused parameters, and interpolating to obtain the target analysis line.

[0076] It should be noted that by using nearest neighbor filtering, weighted fusion, and dynamic interpolation correction, the target analysis line that is compatible with the real-time operating conditions can be accurately identified. When the real-time operating conditions are not in the historical database, multiple similar historical examples can be found from the filter analysis graph, the corresponding filter parameters can be fused, and the filter parameters can be adjusted in real time in combination with the changes in the operating conditions, thereby generating the optimal target analysis line, improving the adaptability and robustness to the real-time operating conditions, and improving the signal filtering effect and the quality of fault feature extraction.

[0077] Furthermore, corresponding filter analysis lines are generated according to the aforementioned operating condition characteristics. The correlation between these lines is established based on the relationships between different operating conditions, and a filter analysis graph is constructed, including:

[0078] S301. Calculate the energy ratio of the working condition feature under the corresponding working condition, and filter out the first working condition feature whose energy ratio is greater than the preset energy threshold.

[0079] S302. Generate corresponding filtering parameters based on the frequency distribution range of the first working condition characteristics, construct a first filtering analysis line, and optimize the filtering process of the fault data by simulating the first filtering analysis line through a preset filtering optimization model to obtain a second filtering analysis line.

[0080] S303. Take the second filter analysis line corresponding to each working condition as a node, analyze the correlation between different working conditions, establish node connections between the corresponding second filter analysis lines, and construct a filter analysis diagram.

[0081] In this embodiment, the energy of each modal component signal is calculated, and its ratio to the total energy of all modal components at that operating point is taken as the energy percentage. This yields the energy percentage of the operating condition characteristic under the corresponding operating condition. An energy threshold is set by averaging the energy data of a few key modal components whose accumulated energy accounts for the majority of the total energy. The first operating condition characteristic with an energy percentage greater than the energy threshold is selected. This first operating condition characteristic reflects the vibration mode that is mainly excited and has the most concentrated energy under this combination of rotational speed and load. By calculating the energy percentage and performing threshold filtering, the dominant vibration mode in each operating condition can be quickly identified, filtering out modal components with weak energy, originating from noise, or secondary excitation. This reduces the amount of data processing and ensures that the constructed filter analysis line can be accurately designed for the most important vibration energy distribution, improving the accuracy and filtering effect of the filter analysis line.

[0082] Specifically, based on the center frequency and bandwidth of the first operating condition, the frequency distribution range is determined, corresponding filtering parameters are generated, the center frequency and bandwidth of the bandpass filter are determined, and the filtering strategy of the bandpass filter is determined as the first filtering analysis line. The filtering process of the first filtering analysis line on the fault data is simulated by a preset filtering optimization model to optimize the parameters. The calculation process of the filtering optimization model is to simulate the filtering of known historical data containing faults using the first filtering analysis line, calculate the fault characteristic indicators of the filtered signal, including but not limited to the envelope spectrum kurtosis, iteratively search around the initial filtering parameters using the gradient descent method, continuously adjust the parameters and re-simulate the filtering and calculate the indicators until the optimal filtering parameters that maximize the target fault characteristic indicators are found, and the optimal filtering parameters are combined to obtain the second filtering analysis line.

[0083] Preferably, the specific working process of the filtering optimization model includes: simulating bandpass filtering of fault data using the first filtering analysis line to obtain the filtered signal; and calculating the envelope spectrum kurtosis of the filtered signal as the objective function. The calculation process of the envelope spectrum kurtosis is as follows: performing a Hilbert transform on the filtered signal to extract the envelope signal, performing a Fourier transform on the envelope signal to obtain the envelope spectrum, and calculating the kurtosis index in the envelope spectrum, i.e., the fourth central moment of the envelope spectrum divided by the square of the second central moment and then subtracting three. The larger the envelope spectrum kurtosis value, the more prominent the periodic impact component in the signal and the more obvious the fault characteristics. The optimization algorithm adopts the gradient descent method and iteratively searches around the initial filtering parameters. Specifically, the optimization algorithm generates a set of candidate filtering parameters, for example, adjusting the center frequency by 50 Hz and the bandwidth by 30 Hz, resimulating the filtering and calculating the envelope spectrum kurtosis. Through multiple iterations, the parameters are continuously adjusted until the envelope spectrum kurtosis value converges to the maximum value. The corresponding filtering parameters at this time are the optimal parameters, and the filtering analysis line composed of the optimal parameters is used as the second filtering analysis line.

[0084] For example, in rolling bearing fault monitoring, the initial filter parameters were set at a center frequency of 1000 Hz and a bandwidth of 200 Hz. After iterative optimization of the filter model, it was found that adjusting the center frequency to 1050 Hz and the bandwidth to 180 Hz resulted in the highest envelope spectrum kurtosis value, indicating that this set of parameters best highlighted the impact component caused by the bearing outer ring fault. The training dataset for the filter optimization model comes from a historical fault database, including vibration signal samples of various working conditions and fault types. Each sample is labeled with the fault type and severity. The goal of model training is to maximize the envelope spectrum kurtosis of various faults using the optimized filter parameters. The training process directly finds the parameter combination that maximizes the objective function through the optimization algorithm. The training parameters mainly include the learning rate, number of iterations, and convergence tolerance of the optimization algorithm. For example, the learning rate can be set to 0.01, and the number of iterations can be set to 100.

[0085] It should be noted that optimizing the filtering parameters through fault feature enhancement can effectively highlight the fault impact component, improve the targeting and effectiveness of the filtering process, enhance the fault characteristics of the signal after processing by the filtering analysis line, thereby providing a higher quality input signal for the fault monitoring process and improving the accuracy of fault analysis results.

[0086] like Figure 2 As shown, the second filter analysis line corresponding to each operating condition is used as a node. The node attributes include operating condition information and filter parameters. The correlation between different operating conditions is analyzed. The Euclidean distance between any two operating condition points in the operating condition space composed of speed and load is calculated to obtain the corresponding similarity. A similarity threshold is set by statistically analyzing the correlation characteristics between operating condition information. Node connections are established between second filter analysis lines with similarity greater than the similarity threshold to construct a filter analysis graph. Figure 2 The numbers in the graph represent the weights of the edges connecting two nodes. By constructing a filter analysis graph, the correlation and proximity between filter strategies under different operating conditions are analyzed, providing a corresponding topological reference for fast and accurate filter strategy matching and interpolation under real-time operating conditions, thereby improving the accuracy and efficiency of filter strategy matching results.

[0087] Furthermore, vibration features of the vibration monitoring signal corresponding to each vibration sensor are extracted, a vibration feature vector is constructed, the correlation of vibration feature vectors between different vibration sensors is analyzed, and a sensor network is constructed, including:

[0088] S401. Extract the vibration features of the vibration monitoring signal corresponding to each vibration sensor and construct a vibration feature vector;

[0089] S402. Take each vibration sensor as a node, calculate the correlation between vibration feature vectors of different vibration sensors, establish node association between vibration sensors with correlation greater than a preset correlation threshold, and obtain the first sensor network.

[0090] S403. Based on the physical positions between vibration sensors, prune the connections of nodes in the first sensor network whose physical distance is greater than the preset propagation distance, and establish connections between vibration sensors whose physical distance is less than or equal to the preset propagation distance and which have not yet been established, to obtain the sensor network.

[0091] In this embodiment, vibration features of the vibration monitoring signal corresponding to each vibration sensor are extracted. Vibration features include, but are not limited to, root mean square value in the time domain, kurtosis index, and centroid of the frequency domain spectrum. The vibration features are arranged in order to construct a vibration feature vector. By constructing a vibration feature vector, the one-sidedness and instability of a single feature parameter are avoided, and an accurate data basis is provided for analyzing the correlation of vibration behavior between different measuring points.

[0092] like Figure 3 As shown, each vibration sensor is treated as a node. By calculating the Pearson correlation of vibration feature vectors between different vibration sensors, a correlation threshold is set according to the accuracy requirements of the fault monitoring process. All node pairs are traversed, and node associations are established between vibration sensors whose Pearson correlation is greater than the correlation threshold to obtain the first sensor network. By constructing the first sensor network, the vibration feature correlation between sensors can be reflected, as well as the similarity between the signals detected by the sensors, providing a preliminary topology for fault analysis.

[0093] Specifically, based on the physical positions of the vibration sensors and the engineering estimation analysis results of the effective propagation range of vibration stress waves in the mechanical structure, the propagation distance is set. For each connection edge in the first sensor network, the physical distance between the two sensors is calculated. If the physical distance is greater than the preset propagation distance, it indicates that the correlation is due to non-structural transmission factors such as electromagnetic interference and common background excitation, rather than direct vibration transmission. The connection of the node is then pruned, and the connection edge is removed from the sensor network. Sensor pairs with physical distances less than or equal to the preset propagation distance are selected. If no connection is established in the sensor network, a connection is added between the corresponding sensor pairs to obtain the sensor network.

[0094] It should be noted that by analyzing the physical location to correct the sensor network, correlations that violate the laws of vibration propagation and mislead fault monitoring results can be removed, thereby improving the reliability of the sensor network. It also completes the structural correlations that are physically closely connected but have low data correlation due to weak signals or measurement noise, thereby enhancing the integrity of the sensor network, improving the accuracy of the sensor network in reflecting the fault vibration propagation path, and improving the accuracy and efficiency of fault location and transmission path analysis.

[0095] Furthermore, target sensors are screened in the sensor network according to the target analysis line, the vibration fault transmission path between target sensors is analyzed, and faulty sensors and corresponding fault vibration signals are screened out, including:

[0096] S501. Analyze the filter frequency band range in the target analysis line, calculate the energy ratio of each sensor vibration signal in the corresponding filter frequency band range, and screen out the target sensors whose energy ratio is greater than the preset ratio threshold.

[0097] S502. Based on the sensor network structure, analyze the vibration fault transmission path between target sensors and screen out the faulty sensors and their corresponding faulty vibration signals.

[0098] In this embodiment, the corresponding filter frequency band range is analyzed in the target analysis line. By performing bandpass filtering on the signal, the sum of the squares of the filtered signal amplitudes is calculated to obtain the signal energy of each sensor vibration signal in the corresponding filter frequency band range. The signal energy is then compared with the total energy of the sensor signal in the entire frequency band to obtain the energy ratio. A ratio threshold is set according to the fault monitoring accuracy requirements, and target sensors with an energy ratio greater than the ratio threshold are selected. By calculating the energy ratio and selecting sensors in the target analysis line frequency band, the set of sensors related to the fault under the current working condition can be quickly and accurately selected. Centralized processing of the sensor set can improve analysis efficiency, eliminate interference from normal measuring points, and improve the accuracy of the fault analysis process by focusing on the corresponding sensor set during the fault transmission path analysis.

[0099] Specifically, based on the sensor network structure, the centrality characteristics of each target sensor are calculated, the vibration fault transmission path between target sensors is analyzed, and faulty sensors and their corresponding faulty vibration signals are screened out. By combining network centrality analysis, path correlation analysis, and feature trend analysis, the fault source and fault propagation path can be accurately located. Centrality is used to locate the fault source, correlation is used to filter invalid transmission paths, and feature trends are used to determine the fault source. This avoids the problem of ambiguity in location caused by the complexity of vibration transmission, and can quickly distinguish faulty sensors from affected sensors, thus improving the accuracy of the fault monitoring process.

[0100] Furthermore, based on the sensor network structure, the vibration fault transmission path between target sensors is analyzed, and faulty sensors and their corresponding fault vibration signals are screened out, including:

[0101] S601. Based on the sensor network structure, calculate the centrality feature of each target sensor, filter out the candidate fault sensor with the highest centrality feature value, and filter out the first fault transmission path connected to the candidate fault sensor according to the connection relationship in the sensor network.

[0102] S602. Calculate the vibration characteristic correlation between adjacent nodes in the first fault transmission path, remove path segments whose vibration characteristic correlation is less than a preset vibration correlation threshold, and obtain the second fault transmission path.

[0103] S603. Analyze the characteristic change trend between adjacent nodes in the second fault transmission path, and screen out the fault sensor and the corresponding fault vibration signal.

[0104] In this embodiment, based on the sensor network structure, the centrality feature of each target sensor is calculated. The centrality feature includes, but is not limited to, betweenness centrality. The betweenness centrality is obtained by statistically analyzing the proportion of paths passing through the node in all shortest paths in the sensor network. The higher the betweenness centrality of a node, the more important its bridging role in the network, and the greater the probability that vibration energy or fault information will flow to other parts through that node. The sensor node with the highest centrality feature value is selected as a candidate fault sensor. Starting from the candidate fault sensor, paths directly or indirectly connected to the candidate fault sensor are selected according to the connection relationships in the sensor network to obtain the first fault transmission path. By calculating the network centrality feature and selecting candidate fault sources, the source node that causes global vibration mode anomalies can be selected by combining the network's structural characteristics, thereby improving the accuracy and efficiency of the fault analysis process.

[0105] Specifically, for each directly connected node pair in the first fault propagation path, the Pearson correlation of the vibration feature vectors between adjacent nodes is calculated to obtain the vibration feature correlation. A vibration correlation threshold is set according to the accuracy requirements of the fault monitoring process. All adjacent node pairs in the first fault propagation path are traversed, and path segments with vibration feature correlation less than the vibration correlation threshold are removed to obtain the second fault propagation path. By analyzing the vibration feature correlation, connection paths in the network topology that are irrelevant to the current fault vibration propagation can be effectively removed, improving the consistency between the fault propagation path and the vibration behavior, improving the fault analysis effect and reliability of the fault propagation path, and enhancing the accuracy of the fault analysis results.

[0106] Specifically, the characteristic change trends between adjacent nodes are analyzed in the second fault transmission path, and the rate of change of the envelope amplitude of the vibration signal between adjacent nodes is analyzed. By identifying sensor nodes located at the starting point of the characteristic trend and whose characteristic values ​​show a significant jump relative to downstream adjacent nodes, the faulty sensor is determined, and the vibration signal collected by the faulty sensor is extracted as the fault vibration signal. By analyzing the characteristic change trends on the fault transmission path, the fault starting point can be quickly and accurately identified, the faulty sensor can be accurately located, and accurate signal data can be provided for the fault monitoring and analysis process, thereby improving the efficiency and accuracy of the fault analysis process.

[0107] Furthermore, based on the real-time rotational speed signal, the time-domain vibration signal is converted into an angular-domain order signal. Combined with the filtering analysis process in the target analysis line, the fault vibration signal is filtered and reconstructed to obtain an updated vibration signal, including:

[0108] S701. Based on the real-time rotational speed signal, the time-domain vibration signal is converted into an angular-domain order signal to obtain the fault angular-domain signal corresponding to the fault vibration signal.

[0109] S702. Based on the filtering analysis process in the target analysis line, the fault angle domain signal is filtered and reconstructed to obtain the updated vibration signal.

[0110] In this embodiment, the vibration acceleration or velocity signals from the fault sensor and the real-time rotational speed signals provided by the key phase sensor or rotary encoder are synchronously acquired. In the angular domain, equal angular intervals are used as new sampling points. Based on the correspondence between the time-domain signal and the instantaneous phase, the vibration amplitude at the equal angular interval points is calculated using a linear interpolation algorithm. The time-domain vibration signal is converted into an angular domain order signal. After signal conversion, the fault angular domain signal corresponding to the original fault vibration signal is obtained. By performing a coordinate transformation from the time domain to the angular domain, the fault characteristics can be transformed from ambiguous frequency bands in the time-frequency signal into clear spectral lines in the angular domain, improving the accuracy of fault characteristic analysis.

[0111] Specifically, based on the filtering analysis process in the target analysis line, the fault angular domain signal is bandpass filtered and reconstructed. Based on the frequency parameters in the target analysis line, the fault angular domain signal is bandpass filtered to extract the first filtered signal including potential fault components. The first filtered signal is subjected to Hilbert transform to extract the envelope signal and undergoes order ratio analysis to obtain the envelope order ratio spectrum. The resonance frequency band with concentrated energy at the corresponding fault characteristic order ratio is identified. The resonance frequency band is used as the filtering parameter to perform secondary filtering and reconstruction on the first filtered signal to obtain a new vibration signal with high signal-to-noise ratio and significantly enhanced fault impact components.

[0112] It should be noted that by performing secondary filtering and reconstruction on the fault angular domain signal based on envelope order ratio spectral analysis, the high-frequency resonance components excited by the fault impact can be accurately separated. Through envelope demodulation, these components are transformed into easily identifiable low-frequency order ratio spectral lines. Through resonance band filtering and reconstruction, random vibrations and background noise unrelated to the fault can be effectively suppressed, and the signal-to-noise ratio of fault features in the updated vibration signal can be improved, thereby improving the signal quality during the fault monitoring process.

[0113] Furthermore, based on the filtering analysis process in the target analysis line, the fault angle domain signal is filtered and reconstructed to obtain an updated vibration signal, including:

[0114] S801. Filter the fault angle domain signal according to the frequency parameters in the target analysis line to obtain the first filtered signal, and extract the envelope signal through Hilbert transform to perform order ratio spectrum analysis to obtain the envelope order ratio spectrum.

[0115] S802. Analyze the energy distribution of the corresponding fault features in the envelope order ratio spectrum, select the resonance frequency bands with energy distribution greater than the preset energy distribution threshold, and filter and reconstruct the first filtered signal according to the resonance frequency band to obtain the updated vibration signal.

[0116] In this embodiment, the center frequency and bandwidth of the bandpass filter are set according to the frequency parameters in the target analysis line. Bandpass filtering is performed on the fault angular domain signal to obtain the first filtered signal. Through Hilbert transform, the analytic signal is constructed and its modulus is calculated to obtain the envelope signal. This effectively removes high-frequency carrier components while retaining the amplitude information corresponding to the fault impact period. Order ratio spectrum analysis is performed on the envelope signal, specifically a Fourier transform is applied to the angular domain envelope signal to obtain the envelope order ratio spectrum. Through envelope order ratio spectrum analysis based on Hilbert transform, the spectral structure of the original vibration wave can be transferred to the envelope spectrum, which directly reflects the fault impact sequence. This effectively displays the fault signal on the order ratio spectrum, improving the fault monitoring capability.

[0117] Specifically, the energy distribution corresponding to fault characteristics in the envelope order spectrum is analyzed to locate the characteristic order positions of known faults. The energy distribution of the characteristic order spectral lines and their sidebands is analyzed, and energy values ​​are calculated. An energy distribution threshold is set based on the analytical precision required for fault monitoring. Resonance frequency bands with energy distributions greater than the preset threshold are selected. The first filtered signal is then reconstructed using these resonance frequency bands as filtering parameters to obtain the updated vibration signal. By selecting and reconstructing resonance frequency bands based on the envelope spectrum energy distribution, combined with the actual energy performance of fault characteristics, the most effective analysis frequency band is confirmed. This ensures that the reconstructed updated vibration signal can contain the true fault information to the greatest extent possible, suppressing irrelevant noise and interference components, improving the signal quality of the updated vibration signal, and enhancing the accuracy of fault monitoring results.

[0118] Furthermore, based on the updated vibration signal, a first sensor associated with the fault sensor is selected from the vibration fault transmission path using a preset fault analysis model. The effect of the first sensor on fault propagation and suppression is analyzed to obtain fault analysis results, including:

[0119] S901. Based on the updated vibration signal, the first sensor associated with the fault sensor is selected in the vibration fault transmission path through the preset fault analysis model to obtain the first sensor set.

[0120] S902. Analyze the impact response characteristics of each sensor signal in the first sensor set, analyze the propagation and suppression effect of the first sensor on the fault, calculate the transmission delay and energy change of the fault impact between the sensors, and construct the fault propagation feature vector.

[0121] S903. Analyze the fault situation based on the fault propagation feature vector to obtain the fault analysis results.

[0122] In this embodiment, based on the updated vibration signal, a preset fault analysis model is used to analyze and filter the vibration fault propagation path. The fault analysis model's analysis process includes: starting from the located fault sensor, traversing along the fault propagation path in the direction of fault vibration propagation, analyzing whether the signals of other sensors along the path also show a synchronous response with a significant increase in energy at the fault characteristic order, filtering out sensor nodes affected by the fault vibration and directly associated with the fault sensor as the first sensor, and obtaining the first sensor set; by combining the fault analysis model with the propagation path and signal characteristics to filter sensors, the fault signal area can be quickly filtered out based on the sensor association, providing an accurate regional reference for analyzing the dynamic propagation process of the fault, and improving the accuracy and efficiency of the fault analysis results.

[0123] Specifically, the impact response characteristics of each sensor signal in the first sensor set are analyzed, and the propagation delay and energy change are calculated. The propagation delay is calculated by measuring the time difference from the fault sensor to each first sensor through cross-correlation analysis. The energy change is calculated by comparing the ratio of impact energy exhibited at the fault sensor and each first sensor when the fault arrives. Combining the patterns of propagation delay and energy change, the propagation or suppression effect of each first sensor in the fault event is analyzed. The calculated delay values, energy change ratios, and other indicators are sequentially combined into a fault propagation feature vector. By calculating the fault propagation delay and energy change and constructing the fault propagation feature vector, data support is provided for comparing the characteristics of different fault events or different propagation paths, improving the accuracy of fault analysis results.

[0124] Specifically, based on fault propagation feature vector analysis, the main propagation path and velocity of fault vibration are inferred from the time delay sequence; the distribution of structural damping is assessed by analyzing the energy attenuation gradient; the fault type and severity are confirmed by matching the current propagation characteristics with template features from historical case libraries or typical fault modes; and the comprehensive analysis results yield the fault analysis outcome. Fault analysis based on fault propagation feature vectors allows for the analysis of the fault's propagation range, impact, and system structural response, thereby enabling more targeted fault repair decisions and improving fault response speed and repair effectiveness.

[0125] like Figure 4 As shown, a vibration monitoring signal filtering and fault monitoring system is used to implement a vibration monitoring signal filtering and fault monitoring method, including:

[0126] The target analysis line construction module analyzes the filter analysis lines corresponding to different speeds and load conditions of historical vibration monitoring signals based on the pre-acquired initial vibration monitoring signals, constructs a filter analysis diagram, and matches the target analysis line in the filter analysis diagram according to the real-time operating conditions.

[0127] The sensor network construction module extracts the vibration characteristics of the vibration monitoring signal corresponding to each vibration sensor, constructs a vibration feature vector, analyzes the correlation of vibration feature vectors between different vibration sensors, and constructs a sensor network.

[0128] The fault analysis module filters target sensors in the sensor network according to the target analysis line, analyzes the vibration fault transmission path between target sensors, and filters out faulty sensors and corresponding faulty vibration signals.

[0129] The signal filtering module converts the time-domain vibration signal into an angular-domain order ratio signal based on the real-time rotational speed signal. Combined with the filtering analysis process in the target analysis line, the fault vibration signal is filtered and reconstructed to obtain an updated vibration signal.

[0130] The fault monitoring module, based on the updated vibration signal, uses a preset fault analysis model to select the first sensor associated with the fault sensor in the vibration fault transmission path, analyzes the first sensor's effect on fault propagation and suppression, and obtains fault analysis results for fault monitoring.

[0131] In this embodiment, the target analysis line construction module constructs a filter analysis map by analyzing historical data and automatically matches the optimal target analysis line based on real-time rotational speed and load. This enables adaptive tracking of filter parameters to operating conditions, avoiding the failure of fixed-parameter filtering under real-time changing conditions. It provides a precise and dynamically optimized filtering benchmark for signal processing, improving the accuracy of signal filtering and fault identification. The sensor network construction module extracts vibration feature vectors from each measuring point and calculates their correlations. Combined with physical locations, it constructs a sensor network to monitor multiple measuring points in association, reflecting the mutual influence of vibration behavior between measuring points and improving the accuracy of fault source location and fault propagation path identification.

[0132] Specifically, the fault analysis module analyzes network structure and signal characteristics to filter out the transmission path of vibration faults, obtaining fault sensors and their signals. By combining adaptive filtering with network analysis, it can infer the fault origin and propagation path while locating abnormal sensors, effectively distinguishing between root cause faults and fault propagation effects, thus improving the accuracy of fault location and fault analysis results. The signal filtering module converts time-domain signals into angular-domain order ratio signals to stabilize fault characteristics, and performs secondary filtering and reconstruction by combining target analysis lines and envelope order ratio analysis. This can eliminate the influence of speed fluctuations. By extracting and reconstructing the resonant frequency band, it can enhance the signal-to-noise ratio of periodic impact fault components, providing accurate vibration signals for fault monitoring.

[0133] Specifically, the fault monitoring module uses a fault analysis model to evaluate the propagation and suppression effects of associated sensors on faults along the identified transmission paths and analyzes the fault analysis results. It can comprehensively analyze the severity, scope of impact, and development trend of faults, thereby achieving predictive maintenance and improving the accuracy and efficiency of maintenance strategy formulation.

[0134] The above description is merely a preferred embodiment of this application. The scope of protection of this application is not limited to the above embodiments. All technical solutions falling within the scope of this application's concept are within the scope of protection of this application. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of this application should also be considered within the scope of protection of this application.

Claims

1. A method for filtering vibration monitoring signals and monitoring faults, characterized in that, include: Based on the pre-acquired initial vibration monitoring signal, analyze the filter analysis lines corresponding to different speeds and load conditions of the historical vibration monitoring signals, construct a filter analysis diagram, and match the target analysis line in the filter analysis diagram according to the real-time operating conditions; The vibration characteristics of the vibration monitoring signal corresponding to each vibration sensor are extracted, a vibration feature vector is constructed, the correlation of vibration feature vectors between different vibration sensors is analyzed, and a sensor network is constructed. According to the target analysis line, target sensors are screened in the sensor network, the vibration fault transmission path between target sensors is analyzed, and faulty sensors and corresponding fault vibration signals are screened out. Based on the real-time rotational speed signal, the time-domain vibration signal is converted into an angular domain order ratio signal. Combined with the filtering analysis process in the target analysis line, the fault vibration signal is filtered and reconstructed to obtain an updated vibration signal. Based on the updated vibration signal, a first sensor associated with the fault sensor is selected in the vibration fault transmission path through a preset fault analysis model. The effect of the first sensor on the propagation and suppression of the fault is analyzed to obtain the fault analysis results for fault monitoring.

2. The vibration monitoring signal filtering and fault monitoring method according to claim 1, characterized in that, The process of analyzing the filter analysis lines corresponding to different speeds and load conditions based on the pre-acquired initial vibration monitoring signal, constructing a filter analysis diagram, and matching the target analysis line in the filter analysis diagram according to the real-time operating conditions includes: Based on the pre-acquired initial vibration monitoring signal, modal decomposition is performed on the historical vibration monitoring signal to extract the operating condition characteristics corresponding to different speed and load conditions. Generate corresponding filter analysis lines according to the operating condition characteristics, establish the correlation between corresponding filter analysis lines by combining the correlation between different operating conditions, and construct a filter analysis diagram. The real-time operating conditions are mapped onto the filter analysis graph, and the first analysis line with a path distance less than a preset distance threshold is selected. The first analysis line is then merged and dynamically interpolated and corrected based on the changes in operating conditions to obtain the target analysis line.

3. The vibration monitoring signal filtering and fault monitoring method according to claim 2, characterized in that, Based on the described operating condition characteristics, corresponding filter analysis lines are generated. The correlation between these lines is established by combining the relationships between different operating conditions, and a filter analysis graph is constructed, including: Calculate the energy proportion of the working condition feature under the corresponding working condition, and filter out the first working condition feature whose energy proportion is greater than the preset energy threshold. Based on the frequency distribution range of the first working condition characteristics, corresponding filtering parameters are generated, a first filtering analysis line is constructed, and the filtering process of the fault data by the first filtering analysis line is simulated by a preset filtering optimization model to optimize the parameters and obtain a second filtering analysis line. Using the second filter analysis line corresponding to each working condition as a node, the correlation between different working conditions is analyzed, and node connections are established between the corresponding second filter analysis lines to construct a filter analysis graph.

4. The vibration monitoring signal filtering and fault monitoring method according to claim 1, characterized in that, The process of extracting vibration features from the vibration monitoring signal corresponding to each vibration sensor, constructing a vibration feature vector, analyzing the correlation of vibration feature vectors between different vibration sensors, and constructing a sensor network includes: Extract the vibration features of the vibration monitoring signal corresponding to each vibration sensor and construct a vibration feature vector; Each vibration sensor is treated as a node. By calculating the correlation between vibration feature vectors of different vibration sensors, node associations are established between vibration sensors whose correlation is greater than a preset correlation threshold, thus obtaining the first sensor network. Based on the physical location between vibration sensors, the connections of nodes in the first sensor network whose physical distance is greater than a preset propagation distance are pruned, and connections are established between vibration sensors whose physical distance is less than or equal to the preset propagation distance and which have not yet been established, thus obtaining the sensor network.

5. The vibration monitoring signal filtering and fault monitoring method according to claim 1, characterized in that, According to the target analysis line, target sensors are screened in the sensor network, the vibration fault transmission path between target sensors is analyzed, and faulty sensors and corresponding fault vibration signals are screened out, including: Analyze the filter frequency band range in the target analysis line, calculate the energy ratio of each sensor vibration signal in the corresponding filter frequency band range, and screen out target sensors whose energy ratio is greater than a preset ratio threshold. Based on the sensor network structure, the vibration fault transmission path between target sensors is analyzed, and faulty sensors and their corresponding fault vibration signals are screened out.

6. The vibration monitoring signal filtering and fault monitoring method according to claim 5, characterized in that, The method based on the sensor network structure analyzes the vibration fault transmission path between target sensors and filters out faulty sensors and their corresponding faulty vibration signals, including: Based on the sensor network structure, the centrality feature of each target sensor is calculated, and the candidate fault sensor with the highest centrality feature value is selected. According to the connection relationship in the sensor network, the first fault transmission path connected to the candidate fault sensor is selected. Calculate the vibration characteristic correlation between adjacent nodes in the first fault transmission path, remove path segments whose vibration characteristic correlation is less than a preset vibration correlation threshold, and obtain the second fault transmission path. In the second fault transmission path, the characteristic change trend between adjacent nodes is analyzed to screen out fault sensors and corresponding fault vibration signals.

7. The vibration monitoring signal filtering and fault monitoring method according to claim 1, characterized in that, The process of converting the time-domain vibration signal into an angular-domain order ratio signal based on the real-time rotational speed signal, and combining this with the filtering analysis process in the target analysis line, to filter and reconstruct the fault vibration signal and obtain an updated vibration signal includes: Based on the real-time rotational speed signal, the time-domain vibration signal is converted into an angular domain order signal to obtain the fault angular domain signal corresponding to the fault vibration signal; Based on the filtering analysis process in the target analysis line, the fault angular domain signal is filtered and reconstructed to obtain the updated vibration signal.

8. The vibration monitoring signal filtering and fault monitoring method according to claim 7, characterized in that, The step of filtering and reconstructing the fault angular domain signal according to the filtering analysis process in the target analysis line to obtain the updated vibration signal includes: The fault angle domain signal is filtered according to the frequency parameters in the target analysis line to obtain the first filtered signal. The envelope signal is then extracted by Hilbert transform and the order ratio spectrum is analyzed to obtain the envelope order ratio spectrum. Analyze the energy distribution of the corresponding fault features in the envelope order ratio spectrum, select the resonant frequency bands with energy distributions greater than the preset energy distribution threshold, and filter and reconstruct the first filtered signal according to the resonant frequency bands to obtain the updated vibration signal.

9. The vibration monitoring signal filtering and fault monitoring method according to claim 1, characterized in that, Based on the updated vibration signal, a first sensor associated with the fault sensor is selected from the vibration fault transmission path using a preset fault analysis model. The effect of the first sensor on fault propagation and suppression is analyzed to obtain the fault analysis results, including: Based on the updated vibration signal, the first sensor associated with the fault sensor is selected in the vibration fault transmission path through a preset fault analysis model, and the first sensor set is obtained. Analyze the impact response characteristics of each sensor signal in the first sensor set, analyze the propagation and suppression effects of the first sensor on the fault, calculate the transmission delay and energy change of the fault impact between the sensors, and construct the fault propagation feature vector. Based on the fault propagation feature vector, the fault situation is analyzed to obtain the fault analysis results.

10. A vibration monitoring signal filtering and fault monitoring system, characterized in that, A method for implementing vibration monitoring signal filtering and fault monitoring as described in any one of claims 1 to 9, comprising: The target analysis line construction module analyzes the filter analysis lines corresponding to different speeds and load conditions of historical vibration monitoring signals based on the pre-acquired initial vibration monitoring signals, constructs a filter analysis diagram, and matches the target analysis line in the filter analysis diagram according to the real-time operating conditions. The sensor network construction module extracts the vibration characteristics of the vibration monitoring signal corresponding to each vibration sensor, constructs a vibration feature vector, analyzes the correlation of vibration feature vectors between different vibration sensors, and constructs a sensor network. The fault analysis module filters target sensors in the sensor network according to the target analysis line, analyzes the vibration fault transmission path between target sensors, and filters out faulty sensors and corresponding faulty vibration signals. The signal filtering module converts the time-domain vibration signal into an angular-domain order ratio signal based on the real-time rotational speed signal. Combined with the filtering analysis process in the target analysis line, the fault vibration signal is filtered and reconstructed to obtain an updated vibration signal. The fault monitoring module, based on the updated vibration signal, uses a preset fault analysis model to select the first sensor associated with the fault sensor in the vibration fault transmission path, analyzes the first sensor's effect on fault propagation and suppression, and obtains fault analysis results for fault monitoring.

Images