Multi-parameter fusion fault diagnosis method and system for non-all-plastic hook-type commutators

Abstract

This invention discloses a multi-parameter fusion method and system for fault diagnosis of non-all-plastic hook-type commutators, relating to the field of data processing technology. The method includes: acquiring design drawings of the non-all-plastic hook-type commutator and identifying M key areas; deploying sensors and establishing a monitoring network to collect multi-source monitoring data streams from the M key areas; performing multi-parameter fusion on the data streams based on the application scenario to form a fused monitoring data stream; performing modal decomposition and fault identification on the fused data stream, extracting fault feature sets, and constructing a fault diagnosis rule space; and generating commutator fault diagnosis results based on rule space matching analysis of the fault feature sets. This achieves the technical effect of fusing multiple parameters, improving information utilization efficiency, and enhancing the accuracy and response speed of fault identification.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a fault diagnosis method and system for non-all-plastic hook-type commutators with multi-parameter fusion. Background Technology

[0002] As a key component in motors and power generation equipment, the non-all-plastic hook-type commutator's operating condition directly affects the overall performance and operational stability of the equipment. Commutator design not only emphasizes structural optimization and performance improvement but also requires efficient online monitoring and fault diagnosis capabilities to identify potential problems in the early stages of equipment operation. Currently, fault diagnosis for commutators mainly relies on local monitoring techniques and single signal feature analysis methods. Traditional solutions typically focus on monitoring and analyzing only one type of parameter (such as temperature, vibration, and sound), easily overlooking the interconnected characteristics between key components and the complexity of multi-factor interactions. This fails to fully utilize the global multi-parameter information during commutator operation, resulting in shortcomings in fault mode identification and early warning. Furthermore, existing fault diagnosis methods are relatively simplistic in extracting and matching diagnostic rules for complex application scenarios, making it difficult to meet the detection needs under multiple operating conditions and intertwined fault modes. Summary of the Invention

[0003] This invention provides a multi-parameter fusion method and system for diagnosing faults in non-all-plastic hook-type commutators, in order to solve the technical problems of poor detection efficiency in the prior art when information modes are single and multiple operating conditions and fault modes are intertwined. It achieves the technical effect of fusion of multiple parameters, improving information utilization efficiency, and improving the accuracy and response speed of fault identification.

[0004] In a first aspect, the present invention provides a multi-parameter fusion method for diagnosing faults in non-all-plastic hook-type commutators, wherein the multi-parameter fusion method for diagnosing faults in non-all-plastic hook-type commutators includes:

[0005] Obtain the design drawing information of the non-all-plastic hook-type commutator, and based on the design drawing information, identify the key parts of the non-all-plastic hook-type commutator to obtain M key area information of the commutator.

[0006] Sensors are deployed and analyzed on the key areas of the M commutators respectively to build a commutator monitoring sensor network. Multi-source monitoring data streams of the M key areas are collected through the commutator monitoring sensor network.

[0007] According to the application scenario information of the non-all-plastic hook-type commutator, the multi-source monitoring data streams of the M key areas are fused with multiple parameters to obtain the fused monitoring data streams of the M key areas.

[0008] Modal decomposition and fault identification are performed on the fused monitoring data streams of the M key areas to obtain fault feature sets of the M key areas, and at the same time, a commutator fault diagnosis rule space is constructed.

[0009] Based on the commutator fault diagnosis rule space, the fault feature sets of the M key areas are integrated and matched for analysis to generate commutator fault diagnosis results.

[0010] In one feasible implementation, obtaining the information on the M key commutator regions includes:

[0011] Based on the design drawings, the non-all-plastic hook-type commutator is disassembled into structural components to obtain a set of commutator structural components.

[0012] Each structural component in the commutator structural component set is labeled with its application function to determine the commutator component application function set.

[0013] The commutator structural component set is integrated with adjacent components according to the commutator component application function set to obtain N commutator functional region sets.

[0014] The N commutator functional regions are evaluated and filtered to obtain M key commutator region information, where N≥M.

[0015] In one feasible implementation, obtaining the information on the M key commutator regions includes:

[0016] Historical fault data is crawled based on the N commutator functional area sets to obtain a fault database for non-all-plastic hook-type commutators.

[0017] The fault database of the non-all-plastic hook-type commutator is mapped and distributed according to the N commutator functional area sets to obtain N functional area fault datasets.

[0018] Based on the fault datasets of the N functional areas, frequency statistics and probability calculations are performed to obtain the probability of fault occurrence in the N functional areas.

[0019] Based on the failure probability of the N functional areas, a failure probability threshold is preset, and the N commutator functional areas are compared and filtered based on the failure probability threshold to obtain the M key commutator area information.

[0020] In one feasible implementation, the construction of the commutator monitoring sensor network includes:

[0021] The monitoring requirements of the M key areas of the commutator are analyzed in sequence to determine the monitoring type and monitoring value range of the M key areas.

[0022] Based on the monitoring types and numerical ranges of the M key areas, select the monitoring sensor models for the M key areas.

[0023] Based on the monitoring sensor models of the M key areas, a monitoring coverage analysis is performed on the information of the key areas of the M commutators to obtain the deployment parameters of the M key area sensors.

[0024] Based on the sensor models and deployment parameters of the M key area monitoring sensors, sensors are deployed on the M key area information of the commutator to build the commutator monitoring sensor network.

[0025] In one feasible implementation, obtaining the fused monitoring data streams of M key areas includes:

[0026] Based on the application scenario information of the non-all-plastic hook-type commutator, the influence degree of each regional data monitoring source in the M key regional multi-source monitoring data streams is evaluated to obtain the M regional data monitoring source influence factor set.

[0027] Based on the set of influencing factors of the M regional data monitoring sources, determine the set of confidence coefficients of the M regional data monitoring sources.

[0028] The M key areas multi-source monitoring data streams are integrated according to time sequence information to obtain M key areas multi-source time sequence data streams.

[0029] Based on the set of confidence coefficients of the M regional data monitoring sources, multi-parameter fusion is performed on the multi-source time-series data streams of the M key regions to obtain the fused monitoring data streams of the M key regions.

[0030] In one feasible implementation, obtaining the M key area fault feature sets includes:

[0031] Empirical mode decomposition (EMD) is performed on the fused monitoring data streams of the M key areas to obtain the IMF components of the M key areas.

[0032] Based on the performance standards of the M key commutator regions, the normal operating thresholds of the M key regions are determined.

[0033] Based on the normal operation threshold of the M key regions, fault features are extracted from the IMF components of the M key regions to obtain the fault feature set of the M key regions.

[0034] In one feasible implementation, constructing the commutator fault diagnosis rule space includes:

[0035] The non-all-plastic hook-type commutator fault database is spatially divided according to the information of the M key areas of the commutator to obtain the fault data space of the M key areas.

[0036] Feature analysis and extraction are performed on the fault data spaces of the M key areas respectively to determine the fault feature spaces of the M key areas.

[0037] Fault diagnosis rule mining is performed within the fault feature space of the M key regions to construct the commutator fault diagnosis rule space.

[0038] In one feasible implementation, constructing the commutator fault diagnosis rule space includes:

[0039] Based on the fault feature space of the M key areas, the combination of fault features of the M key areas and the corresponding commutator diagnostic information are obtained.

[0040] Fault correlation topology analysis is performed on the fault feature combinations of the M key areas to obtain the commutator fault cascade feature set.

[0041] The commutator fault cascade feature set and the commutator diagnostic information are logically mapped and associated to construct the commutator fault diagnosis rule space.

[0042] In one feasible implementation, generating the commutator fault diagnosis result includes:

[0043] Based on the commutator fault diagnosis rule space, the fault feature sets of the M key areas are matched and analyzed to activate the matching commutator diagnosis rule set.

[0044] The matching commutator diagnostic rule set is integrated and output to generate the commutator fault diagnosis result.

[0045] Secondly, the present invention also provides a multi-parameter fusion fault diagnosis system for non-all-plastic hook-type commutators, wherein the multi-parameter fusion fault diagnosis system for non-all-plastic hook-type commutators includes:

[0046] The region identification module is used to obtain the design drawing information of the non-all-plastic hook-type commutator, and to identify the key parts of the non-all-plastic hook-type commutator based on the design drawing information, thereby obtaining M key region information of the commutator.

[0047] The regional monitoring module is used to deploy sensors on the key areas of the M commutators respectively, build a commutator monitoring sensor network, and collect multi-source monitoring data streams of the M key areas through the commutator monitoring sensor network.

[0048] The data stream fusion module is used to perform multi-parameter fusion of the multi-source monitoring data streams of the M key areas according to the application scenario information of the non-all-plastic hook-type commutator, so as to obtain the M key area fused monitoring data streams.

[0049] The fault feature acquisition module is used to perform modal decomposition and fault identification on the fused monitoring data streams of the M key areas to obtain fault feature sets of the M key areas, and at the same time construct the commutator fault diagnosis rule space.

[0050] The fault diagnosis execution module is used to perform integrated matching analysis on the fault feature sets of the M key areas based on the commutator fault diagnosis rule space, and generate commutator fault diagnosis results.

[0051] This invention discloses a multi-parameter fusion method and system for fault diagnosis of non-all-plastic hook-type commutators, comprising: acquiring design drawing information of the non-all-plastic hook-type commutator; identifying key components of the non-all-plastic hook-type commutator based on the design drawing information to obtain M key commutator region information; deploying sensors on the M key commutator regions to build a commutator monitoring sensor network; collecting multi-source monitoring data streams of the M key regions through the sensor network; and fusing the multi-parameter data streams of the M key regions according to the application scenario information of the non-all-plastic hook-type commutator to obtain fused monitoring data of the M key regions. The invention discloses a multi-parameter fusion method and system for commutator fault diagnosis. This method involves modal decomposition and fault identification of the fused monitoring data streams from M key regions to obtain fault feature sets for the M key regions, and constructing a commutator fault diagnosis rule space. Based on this rule space, integrated matching analysis is performed on the fault feature sets of the M key regions to generate commutator fault diagnosis results. This invention solves the technical problem of poor detection efficiency under conditions of single information modes and intertwined multiple operating conditions and fault modes, achieving the technical effects of fusing multiple parameters, improving information utilization efficiency, and enhancing the accuracy and response speed of fault identification. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating the multi-parameter fusion fault diagnosis method for non-all-plastic hook-type commutators of the present invention.

[0053] Figure 2 This is a schematic diagram of the multi-parameter fusion non-all-plastic hook-type commutator fault diagnosis system of the present invention.

[0054] Explanation of reference numerals in the attached diagram: Area identification module 11, Area monitoring module 12, Data stream fusion module 13, Fault feature acquisition module 14, Fault diagnosis execution module 15. Detailed Implementation

[0055] The above technical solutions will now be described in detail with reference to the accompanying drawings and specific embodiments to provide a better understanding of them. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments used only to explain the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Furthermore, it should be noted that, for ease of description, only the parts related to the present invention are shown in the drawings, not all of them.

[0056] Example 1, as Figure 1 This is a flowchart illustrating the multi-parameter fusion-based fault diagnosis method for non-all-plastic hook-type commutators according to the present invention. The method includes:

[0057] S100: Obtain the design drawing information of the non-all-plastic hook-type commutator, and identify the key parts of the non-all-plastic hook-type commutator based on the design drawing information to obtain M key area information of the commutator.

[0058] Specifically, firstly, the interactive production management system obtains the design data of the target non-all-plastic hook-type commutator and extracts the design drawings, which contain the detailed structure of the commutator and the location information of each component. Analysis of these drawings allows for the precise identification of critical parts of the commutator, such as areas prone to failure or significantly impacting its performance. Accurate identification of these critical parts provides a basis for subsequent sensor deployment, helping to ensure the quality of monitoring data.

[0059] In some embodiments, obtaining the information of the M key commutator regions includes:

[0060] Based on the design drawings, the non-all-plastic hook-type commutator is disassembled into structural components to obtain a set of commutator structural components. Each structural component in the set is then labeled with its application function to determine the set of commutator component application functions. Adjacent components in the set of commutator structural components are integrated according to the set of commutator component application functions to obtain N sets of commutator functional regions. These N sets of commutator functional regions are then evaluated and filtered to obtain M key commutator region information, where N ≥ M.

[0061] Specifically, firstly, based on the design drawings, the target non-all-plastic hook-type commutator is structurally disassembled into components, forming a complete set of commutator structural components. The design drawings include component labels for the target non-all-plastic hook-type commutator, theoretically breaking it down into independent structural components such as commutator segments, hooks, and insulation materials. Then, each component is labeled with its specific function in actual commutator operation; for example, the main function of the commutator segment is to achieve current commutation, and the function of the hook is to fix the commutator. Through this process, the role of each component in the overall function of the commutator can be clearly defined.

[0062] Furthermore, based on the functional set of the commutator components, adjacent components of the commutator structure are integrated to obtain N sets of commutator functional regions. In other words, adjacent components with similar or related functions are integrated together to form a larger set of functional regions. For example, the commutator segment, its adjacent insulating sheet, and hook are integrated into a single functional region because they cooperate with each other during the commutation process to perform specific functions.

[0063] Finally, the N commutator functional area sets are evaluated and filtered to obtain M key commutator area information, where N≥M. Specifically, the evaluation and filtering refers to selecting the M most critical functional areas from the N functional area sets based on certain criteria, such as the functionality of components during commutator operation and the probability of failure, as the key areas for subsequent monitoring. Preferably, components that are adjacent in location and have the same application function are integrated into the same functional area.

[0064] By following the steps above, the complex structure of the commutator can be decomposed into several key areas, providing clear targets for subsequent sensor deployment and fault monitoring. This helps improve the targeting and accuracy of fault diagnosis, while also optimizing the configuration of monitoring resources and avoiding unnecessary monitoring point settings, thereby reducing the complexity and cost of the system.

[0065] In some implementations, obtaining the information of the M key commutator regions includes:

[0066] Historical fault data is crawled based on the N commutator functional area sets to obtain a non-all-plastic hook-type commutator fault database; the non-all-plastic hook-type commutator fault database is mapped and distributed according to the N commutator functional area sets to obtain N functional area fault datasets; frequency statistics and probability calculations are performed on the N functional area fault datasets to obtain the N functional area fault occurrence probabilities; based on the N functional area fault occurrence probabilities, a fault probability threshold is preset, and the N commutator functional area sets are compared and filtered based on the fault probability threshold to obtain the M commutator key area information.

[0067] Specifically, firstly, it refers to collecting fault data related to the commutator's functional areas from various data sources (such as equipment maintenance records, fault reports, online databases, etc.), crawling historical fault data, and obtaining information such as the type of fault, time of occurrence, and location of occurrence in the non-all-plastic hook-type commutator fault database.

[0068] Table 1. Exemplary historical fault data

[0069]

[0070]

[0071] Then, the fault data is classified according to the commutator's functional areas, so that each functional area corresponds to an independent fault dataset, forming a functional area fault dataset. Through this classification, the fault data can be matched with specific functional areas, which facilitates subsequent targeted analysis.

[0072] For example, the functional area fault dataset includes:

[0073] Contact fault dataset: F001, F003, F005, F006.

[0074] Spring fault dataset: F002.

[0075] Shaft fault dataset: F004.

[0076] Furthermore, frequency statistics and probability calculations are performed based on the fault datasets of N functional areas to obtain the probability of fault occurrence in each of the N functional areas. This step, through quantitative analysis, clarifies the likelihood of fault occurrence in each functional area.

[0077] Table 2. Examples of Failure Probability in Each Functional Area

[0078] Functional Area Number of faults Failure probability Contact 4 66.67％ spring 1 16.67％ pivot 1 16.67％

[0079] Next, based on the failure probability of N functional areas, a preset failure probability threshold is established. Then, the N commutator functional areas are compared and filtered based on this threshold to obtain information on M critical commutator areas. For example, if an area with a failure probability greater than 30% is defined as a critical area, then: contacts (66.67%) > 30%, contacts are critical areas; springs (16.67%) < 30%, springs are non-critical areas; and shafts (16.67%) < 30%, shafts are non-critical areas.

[0080] The above steps, through the analysis of historical data, further refine the functional area set into key areas, providing more targeted target areas for subsequent monitoring and fault diagnosis. In particular, the data-driven screening method better reflects the fault characteristics of the commutator in actual operation, improving the accuracy and efficiency of fault diagnosis.

[0081] S200: Sensors are deployed and analyzed on the information of the key areas of the M commutators respectively to build a commutator monitoring sensor network, and multi-source monitoring data streams of the M key areas are collected through the commutator monitoring sensor network.

[0082] Specifically, based on the information of M key commutator areas, the M key commutator areas are located, and corresponding monitoring sensors are deployed to form a commutator monitoring sensor network. Sensor analysis refers to the process of determining the categories of indicators that need to be monitored for each of the M key commutator areas based on the information of the M key commutator areas, and then selecting appropriate sensor types and models, as well as determining the specific installation locations, quantities, and layout of the sensors. This ensures that the sensors can effectively monitor the operating status of the key areas while avoiding mutual interference between sensors.

[0083] For example, in the commutator assembly area, it may be necessary to monitor parameters such as temperature, current, and vibration; in the hook area, it may be necessary to monitor parameters such as stress and displacement. Therefore, an appropriate sensor model should be selected based on the monitoring requirements.

[0084] Specifically, the commutator monitoring sensor network is a network system composed of multiple sensors distributed in key areas of the commutator to collect various physical quantity data during commutator operation in real time, such as temperature, vibration, current, and voltage. The multi-source monitoring data stream is a collection of monitoring data strings from different types of sensors and different monitoring areas. Typically, these data have different physical dimensions, sampling frequencies, and data formats, requiring integration and processing before they can be used for subsequent fault diagnosis.

[0085] By strategically deploying multiple sensors in key areas, various physical quantity changes during commutator operation can be comprehensively monitored, more accurately reflecting the actual operating status of the commutator and providing reliable data support for subsequent steps such as multi-parameter fusion, mode decomposition, and fault identification.

[0086] In some embodiments, the construction of the commutator monitoring sensor network includes:

[0087] The monitoring requirements of the M key commutator areas are analyzed sequentially to determine the monitoring types and numerical ranges for each of the M key areas. Based on these monitoring types and ranges, the models of the M key area monitoring sensors are selected. A monitoring coverage analysis is then performed on the M key commutator areas based on these sensor models to obtain sensor deployment parameters. Finally, sensors are deployed on the M key commutator areas based on these sensor models and deployment parameters to establish the commutator monitoring sensor network.

[0088] Specifically, firstly, a monitoring requirement analysis is conducted on the key areas of the M commutators to determine: the monitoring type (such as temperature, vibration, current, etc.) and the monitoring value range (normal / abnormal thresholds for each physical quantity).

[0089] Table 3 Examples of Monitoring Requirements

[0090]

[0091]

[0092] Then, based on the determined monitoring type and monitoring value range, the sensor selection library or corresponding equipment selection manual is traversed to select a matching sensor model. For example, for temperature monitoring, a thermocouple or infrared temperature sensor can be selected; for vibration monitoring, an acceleration sensor or velocity sensor can be selected.

[0093] Table 4. Exemplary sensor selection results

[0094] Monitoring type Sensor model Measurement range accuracy Temperature sensor K-type thermocouple 0℃-400℃ ±1℃ Current sensor Hall current sensor 0A-100A ±0.5A Vibration sensor MEMS accelerometer 0g-10g ±0.1g Torque sensor Torque measurement module 0Nm-50Nm ±0.2Nm

[0095] Next, a monitoring coverage analysis is conducted to determine the deployment locations and number of sensors. The purpose of the monitoring coverage analysis is to ensure that the sensors can fully cover key areas and avoid monitoring blind spots. For example, in the commutator segment area, multiple temperature sensors can be evenly arranged along the arrangement direction of the commutator segments to monitor temperature changes at different locations.

[0096] For example, in the commutator segment group area, assuming the commutator segment is 10cm long and 2cm wide, a temperature sensor can be arranged every 2cm along the length direction to ensure that the temperature change of each commutator segment and its adjacent area can be monitored; thus, the sensor deployment parameters for this area are obtained, with 6 sensors and a spacing of 2cm.

[0097] Furthermore, based on the monitoring sensor model and sensor deployment parameters, sensors are deployed in key areas to build a commutator monitoring sensor network. For example, in the commutator segment area, thermocouple sensors, current transformers, and accelerometers are installed according to the above deployment parameters and connected to the data acquisition system to ensure that the sensors can work properly and transmit the collected data to the system.

[0098] Through the above process, reasonable sensor selection and deployment are achieved, which enables the collected data to comprehensively and accurately reflect the operating status of key areas of the commutator, providing reliable data support for subsequent steps such as multi-parameter fusion, mode decomposition and fault identification.

[0099] S300: According to the application scenario information of the non-all-plastic hook-type commutator, the multi-source monitoring data streams of the M key areas are fused with multiple parameters to obtain the fused monitoring data streams of the M key areas.

[0100] Specifically, application scenario information refers to the specific environment and operating conditions of the commutator in actual use, such as the motor's speed range, load type, ambient temperature, and humidity. This information affects the commutator's operating characteristics and fault modes.

[0101] Specifically, multi-parameter fusion is performed on the multi-source monitoring data streams of each key area according to application scenario information. For example, based on the motor's speed range and load type, the weights of parameters such as temperature, vibration, and current are determined under different speeds and loads. Under low speed and high load conditions, the weight of vibration parameters may be higher; under high speed and low load conditions, the weight of temperature parameters may be higher.

[0102] The above steps combine the collected multi-source monitoring data with the actual application scenario of the commutator to perform parameter fusion, forming a data stream that can comprehensively reflect the operating status of the commutator, providing a more suitable and accurate information foundation for subsequent modal decomposition and fault identification.

[0103] In some embodiments, obtaining the fused monitoring data streams of M key areas includes:

[0104] According to the application scenario information of the non-all-plastic hook-type commutator, the influence degree of each regional data monitoring source in the M key regional multi-source monitoring data streams is evaluated to obtain the influence factor set of the M regional data monitoring sources; based on the influence factor set of the M regional data monitoring sources, the reliability coefficient set of the M regional data monitoring sources is determined; the M key regional multi-source monitoring data streams are integrated according to time series information to obtain the M key regional multi-source time series data streams; based on the reliability coefficient set of the M regional data monitoring sources, the M key regional multi-source time series data streams are fused using multiple parameters to obtain the M key regional fused monitoring data streams.

[0105] Specifically, firstly, based on the commutator's application scenario information, the impact of data monitoring sources in each region on the commutator's operating status is assessed. In other words, under different motor speeds and load conditions, the impact of monitoring sources such as temperature, vibration, and current on the commutator's operating status varies. For example, methods such as correlation coefficients and principal component analysis can be used to analyze the contribution of each monitoring source to commutator anomalies in historical monitoring data associated with application scenario information, generating M sets of influence factors for regional data monitoring sources. For instance, for the commutator segment group region, under high motor speed and low load conditions, the influence of temperature monitoring sources may be relatively high, with an influence factor set at 0.6; the influence of vibration monitoring sources is second highest, with an influence factor set at 0.3; and the influence of current monitoring sources is relatively low, with an influence factor set at 0.1.

[0106] Then, based on the set of influencing factors from the M regional data monitoring sources, a set of reliability coefficients for the M regional data monitoring sources is determined. The reliability coefficients can be determined based on the normalization of the influencing factors; for example, each influencing factor can be divided by the sum of similar influencing factors in the M regional data monitoring sources to obtain the reliability coefficient of each monitoring source. For example, the calculation formula can be expressed as:

[0107]

[0108] Among them, C i Let I be the confidence coefficient of indicator item I in the i-th monitoring source. i Let I be the influence factor of indicator item I in the i-th monitoring source, and N be the number of monitoring sources with indicator item I among the M regional data monitoring sources; j Let I be the influence factor of indicator item I in the j-th monitoring source.

[0109] Furthermore, the multi-source monitoring data streams of M key areas are integrated according to time sequence information, that is, the M multi-source monitoring data streams of key areas are aligned by timestamp to obtain M multi-source time-series data streams of key areas. For example, for the commutator group area, the monitoring data such as temperature, vibration, and current are arranged in chronological order, with values ​​of these three data points at each time point, forming a multi-source time-series data stream. Optionally, compensation for missing values ​​(interpolation method) is involved in the integration process.

[0110] Furthermore, based on the set of confidence coefficients for the M regional data monitoring sources, multi-parameter fusion is performed on the multi-source time-series data streams of the M key regions. For example, weighted fusion is performed according to the confidence coefficients to obtain the fused monitoring data streams of the M key regions. The resulting fused monitoring data streams can comprehensively reflect the operating status of the commutator at different time points, enabling the fault diagnosis method to adapt to various different operating conditions and environments, and possessing greater versatility and flexibility.

[0111] S400: Perform modal decomposition and fault identification on the fused monitoring data streams of the M key areas to obtain fault feature sets of the M key areas, and at the same time construct the commutator fault diagnosis rule space.

[0112] Specifically, mode decomposition is used to decompose complex data into multiple intrinsic mode functions (IMFs), thereby better reflecting the local characteristics of the signal. Hierarchical methods include empirical mode decomposition (EMD) and ensemble empirical mode decomposition (EEMD). The fault feature set is a collection of fault-related feature information extracted by analyzing the decomposed IMF components. For example, it could include energy, kurtosis, and root mean square values ​​in a specific frequency band. The fault diagnosis rule space is a rule base built based on historical fault data and expert knowledge, containing mapping relationships between various fault modes and corresponding features, used to guide fault identification and diagnosis.

[0113] In some embodiments, obtaining the M key area fault feature sets includes:

[0114] Empirical mode decomposition (EMD) is performed on the fused monitoring data streams of the M key areas to obtain the IMF components of the M key areas; normal operating thresholds for the M key areas are determined based on the performance operating standards of the M commutator key areas; and fault features are extracted from the IMF components of the M key areas according to the normal operating thresholds of the M key areas to obtain the fault feature set of the M key areas.

[0115] Specifically, firstly, modal decomposition is performed on the fused monitoring data streams of the M key regions. For example, for the fused data stream of the commutator segment group region, the Empirical Mode Decomposition (EMD) method is used to decompose it into multiple IMF components, which are arranged sequentially from high frequency to low frequency, each corresponding to a different modal characteristic in the signal. Then, based on the performance operating standards of non-all-plastic hook-type commutators, the normal operating thresholds for the M key regions are determined, i.e., the signal variation range of the equipment in a healthy state. For example, the mean (μ) and standard deviation (σ) of the IMF components are calculated, and the normal operating thresholds are set as: [μ-nσ, μ+nσ], where n can be 2 or 3, corresponding to 95% or 99% confidence intervals.

[0116] Next, fault features are extracted from the decomposed IMF components according to these normal operating thresholds, resulting in M ​​key area fault feature sets. For example, the fault feature extraction methods include: statistical analysis, which calculates the peak value, mean, variance, skewness, kurtosis, and other characteristic values ​​of each IMF component and compares them with the normal operating thresholds; if the values ​​exceed the thresholds, they are marked as abnormal features; time-frequency analysis, which performs Hilbert transform on the IMF components to calculate the instantaneous frequency and amplitude, and detects whether abnormal frequency shifts occur; and energy distribution analysis, which calculates the energy proportion of the IMF components; if the energy of low-frequency components suddenly increases, it may indicate component wear or aging.

[0117] Through the above modal decomposition, complex fused monitoring data can be decomposed into IMF components that are easier to analyze, thereby more accurately identifying fault characteristics; in particular, the construction of the fault feature set provides a direct basis for subsequent fault diagnosis.

[0118] In some embodiments, constructing the commutator fault diagnosis rule space includes:

[0119] The non-all-plastic hook-type commutator fault database is spatially divided according to the information of the M key areas of the commutator to obtain M key area fault data spaces; feature analysis and extraction are performed on the M key area fault data spaces to determine the M key area fault feature spaces; fault diagnosis rule mining is performed in the M key area fault feature spaces to construct the commutator fault diagnosis rule space.

[0120] Specifically, firstly, the fault database of non-all-plastic hook-type commutators is spatially divided according to M key commutator regions, resulting in M ​​key region fault data spaces, each containing historical fault data for that key region. Then, feature analysis and extraction are performed on each of the M key region fault data spaces (e.g., calculating statistical features such as mean, standard deviation, peak value, skewness, and kurtosis; using dimensionality reduction (principal component analysis)) to determine the M key region fault feature spaces. For example, for the fault data space of the commutator segment group region, analysis reveals that the energy values ​​of certain specific frequency bands change significantly when a fault occurs; these frequency band energy values ​​are extracted as fault features, forming the fault feature space for that region. Finally, fault diagnosis rule mining is performed within the M key region fault feature spaces. For each key region, the set of parameters most representative of its fault characteristics is extracted, forming the M key region fault feature spaces.

[0121] For example, it was found that when a certain fault feature in the commutator segment group exceeds a certain threshold, it often corresponds to a specific fault type, such as poor contact. Therefore, within the fault feature space of M key areas, data mining methods are used to extract fault diagnosis rules and construct a commutator fault diagnosis rule space, including commutator fault diagnosis rule spaces based on decision trees, Apriori algorithm, fuzzy logic, or deep learning (LSTM or CNN model).

[0122] For example, the resulting fault diagnosis rule space includes:

[0123] Contact fault rule space: Rule 1: Temperature > 85℃ and current > 22A → Contact overheating fault; Rule 2: Current fluctuation > 5A and temperature rise rate > 2℃ / s → Contact aging.

[0124] Spring failure rule space: Rule 1: Vibration > 1.8g → Spring mechanical fatigue; Rule 2: Abnormal increase in IMF high-frequency component energy → Spring breakage risk.

[0125] Shaft failure rule space: Rule 1: Torque change rate > 2 Nm / s and vibration > 1.5g → bearing wear; Rule 2: Running time > 5000h and frequent speed fluctuations → shaft fatigue.

[0126] By systematically organizing and analyzing historical fault data, the association rules between fault characteristics and fault types are mined, and the resulting fault diagnosis rule space provides intelligent decision support for subsequent fault diagnosis. This helps to quickly match real-time monitored fault characteristics with known fault modes, thereby achieving accurate and efficient fault diagnosis.

[0127] In some embodiments, constructing the commutator fault diagnosis rule space includes:

[0128] Based on the fault feature space of the M key areas, the fault feature combination of the M key areas and the corresponding commutator diagnostic information are obtained; fault association topology analysis is performed on the fault feature combination of the M key areas to obtain the commutator fault cascade feature set; the commutator fault cascade feature set and the commutator diagnostic information are logically mapped and associated to construct the commutator fault diagnosis rule space.

[0129] Specifically, firstly, based on the fault feature space of M key areas, historical diagnostic records or logs are interacted with to obtain the combination of fault features of the M key areas and the corresponding commutator diagnostic information, which serves as samples for subsequent analysis. Then, fault correlation topology analysis is performed on the combination of fault features of the M key areas to obtain a set of cascaded commutator fault features. In other words, during commutator operation, faults in different key areas may influence each other and cascade, for example: contact overheating → increased contact resistance → abnormal current → risk of burn-off; spring fatigue → unstable commutator commutation → increased shaft load; shaft wear → increased vibration → further accelerated spring aging. Organizing these cascaded features into a set can more comprehensively reflect the occurrence and development process of faults.

[0130] Furthermore, the commutator fault cascade feature set and commutator diagnostic information are logically mapped and associated to construct a commutator fault diagnosis rule space. This allows the type, severity, and potential impact of a fault to be quickly determined based on its position and association in the cascade feature set, combined with the corresponding diagnostic information, when a certain combination of fault features is detected.

[0131] S500: Based on the commutator fault diagnosis rule space, perform integrated matching analysis on the fault feature sets of the M key areas to generate commutator fault diagnosis results.

[0132] In some embodiments, generating commutator fault diagnosis results includes:

[0133] Based on the commutator fault diagnosis rule space, the fault feature sets of the M key areas are matched and analyzed to activate the matching commutator diagnosis rule set; the matching commutator diagnosis rule set is integrated and output to generate the commutator fault diagnosis result.

[0134] Specifically, firstly, based on the key region categories and locations of the fault feature sets of M key regions, the corresponding rule sets are matched in the commutator fault diagnosis rule space. The fault feature sets of the M key regions are then scanned, and each feature is matched against the diagnostic rule base to obtain a matching commutator diagnosis rule set. Next, the matching commutator diagnosis rule sets are integrated: if multiple rules involve the same key region, feature fusion is performed to reduce redundant information; if multiple faults are cascaded, fault propagation analysis is performed to determine the dominant fault. Finally, the integrated commutator fault diagnosis results are output. For example, a structured commutator fault diagnosis result is generated, including equipment name, diagnosis time, and fault type.

[0135] In summary, the multi-parameter fusion fault diagnosis method for non-all-plastic hook-type commutators provided by this invention has the following technical effects:

[0136] By acquiring the design drawings of non-all-plastic hook-type commutators, key components of the commutator are identified based on these drawings, resulting in M ​​key commutator regions. Sensors are deployed and analyzed in each of these M key regions to build a commutator monitoring sensor network. This network collects multi-source monitoring data streams from the M key regions. Based on the application scenario of the non-all-plastic hook-type commutator, multi-parameter fusion is performed on the multi-source monitoring data streams from the M key regions to obtain a fused monitoring data stream. Modal decomposition and fault identification are performed on the fused monitoring data streams from the M key regions to obtain fault feature sets for the M key regions, and a commutator fault diagnosis rule space is constructed. Based on the commutator fault diagnosis rule space, integrated matching analysis is performed on the fault feature sets of the M key regions to generate commutator fault diagnosis results. This achieves the technical effect of fusing multiple parameters, improving information utilization efficiency, and enhancing the accuracy and response speed of fault identification.

[0137] Example 2, as Figure 2 This is a schematic diagram of the multi-parameter fusion non-all-plastic hook-type commutator fault diagnosis system of the present invention. For example, Figure 1 The flowchart of the multi-parameter fusion non-all-plastic hook-type commutator fault diagnosis method of the present invention can be seen as follows: Figure 2 The structure shown is implemented.

[0138] Based on the same concept as the multi-parameter fusion non-all-plastic hook commutator fault diagnosis method in the above embodiments, the present invention also provides a multi-parameter fusion non-all-plastic hook commutator fault diagnosis system comprising:

[0139] The area identification module 11 is used to obtain the design drawing information of the non-all-plastic hook-type commutator, and to identify the key parts of the non-all-plastic hook-type commutator based on the design drawing information to obtain M key area information of the commutator.

[0140] The area monitoring module 12 is used to deploy sensors on the key area information of the M commutators respectively, build a commutator monitoring sensor network, and collect multi-source monitoring data streams of the M key areas through the commutator monitoring sensor network.

[0141] The data stream fusion module 13 is used to perform multi-parameter fusion of the multi-source monitoring data streams of the M key areas according to the application scenario information of the non-all-plastic hook-type commutator, so as to obtain the M key area fused monitoring data streams.

[0142] The fault feature acquisition module 14 is used to perform modal decomposition and fault identification on the fused monitoring data streams of the M key areas to obtain fault feature sets of the M key areas, and at the same time construct the commutator fault diagnosis rule space.

[0143] The fault diagnosis execution module 15 is used to perform integrated matching analysis on the fault feature sets of the M key areas based on the commutator fault diagnosis rule space, and generate commutator fault diagnosis results.

[0144] In some embodiments, the region identification module 11 includes:

[0145] The structural component disassembly unit is used to disassemble the non-all-plastic hook-type commutator into structural components based on the design drawing information, thereby obtaining a set of commutator structural components.

[0146] The component application function marking unit is used to mark the application functions of each structural component in the commutator structural component set to determine the commutator component application function set.

[0147] The functional area set integration unit is used to integrate adjacent components of the commutator structural component set according to the commutator component application function set, to obtain N commutator functional area sets.

[0148] The key area information filtering unit is used to evaluate and filter the N commutator functional area sets to obtain M key area information of the commutator, where N≥M.

[0149] In some implementations, the key area information filtering unit in the area identification module 11 includes:

[0150] The historical fault data crawling unit is used to crawl historical fault data based on the N commutator functional area sets to obtain a non-all-plastic hook-type commutator fault database.

[0151] The functional area fault dataset generation unit is used to map and distribute the non-all-plastic hook-type commutator fault database according to the N commutator functional area sets to obtain N functional area fault datasets.

[0152] The functional area failure probability calculation unit is used to perform frequency statistics and probability calculation based on the N functional area failure datasets to obtain the failure probability of N functional areas.

[0153] The key area information filtering unit is used to preset a fault probability threshold based on the fault occurrence probability of the N functional areas, and compare and filter the N commutator functional areas based on the fault probability threshold to obtain the M key area information of the commutator.

[0154] In some embodiments, the area monitoring module 12 includes:

[0155] The monitoring demand analysis unit is used to sequentially perform monitoring demand analysis on the information of the M key areas of the commutator, and determine the monitoring type and monitoring value range of the M key areas.

[0156] The monitoring sensor model selection unit is used to select the monitoring sensor models for the M key areas based on the monitoring types and monitoring value ranges of the M key areas.

[0157] The sensor deployment parameter acquisition unit is used to perform monitoring coverage analysis on the information of the M key areas of the commutator based on the monitoring sensor models of the M key areas, and obtain the sensor deployment parameters of the M key areas.

[0158] The commutator monitoring sensor network construction unit is used to deploy sensors on the information of the M key commutator areas based on the sensor models and sensor deployment parameters of the M key areas, and to build the commutator monitoring sensor network.

[0159] In some embodiments, the data stream fusion module 13 includes:

[0160] The regional data monitoring source impact factor assessment unit is used to assess the impact of each regional data monitoring source in the M key regional multi-source monitoring data streams according to the application scenario information of the non-all-plastic hook-type commutator, and obtain the M regional data monitoring source impact factor sets.

[0161] The regional data monitoring source credibility coefficient determination unit is used to determine the M regional data monitoring source credibility coefficient sets based on the M regional data monitoring source influence factor sets.

[0162] The key area multi-source time-series data stream integration unit is used to integrate the M key area multi-source monitoring data streams according to time-series information to obtain M key area multi-source time-series data streams.

[0163] The key area fusion monitoring data stream generation unit is used to perform multi-parameter fusion of the M key area multi-source time series data streams based on the confidence coefficient set of the M area data monitoring sources to obtain the M key area fusion monitoring data streams.

[0164] In some embodiments, the fault feature acquisition module 14 includes:

[0165] The key area empirical mode decomposition unit is used to perform empirical mode decomposition on the M key area fused monitoring data streams respectively to obtain M key area IMF components.

[0166] The critical area normal operation threshold determination unit is used to determine the normal operation threshold of the M critical areas based on the performance operating standards of the M commutator critical area information.

[0167] The critical area fault feature extraction unit is used to extract fault features from the IMF components of the M critical areas according to the normal operation threshold of the M critical areas, so as to obtain the fault feature set of the M critical areas.

[0168] In some embodiments, the fault feature acquisition module 14 includes:

[0169] The critical area fault data space partitioning unit is used to partition the non-all-plastic hook-type commutator fault database according to the M commutator critical area information to obtain M critical area fault data spaces.

[0170] The critical area fault feature space determination unit is used to perform feature analysis and extraction on the fault data spaces of the M critical areas respectively, and determine the fault feature spaces of the M critical areas.

[0171] The commutator fault diagnosis rule space construction unit is used to mine fault diagnosis rules within the fault feature spaces of the M key regions and construct the commutator fault diagnosis rule space.

[0172] In some embodiments, the fault feature acquisition module 14 includes:

[0173] The critical area fault feature combination and diagnostic information acquisition unit is used to obtain the M critical area fault feature combinations and corresponding commutator diagnostic information based on the M critical area fault feature spaces.

[0174] The fault association topology analysis unit is used to perform fault association topology analysis on the combination of fault features in the M key areas to obtain the commutator fault cascade feature set.

[0175] The fault diagnosis rule space construction unit is used to logically map and associate the commutator fault cascade feature set and the commutator diagnostic information to construct the commutator fault diagnosis rule space.

[0176] In some embodiments, the fault diagnosis execution module 15 includes:

[0177] The commutator fault diagnosis rule matching unit is used to perform matching analysis on the fault feature sets of the M key areas based on the commutator fault diagnosis rule space, and activate the matching commutator diagnosis rule set.

[0178] The commutator fault diagnosis result generation unit is used to integrate and output the matching commutator diagnosis rule set to generate the commutator fault diagnosis result.

[0179] It should be understood that the focus of the embodiments mentioned in this specification is their difference from other embodiments. The specific embodiments in the aforementioned Embodiment 1 are also applicable to the multi-parameter fusion non-all-plastic hook commutator fault diagnosis system described in Embodiment 2. For the sake of brevity, they will not be further elaborated here.

[0180] It should be understood that the embodiments disclosed in this invention and the above description enable those skilled in the art to implement this invention. However, this invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention, and should all be included within the protection scope of this invention.

Claims

1. A multi-parameter fusion non-full plastic hook type commutator fault diagnosis method, characterized in that, The method comprises the following steps: Obtain the design drawing information of the non-full-plastic hook-type commutator, identify the key parts of the non-full-plastic hook-type commutator based on the design drawing information, and obtain M commutator key region information; Deploy sensors on the M commutator key region information respectively, build a commutator monitoring sensor network, and collect M key region multi-source monitoring data streams through the commutator monitoring sensor network; According to the application scene information of the non-full-plastic hook-type commutator, perform multi-parameter fusion on the M key region multi-source monitoring data streams respectively, and obtain M key region fusion monitoring data streams; Modal decomposition and fault identification are performed on the M key region fusion monitoring data streams to obtain M key region fault feature sets, and a commutator fault diagnosis rule space is constructed; Based on the commutator fault diagnosis rule space, the M key region fault feature sets are integrated and matched to generate a commutator fault diagnosis result; The M commutator key region information is obtained by: Based on the design drawing information, the non-full-plastic hook-type commutator is disassembled into structural components to obtain a commutator structural component set; The application functions of each structural component in the commutator structural component set are labeled respectively to determine a commutator component application function set; According to the commutator component application function set, the adjacent components of the commutator structural component set are integrated to obtain N commutator function region sets; The N commutator function region sets are evaluated and screened to obtain M commutator key region information, wherein N≥M; The M commutator key region information is obtained by: Based on the N commutator function region sets, historical fault data is crawled to obtain a non-full-plastic hook-type commutator fault database; According to the N commutator function region sets, the non-full-plastic hook-type commutator fault database is mapped and distributed to obtain N function region fault data sets; Based on the N function region fault data sets, frequency statistics and probability calculation are performed to obtain N function region fault occurrence probabilities; According to the N function region fault occurrence probabilities, a fault probability threshold is preset, and the N commutator function region sets are compared and screened based on the fault probability threshold to obtain the M commutator key region information; The commutator fault diagnosis rule space is constructed by: The non-full-plastic hook-type commutator fault database is spatially divided according to the M commutator key region information to obtain M key region fault data spaces; Feature analysis and extraction are performed on the M key region fault data spaces respectively to determine M key region fault feature spaces; Fault diagnosis rules are mined in the M key region fault feature spaces to construct a commutator fault diagnosis rule space; The commutator fault diagnosis rule space is constructed by: According to the M key region fault feature spaces, M key region fault feature combinations and corresponding commutator diagnosis information are obtained; Fault correlation topology analysis is performed on the M key region fault feature combinations to obtain a commutator fault cascading feature set; The commutator fault diagnosis rule space is constructed by logically mapping and associating the commutator fault cascade feature set and the commutator diagnosis information.

2. The multi-parameter fused non-plastic hook type commutator fault diagnostic method of claim 1, wherein, The monitoring sensor network of the commutator comprises: The monitoring demand of the M commutator key area information is analyzed in sequence, and the M key area monitoring types and the M key area monitoring numerical value ranges are determined; According to the M key area monitoring types and the M key area monitoring numerical value ranges, the M key area monitoring sensor models are selected; Based on the M key area monitoring sensor models, the monitoring coverage of the M commutator key area information is analyzed, and the M key area sensor deployment parameters are obtained; Based on the M key area monitoring sensor models and the M key area sensor deployment parameters, the sensor deployment is performed on the M commutator key area information, and the monitoring sensor network of the commutator is built.

3. The multi-parameter fused non-plastic hook type commutator fault diagnostic method of claim 1, wherein, The M key area fusion monitoring data streams are obtained by: According to the application scene information of the non-full-plastic hook type commutator, the influence degree of each area data monitoring source in the M key area multi-source monitoring data streams is evaluated, and a set of M area data monitoring source influence factors is obtained; According to the M area data monitoring source influence factor set, a set of M area data monitoring source credibility coefficients is determined; The M key area multi-source time sequence data streams are obtained by integrating the M key area multi-source monitoring data streams according to the time sequence information; Based on the M area data monitoring source credibility coefficient set, the M key area multi-source time sequence data streams are subjected to multi-parameter fusion, and the M key area fusion monitoring data streams are obtained.

4. The multi-parameter fused non-plastic hook type commutator fault diagnostic method of claim 1, wherein, The M key area fault feature sets are obtained by: The M key area IMF components are obtained by respectively performing empirical mode decomposition on the M key area fusion monitoring data streams; According to the performance working standards of the M commutator key area information, M key area normal working thresholds are determined; According to the M key area normal working thresholds, the M key area IMF components are subjected to fault feature extraction, and the M key area fault feature sets are obtained.

5. The multi-parameter fused non-plastic hook type commutator fault diagnostic method of claim 1, wherein, The commutator fault diagnosis result is generated by: Based on the commutator fault diagnosis rule space, the M key area fault feature sets are subjected to matching analysis, and a set of matching commutator diagnosis rules are activated; The matching commutator diagnosis rule set is integrated and output, and the commutator fault diagnosis result is generated.

6. A multi-parameter fused non-full plastic hanger commutator fault diagnostic system, characterized in that, The non-full-plastic hook type commutator fault diagnosis method for performing multi-parameter fusion according to any one of claims 1-5 comprises: A region identification module is configured to obtain design drawing information of a non-full-plastic hook type commutator, identify key positions of the non-full-plastic hook type commutator based on the design drawing information, and obtain M commutator key area information; A region monitoring module is configured to perform sensor analysis and deployment on the M commutator key area information respectively, build a commutator monitoring sensor network, and collect and obtain M key area multi-source monitoring data streams through the commutator monitoring sensor network; a data stream fusion module configured to perform multi-parameter fusion on the M key area multi-source monitoring data streams according to application scene information of the non-full plastic hook commutator, to obtain M key area fusion monitoring data streams; a fault feature acquisition module configured to perform modal decomposition and fault identification on the M key area fusion monitoring data streams, to obtain M key area fault feature sets, and to construct a commutator fault diagnosis rule space; a fault diagnosis execution module configured to perform integrated matching analysis on the M key area fault feature sets based on the commutator fault diagnosis rule space, to generate a commutator fault diagnosis result.

Images