A method and system for relay condition monitoring and diagnosis

Abstract

This invention belongs to the field of equipment condition monitoring technology. It discloses a method and system for relay condition monitoring and diagnosis. By collecting vibration time-domain signals from various monitoring points during relay operation, a time-delay feature matrix is ​​constructed using piecewise cross-correlation calculations. Combined with spatial divergence analysis of the time-delay gradient vector field, candidate regions for near-field sound sources are identified. A sound-vibration coupling propagation model is introduced to distinguish between near-field sound source excitation and structurally transmitted vibration components. Spatial constraints and frequency domain sparsity constraints are used to decouple multi-component signals, effectively separating the relay body vibration signal. The body vibration signal is decomposed into time-frequency components to extract transient impact features of contact action and steady-state vibration features of coil excitation, constructing a multi-dimensional state feature vector. Condition diagnosis is achieved through similarity matching with a preset feature library. This invention improves the sensitivity and accuracy of early relay fault detection, providing a reliable basis for preventative maintenance.

Description

Technical Field

[0001] This invention relates to the field of equipment condition monitoring technology, and more specifically, to a relay condition monitoring and diagnostic method and system. Background Technology

[0002] Existing relay condition monitoring and diagnostic technologies have limitations when facing complex industrial environments, particularly in terms of near-field sound source isolation and structural vibration decoupling. In industrial settings, relays typically operate within high-density distribution cabinets or automated control systems. Mechanical vibrations, electromagnetic interference, and environmental noise from surrounding equipment continuously superimpose onto the relay's signal, creating a complex, multi-source, aliased vibration field. Traditional monitoring methods rely on single-point or limited sensor acquisition of mixed vibration signals, lacking effective mathematical models to separate the near-field impact vibrations generated by contact actuation from the structural vibrations caused by coil excitation, leading to unclear signal interpretation. Especially in high-frequency relay systems, the transient impacts from contact bounce and the continuous vibrations caused by coil electromagnetic forces overlap in both the time and frequency domains, making it difficult for traditional spectral analysis methods to distinguish their respective characteristics. Furthermore, during vibration propagation on the relay housing, structural transmission at different frequency bands exhibits significant dispersion—high-frequency components attenuate rapidly while low-frequency components propagate further. However, current technologies lack systematic modeling of this frequency-dependent attenuation pattern, failing to accurately characterize the differences in vibration propagation paths. In practical applications, such as protective relays in power systems, misjudging faults can lead to serious safety accidents and economic losses. On intelligent manufacturing production lines, relay failure often becomes a critical bottleneck for system downtime. However, due to the inability to effectively isolate near-field sound sources and decouple structurally transmitted vibrations, existing monitoring systems suffer from low detection sensitivity and high misjudgment rates when facing early and complex faults. In particular, gradual faults such as increased contact roughness and slight reduction in spring elasticity are difficult to detect in a timely manner.

[0003] In view of this, the present invention proposes a relay condition monitoring and diagnosis method and system to solve the above problems. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a relay condition monitoring and diagnosis method, comprising:

[0005] The vibration time-domain signals of each monitoring point during the operation of the relay are collected, and the spatial coordinates of each monitoring point are recorded.

[0006] Segmented cross-correlation operation is performed on the vibration time-domain signal to extract the arrival time delay of vibration waves between any two monitoring points and construct the time delay feature matrix between monitoring points.

[0007] Calculate the time delay gradient vector for each monitoring point in the time delay feature matrix, and identify near-field sound source candidate regions on the relay housing based on the spatial divergence distribution of the time delay gradient vector.

[0008] Obtain the structural conduction path length between the near-field sound source candidate region and each monitoring point, and calculate the frequency dependence attenuation coefficient of the structural conduction path under different frequency bands by combining the spectral energy distribution of the vibration time-domain signal.

[0009] Based on the frequency-dependent attenuation coefficient and time delay feature matrix, a sound-vibration coupled propagation model is constructed. The sound-vibration coupled propagation model includes the propagation path difference characteristics of the near-field sound source excitation component and the structure-transmitted vibration component.

[0010] By introducing spatial constraints and frequency domain sparsity constraints into the acoustic-vibration coupling propagation model, a multi-component decoupling objective function is established. An iterative optimization algorithm is used to solve the multi-component decoupling objective function, and the vibration signal of the relay body is obtained.

[0011] Time-frequency decomposition is performed on the vibration signal of the relay body to extract the transient impact characteristics during the contact action stage and the steady-state vibration characteristics during the coil excitation stage, and a multi-dimensional state feature vector is constructed.

[0012] The multidimensional state feature vector is matched with a preset relay state feature library for similarity. Based on the matching results, the current operating state of the relay is determined and a state diagnosis conclusion is output.

[0013] A relay condition monitoring and diagnostic system, used to implement a relay condition monitoring and diagnostic method, includes:

[0014] The data acquisition module is used to acquire the vibration time-domain signals of each monitoring point during the operation of the relay and record the spatial coordinates of each monitoring point;

[0015] The matrix construction module is used to perform piecewise cross-correlation calculations on vibration time-domain signals, extract the arrival time delay of vibration waves between any two monitoring points, and construct the time delay feature matrix between monitoring points.

[0016] The sound source identification module is used to calculate the time delay gradient vector of each monitoring point in the time delay feature matrix, and to identify near-field sound source candidate regions on the relay housing based on the spatial divergence distribution of the time delay gradient vector.

[0017] The attenuation coefficient calculation module is used to obtain the structural conduction path length between the near-field sound source candidate region and each monitoring point, and calculate the frequency-dependent attenuation coefficient of the structural conduction path under different frequency bands by combining the spectral energy distribution of the vibration time-domain signal.

[0018] The coupling model construction module is used to construct an acoustic-vibration coupling propagation model based on the frequency-dependent attenuation coefficient and the time delay feature matrix.

[0019] The signal decoupling module is used to establish a multi-component decoupling objective function by introducing spatial constraints and frequency domain sparsity constraints into the acoustic-vibration coupling propagation model, and to solve the multi-component decoupling objective function by using an iterative optimization algorithm to separate the vibration signal of the relay body.

[0020] The feature extraction module is used to perform time-frequency decomposition on the vibration signal of the relay body, extract the transient impact features during the contact action stage and the steady-state vibration features during the coil excitation stage, and construct a multi-dimensional state feature vector.

[0021] The status diagnosis module is used to perform similarity matching between the multi-dimensional status feature vector and the preset relay status feature library, determine the current operating status of the relay based on the matching results, and output the status diagnosis conclusion.

[0022] The technical effects and advantages of the relay condition monitoring and diagnosis method and system of the present invention are as follows:

[0023] This invention enhances the monitoring system's ability to detect early deterioration states, enabling maintenance personnel to identify potential problems before they reach critical stages, thus preventing production line downtime and system paralysis caused by sudden relay failures. In power systems, this invention effectively prevents safety accidents and cascading failures caused by protection failures, significantly reducing operational risks and economic losses. This invention exhibits extremely high detection sensitivity for gradual faults such as minor contact wear and slight spring fatigue, making preventative maintenance strategies truly feasible. In smart manufacturing environments, this invention reduces unnecessary maintenance interventions and unplanned downtime, while avoiding premature replacement of relay components still in good condition, achieving optimal allocation of maintenance resources. For large-scale automated systems with numerous relay groups, the condition assessment and lifespan prediction functions of this invention qualitatively improve overall system reliability, allowing maintenance teams to develop scientific replacement strategies based on diagnostic results, balancing immediate maintenance costs with long-term operational risks. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of a relay status monitoring and diagnosis method according to the present invention;

[0025] Figure 2 This is a schematic diagram of a relay status monitoring and diagnosis system according to the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] This application provides a relay condition monitoring and diagnosis method and system. The execution entities of the system include, but are not limited to, relay condition monitoring systems, vibration analysis platforms, industrial equipment diagnostic centers, intelligent maintenance systems, etc., which can be regarded as general computing nodes of this application. The diagnostic system includes, but is not limited to, at least one of the following: vibration-based condition monitoring systems, relay life prediction systems, and multi-sensor fusion diagnostic platforms.

[0028] Please see Figure 1 In this embodiment of the invention, a specific implementation of a relay status monitoring and diagnosis method includes:

[0029] Vibration time-domain signals from various monitoring points during relay operation are collected, and the spatial coordinates of each point are recorded. The vibration time-domain signals are acquired in real-time by multiple accelerometers arranged on the relay housing surface, encompassing the complete vibration response during key operational phases such as relay contact action, coil excitation, and mechanical transmission. The monitoring points are arranged according to the principles of critical location coverage and signal integrity, typically including important locations such as the area near the contacts, the area surrounding the coil, and the surfaces of mechanical connecting components. Spatial coordinates are recorded using a three-dimensional Cartesian coordinate system, with a reference coordinate system established using a fixed point on the relay housing as the origin to ensure accurate spatial positioning and provide a spatial basis for subsequent time delay analysis and sound source identification.

[0030] Piecewise cross-correlation is performed on the vibration time-domain signal to extract the arrival time delay of the vibration wave between any two monitoring points, constructing a time delay feature matrix between the monitoring points. Piecewise cross-correlation is a crucial step in extracting time delay information, determining the propagation time delay by comparing the phase difference of vibration signals from different monitoring points. The calculation process first divides the signal into multiple time periods according to the relay's operating cycle, then calculates the normalized cross-correlation function between each pair of monitoring points, extracting the peak position as the initial time delay estimate, and finally obtaining the sub-sampling precision time delay value through interpolation fitting. The time delay feature matrix is ​​a key data structure describing the vibration propagation law; each element in the matrix corresponds to the precise time delay between a pair of monitoring points, providing fundamental data for subsequent gradient analysis.

[0031] The time delay gradient vector of each monitoring point in the time delay feature matrix is ​​calculated. Based on the spatial divergence distribution of the time delay gradient vector, candidate regions for near-field sound sources on the relay housing are identified. The time delay gradient vector reflects the direction and velocity of the vibration wavefront propagation and is an important clue for identifying the vibration source. The calculation process first selects a reference point and extracts its time delay relationship with other points. Then, the gradient vector is obtained through spatial gradient fitting, and the divergence distribution of the gradient field is calculated. Regions with large divergence values ​​usually correspond to the location of the vibration source. By setting an appropriate threshold, candidate regions for near-field sound sources can be identified, providing a basis for the subsequent construction of the propagation model.

[0032] The structural conduction path length between the candidate near-field sound source region and each monitoring point is obtained. Combined with the spectral energy distribution of the vibration time-domain signal, the frequency-dependent attenuation coefficient of the structural conduction path at different frequency bands is calculated. The structural conduction path is the physical channel through which vibration propagates within the relay housing, and its length is closely related to signal attenuation. The acquisition process first calculates the shortest conduction path based on the relay's structural geometry. Then, by analyzing the energy attenuation characteristics of the vibration signal at different frequency bands, a quantitative relationship between attenuation and distance is established, and the frequency-dependent attenuation coefficient is extracted. This coefficient reflects the attenuation characteristics of vibrations propagating in the structure at different frequencies and is a key parameter for constructing an accurate propagation model.

[0033] Based on the frequency-dependent attenuation coefficient and time delay characteristic matrix, a sound-vibration coupled propagation model is constructed. This model incorporates the differences in the propagation paths of the near-field sound source excitation component and the structurally conducted vibration component. The sound-vibration coupled propagation model is a mathematical expression describing the vibration propagation mechanism in a relay, integrating both acoustic propagation and structural vibration mechanisms. The construction process first estimates the vibration wave propagation velocity and establishes the frequency domain transfer function. Then, the multi-source assumption and spatial sparsity constraint are introduced to form a complete coupled model. This model can distinguish between vibrations generated by near-field sound sources (such as contact impact) and vibrations generated by structural conduction (such as coil excitation), providing a theoretical basis for signal separation.

[0034] By introducing spatial constraints and frequency-domain sparsity constraints into the acoustic-vibration coupling propagation model, a multi-component decoupling objective function is established. An iterative optimization algorithm is used to solve this objective function, resulting in the separation of the relay's vibration signal. Multi-component decoupling is a crucial step in extracting a pure relay signal. The optimization algorithm separates the mixed vibration into components from different sources. The decoupling process first determines the spatial constraint range of the relay's vibration, identifies the dominant frequency to construct frequency-domain sparsity constraints, and then establishes a decoupling function with the objective of minimizing reconstruction error. This function is iteratively solved using the alternating direction multiplier method, ultimately extracting the relay's vibration signal. This signal eliminates environmental noise and propagation distortion, directly reflecting the internal mechanical state of the relay, laying the foundation for subsequent feature extraction.

[0035] Time-frequency decomposition is performed on the vibration signal of the relay body to extract transient impact features during the contact operation phase and steady-state vibration features during the coil excitation phase, constructing a multi-dimensional state feature vector. Time-frequency decomposition is an effective method for analyzing non-stationary vibration signals, capable of simultaneously acquiring feature information in both the time and frequency domains. The decomposition process employs continuous wavelet transform for time-frequency analysis, identifying the time boundary of contact operation and extracting transient impact feature parameters; simultaneously, it analyzes the steady-state frequency band corresponding to coil excitation, extracting frequency statistical features; and obtains modal energy distribution features through empirical mode decomposition. These multi-dimensional features collectively constitute the state feature vector, comprehensively describing the relay's operating state characteristics and providing a feature basis for state identification.

[0036] The system performs similarity matching between multidimensional state feature vectors and a pre-defined relay state feature library. Based on the matching results, it determines the current operating state of the relay and outputs a state diagnosis conclusion. Similarity matching is the final step in state determination, identifying the operating state by comparing the differences between the current features and known states. The matching process first normalizes the feature vectors, then calculates the Mahalanobis distance with each state reference vector in the feature library, converts this distance into similarity values, sorts and compares them, and selects the state with the highest similarity as the diagnostic result. The system can identify normal states and various fault states, such as contact wear, spring fatigue, and coil aging, and provides diagnostic confidence assessments to ensure the reliability of the diagnostic results.

[0037] In this embodiment of the invention, the detailed implementation steps for constructing the time delay feature matrix between monitoring points include:

[0038] The vibration time-domain signal is divided into multiple time periods according to the relay's operating cycle, with each time period corresponding to a complete relay operation. Time period division is a fundamental step in signal analysis, segmenting continuous monitoring data into meaningful units for refined analysis. The division process first identifies the relay's operating start point using an energy detection algorithm, then sets the window length based on the typical operating duration to ensure each time period includes the complete on / off process. For frequently operating relays, an adaptive window technique is used to dynamically adjust the segment length to ensure all feature information is captured. The divided time periods typically include a pre-operation phase (coil energization begins), a contact operation phase (contact or separation), and a holding phase (stable connection or disconnection), comprehensively recording the dynamic characteristics of the entire relay operation process.

[0039] Vibration signals from any two monitoring points within the same time period are selected, and their normalized cross-correlation function is calculated. Normalized cross-correlation is a classic method for measuring the similarity between two signals and can effectively extract phase difference information. The calculation process first preprocesses the signals, including detrending, filtering, and normalization, to eliminate the influence of baseline drift and amplitude differences. Then, the cross-correlation function is calculated using time-domain convolution or frequency-domain multiplication methods to obtain a cross-correlation sequence representing the phase relationship between the two signals. Normalization ensures that the result is not affected by signal amplitude and only reflects the phase relationship. The calculation formula is as follows:

[0040] ;

[0041] in, For the normalized cross-correlation function, and The vibration signals are from two monitoring points. This is the time lag. This represents the summation operation over the entire time period.

[0042] The peak position is searched within the normalized cross-correlation function, and the time lag corresponding to the peak is recorded as the initial time delay estimate between the two monitoring points. Peak search is a crucial step in determining the initial time delay, as the peak position directly corresponds to the time difference of the maximum correlation. The search process employs a peak detection algorithm. First, the maximum cross-correlation value is located globally to determine a rough time delay position. Then, the reasonableness of the peak is verified within physical constraints, and possible spurious peaks are eliminated. Finally, the time lag corresponding to the peak is recorded as the initial estimate. For cases with low signal-to-noise ratios, a multi-peak analysis strategy is adopted, combining amplitude and width information to comprehensively determine the true peak, thereby improving the noise resistance of the time delay estimate.

[0043] A search window is set near the initial time delay estimate, and parabolic interpolation is performed on the normalized cross-correlation function to obtain a subsampling-accurate time delay correction value. Interpolation fitting is an important technique for improving time delay accuracy, overcoming the limitations of the original sampling rate through subsampling processing. The fitting process sets a window of ± several sampling points near the initial peak, typically 3-5 points, and then uses a quadratic parabolic model to perform least-squares fitting on the peak region to calculate the theoretical peak position. This method can achieve time delay accuracy far superior to the sampling interval, making it particularly suitable for high-precision time delay analysis. The difference between the fitted peak position and the initial integer sampling point position is the time delay correction value, usually in decimal form, representing the fine adjustment amount at the subsampling level.

[0044] The initial delay estimate is added to the delay correction value to obtain the precise delay between the two monitoring points. Precise delay calculation is the final step in delay estimation, integrating information from both integer sampling precision and sub-sampling precision levels. The calculation process is simple and direct: the initial delay estimate (integer sample count) is added to the delay correction value (fractional part) to obtain the final precise delay value. For a sampling rate of... The system's latency resolution can reach Horizontal, among which As an interpolation factor, it can typically achieve an accuracy improvement of 1 / 10 to 1 / 20 of the original sampling interval, providing a high-precision data foundation for subsequent gradient analysis.

[0045] The process iterates through all monitoring point pairs, filling the matrix with the precise time delays between each pair to construct a time delay feature matrix. This matrix is ​​symmetric with zero elements on its diagonal. The time delay feature matrix is ​​a complete data structure describing the system's time delay relationships, containing the time propagation relationships between all monitoring point pairs. The construction process involves a double loop iterating through all possible combinations of monitoring point pairs, calculating and recording the precise time delay between each pair, and filling the corresponding matrix element positions. Since the time delay from point A to point B is antisymmetric to the time delay from point B to point A (equal values, opposite signs), the matrix possesses the antisymmetric property. Meanwhile, the time delay from a point to itself is zero, so all diagonal elements are zero. The resulting time delay feature matrix is ​​an n×n matrix (n is the number of monitoring points), which fully records the time delay relationship of vibration propagation, providing a data foundation for sound source localization and propagation path analysis.

[0046] In this embodiment of the invention, the detailed implementation steps for identifying near-field sound source candidate regions on the relay housing include:

[0047] A monitoring point in the time delay feature matrix is ​​selected as a reference point. The time delay values ​​between this reference point and all other monitoring points are extracted to form a time delay distribution sequence. Reference point selection is the initial step in gradient analysis and affects the accuracy and efficiency of subsequent calculations. Priority is given to monitoring points with high signal quality, stable location, and wide coverage, typically those close to the core components of the relay. The time delay distribution sequence extraction involves extracting the matrix row (or column) data corresponding to the reference point to form a one-dimensional sequence, recording the propagation time relationship of vibration from the reference point to all other points. For The monitoring point system was selected. When point is used as the reference point, the time delay distribution sequence is: ,Include This sequence of elements forms the basis for spatial gradient calculations and directly affects the accuracy of the gradient field.

[0048] Based on the time delay distribution sequence and the spatial coordinates of each monitoring point, the least squares method is used to fit the gradient field of the time delay with respect to spatial location, obtaining the time delay gradient vector at the reference point. Gradient field fitting is a key step in extracting continuous gradient information from discrete time delay data, obtaining gradient information by establishing a functional relationship between time delay and spatial location. The fitting process first establishes a spatial linear regression model of the time delay, using the time delay value as the dependent variable and the spatial coordinate increment as the independent variable; then, a design matrix is ​​constructed, with each row containing the three-dimensional coordinate increment of a monitoring point relative to the reference point. Finally, the overdetermined equations are solved using the least squares method to obtain the partial derivatives of the time delay with respect to the x, y, and z directions, which are then combined to form the gradient vector. The linear regression equation is:

[0049] ;

[0050] in, For position The time delay at the location, For reference point The time delay at that point is always zero. Let be the time delay gradient vector to be solved. The least squares solution is:

[0051] ;

[0052] in, To design the matrix, its row vectors represent the coordinate increments of each point relative to a reference point. This represents the time delay vector at each point. The resulting gradient vector... The direction of the wavefront points to the direction in which the vibration wave propagates the fastest, and its magnitude reflects the steepness of the wavefront, making it an important indicator for sound source localization.

[0053] Repeat the above steps for all monitoring points to obtain the time delay gradient vector at each monitoring point, thus constructing a time delay gradient vector field for each monitoring point. Constructing the gradient vector field is a crucial step in extending single-point gradient analysis to global analysis, forming a complete vector field by calculating the gradient at each monitoring point. The construction process iteratively sets each monitoring point as a reference point, repeatedly performing time delay sequence extraction and gradient fitting operations, calculating the gradient vector at that point, until the gradient information for all monitoring points is obtained. The final gradient vector field is a set of vectors. Each vector corresponds to a monitoring point location and points in the direction of vibration propagation at that location. This vector field comprehensively describes the spatial distribution characteristics of vibration propagation, providing complete vector data for subsequent divergence calculations.

[0054] The divergence value of the time-delay gradient vector field at each monitoring point is calculated. The divergence value is obtained by summing the spatial partial derivatives of the time-delay gradient vector. Divergence calculation is a core step in extracting sound source information from the gradient vector field, physically corresponding to the source-sink distribution in the flow field. The calculation process uses numerical differentiation methods to calculate the spatial rate of change of the gradient vector along the x, y, and z directions at each monitoring point, and then sums them to obtain the divergence value. The divergence calculation formula is:

[0055] ;

[0056] Since the monitoring points are discretely distributed, the partial derivatives are calculated using the difference approximation in actual calculations, and the central difference scheme is usually used to improve accuracy. The divergence value has a clear physical meaning: a positive value indicates that the point is a wave source (divergence point), a negative value indicates that the point is a convergence point, and a zero value indicates an ordinary point in the propagation process. The location of the sound source usually corresponds to the region of positive maxima in the divergence field, which is a direct basis for identifying the vibration source.

[0057] The spatial distribution of statistical divergence values ​​is analyzed, and monitoring points with divergence values ​​greater than a preset divergence threshold, along with their neighborhoods, are marked as near-field sound source candidate regions. The preset divergence threshold is determined based on the mean and standard deviation of the statistical distribution of divergence values. Marking near-field sound source candidate regions is the final step in sound source localization, identifying possible sound source locations through thresholded divergence fields. The statistical process first analyzes the distribution characteristics of the divergence values ​​and calculates the mean. and standard deviation Then, a divergence threshold is set based on statistical characteristics, typically using... Form, in which The multiplier is adjusted according to the detection sensitivity requirements, typically 2-3. Finally, points with divergence values ​​exceeding the threshold are marked as candidate sound source points, and their spatial neighborhoods (usually spherical regions) are also marked to form complete candidate regions. For multi-source cases, clustering algorithms are used to merge nearby candidate points into independent candidate regions, avoiding duplicate marking. These candidate regions are the focus of subsequent analysis, providing sound source location constraints for propagation model construction.

[0058] In this embodiment of the invention, the detailed implementation steps for calculating the frequency-dependent attenuation coefficient of the structural conduction path under different frequency bands include:

[0059] Based on the geometry of the relay housing and the location of near-field sound source candidate regions, the shortest structural transmission path from the near-field sound source candidate regions to each monitoring point is calculated and denoted as the path length value. The structural transmission path is the physical channel through which vibration propagates in a solid structure, and the path length directly affects propagation attenuation. The calculation process first establishes a three-dimensional geometric model of the relay housing, including the outer shell, support, and internal structure, among other main components. Then, the near-field sound source candidate regions and monitoring points are mapped onto the model and marked as nodes. Finally, an improved Dijkstra algorithm is used to search for the shortest path on the geometric model, considering structural continuity and material property constraints. For complex structures, meshing is used to discretize the continuous geometry into a node network to improve computational efficiency. The final path length value is the basic data for attenuation analysis, reflecting the actual propagation distance of vibration from the sound source to the monitoring point, usually recorded in millimeters, providing a spatial reference for subsequent attenuation law analysis.

[0060] A short-time Fourier transform (STFT) is performed on the vibration time-domain signal to obtain its time-frequency spectrum. The instantaneous energy sequence of a preset frequency band is then extracted from this spectrum. The STFT is a classic method for analyzing the time-frequency characteristics of non-stationary signals, demonstrating the variation of signal frequency components over time. The transformation process first selects an appropriate window function (usually the Hanning window) and window length to segment the signal. Then, an FFT is performed on each signal segment within the window to obtain the local spectrum. Finally, the spectra of all windows are arranged in chronological order to form the complete time-frequency spectrum. The preset frequency band is selected based on the typical frequency range of relay vibration, typically including the low-frequency band (50-500Hz, corresponding to structural resonance), the mid-frequency band (500-2000Hz, corresponding to mechanical action), and the high-frequency band (2000-5000Hz, corresponding to contact impact). The instantaneous energy sequence is obtained by integrating the corresponding frequency band of the time-frequency spectrum, reflecting the change of energy in a specific frequency band over time, and is the key input data for attenuation analysis.

[0061] Monitoring points located within the candidate region of the near-field sound source are selected as reference energy points. The energy ratio of other monitoring points in the same frequency band to that of the reference energy point is calculated and denoted as the energy attenuation ratio. Energy attenuation ratio analysis is a direct method for studying the attenuation law of vibration propagation. It quantitatively describes the attenuation process by comparing energy levels at different distances. The analysis process first selects the point with the best signal quality within the candidate region as the reference point, representing the energy level at the sound source. Then, the ratio of the energy at other points in the same frequency band to that of the reference point is calculated to obtain the energy attenuation ratio. Finally, the attenuation ratio is paired with the corresponding path length to form the relationship data between attenuation and distance. The energy attenuation ratio is a dimensionless value, typically ranging from 0 to 1; the smaller the value, the more severe the attenuation. This attenuation analysis method based on measured data avoids the simplification assumptions of theoretical models and can accurately reflect the complex attenuation phenomena in actual structures, providing reliable experimental data for solving the attenuation coefficient.

[0062] A logarithmic linear relationship between the energy attenuation ratio and the path length is established. Regression analysis is used to determine the slope of this relationship, which is then denoted as the frequency-dependent attenuation coefficient for that frequency band. Logarithmic linear regression is a classic method for extracting attenuation coefficients, based on the physical law that vibration energy decays exponentially during propagation. The regression process first converts the energy attenuation ratio into a linear relationship with distance by taking its natural logarithm; then, least squares regression is used to fit the optimal line; finally, the slope of the line is extracted as the attenuation coefficient. The logarithmic linear relationship model is as follows:

[0063] ;

[0064] in, The energy decay ratio, This is the path length value. The attenuation coefficient to be solved is denoted as . The value is usually positive, with units of 1 / mm or 1 / m. A larger value indicates faster attenuation. This coefficient directly reflects the propagation loss characteristics of vibration in a specific frequency band and structure. It is a core parameter of the propagation model and is crucial for predicting long-distance vibration responses.

[0065] The above operations are repeated for multiple preset frequency bands to obtain the frequency-dependent attenuation coefficients for different frequency bands, constructing a frequency-dependent attenuation coefficient spectrum. The attenuation coefficient spectrum is a complete dataset describing the attenuation variation with frequency, revealing the frequency-selective characteristics of structural propagation. The construction process iteratively performs energy attenuation analysis and regression calculations for each preset frequency band to obtain the attenuation coefficient for that band, ultimately forming a frequency-attenuation coefficient correspondence. For complex structures, the attenuation coefficient typically increases with frequency, exhibiting the characteristics of rapid attenuation at high frequencies and long propagation at low frequencies, which is consistent with the viscoelasticity of materials and the phonon scattering mechanism of structures. The complete attenuation coefficient spectrum is usually recorded in tabular or function curve form, providing a parameter basis for predicting the propagation of different frequency components and is a key element in constructing an accurate propagation model.

[0066] In this embodiment of the invention, the detailed implementation steps for constructing the acoustic-vibration coupling propagation model include:

[0067] Based on the time delay feature matrix and the spatial coordinates of the monitoring points, the average propagation velocity of the vibration wave in the relay housing is estimated. Propagation velocity estimation is a fundamental step in constructing the propagation model and directly affects the accuracy of the spatiotemporal mapping. The estimation process is based on the ratio of time delay to distance. First, the time delay values ​​of multiple pairs of monitoring points are extracted from the time delay feature matrix; then, the Euclidean distance between these point pairs is calculated; finally, the local propagation velocity is calculated using the ratio of distance to time delay, and the average value is taken as the overall estimate. To improve accuracy, a weighted averaging method is used, giving higher weight to measurements with high signal-to-noise ratio and short distances. For complex structures, there may be directional differences, requiring separate estimation of the propagation velocity in different directions. The estimation formula is:

[0068] ;

[0069] in, For average propagation speed, For the first Regarding the distance between monitoring points, For the corresponding delay value, This represents the weighting coefficient. Propagation speed is typically expressed in m / s. The elastic wave velocity in a typical metallic structure ranges from 3000 to 6000 m / s, with the specific value influenced by material properties and structural form. Accurate velocity estimation provides a spatiotemporal conversion benchmark for subsequent phase calculations.

[0070] Based on the average propagation velocity and frequency-dependent attenuation coefficient, a frequency domain transfer function is constructed from the near-field sound source candidate region to each monitoring point. The frequency domain transfer function includes a phase delay term and an amplitude attenuation term. The frequency domain transfer function is a mathematical expression describing the phase and amplitude transformations of vibration propagation from the sound source to the monitoring point, and is a core component of the sound-vibration propagation model. The construction process first calculates the phase delay based on the average propagation velocity and geometric distance, expressed as a linear function of frequency; then, it calculates the amplitude attenuation based on the frequency-dependent attenuation coefficient, expressed as an exponential function of distance and frequency; finally, the two parts are combined to form the complete transfer function. The transfer function is expressed in complex form, with the real and imaginary parts corresponding to the amplitude and phase information, respectively. For a distance of... , frequency is The transfer function expression is:

[0071] ;

[0072] Among them, the first item This is the amplitude attenuation term. For frequency The attenuation coefficient at the location; the second term For phase delay term, The propagation speed is given by the transfer function. This transfer function comprehensively describes the transformation law of vibration during propagation, providing a propagation model basis for subsequent signal separation.

[0073] Assuming multiple independent near-field sound sources and at least one structurally conducted excitation source exist on the relay housing, the vibration signal at each monitoring point is represented as a weighted superposition of all sound source points and the structurally conducted excitation source. The multi-source assumption is a key approach to handling complex vibration fields, decomposing the observed mixed signal into contributions from multiple independent sources. The model assumes that relay vibration contains two types of sources: near-field sound sources (such as local impacts from contact collisions) and structurally conducted excitation sources (such as overall vibrations generated by coil excitation). The superposition model is based on linear system theory, representing the signal at each monitoring point as a weighted sum of the source signals modulated by their corresponding transfer functions. For n near-field sound sources and 1 structurally conducted source, the... The vibration signal model for each monitoring point is as follows:

[0074] ;

[0075] in, For monitoring points The vibration signal spectrum For the first The source signal spectrum of a near-field sound source The source signal spectrum of the structural conduction excitation source. and They are the corresponding transfer functions respectively. This superposition model is the theoretical basis for signal separation, which transforms complex mixing problems into solvable linear equations.

[0076] Introduce the spatial sparsity hypothesis of sound source points, construct a sparse constraint matrix for the sound source positions, and embed the sparse constraint matrix into the mathematical expression of weighted superposition. Spatial sparsity constraint is an important technique for solving underdetermined separation problems, based on the physical property that sound sources are discretely distributed in space. The constraint construction first establishes a dense grid lattice on the relay housing, and regards each grid point as a potential sound source position; then introduces the L1 norm regularization term to promote the contribution of most grid points to be zero, and only retains the contribution of a few real sound source points; finally, embeds the constraint into the superposition model to form a sparse representation problem. In practical implementation, the space is discretized into m grid points, where only a few k points (k << m) are real sound sources, and the model becomes:

[0077] , the constraint condition:

[0078] where, represents the L0 norm of vector S, that is, the number of non-zero elements. Since the L0 problem is difficult to solve directly, it is usually transformed into L1 constraint for approximation. This sparse constraint greatly reduces the number of parameters to be solved and improves the solvability and stability of the separation problem.

[0079] Combine the mathematical expression of weighted superposition, the frequency-domain transfer function and the sparse constraint matrix to form an acoustic-vibration coupling propagation model. The acoustic-vibration coupling propagation model can describe the contribution degree distribution of near-field sound source components and structure-borne vibration components at each monitoring point. The acoustic-vibration coupling propagation model is a complete mathematical expression integrating the propagation characteristics of multiple sources, which combines independent components into a unified analysis framework. The model integrates the linear superposition principle, the frequency-domain transfer function and the spatial sparsity constraint to form a compact expression in matrix form:

[0080] ;

[0081] where, is the observation signal matrix of all monitoring points, <000023> is the near-field sound source transfer function matrix, is the near-field sound source signal matrix, is the structure-borne transfer function matrix, is the structure-borne source signal matrix, is the noise term. This model completely describes the characteristic differences of different types of vibration sources in the propagation process, distinguishes the local near-field effect and the global structure-borne effect, provides a theoretical framework for signal separation and source identification, and is the basic model for subsequent multi-component decoupling.

[0082] In this embodiment of the invention, the detailed implementation steps for separating the vibration signal of the relay body include:

[0083] Based on the structural characteristics of the relay, the main location of the relay body vibration is determined, and this location is set as the spatial constraint range of the structural conduction excitation source. Spatial constraints are introduced as prior knowledge to improve separation accuracy, based on a physical understanding of the relay structure and working principle. Constraint setting first analyzes the location and function of the relay's core components, identifying the parts that generate the main structural vibration, typically the coil assembly and the iron core. Then, a precise three-dimensional coordinate range is determined using CAD models or physical measurements. Finally, this area is defined as the spatial constraint of the structural conduction excitation source. The constraint form is usually a spherical or ellipsoidal region in three-dimensional space, centered at the geometric center of the coil, with a radius covering the main body of the coil and iron core. This spatial constraint directly limits the solution space of the decoupling algorithm, preventing the structural vibration source from being incorrectly assigned to other locations, thus improving the physical rationality and stability of the separation.

[0084] The energy distribution of the vibration time-domain signal in the frequency domain is statistically analyzed. Dominant frequency components with energy proportions exceeding a preset threshold are identified, and frequency-domain sparsity constraints are constructed. These constraints limit the frequency spectrum of the relay body vibration signal to have non-zero values ​​only at the dominant frequency components. Frequency-domain sparsity constraints are important constraints utilizing the spectral characteristics of vibration signals, based on the physical fact that different vibration sources have different spectral characteristics. The constraint construction first performs an FFT transform on the vibration signal to obtain the power spectral density; then, it sorts the signals by energy contribution and calculates the cumulative energy distribution; next, it selects the set of frequency points where the cumulative energy reaches 80%-90% of the total energy, defining them as dominant frequency components; finally, it constructs a frequency-domain mask function, setting the value to 1 at the dominant frequency and 0 at other frequencies. This constraint limits the frequency domain distribution of the body vibration signal. Typically, body vibrations (such as coil excitation vibration) have a relatively concentrated spectrum, mainly containing the fundamental frequency and its harmonic components. Their frequency domain characteristics are significantly different from broadband impact vibrations and random noise, providing an important frequency domain distinction for signal separation.

[0085] A multi-component decoupling objective function is established with the goal of minimizing the reconstruction error, where the reconstruction error is the mean square error between the measured vibration signals at each monitoring point and the reconstructed signal from the acoustic-vibration coupling propagation model. The objective function design is the core of the optimization solution, and the evaluation criteria for separation quality are clearly defined. The design process first defines the reconstruction error index, using the Euclidean distance of the frequency domain signal to measure the difference between the observed signal and the model-reconstructed signal; then, minimizing the error is taken as the main optimization objective, forming the basic objective function framework; finally, constraint terms and regularization terms are added to refine the objective function structure. The basic form is the least squares error:

[0086] ;

[0087] in, This represents the Frobenius norm of the matrix, which is the sum of squares of all its elements. This error term directly measures how well the model fits the observed data and is a fundamental measure of decoupling quality. The smaller the error, the more accurate the separation and the more complete the reconstruction.

[0088] In the multi-component decoupling objective function, a regularization term corresponding to the spatial constraint range and an L1 norm penalty term corresponding to the frequency domain sparsity constraint are added. Constraint introduction is a key step in integrating prior knowledge into the optimization process, achieving the constraint effect through mathematically formalized regularization terms. The introduction process first transforms the spatial constraint into a quadratic regularization term, causing the solution to concentrate within the constraint region; then, the frequency domain sparsity constraint is transformed into an L1 norm penalty term, causing the spectrum to approach zero at non-dominant frequencies; finally, the weight coefficients of each term are adjusted to balance the fitting accuracy and constraint strength. The complete objective function is:

[0089] The first term is the reconstruction error, and the second term is... For the sparsity constraint (L1 norm) of the near-field source, the third term Spatial constraints (weighted L2 norm) for structural conduction sources. and These are the weighting coefficients. Let be the spatial weight matrix. This composite objective function comprehensively considers data fitting and prior constraints, balancing model complexity and generalization ability, and providing a clear optimization direction for the solution algorithm.

[0090] The alternating direction multiplier method (ADMM) is used to iteratively solve the multi-component decoupling objective function. In each iteration, the estimated values ​​of the near-field sound source component and the structure-transmitted vibration component are updated alternately until the reconstruction error converges to a preset error threshold. ADMM is an efficient algorithm for solving constrained optimization problems, particularly suitable for handling composite objective functions. The solution process first initializes the near-field sound source components. and structural conduction components The matrix is ​​zero; then the iteration loop begins, with each iteration consisting of two steps: fixed... renew Then fix renew The reconstruction error is continuously reduced during the iteration process until the error change is less than a preset threshold or the maximum number of iterations is reached. Each update step employs convex optimization techniques. The update implements L1 constraints using a soft threshold operator. The update achieves quadratic programming through a closed-form solution. The ADMM algorithm effectively handles non-smooth constraints, avoids the limitations of traditional gradient methods, converges quickly, is insensitive to initial values, and is particularly suitable for large-scale separation problems. After iterative convergence, the optimal estimates of the near-field sound source component and the structure-transmitted component are obtained, achieving effective separation of the mixed vibration signals.

[0091] The structurally conducted vibration components are extracted from the converged solution results and denoted as the relay body vibration signal. The extraction of the body vibration signal is the final step in the decoupling process, converting the optimization solution results into actual time-domain signals. The extraction process first obtains the structurally conducted vibration components from the optimization results. Frequency domain estimation; then through the frequency domain transfer function The vibration spectrum of the relay is projected onto each monitoring point to obtain the vibration spectrum at that point. Finally, an inverse Fourier transform is used to convert the signal back to the time domain, yielding the complete time-domain signal of the relay vibration. For multiple monitoring points, the point with the highest signal-to-noise ratio is typically selected as the final output, or a weighted average is used to improve robustness. The extracted relay vibration signal eliminates near-field interference and environmental noise, accurately reflecting the internal mechanical state of the relay. It mainly includes vibration components generated by coil excitation and mechanical transmission, providing a high-quality signal foundation for subsequent state feature extraction and serving as key input data for state monitoring and fault diagnosis.

[0092] In this embodiment of the invention, the detailed implementation steps for constructing a multidimensional state feature vector include:

[0093] A continuous wavelet transform (CWT) is performed on the vibration signal of the relay body to obtain a time-frequency distribution map. The start and end times of relay contact action are identified based on the energy ridges in the time-frequency distribution map. Continuous wavelet transform (CWT) is an efficient tool for analyzing non-stationary signals, providing better time-frequency resolution than STFT. The transform process first selects a wavelet basis function suitable for transient signal analysis, typically Morlet or Gabor wavelets. Then, through scaling and translation operations, the convolution of the signal with the wavelet basis function is calculated to obtain the time-frequency coefficient matrix. Finally, the coefficient matrix is ​​converted into an energy density map, visually displaying the distribution of signal energy in the time-frequency plane. The energy ridges, which are the lines connecting energy peaks in the time-frequency map, reflect the time-varying characteristics of the main components of the signal and are particularly suitable for identifying the time boundaries of transient events. Contact action is usually manifested as a significant energy abrupt change in the time-frequency map, with the start time corresponding to a point where energy rises rapidly and the end time corresponding to energy falling back to the background level. Accurate identification of this time boundary provides a time window for subsequent feature extraction, ensuring that feature calculation focuses on the key action phase.

[0094] The vibration signal segment between the start and end times is extracted and denoted as the contact action phase signal. The peak factor and impulse factor of the contact action phase signal are calculated as transient impact characteristic parameters. These transient impact characteristic parameters are important indicators describing the transient characteristics of contact action, reflecting the impact intensity and waveform characteristics. Parameter calculation first extracts the contact action phase signal based on the identified time boundary; then, the peak factor, the ratio of the signal peak value to the root mean square value, is calculated, reflecting the steepness of the signal peak; simultaneously, the impulse factor, the ratio of the signal peak value to the average absolute value, reflects the impact intensity of the signal. These two dimensionless parameters together describe the transient characteristics of contact action and are highly sensitive to changes in contact state and action speed. Normal contact action exhibits moderate peak and impulse factors. Contact wear or spring fatigue faults usually cause significant changes in these parameters; for example, poor contact increases the peak factor, while spring weakening decreases the impulse factor. These transient characteristic parameters are important indicators of the relay's mechanical state, providing direct characteristic evidence for fault diagnosis.

[0095] In the time-frequency distribution diagram, the steady-state vibration frequency band corresponding to the coil excitation is identified, and the time-frequency coefficients within this band are extracted. The statistical moment characteristics of these time-frequency coefficients, including frequency center, frequency variance, and frequency skewness, are calculated as steady-state vibration characteristic parameters. These steady-state vibration characteristic parameters are key indicators describing the coil excitation response and reflect the spectral characteristics of electromagnetic excitation. Parameter extraction first identifies the coil excitation frequency band in the time-frequency diagram, typically a sustained stable energy distribution in the low-frequency region (50-500Hz); then, the time-frequency coefficient matrix of this band is extracted; finally, the statistical moment characteristics of the spectrum are calculated. The frequency center reflects the dominant frequency position of energy concentration, the frequency variance describes the dispersion of the spectrum, and the frequency skewness characterizes the asymmetry of the spectrum distribution. These characteristics collectively constitute the spectral profile of the coil excitation response, which is highly sensitive to changes in coil condition. A normal coil exhibits a stable frequency center and appropriate spectral width; coil aging or insufficient excitation can lead to frequency center shift, spectral broadening, or changes in skewness. These steady-state characteristics are important indicators of the coil's electrical condition, providing frequency domain feature support for diagnosing coil-related faults.

[0096] Empirical Mode Decomposition (EMD) is performed on the vibration signal of the relay body to obtain multiple intrinsic mode function (IMF) components. The energy proportion of each IMF component is calculated, and the average instantaneous frequency of the three IMF components with the largest energy proportions is selected as the modal characteristic parameters. Modal characteristic parameters are deep features describing the intrinsic structure of the signal and reflect the inherent characteristics of the vibration system. Parameter extraction first involves performing EMD on the signal, adaptively decomposing the complex signal into multiple IMFs, each representing a vibration mode at a different scale. Then, the energy proportion of each IMF component is calculated to identify the dominant mode. Finally, the average instantaneous frequency of the main IMFs is calculated as the modal characteristic parameters. The instantaneous frequency is obtained through Hilbert transform, reflecting the frequency variation characteristics of the mode over time. These modal characteristics are directly related to the mechanical structural characteristics of the relay and are highly sensitive to structural changes and mechanical loosening. A normal relay exhibits a stable modal energy distribution and instantaneous frequency, while mechanical faults can lead to changes in energy distribution or instantaneous frequency drift. These modal features provide a structural dynamics perspective for fault diagnosis, and are particularly suitable for the detection and identification of early mechanical faults.

[0097] Transient impact characteristic parameters, steady-state vibration characteristic parameters, and modal characteristic parameters are arranged sequentially to form a multi-dimensional state feature vector. This multi-dimensional state feature vector is a set of features describing the comprehensive state of the relay, integrating multi-dimensional information from the time domain, frequency domain, and modal domain. The vector construction organizes features according to functional modules, including transient impact features (such as peak factor and impulse factor), steady-state vibration features (such as frequency center, frequency variance, and frequency skewness), and modal features (such as the instantaneous average frequency of the main IMFs). This multi-dimensional feature structure achieves a comprehensive characterization of different aspects of the relay's state: transient features reflect the contact operation state, steady-state features reflect the coil electrical state, and modal features reflect the mechanical structure state. This comprehensive feature vector can capture feature changes caused by different types of faults, providing a rich feature foundation for subsequent state identification and fault diagnosis, greatly improving the accuracy and comprehensiveness of state monitoring.

[0098] In this embodiment of the invention, the detailed implementation steps for determining the current operating state of the relay include:

[0099] Normalization is performed on multidimensional state feature vectors to eliminate the influence of differences in the dimensions of different feature parameters. Normalization is a fundamental step in feature processing, ensuring that features with different dimensions have equal weight in subsequent analysis. The process first statistically analyzes the distribution of each feature in the training set, calculating the mean and standard deviation. Then, Z-score standardization is used, subtracting the mean from each feature and dividing by the standard deviation to convert it into a standard distribution with a mean of 0 and a standard deviation of 1. For some non-normally distributed features, quantile normalization is used, based on the cumulative distribution function, to ensure a more uniform distribution of the normalized features. Special attention is paid to outlier handling during normalization, using quantile truncation or logarithmic transformation techniques to reduce the impact of extreme values. The normalized feature vectors have similar numerical ranges across dimensions, eliminating the original dimensional differences (e.g., peak factors are typically 3-5, while frequency centers may be in the hundreds of hertz), providing a fair basis for subsequent distance calculations and improving the accuracy and robustness of feature matching.

[0100] Reference feature vectors for each known state category are extracted from the relay state feature library. This library contains reference feature vectors corresponding to normal state, contact wear state, spring fatigue state, and coil aging state. Reference feature vector extraction is a prerequisite for state matching, providing standard feature templates for known states. The feature library is constructed based on extensive experimental and field data. Through vibration signal analysis of relays in different states, feature vectors for typical states are extracted, validated, and then stored in the feature library. The feature library contains multiple state categories: normal state represents the ideal operating state of the relay; contact wear state reflects contact abnormalities caused by contact surface erosion and increased roughness; spring fatigue state describes the weakening of mechanical force due to the failure of elastic elements; and coil aging state manifests as insulation degradation, inter-turn short circuits, and other electrical performance degradation. Each state typically contains multiple feature vector samples, reflecting different degrees and manifestations of the same state, enhancing the coverage and adaptability of the feature library. These reference feature vectors serve as benchmarks for state identification, providing comparison standards for classifying unknown states and forming the knowledge base for state monitoring.

[0101] The Mahalanobis distance between a multidimensional state feature vector and each reference feature vector is calculated, taking into account the covariance relationship between the dimensions of the feature vectors. Mahalanobis distance is a high-level metric for measuring the similarity of feature vectors, considering both correlation and scale differences between features. The calculation process first estimates the covariance matrix of the feature space, describing the statistical association between each feature dimension; then, it calculates the Mahalanobis distance between feature vectors based on the covariance matrix, achieving a distance metric that considers correlation. Compared to Euclidean distance, Mahalanobis distance achieves rotation and scaling transformations of the feature space through the covariance matrix, eliminating the influence of correlation between features and providing a more reasonable similarity assessment. For highly correlated feature dimensions, Mahalanobis distance automatically reduces their influence weight, avoiding the amplification effect of redundant information; for highly discriminative feature dimensions, it automatically increases their contribution ratio, enhancing discriminative ability. This intelligently weighted distance metric is particularly suitable for the similarity assessment of multidimensional features, providing a precise mathematical foundation for state recognition.

[0102] The Mahalanobis distance is converted into a similarity value, which is obtained by taking the negative exponential function of the Mahalanobis distance. Similarity conversion is a process that transforms distance metrics into more intuitive similarity metrics, facilitating comparison and threshold determination. The conversion uses a negative exponential function to map unbounded distance values ​​into similarity values ​​within the interval [0,1]. The conversion formula is as follows:

[0103] ;

[0104] in, The similarity value is... The Mahalanobis distance, The scaling parameter is used to adjust the sensitivity of the transformation. A similarity value of 1 indicates perfect similarity (distance 0), while a similarity value close to 0 indicates extreme dissimilarity (large distance). This transformation maintains the monotonic relationship of the original distances while providing a more standardized and intuitive similarity metric, facilitating the setting of a unified judgment threshold and simplifying subsequent decision-making logic. The scaling parameter σ is typically determined through cross-validation, choosing a value that maximizes the discriminative power between categories; a typical value is the median of the intra-class distances within the training set. This similarity metric provides a probabilistic basis for the final state judgment, improving the interpretability and confidence assessment capabilities of the diagnosis.

[0105] The state category corresponding to the reference feature vector with the highest similarity value is selected as the preliminary diagnostic result, and this similarity value is recorded as the diagnostic confidence level. State selection is the core decision-making step in the diagnostic process, determining the most likely state category based on the principle of maximum similarity. The selection process first compares the similarity values ​​of the current feature vector with all reference feature vectors to identify the maximum similarity and its corresponding state category; then, this category is used as the preliminary diagnostic result, and the maximum similarity value is recorded as the diagnostic confidence level to quantify the reliability of the result. To improve robustness, for multiple states with close similarity, an ensemble decision-making method can be used to comprehensively consider the weighted voting results of multiple highly similar states. Diagnostic confidence level is an important indicator for result evaluation, directly reflecting the degree of matching between the current state and the known pattern, providing a reference for subsequent decision-making processes, especially providing a risk assessment basis at fuzzy state boundaries. This state selection strategy based on maximizing similarity inherits the simplicity and efficiency of traditional pattern matching, and increases the flexibility of decision-making through confidence level evaluation, adapting to the diagnostic needs of different scenarios.

[0106] If the diagnostic confidence level is lower than a preset confidence threshold, it is marked as a suspected abnormal state, triggering a manual review process. If the diagnostic confidence level is higher than the preset confidence threshold, a preliminary diagnostic result is output as the final status diagnosis conclusion. Confidence level judgment is a safety assurance mechanism for the diagnostic process, identifying uncertain situations and avoiding erroneous diagnoses through threshold judgment. The judgment process first sets an appropriate confidence threshold, usually determined based on the statistical accuracy of historical diagnoses; then, it compares the current diagnostic confidence level with the threshold and makes an appropriate decision. For high-confidence results, the preliminary diagnosis is directly adopted as the final conclusion, improving processing efficiency; for low-confidence results, it is marked as a suspected state and triggers manual review, making a judgment based on expert experience and more evidence to avoid the risk of system misjudgment. The confidence threshold setting balances automation efficiency and safety reliability, typically between 0.7 and 0.85, adjusted according to the risk tolerance of the application scenario. This dual-threshold decision mechanism ensures efficient automatic diagnosis in most cases while reserving a safe channel for expert intervention in special and complex situations, improving the overall reliability and adaptability of the system, making it particularly suitable for status monitoring applications of critical equipment.

[0107] In this embodiment of the invention, the detailed implementation steps for obtaining the time delay gradient vector at the reference point by fitting the gradient field of the time delay with respect to spatial location using the least squares method based on the time delay distribution sequence and the spatial coordinates of each monitoring point include:

[0108] Using the time delay values ​​in the time delay distribution sequence as the dependent variable and the spatial coordinate increments of each monitoring point relative to the reference point as the independent variable, a spatial linear regression model of time delay is established. The spatial linear regression model is a fundamental mathematical tool for mapping discrete time delay data to a continuous gradient field. Model establishment first determines the reference coordinate system, typically a local Cartesian coordinate system with the reference point as the origin; then, the coordinate increment vector of each monitoring point relative to the reference point is calculated. That is, the difference between the coordinates of the monitoring point and the coordinates of the reference point; finally, the time delay is established. The linear mapping relationship between time delay and spatial location. The basic form of the linear regression model expresses the linear relationship between time delay and coordinate increment, where the coefficients to be solved are the partial derivatives of time delay with respect to each direction. This linear model expresses the rate of change of time delay in three-dimensional space as partial derivatives with respect to each direction, intuitively reflecting the variation law of time delay in space, and laying the mathematical foundation for solving the gradient vector.

[0109] The design matrix for constructing the spatial linear regression model is used. Each row of the design matrix contains the components of the spatial coordinate increment of a monitoring point. The design matrix is ​​a key data structure for solving the least squares method, transforming the problem into a standard matrix form for easier computation. The matrix construction process first determines the matrix dimension. For n monitoring points (excluding the reference point itself), the design matrix is ​​an n×3 matrix, where 3 represents the coordinate dimension in three-dimensional space. Then, the matrix elements are filled, with the i-th row corresponding to the coordinate increment of the i-th monitoring point. Simultaneously, a dependent variable vector is constructed, containing the time delay values ​​of all monitoring points. Each column of the matrix corresponds to a spatial direction, and each row corresponds to an observation point, organizing the discrete spatial-time delay correspondence into a standardized matrix form, facilitating subsequent least-squares solutions. This matrix design fully utilizes the redundant information from multi-point measurements, improving the noise resistance and accuracy of gradient estimation.

[0110] The least squares method is used to solve the spatial linear regression model, obtaining the partial derivatives of time delay with respect to each direction of the spatial coordinates. The least squares method is a classic approach for handling overdetermined equation systems, finding the optimal fitting parameters by minimizing the sum of squared residuals. The solution process expresses the linear regression problem as a matrix equation, where the parameter vector to be solved contains the partial derivatives of time delay with respect to the x, y, and z directions. When the number of monitoring points is greater than 3, the equation system is usually overdetermined, and no exact solution exists. In this case, the least squares criterion is used to find the solution that minimizes the sum of squared residuals. To improve computational stability, numerical methods such as QR decomposition and SVD decomposition are commonly used in practical implementations, especially when the monitoring points are unevenly distributed, which may lead to ill-conditioned matrix problems. The least squares method not only provides estimates of the partial derivatives but also evaluates the fitting quality through residual analysis, providing a reference for the reliability of the gradient results. The final obtained partial derivatives directly reflect the rate of change of time delay in the three orthogonal directions and are the fundamental components for constructing the gradient vector.

[0111] Combining partial derivatives into a vector form yields the time delay gradient vector at the reference point. The direction of the time delay gradient vector indicates the direction of fastest wave propagation, and its magnitude reflects the steepness of the wavefront. Gradient vector combination is the final step in the analysis process, integrating the partial derivatives in the three directions into a unified vector expression. The combination process directly arranges the obtained partial derivatives according to spatial directions, forming a three-dimensional vector. This vector has a clear physical meaning: its direction is perpendicular to the time delay isosurface, pointing in the direction of the fastest increase in time delay, i.e., the direction of wave propagation; the vector magnitude represents the rate of change of time delay per unit distance, reflecting the steepness of the wavefront and inversely proportional to the wave velocity. For a point source, the gradient direction should theoretically point to the source location; for a plane wave, the gradient direction is perpendicular to the wavefront. The magnitude of the gradient vector is related to the intensity and distance of the vibration source; generally, the closer to the source, the larger the gradient magnitude. In practical applications, by analyzing the gradient vector field at multiple reference points, the vibration propagation path can be traced, the vibration source location identified, and crucial information provided for subsequent source localization and propagation analysis. This spatial gradient-based analysis method makes full use of the geometric characteristics of vibration propagation, achieving an effective mapping from time delay distribution to propagation direction, and is the mathematical basis for sound source identification.

[0112] The foregoing has described a relay condition monitoring and diagnosis method according to an embodiment of this application. The following describes a relay condition monitoring and diagnosis system according to an embodiment of this application. Please refer to [link to relevant documentation]. Figure 2 One embodiment of a relay condition monitoring and diagnostic system in this application includes:

[0113] The data acquisition module is used to acquire the vibration time-domain signals of each monitoring point during the operation of the relay and record the spatial coordinates of each monitoring point;

[0114] The matrix construction module is used to perform piecewise cross-correlation calculations on vibration time-domain signals, extract the arrival time delay of vibration waves between any two monitoring points, and construct the time delay feature matrix between monitoring points.

[0115] The sound source identification module is used to calculate the time delay gradient vector of each monitoring point in the time delay feature matrix, and to identify near-field sound source candidate regions on the relay housing based on the spatial divergence distribution of the time delay gradient vector.

[0116] The attenuation coefficient calculation module is used to obtain the structural conduction path length between the near-field sound source candidate region and each monitoring point, and calculate the frequency-dependent attenuation coefficient of the structural conduction path under different frequency bands by combining the spectral energy distribution of the vibration time-domain signal.

[0117] The coupling model construction module is used to construct an acoustic-vibration coupling propagation model based on the frequency-dependent attenuation coefficient and the time delay feature matrix.

[0118] The signal decoupling module is used to establish a multi-component decoupling objective function by introducing spatial constraints and frequency domain sparsity constraints into the acoustic-vibration coupling propagation model, and to solve the multi-component decoupling objective function by using an iterative optimization algorithm to separate the vibration signal of the relay body.

[0119] The feature extraction module is used to perform time-frequency decomposition on the vibration signal of the relay body, extract the transient impact features during the contact action stage and the steady-state vibration features during the coil excitation stage, and construct a multi-dimensional state feature vector.

[0120] The status diagnosis module is used to perform similarity matching between the multidimensional status feature vector and the preset relay status feature library, determine the current operating status of the relay based on the matching results, and output the status diagnosis conclusion.

[0121] The modules are connected via wired and / or wireless means to enable data transmission between them.

[0122] This invention achieves comprehensive monitoring and accurate diagnosis of relay conditions through vibration signal acquisition, time delay feature extraction, near-field sound source identification, sound-vibration coupling model construction, and state feature matching. The sound-vibration coupling analysis method of this invention can accurately separate the relay body vibration signal and effectively identify various fault states.

[0123] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, 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. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0124] It should be noted that all formulas in this manual are calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0125] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for relay condition monitoring and diagnosis, characterized in that, include: The vibration time-domain signals of each monitoring point during the operation of the relay are collected, and the spatial coordinates of each monitoring point are recorded. Perform piecewise cross-correlation on the vibration time-domain signal, extract the arrival time delay of vibration waves between any two monitoring points, and construct the time delay feature matrix between monitoring points; Calculate the time delay gradient vector for each monitoring point in the time delay feature matrix, and identify near-field sound source candidate regions on the relay housing based on the spatial divergence distribution of the time delay gradient vector; Obtain the structural conduction path length between the near-field sound source candidate region and each monitoring point, and calculate the frequency dependence attenuation coefficient of the structural conduction path under different frequency bands by combining the spectral energy distribution of the vibration time-domain signal. Based on the frequency-dependent attenuation coefficient and the time delay feature matrix, an acoustic-vibration coupled propagation model is constructed. By introducing spatial constraints and frequency domain sparsity constraints into the acoustic-vibration coupling propagation model, a multi-component decoupling objective function is established. An iterative optimization algorithm is used to solve the multi-component decoupling objective function, and the vibration signal of the relay body is separated. The vibration signal of the relay body is decomposed into time and frequency, and the transient impact characteristics of the contact action stage and the steady-state vibration characteristics of the coil excitation stage are extracted to construct a multi-dimensional state feature vector. The multidimensional state feature vector is matched with a preset relay state feature library for similarity. Based on the matching result, the current operating state of the relay is determined and a state diagnosis conclusion is output.

2. The relay condition monitoring and diagnosis method according to claim 1, characterized in that, The step of performing piecewise cross-correlation on the vibration time-domain signal, extracting the arrival time delay of the vibration wave between any two monitoring points, and constructing a time delay feature matrix between monitoring points includes: The vibration time-domain signal is divided into multiple time periods according to the relay's operating cycle; Select vibration signals from any two monitoring points within the same time period and calculate their normalized cross-correlation function. Search for the peak position in the normalized cross-correlation function, and record the time lag corresponding to the peak as the initial time delay estimate between the two monitoring points; A search window is set near the initial time delay estimate, and parabolic interpolation fitting is performed on the normalized cross-correlation function to obtain the time delay correction value with subsampling accuracy. The initial delay estimate is added to the delay correction value to obtain the precise delay between the two monitoring points; Traverse all monitoring point pairs and fill the matrix with the precise time delay between each monitoring point pair to construct the time delay feature matrix.

3. The relay condition monitoring and diagnosis method according to claim 1, characterized in that, The step of calculating the time delay gradient vector for each monitoring point in the time delay feature matrix, and identifying near-field sound source candidate regions on the relay housing based on the spatial divergence distribution of the time delay gradient vector, includes: Select one monitoring point from the time delay feature matrix as a reference point, and extract the time delay values ​​between the reference point and all other monitoring points to form a time delay distribution sequence; Based on the time delay distribution sequence and the spatial coordinates of each monitoring point, the gradient field of the time delay with respect to the spatial position is fitted using the least squares method to obtain the time delay gradient vector at the reference point. Repeat the above operation for all monitoring points to obtain the time delay gradient vector at each monitoring point and construct the time delay gradient vector field of the monitoring points. Calculate the divergence value of the time-delay gradient vector field at each monitoring point; The spatial distribution of the divergence values ​​is statistically analyzed, and monitoring points with divergence values ​​greater than a preset divergence threshold and their neighborhoods are marked as candidate regions for near-field sound sources.

4. The relay condition monitoring and diagnosis method according to claim 1, characterized in that, The step of obtaining the structural conduction path length between the near-field sound source candidate region and each monitoring point, and calculating the frequency-dependent attenuation coefficient of the structural conduction path at different frequency bands in combination with the spectral energy distribution of the vibration time-domain signal, includes: Based on the geometry of the relay housing and the location of the near-field sound source candidate region, the shortest structural conduction path from the near-field sound source candidate region to each monitoring point is calculated and denoted as the path length value. Perform a short-time Fourier transform on the vibration time-domain signal to obtain the time spectrum of the vibration signal, and extract the instantaneous energy sequence of the preset frequency band from the time spectrum; A monitoring point located within the near-field sound source candidate region is selected as a reference energy point. The energy ratio of other monitoring points in the same frequency band to that of the reference energy point is calculated and denoted as the energy attenuation ratio. Establish a logarithmic linear relationship between the energy attenuation ratio and the path length value, and use regression analysis to solve for the slope of the logarithmic linear relationship. Record the slope as the frequency-dependent attenuation coefficient in this frequency band.

5. The relay condition monitoring and diagnosis method according to claim 1, characterized in that, The construction of the acoustic-vibration coupled propagation model based on the frequency-dependent attenuation coefficient and the time delay feature matrix includes: Based on the time delay feature matrix and the spatial coordinates of the monitoring point, the average propagation speed of the vibration wave in the relay housing is estimated. Based on the average propagation speed and the frequency-dependent attenuation coefficient, a frequency domain transfer function is constructed from the near-field sound source candidate region to each monitoring point; Assuming there are multiple independent near-field sound sources and at least one structural conduction excitation source on the relay housing, the vibration signal of each monitoring point is represented as a weighted superposition of all sound sources and structural conduction excitation sources; By introducing the spatial sparsity assumption of sound source points, a sparse constraint matrix of sound source locations is constructed, and the sparse constraint matrix is ​​embedded in the weighted superposition mathematical expression. The weighted superposition mathematical expression, the frequency domain transfer function, and the sparse constraint matrix are combined to form the acoustic-vibration coupled propagation model.

6. The relay condition monitoring and diagnosis method according to claim 1, characterized in that, The process involves introducing spatial constraints and frequency domain sparsity constraints into the acoustic-vibration coupling propagation model to establish a multi-component decoupling objective function. An iterative optimization algorithm is then used to solve this multi-component decoupling objective function, resulting in the separation of the relay body vibration signal, including: Based on the structural characteristics of the relay, the main location where the relay body vibration is generated is determined, and this location is set as the spatial constraint range of the structural conduction excitation source; The energy distribution of the vibration time-domain signal in the frequency domain is statistically analyzed, the dominant frequency components with an energy proportion greater than a preset proportion threshold are identified, and frequency domain sparsity constraints are constructed. Establish a multi-component decoupling objective function with the goal of minimizing reconstruction error; Add the regularization term corresponding to the spatial constraint range and the L1 norm penalty term corresponding to the frequency domain sparsity constraint to the multi-component decoupling objective function; The multi-component decoupling objective function is solved iteratively using the alternating direction multiplier method. In each iteration, the estimated values ​​of the near-field sound source component and the structure-transmitted vibration component are updated alternately until the reconstruction error converges to a preset error threshold. The structural vibration component is extracted from the converged solution and denoted as the vibration signal of the relay body.

7. The relay condition monitoring and diagnosis method according to claim 1, characterized in that, The process involves performing time-frequency decomposition on the vibration signal of the relay body, extracting the transient impact characteristics during the contact operation phase and the steady-state vibration characteristics during the coil excitation phase, and constructing a multi-dimensional state feature vector, including: A continuous wavelet transform is performed on the vibration signal of the relay body to obtain a time-frequency distribution map. The start and end times of the relay contact action are identified based on the energy ridge line of the time-frequency distribution map. The vibration signal segment between the start time and the end time is extracted and denoted as the contact action stage signal. The peak factor and impulse factor of the contact action stage signal are calculated as transient impact characteristic parameters. In the time-frequency distribution diagram, identify the steady-state vibration frequency band corresponding to the coil excitation, extract the time-frequency coefficients in the frequency band, and calculate the statistical moment characteristics of the time-frequency coefficients, including the frequency center, frequency variance, and frequency skewness, as steady-state vibration characteristic parameters. Empirical mode decomposition is performed on the vibration signal of the relay body to obtain multiple intrinsic mode function components. The energy ratio of each intrinsic mode function component is calculated, and the average instantaneous frequency of the three intrinsic mode function components with the largest energy ratio is selected as the modal characteristic parameter. The transient impact characteristic parameters, the steady-state vibration characteristic parameters, and the modal characteristic parameters are arranged in order to form the multidimensional state characteristic vector.

8. The relay status monitoring and diagnosis method according to claim 1, characterized in that, The step of performing similarity matching between the multidimensional state feature vector and a preset relay state feature library, and determining the current operating state of the relay based on the matching result, includes: The multidimensional state feature vector is normalized to eliminate the influence of differences in the dimensions of different feature parameters; Reference feature vectors for each known state category are extracted from the relay state feature library, which includes reference feature vectors corresponding to normal state, contact wear state, spring fatigue state, and coil aging state. Calculate the Mahalanobis distance between the multidimensional state feature vector and each reference feature vector; Convert the Mahalanobis distance into similarity values; The state category corresponding to the reference feature vector with the highest similarity value is selected as the preliminary diagnosis result, and the similarity value is recorded as the diagnosis confidence. If the diagnostic confidence level is lower than the preset confidence threshold, it is marked as a suspected abnormal state, triggering a manual review process; if the diagnostic confidence level is higher than the preset confidence threshold, the preliminary diagnostic result is output as the final state diagnostic conclusion.

9. The relay condition monitoring and diagnosis method according to claim 3, characterized in that, The step of fitting the gradient field of the time delay with respect to spatial location using the least squares method based on the time delay distribution sequence and the spatial coordinates of each monitoring point to obtain the time delay gradient vector at the reference point includes: Using the time delay values ​​in the time delay distribution sequence as the dependent variable and the spatial coordinate increments of each monitoring point relative to the reference point as the independent variable, a spatial linear regression model of time delay is established. Construct the design matrix of the spatial linear regression model, wherein each row of the design matrix contains the components of the spatial coordinate increment of a monitoring point; The spatial linear regression model is solved using the least squares method to obtain the partial derivatives of the time delay with respect to each direction of the spatial coordinates. The partial derivatives are combined into a vector form to obtain the time delay gradient vector at the reference point.

10. A relay condition monitoring and diagnostic system, used to implement the relay condition monitoring and diagnostic method according to any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to acquire the vibration time-domain signals of each monitoring point during the operation of the relay and record the spatial coordinates of each monitoring point; The matrix construction module is used to perform piecewise cross-correlation calculation on the vibration time-domain signal, extract the arrival time delay of vibration waves between any two monitoring points, and construct the time delay feature matrix between monitoring points. The sound source identification module is used to calculate the time delay gradient vector of each monitoring point in the time delay feature matrix, and to identify near-field sound source candidate regions on the relay housing based on the spatial divergence distribution of the time delay gradient vector. The attenuation coefficient calculation module is used to obtain the structural conduction path length between the near-field sound source candidate region and each monitoring point, and calculate the frequency-dependent attenuation coefficient of the structural conduction path under different frequency bands by combining the spectral energy distribution of the vibration time-domain signal. The coupling model construction module is used to construct an acoustic-vibration coupling propagation model based on the frequency-dependent attenuation coefficient and the time delay feature matrix. The signal decoupling module is used to establish a multi-component decoupling objective function by introducing spatial constraints and frequency domain sparsity constraints into the acoustic-vibration coupling propagation model, and to solve the multi-component decoupling objective function by using an iterative optimization algorithm to separate the vibration signal of the relay body. The feature extraction module is used to perform time-frequency decomposition on the vibration signal of the relay body, extract the transient impact features of the contact action stage and the steady-state vibration features of the coil excitation stage, and construct a multi-dimensional state feature vector. The status diagnosis module is used to perform similarity matching between the multidimensional status feature vector and a preset relay status feature library, determine the current operating status of the relay based on the matching result, and output the status diagnosis conclusion.

Images