Actuator sealing test method and system

Abstract

The present invention discloses an actuator sealing test method and system, which specifically relates to the technical field of sealing performance detection of industrial equipment, and is used to solve the positioning failure problem caused by signal aliasing in multiple leakage scenarios of existing methods; by pressurizing the sealing cavity and synchronously collecting dynamic pressure and vibration signals, the pressure attenuation spectrum characteristics and leakage impact spectrum peaks are extracted respectively; the path loss value is calculated based on the amplitude gradient and propagation time difference to screen candidate leakage nodes; a spatial grid model is constructed to solve the three-dimensional coordinates of the leakage source; after projecting the coordinates onto the surface of the structural model, the pressure modal coherence and vibration impact energy are coupled and analyzed to generate a detection map; finally, a multi-leak point identification map is output through spatial clustering and intensity grading; the precise separation and spatial positioning of multiple leakage signals are achieved, and the reliability of complex sealing structure detection is significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial equipment sealing performance detection technology, and more particularly, to an actuator sealing performance testing method and system. Background Art

[0002] In the manufacturing and maintenance of industrial equipment, the sealing performance of actuators is a key indicator for reliable operation. Currently, actuator sealing testing relies primarily on conventional techniques such as pressure decay, bubble detection, or tracer gas detection. These methods effectively determine whether a sealing system is leaking by monitoring pressure changes, observing bubble formation, or detecting tracer gas concentration. However, these techniques typically rely on a single signal dimension (such as overall pressure or total gas concentration) for leak detection, and their application scenarios are primarily focused on detecting a single leak source.

[0003] When there are multiple leakage points in the actuator sealing structure, the signals generated by different leakage sources will superimpose on each other, resulting in the detection data being unable to accurately reflect the actual status of each leakage point. The existing methods lack the ability to separate multiple leakage signals, making it difficult to locate the specific leakage location, which can easily lead to misjudgment or repeated testing, significantly reducing the detection efficiency and the reliability of maintenance guidance. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an actuator sealing test method and system to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for testing the sealing performance of an actuator comprises the following steps:

[0007] S1. Pressurize the sealed cavity of the actuator to be tested to a preset pressure value, and simultaneously collect the dynamic pressure signal in the sealed cavity and the vibration waveform signal of the housing;

[0008] S2. Performing time-frequency analysis on the dynamic pressure signal to extract a first spectrum peak containing pressure decay characteristics, and performing spectrum analysis on the vibration waveform signal to extract a second spectrum peak related to leakage impact;

[0009] S3, calculating a path loss value based on the amplitude gradient of the first spectrum peak and the propagation time difference of the second spectrum peak, and screening sensor groups with path loss values ​​less than a loss threshold as candidate leakage nodes;

[0010] S4, constructing a spatial grid model with the candidate leakage nodes as vertices, and solving the initial three-dimensional coordinate set of the leakage source by the arrival time difference of the vibration waveform signal;

[0011] S5. Projecting the initial three-dimensional coordinate set onto the surface of the actuator three-dimensional structural model, and generating a detection map based on the coupled analysis of the modal spatial coherence of the pressure signal and the transient impact energy of the vibration;

[0012] S6. Based on the coordinate distribution and energy proportion of the leakage area in the detection map, a detection identification map of multiple leakage points is generated through spatial density clustering and leakage intensity classification analysis.

[0013] In a preferred embodiment, pressurizing the sealed cavity of the actuator to be tested to a preset pressure value, and synchronously collecting the dynamic pressure signal in the sealed cavity and the vibration waveform signal of the housing, includes:

[0014] After fixing the actuator to be tested to the test bench, apply increasing pressure in stages until the sealed chamber reaches the preset pressure value;

[0015] When the pressure in the sealed cavity reaches a preset pressure value, the dynamic pressure signal in the sealed cavity is continuously collected at a preset sampling frequency by the pressure sensor group, and the vibration waveform signal of the shell is collected at a synchronous sampling clock by the vibration sensor array attached to the surface of the shell;

[0016] The dynamic pressure signal and the vibration waveform signal are aligned in the time domain to generate a multi-channel signal dataset with consistent timestamps.

[0017] In a preferred embodiment, performing time-frequency analysis on the dynamic pressure signal to extract a first spectrum peak containing pressure decay characteristics, and performing spectrum analysis on the vibration waveform signal to extract a second spectrum peak related to leakage impact, includes:

[0018] Perform bandpass filtering on the dynamic pressure signal in the multi-channel signal data set to retain the signal components within the preset frequency band;

[0019] performing time-frequency analysis on the filtered dynamic pressure signal, extracting a time-frequency spectrum through short-time Fourier transform, and identifying a frequency component in the time-frequency spectrum whose amplitude decay rate exceeds a first threshold as a first spectrum peak;

[0020] Performing windowing processing on the vibration waveform signal in the multi-channel signal data set, generating a power spectrum density map through fast Fourier transform, and locating the frequency component in the power spectrum density map where the energy mutation amplitude exceeds a second threshold as the second spectrum peak;

[0021] The first spectrum peak is waveform matched with the preset leakage attenuation feature template, and candidate first spectrum peaks whose matching degree exceeds the third threshold are screened; the second spectrum peak is correlated with the preset mechanical shock feature template, and candidate second spectrum peaks whose correlation coefficient exceeds the fourth threshold are screened.

[0022] In a preferred embodiment, the path loss value is calculated based on the propagation time difference between the amplitude gradient of the first spectrum peak and the second spectrum peak, and the sensor group having a path loss value less than a loss threshold is screened as a candidate leakage node, including:

[0023] Calculate the amplitude attenuation coefficient of the received signal of each sensor on the leakage propagation path based on the frequency distribution and amplitude gradient of the candidate first spectrum peak;

[0024] Calculate the propagation speed of the leakage impulse signal between the sensors based on the arrival time difference of the candidate second spectrum peak and the sensor spacing;

[0025] Generate a path loss value based on the product of the amplitude attenuation coefficient and the propagation speed;

[0026] Compare the path loss value with the preset loss threshold, and select the sensor group whose path loss value is less than the loss threshold as the candidate leakage node;

[0027] The spatial distribution of candidate leakage nodes is topologically verified, and after removing isolated nodes, sensor groups that meet the adjacency density conditions are retained as valid candidate leakage nodes.

[0028] In a preferred embodiment, a spatial grid model with candidate leakage nodes as vertices is constructed, and the initial three-dimensional coordinate set of the leakage source is solved by the arrival time difference of the vibration waveform signal, including:

[0029] Construct a spatial grid model with the three-dimensional spatial coordinates of valid candidate leakage nodes as vertices, and the edges between vertices are the actual distances between adjacent nodes;

[0030] Calculate the propagation time difference vector of the leakage shock wave according to the arrival time difference of the vibration waveform signal between each valid candidate leakage node;

[0031] Based on the topological relationship of vertices in the spatial grid model and the propagation time difference vector, a set of geometric constraint equations for solving the leakage source location is established;

[0032] The least squares method is used to iteratively optimize the solution of the geometric constraint equations to generate an initial three-dimensional coordinate set of the leakage source that meets the preset error threshold.

[0033] In a preferred embodiment, the initial three-dimensional coordinate set is projected onto the surface of the actuator three-dimensional structural model, and a detection map is generated based on the coupled analysis of the modal spatial coherence of the pressure signal and the transient impact energy of the vibration, including:

[0034] Orthogonally project the initial three-dimensional coordinate set of the leakage source along the normal direction of the surface of the three-dimensional structure model of the actuator to generate the surface projection coordinates of the candidate leakage point;

[0035] Extract the pressure signal modal space coherence matrix corresponding to the projection area of ​​the candidate leak point, and calculate the covariance ratio of the main diagonal elements to the off-diagonal elements in the pressure signal modal space coherence matrix as the modal coherence coefficient;

[0036] The vibration transient impact energy value corresponding to the projection area of ​​the candidate leakage point is simultaneously extracted, and the energy correction factor is calculated based on the propagation attenuation model of the leakage shock wave;

[0037] The modal coherence coefficient and the energy correction factor are fused according to preset weights to generate a leakage intensity factor; a two-dimensional detection map is generated based on the distribution density of the surface projection coordinates of the candidate leakage points and the leakage intensity factor.

[0038] In a preferred embodiment, the color depth of the coordinate points in the two-dimensional detection map represents the leakage intensity level.

[0039] In a preferred embodiment, based on the coordinate distribution and energy proportion of the leakage area in the detection map, a detection identification map of multiple leakage points is generated through spatial density clustering and leakage intensity classification analysis, including:

[0040] Based on the coordinate distribution of the leakage area in the two-dimensional detection map, the spatial density clustering algorithm is used to divide the spatial clusters of candidate leakage points;

[0041] Extract the maximum and mean values ​​of the leakage intensity factor within each spatial cluster, and map the leakage intensity factor to a preset classification interval based on a dynamic interval partitioning strategy;

[0042] A unique identifier is assigned to each spatial cluster according to the leakage intensity classification interval, and the graphic size of the identifier is proportional to the total value of the leakage intensity factor in the cluster;

[0043] The identifiers are superimposed on the surface of the actuator's three-dimensional structural model to generate a detection identification map of multiple leakage points. The colors of different identifiers in the detection identification map distinguish the leakage intensity levels.

[0044] In a preferred embodiment, the neighborhood radius when the spatial density clustering algorithm is used to divide the spatial clusters of the candidate leakage points is adaptively adjusted according to the energy proportion of the leakage area.

[0045] In another aspect, the present invention provides an actuator sealing test system, comprising the following modules:

[0046] A pressurization acquisition module is used to pressurize the sealed cavity of the actuator to be tested to a preset pressure value and simultaneously acquire the dynamic pressure signal in the sealed cavity and the vibration waveform signal of the shell;

[0047] A signal analysis module is used to perform time-frequency analysis on the dynamic pressure signal to extract the first spectrum peak containing the pressure decay feature, and to perform spectrum analysis on the vibration waveform signal to extract the second spectrum peak related to the leakage impact;

[0048] a feature processing module, configured to calculate a path loss value based on the amplitude gradient of the first spectrum peak and the propagation time difference of the second spectrum peak, and screen sensor groups whose path loss values ​​are less than a loss threshold as candidate leakage nodes;

[0049] The model building module is used to construct a spatial grid model with candidate leakage nodes as vertices and solve the initial three-dimensional coordinate set of the leakage source through the arrival time difference of the vibration waveform signal;

[0050] A map generation module is used to project the initial three-dimensional coordinate set onto the surface of the actuator's three-dimensional structural model and generate a detection map based on the coupled analysis of the pressure signal modal spatial coherence and the vibration transient impact energy;

[0051] The identification generation module is used to generate a detection identification map of multiple leakage points based on the coordinate distribution and energy proportion of the leakage area in the detection map, through spatial density clustering and leakage intensity classification analysis.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. Through multi-dimensional signal fusion and spatial correlation analysis, accurate identification and visual positioning of multiple leakage points of the actuator are achieved. Based on the joint feature extraction of dynamic pressure signals and vibration waveforms, the signal interference superimposed by different leakage sources can be effectively separated. By screening candidate nodes through the path loss model and combining it with three-dimensional grid modeling technology, the positioning accuracy of the leakage source is significantly improved. Compared with the traditional single-dimensional detection method, the physical correlation between pressure attenuation characteristics and vibration impact propagation is used to construct a coupled distribution model of leakage energy in space, which greatly improves the spatial resolution of weak leakage signals in complex structures. The generation of the detection map combines the dynamic weight of modal coherence and transient energy, intuitively presenting the leakage intensity level and location distribution, and providing highly reliable data support for maintenance decisions.

[0054] 2. Adaptive threshold adjustment and a hierarchical clustering algorithm address the high misjudgment rate and ambiguous positioning issues in multi-leak point detection. Spatial density analysis and path loss calculation based on the sensor network automatically eliminate noise interference and identify true leak clusters, effectively avoiding repeated testing. The detection identification diagram integrates leak intensity, spatial distribution, and energy contribution information into a single visual interface through multi-dimensional encoding of color and size, significantly improving the interpretability of detection results. While maintaining ease of implementation in industrial sites, this technology achieves a technological leap from single leak determination to precise location of multiple leak sources. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a flow chart of a method for testing the sealing performance of an actuator according to the present invention;

[0056] Figure 2 This is a structural schematic diagram of an actuator sealing test system of the present invention. DETAILED DESCRIPTION

[0057] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0058] Example 1: Figure 1 The present invention provides a method for testing the sealing performance of an actuator, which comprises the following steps:

[0059] S1. Pressurize the sealed cavity of the actuator to be tested to a preset pressure value, and simultaneously collect the dynamic pressure signal in the sealed cavity and the vibration waveform signal of the housing;

[0060] S2. Performing time-frequency analysis on the dynamic pressure signal to extract a first spectrum peak containing pressure decay characteristics, and performing spectrum analysis on the vibration waveform signal to extract a second spectrum peak related to leakage impact;

[0061] S3, calculating a path loss value based on the amplitude gradient of the first spectrum peak and the propagation time difference of the second spectrum peak, and screening sensor groups with path loss values ​​less than a loss threshold as candidate leakage nodes;

[0062] S4, constructing a spatial grid model with the candidate leakage nodes as vertices, and solving the initial three-dimensional coordinate set of the leakage source by the arrival time difference of the vibration waveform signal;

[0063] S5. Projecting the initial three-dimensional coordinate set onto the surface of the actuator three-dimensional structural model, and generating a detection map based on the coupled analysis of the modal spatial coherence of the pressure signal and the transient impact energy of the vibration;

[0064] S6. Based on the coordinate distribution and energy proportion of the leakage area in the detection map, a detection identification map of multiple leakage points is generated through spatial density clustering and leakage intensity classification analysis.

[0065] S1. Pressurize the sealed cavity of the actuator to be tested to a preset pressure value, and simultaneously collect the dynamic pressure signal in the sealed cavity and the vibration waveform signal of the housing. The specific implementation can be as follows:

[0066] After the actuator under test is secured to the test bench, incremental pressure is applied in stages until the sealed chamber reaches a preset pressure. The test bench comprises a rigid support and an adjustable clamp. For example, the rigid support is bolted to the ground, and the adjustable clamp uses a knob to adjust the clamping force to accommodate actuators of different sizes. When applying incremental pressure in stages, a piping system with multiple pressurization stages is used. For example, the piping system includes a low-pressure air pump, a medium-pressure air pump, and a high-pressure air pump connected in series. The low-pressure air pump has an outlet pressure of, for example, 0.1 MPa, the medium-pressure air pump has an outlet pressure of 0.5 MPa, and the high-pressure air pump has an outlet pressure of 1.0 MPa. The preset pressure value is set based on the rated operating pressure of the actuator. For example, for an actuator with a rated operating pressure of 0.8 MPa, the preset pressure value can be set to 1.2 times the rated pressure (e.g., 0.96 MPa). During the pressure application process, the duration of each pressure stage can be set according to actual needs, for example, to 30 seconds to prevent sudden pressure changes from impacting the sealed chamber structure.

[0067] When the pressure in the sealed cavity reaches a preset pressure value, the dynamic pressure signal in the sealed cavity is continuously collected by the pressure sensor group at a preset sampling frequency. The pressure sensor group includes a plurality of pressure sensors, for example, at least three pressure sensors are evenly distributed on the inner wall of the sealed cavity, and the spacing between adjacent pressure sensors is set according to the size of the sealed cavity, for example, not more than one-quarter of the circumference of the sealed cavity to ensure the uniformity of signal coverage. The preset sampling frequency is set according to the frequency characteristics of the leakage signal. For example, when the pressure fluctuation caused by the leakage is mainly concentrated in the low-frequency band, the preset sampling frequency can be set to not less than 400Hz to meet the signal integrity requirements. The acquisition time of the dynamic pressure signal covers at least three complete pressure fluctuation cycles. For example, when the leakage characteristic frequency is 50Hz, the acquisition time can be set to 60 milliseconds to fully capture the periodic characteristics.

[0068] At the same time, the vibration waveform signal of the shell is collected by a synchronous sampling clock through a vibration sensor array attached to the surface of the shell. The vibration sensor array includes a plurality of vibration sensors, for example, at least four vibration sensors are staggered along the axial and circumferential directions of the shell, and the spacing between adjacent vibration sensors is set according to the size of the shell, for example, the axial spacing is one-fifth of the length of the shell, and the circumferential spacing is one-sixth of the circumference of the shell. The synchronous sampling clock is generated by an external clock source, for example, the clock source outputs a pulse signal to simultaneously trigger the sampling action of the pressure sensor group and the vibration sensor array, and the rising edge accuracy of the pulse signal can be set to 1 microsecond to ensure synchronization accuracy. The sampling frequency of the vibration waveform signal is consistent with the preset sampling frequency of the dynamic pressure signal, for example, both are set to 1kHz to ensure the signal alignment requirements of subsequent analysis.

[0069] The dynamic pressure signal and the vibration waveform signal are time-domain aligned to generate a multi-channel signal dataset with consistent timestamps. The time-domain alignment process includes the following steps: extracting the raw sampled data from the pressure sensor group and the vibration sensor array, marking the start timestamp of each signal channel according to the trigger time of the synchronous sampling clock; interpolating the sampling time series of the pressure and vibration signals. For example, when the sampling interval of the pressure signal is 1 millisecond and the sampling interval of the vibration signal is 0.5 milliseconds, the sampling interval of the pressure signal is adjusted to 0.5 milliseconds through interpolation; and aligning the adjusted pressure and vibration signals along the time axis according to the start timestamp to generate a multi-channel signal dataset with a unified time base. The data format of the multi-channel signal dataset can be set as a two-dimensional matrix, for example, the rows of the matrix correspond to the time series, and the columns of the matrix correspond to the sensor channels. The first column is the data of channel 1 of the pressure sensor group, the second column is the data of channel 1 of the vibration sensor array, and so on.

[0070] When applying increasing pressure in stages, if the leakage from the sealed chamber exceeds a temporary threshold, pressurization is suspended and the current pressure value is recorded as the critical leakage pressure. The temporary threshold is set based on the relationship between the volume of the sealed chamber and the leakage rate. For example, for a sealed chamber with a volume of 0.1 cubic meters, the temporary threshold can be set to a pressure drop of no more than 0.01 MPa per second. The critical leakage pressure is used to assess the sealing performance level of the actuator. For example, if the critical leakage pressure reaches 80% of the preset pressure value, the sealing performance is considered acceptable; if it falls below this ratio, it is considered unacceptable.

[0071] The placement of the pressure sensor group and vibration sensor array is optimized based on the actuator's structural characteristics. For example, for an axisymmetric actuator, the pressure sensors can be distributed equidistantly along the circumference of the sealed cavity, for example, with one pressure sensor every 120 degrees. The vibration sensors can be symmetrically arranged along the axial centerline of the housing, for example, with one vibration sensor at each end and in the center. For non-axisymmetric actuators, the pressure sensors can be placed in stress-concentrated areas of the sealed cavity, such as welds or flange connections. The vibration sensors can be placed in areas of the housing where vibration response is significant, such as the vibration peak locations determined through modal analysis.

[0072] The time-domain alignment of the dynamic pressure signal and the vibration waveform signal also includes the removal of abnormal data segments. Abnormal data segments include distorted waveforms caused by transient sensor overload or environmental interference. Abnormal data segments are identified by calculating the signal's average energy and standard deviation. When the signal energy within a time window exceeds a certain multiple of the average energy (for example, three times the standard deviation), the window is identified as an abnormal data segment. Abnormal data segments can be repaired by interpolating adjacent normal data segments. For example, an abnormal data segment lasting five sampling points is padded with the average of the 10 sampling points before and after.

[0073] After generating a multi-channel signal dataset, alignment accuracy is ensured by verifying the timestamp offset of each sensor channel. The permissible timestamp offset can be set to no more than one sampling interval. For example, at a sampling frequency of 1 kHz, the timestamp offset should be no more than 1 millisecond. If the timestamp offset of a sensor channel exceeds this limit, the synchronous sampling clock is retriggered and the data acquisition process is repeated until all channels meet the timestamp offset requirement.

[0074] S2. Performing time-frequency analysis on the dynamic pressure signal to extract a first spectrum peak containing pressure attenuation characteristics, and performing spectrum analysis on the vibration waveform signal to extract a second spectrum peak related to leakage impact, which can be specifically implemented as follows:

[0075] The dynamic pressure signal in the multi-channel signal data set is band-pass filtered to retain the signal components within the preset frequency band. The preset frequency band for band-pass filtering is set according to the typical frequency range of leakage pressure fluctuations. For example, the pressure attenuation characteristics caused by leakage in the sealing chamber are mainly concentrated in the frequency band of 20Hz to 200Hz, and the preset frequency band can be set to 20Hz to 200Hz. The band-pass filter adopts a Butterworth filter, such as a fourth-order Butterworth filter, with the passband ripple set to 1dB and the stopband attenuation set to 40dB to ensure that the target frequency band signal is retained while suppressing out-of-band noise. Only the signal components within the preset frequency band are retained in the filtered dynamic pressure signal, such as the pressure fluctuation data in the range of 20Hz to 200Hz.

[0076] Time-frequency analysis is performed on the filtered dynamic pressure signal. A time-frequency spectrum is extracted using a short-time Fourier transform (SFT). Frequency components in the SFT spectrum whose amplitude decay rate exceeds a first threshold are identified as the first spectral peak. The SFT window length is set based on the non-stationary nature of the signal. For example, when the time scale of leakage pressure fluctuations is 10 milliseconds, the window length is set to 20 milliseconds to balance time-frequency resolution. The SFT spectrum plot has time on the horizontal axis and frequency on the vertical axis, with amplitude represented by light and dark colors. The amplitude decay rate is calculated by calculating the decreasing slope of the amplitude within the adjacent time window at each frequency point in the SFT spectrum. For example, if the amplitude decreases from A1 to A2 from second t to second t+0.1, the decay rate is (A1-A2) / 0.1. The first threshold is set based on statistical results from historical leakage data. For example, an amplitude decay rate exceeding 0.5 MPa per second is considered a valid leak signature, while a value below this value is considered noise.

[0077] The vibration waveform signals in the multi-channel signal dataset are windowed and a power spectrum density plot is generated using a fast Fourier transform (FFT). Frequency components in the FFT plot with energy fluctuations exceeding a second threshold are identified as the second spectral peak. A Hanning window function is used for windowing. For example, the window length is 1024 sampling points, with an overlap ratio of 50% to reduce spectral leakage. The number of FFT points is set based on the required frequency resolution. For example, at a sampling frequency of 1 kHz, a 1024-point FFT can achieve a resolution of approximately 0.98 Hz. The FFT plot has frequency on the horizontal axis and energy amplitude on the vertical axis. The energy fluctuation amplitude is defined as the ratio of the energy value at a local frequency point to the average energy of its adjacent frequency points. For example, if the energy at a frequency point is E, the average energy of the five preceding and succeeding frequency points is E_avg, and the fluctuation amplitude is E / E_avg. The second threshold is set based on typical mechanical shock characteristics. For example, a fluctuation amplitude exceeding three times the energy value is considered a leakage shock component; anything below this threshold is ignored.

[0078] The first spectrum peak is waveform-matched against a preset leakage attenuation feature template, and candidate first spectrum peaks whose matching degree exceeds a third threshold are screened. The leakage attenuation feature template is constructed based on time-frequency spectrum graphs of known leakage conditions. For example, time-frequency spectrum graphs corresponding to different leakage apertures are collected in a standard leakage experiment, and their amplitude attenuation rate and frequency distribution characteristics are extracted as templates. Waveform matching is achieved by calculating a similarity metric, such as cosine similarity, which measures the proximity between the first spectrum peak and the template in the frequency-amplitude space. The third threshold is set based on the matching degree distribution. For example, a cosine similarity exceeding 0.8 is considered a valid match; otherwise, it is rejected.

[0079] The correlation between the second spectrum peak and the preset mechanical shock feature template is calculated, and the candidate second spectrum peaks with correlation coefficients exceeding the fourth threshold are screened. The mechanical shock feature template is constructed based on the difference in vibration signals between normal working conditions and leakage working conditions. For example, the power spectrum density diagram of the vibration signal is collected when there is no leakage as the baseline template, and the difference area between the power spectrum density diagram under the leakage condition and the baseline template is the mechanical shock feature. The Pearson correlation coefficient is used for correlation calculation. For example, the correlation between the energy distribution of the second spectrum peak and the template at the same frequency point is calculated. The fourth threshold is set according to the significance level of the correlation. For example, when the Pearson correlation coefficient exceeds 0.7, it is determined to be a valid leakage shock component, otherwise it is excluded.

[0080] During bandpass filtering, if the signal-to-noise ratio (SNR) of a signal component within a preset frequency band falls below a temporary SNR threshold, the filter's passband is automatically expanded. The temporary SNR threshold is set based on signal quality requirements; for example, a SNR below 10dB is considered low-quality. The passband is expanded in 10Hz increments. For example, if the initial preset frequency band is 20Hz to 200Hz, it is expanded to 10Hz to 210Hz until the SNR reaches the threshold or the maximum expansion limit.

[0081] The short-time Fourier transform window type can be adjusted based on signal characteristics. For example, for transient leakage signals, a Gaussian window is used to improve time-domain resolution; for steady-state leakage signals, a rectangular window is used to reduce computational complexity. The window type is selected based on pre-set rules. For example, if the signal's non-stationarity index exceeds 0.5, a Gaussian window is used; otherwise, a rectangular window is used. The non-stationarity index is calculated as the ratio of the signal's time-domain variance to its frequency-domain variance.

[0082] The calculation of the energy mutation amplitude in the power spectrum density graph also includes verification of the frequency band energy integral. For example, after locating the energy mutation frequency point, the total energy integral value within the frequency band (e.g., within ±5 Hz) is calculated. If the integral value exceeds a preset integral threshold, the second spectrum peak is retained. The preset integral threshold is set based on the sensitivity calibration results of the vibration sensor. For example, a valid energy mutation is determined when the integral value exceeds 10 mV² / Hz.

[0083] The results of waveform matching and correlation calculations are weighted and combined to generate the final candidate spectral peaks. For example, the matching weight of the first spectral peak is set to 0.6, and the correlation weight of the second spectral peak is set to 0.4. Spectral peaks with a total weighted score exceeding 0.75 are considered valid leakage features. The weighting is dynamically adjusted based on the signal-to-noise ratio of the pressure and vibration signals. For example, when the pressure signal-to-noise ratio is lower than that of the vibration signal, the vibration signal weight is increased to 0.6.

[0084] The time alignment of the candidate first and second spectral peaks is verified by checking whether their occurrence time windows overlap. For example, if the candidate first spectral peak occurs within the window from t seconds to t + 0.2 seconds, the candidate second spectral peak's occurrence time window must overlap by at least 50% to be considered a valid association. The time window overlap ratio is set based on the leak propagation delay. For example, if the physical transmission delay between the leak-induced pressure fluctuation and the mechanical shock is less than 0.1 seconds, the overlap ratio threshold is set to 50%.

[0085] S3. Calculate the path loss value based on the amplitude gradient of the first spectrum peak and the propagation time difference of the second spectrum peak, and select sensor groups with path loss values ​​less than a loss threshold as candidate leakage nodes. This can be specifically implemented as follows:

[0086] Based on the frequency distribution and amplitude gradient of the candidate first spectrum peak, the amplitude attenuation coefficient of the received signal at each sensor along the leakage propagation path is calculated. The frequency distribution of the candidate first spectrum peak is determined by statistically analyzing the dominant frequency position of the candidate first spectrum peak detected by each sensor. For example, sensor A detects a dominant frequency of 50 Hz for the candidate first spectrum peak, while sensor B detects a dominant frequency of 52 Hz. This difference in frequency distribution reflects the dispersion characteristics of the leakage signal propagation path. The amplitude gradient is calculated by calculating the ratio of the amplitude difference between the candidate first spectrum peaks of different sensors to the sensor spacing. For example, if the amplitude of the candidate first spectrum peak of sensor A is 1.2 V and that of sensor B is 0.9 V, and the distance between them is 0.5 meters, the amplitude difference is 0.3 V. The amplitude gradient is 0.3 V divided by 0.5 meters, which is 0.6 V / m. The amplitude attenuation coefficient is calculated by multiplying the amplitude gradient by the signal propagation distance. For example, if the amplitude gradient is 0.6 V / m and the propagation distance is 2 meters, the amplitude attenuation coefficient is 0.6 V / m multiplied by 2 meters, which is 1.2 V.

[0087] The propagation velocity of the leakage impulse signal between sensors is calculated based on the arrival time difference of the candidate second spectrum peak and the sensor spacing. The arrival time difference of the candidate second spectrum peak is determined by comparing the time stamps of the same candidate second spectrum peak detected by different sensors. For example, if sensor A detects the candidate second spectrum peak at time t1 and sensor B detects it at time t2, the arrival time difference is t2 minus t1, resulting in Δt. The sensor spacing is calculated based on the three-dimensional coordinate difference of the sensor locations. For example, if sensor A has coordinates (x1, y1, z1) and sensor B has coordinates (x2, y2, z2), the coordinate differences are x2-x1, y2-y1, and z2-z1, respectively. The spacing is the square root of the sum of the squares of the coordinate differences. The propagation velocity is the ratio of the sensor spacing to the arrival time difference. For example, if the spacing is 0.5 meters and Δt is 0.001 seconds, the propagation velocity is 0.5 meters divided by 0.001 seconds, which is 500 meters per second.

[0088] The path loss value is generated by multiplying the amplitude attenuation coefficient by the propagation velocity. The path loss value represents the comprehensive attenuation characteristics of the leakage signal along the propagation path. For example, if the amplitude attenuation coefficient is 1.2V and the propagation velocity is 500 m / s, the path loss value is 1.2V multiplied by 500 m / s, resulting in 600V·m / s. The dimensionless combination of the path loss value reflects the energy dissipation rate of the leakage signal. A larger value indicates a greater distance between the leak point and the sensor or greater dielectric attenuation.

[0089] The path loss value is compared with a preset loss threshold, and sensor groups with path loss values ​​less than the loss threshold are selected as candidate leak nodes. The loss threshold is set based on the typical propagation characteristics of leakage signals. For example, in a standard leakage experiment, the average path loss value at 1 meter from the leak point is 400V·m / s, so the loss threshold can be set to 400V·m / s. During screening, if a sensor group's path loss value is 300V·m / s and less than the threshold of 400V·m / s, it is identified as a candidate leak node. If the path loss value is 600V·m / s and greater than the threshold of 400V·m / s, it is excluded.

[0090] The spatial distribution of candidate leakage nodes is topologically verified. After eliminating isolated nodes, sensor groups that meet the adjacency density conditions are retained as valid candidate leakage nodes. Topological verification includes the following steps: constructing a spatial position matrix of candidate leakage nodes and calculating the three-dimensional Euclidean distance between each node; if the minimum distance between a node and other nodes exceeds the preset adjacency threshold, it is determined to be an isolated node. The adjacency threshold is set according to the sensor deployment density. For example, when the average sensor spacing is 0.5 meters, the adjacency threshold is set to 1.0 meters. The adjacency density condition requires that at least three nodes in the valid candidate leakage nodes have a mutual spacing less than the adjacency threshold. For example, the spacing between nodes A, B, and C is 0.8 meters, 0.7 meters, and 0.9 meters respectively, and all are less than 1.0 meters. They are retained as valid candidate leakage nodes; if the spacing between node D and other nodes is 1.2 meters and greater than 1.0 meters, they are eliminated.

[0091] When calculating the amplitude attenuation coefficient, if the frequency distribution of the candidate first spectrum peaks differs significantly, dispersion compensation is initiated. Dispersion compensation adjusts the amplitude gradient by incorporating a frequency-propagation velocity relationship. For example, if the dominant frequency of the candidate first spectrum peaks differs by more than 10 Hz, a weighted correction is applied to the amplitude gradient based on the experimentally calibrated dispersion curve. The corrected amplitude gradient is then used to recalculate the amplitude attenuation coefficient, improving the accuracy of the path loss value.

[0092] The propagation velocity calculation also includes calibration of the dielectric attenuation factor. This factor is obtained through calibration experiments at known leak points. For example, a standard impulse signal is injected under leak-free conditions and its propagation velocity between sensors is measured as a baseline value. During actual testing, the ratio of the measured velocity to the baseline velocity is used as the attenuation factor. The attenuation factor is used to correct the path loss value. For example, if the baseline velocity is 500 m / s and the measured velocity is 450 m / s, the attenuation factor is 450 m / s divided by 500 m / s, which is 0.9. The corrected path loss value is the original value multiplied by 0.9.

[0093] The path loss screening process includes dynamic threshold adjustment. When the ambient noise level is high, the loss threshold is adjusted upward in real time based on the signal-to-noise ratio. For example, if the base threshold is 400V·m / s and the current noise energy is 6dB higher than the standard operating condition, the threshold is proportionally adjusted to 600V·m / s. Dynamic adjustment is achieved through a table lookup method, and the mapping between noise energy and threshold adjustment coefficients is pre-calibrated through experiments.

[0094] Verification of the spatial distribution of valid candidate leak nodes also includes a directional consistency check. This consistency is achieved by calculating the angular differences between the lines connecting the nodes. For example, if the angular differences between the lines connecting the valid candidate leak nodes A, B, and C are 30°, 35°, and 40°, respectively, the angular differences are considered consistent within a preset threshold. If the angular difference of a node exceeds the threshold, it is identified as an abnormal node and removed. The angular difference threshold is set based on the physical size of the leak source. For example, if the leak aperture is less than 1 mm, the threshold is set to 10°; if the aperture is greater than 5 mm, the threshold is set to 30°.

[0095] S4. Construct a spatial grid model with candidate leakage nodes as vertices, and solve the initial three-dimensional coordinate set of the leakage source by the arrival time difference of the vibration waveform signal. The specific implementation can be as follows:

[0096] A spatial grid model is constructed with the three-dimensional spatial coordinates of valid candidate leakage nodes as vertices, and the edges between vertices are the actual distances between adjacent nodes. The three-dimensional spatial coordinates of valid candidate leakage nodes are determined by the sensor deployment locations. For example, the coordinates of valid candidate leakage node A are determined by the installation location of sensor A, and the coordinates of node B are determined by the installation location of sensor B. The edges between vertices are generated by calculating the square root of the sum of the squared differences between the three-dimensional coordinate components of adjacent nodes. For example, the length of the edge between node A and node B is the square root of the sum of the squared differences between the x-coordinates of node A and the x-coordinates of node B, the squared differences between the y-coordinates of node A and the squared differences between the z-coordinates of node B. The topological relationships of the spatial grid model are represented by an adjacency matrix. For example, if nodes A and B are adjacent, the corresponding positions in the matrix are marked as 1, and if they are not adjacent, they are marked as 0.

[0097] The propagation time difference vector of the leakage shock wave is calculated based on the arrival time difference of the vibration waveform signal at each valid candidate leakage node. The arrival time difference is obtained by comparing the detection timestamps of the same leakage shock wave at different nodes. For example, the timestamp of the vibration waveform signal detected by node A is t1, and the timestamp of the detection by node B is t2. The time difference between node A and node B is the result of subtracting t1 from t2. The propagation time difference vector is composed of the time differences of all adjacent nodes arranged in sequence. For example, if nodes A, B, and C form a triangular mesh, the time difference vector includes the time difference between nodes A and B, the time difference between nodes B and C, and the time difference between nodes A and C. The directionality of the time difference vector is determined by the spatial position relationship of the sensors. For example, if node A is located upstream of the leakage propagation path and node B is located downstream, the time difference between node A and node B is positive.

[0098] Based on the topological relationships of vertices in the spatial grid model and the propagation time difference vector, a set of geometric constraint equations is established to solve the leakage source location. Each equation in the geometric constraint equations represents the relationship between the propagation time difference and spatial distance for a vertex pair. For example, for nodes A and B, the equation is: the propagation velocity multiplied by the time difference equals the three-dimensional spatial distance between nodes A and B. The propagation velocity is set based on the medium properties. For example, the propagation velocity in a metal shell is measured to be 5000 m / s through sound velocity experiments, while the standard sound velocity in the gas medium within the sealed chamber is 340 m / s. The coordinates of the leakage source are introduced as unknowns in the equations. For example, the distance from node A to the leakage source is the square root of the sum of the squared differences between the leakage source's x-coordinate and node A's x-coordinate, y-coordinate, and z-coordinate. The distance from node B to the leakage source is calculated similarly: the distance difference between the two distances is equal to the propagation velocity multiplied by the time difference between nodes A and B.

[0099] The least squares method is used to iteratively optimize the solution to the geometric constraint equations to generate an initial set of three-dimensional coordinates of the leakage source that meets a preset error threshold. The least squares method minimizes the sum of squared residuals of all equations by gradually adjusting the leakage source coordinates. For example, the initial guess coordinates are the actuator's geometric center point, and the coordinates are adjusted with each iteration to reduce the residuals. The preset error threshold is set based on the required positioning accuracy. For example, the maximum allowable residual is 0.1 meter, and iterations are terminated when the sum of squared residuals is less than 0.1. During the iteration process, if the sum of squared residuals of an iteration is less than the threshold, the current coordinates are output. If convergence does not occur after the maximum number of iterations, the solution is discarded. The initial set of three-dimensional coordinates contains all coordinates that pass the threshold verification. For example, after 10 iterations, three sets of coordinates are obtained, two of which have residuals of 0.08 and 0.09 meters, and are retained as valid solutions.

[0100] When constructing the spatial grid model, if the number of valid candidate leak nodes is insufficient, virtual nodes are generated by interpolating from neighboring nodes to supplement the model. The coordinates of the virtual node are calculated by averaging the coordinates of the adjacent nodes. For example, if the coordinates of node A are (1.0, 2.0, 3.0) and the coordinates of node B are (2.0, 3.0, 4.0), the coordinates of the inserted virtual node are (1.5, 2.5, 3.5). The time difference of the virtual node is estimated by linear interpolation of the time differences of adjacent nodes. For example, if the time difference between nodes A and B is 0.001 seconds, the time difference between the virtual node and node A is set to 0.0005 seconds. Virtual nodes are used only for model construction and do not participate in the final output of the leak source coordinates.

[0101] The establishment of the geometric constraint equations also includes dielectric attenuation correction. Dielectric attenuation correction adjusts the propagation velocity by introducing an attenuation factor. For example, at the interface between the sealed cavity and the shell, the propagation velocity is set as the weighted average of the media velocities on both sides based on the acoustic impedance difference at the interface. The attenuation factor is calibrated experimentally. For example, if the measured propagation velocity at the interface is 2500 m / s and the original set velocity is 5000 m / s, the attenuation factor is 2500 / 5000 = 0.5. The corrected propagation velocity is the original velocity multiplied by the attenuation factor.

[0102] During the least-squares iterative optimization process, if the initial guessed coordinates deviate too far from the true leak source, a multi-starting point initialization strategy is employed. This strategy involves uniformly selecting multiple initial coordinate points within the spatial mesh model, such as the model center, vertex positions, and edge midpoints. Iterative optimization is then performed on each of these points, followed by selecting the solution with the smallest residual error. The number of initial points is determined based on the model complexity; for example, a triangular mesh requires at least three initial points, and a tetrahedral mesh requires at least four.

[0103] After the initial set of 3D coordinates of the leakage source is generated, physically infeasible solutions are eliminated through coordinate validity verification. This validation includes boundary checks and propagation consistency checks. The boundary check requires that the coordinates lie within the geometric range of the actuator's 3D structural model, for example, the x-component of the coordinates lies between 0 and the actuator length L. The propagation consistency check requires that the deviation between the theoretical and measured time differences from the leakage source to each node is within the allowable error, for example, no more than 10%. Coordinates that do not meet either condition are rejected.

[0104] The preset error threshold is dynamically adjusted based on the ambient noise level. This dynamic adjustment is implemented using a mapping table that compares noise energy to threshold coefficients. For example, when noise energy is 6dB above standard operating conditions, the error threshold is relaxed from 0.1m to 0.15m; when noise energy is below standard operating conditions, the threshold is tightened to 0.05m. This mapping table is pre-calibrated through experiments, such as testing positioning error under different noise levels to establish a corresponding relationship between noise energy and thresholds.

[0105] S5. Projecting the initial three-dimensional coordinate set onto the surface of the actuator three-dimensional structural model, generating a detection map based on the coupled analysis of the pressure signal modal spatial coherence and the vibration transient impact energy, which can be specifically implemented as follows:

[0106] The initial 3D coordinates of the leak source are orthogonally projected along the surface normal of the actuator's 3D structural model to generate the surface projection coordinates of the candidate leak points. The surface normal direction of the actuator's 3D structural model is determined based on the model's geometric characteristics. For example, the normal direction of a flat surface is perpendicular to the plane, and the normal direction of a curved surface is directed outward along the center of curvature. During the orthogonal projection operation, a perpendicular line is drawn from the initial 3D coordinates to the model surface. The coordinates at the foot of the perpendicular are the surface projection coordinates. For example, a point with initial coordinates (2.0, 3.0, 4.0) is projected to surface coordinates (2.0, 3.0, 3.8). When multiple initial coordinates are projected onto the same surface area, the point with the smallest projection distance is retained as the candidate leak point. For example, if the projection distance between two initial coordinates is less than 0.01 meters, they are merged into a single candidate point.

[0107] The modal spatial coherence matrix of the pressure signal corresponding to the projected area of ​​the candidate leak point is extracted. The covariance ratio between the main diagonal elements and the off-diagonal elements in the pressure signal modal spatial coherence matrix is ​​calculated as the modal coherence coefficient. The pressure signal modal spatial coherence matrix is ​​constructed through cross-correlation analysis of the pressure signals from multiple sensors. For example, the modal coherence matrix elements of sensors A, B, and C represent the modal correlations between sensors A and A, A and B, and A and C, respectively. The main diagonal elements represent the modal correlations within the sensor itself; for example, the modal correlation of sensor A is set to 1.0; the off-diagonal elements represent the cross-sensor modal correlations; for example, the modal correlation between sensors A and B is 0.8. The covariance ratio is calculated as: the sum of the covariances of the main diagonal elements divided by the sum of the covariances of the off-diagonal elements. For example, if the sum of the main diagonal covariances is 1.5 and the sum of the off-diagonal covariances is 0.6, the ratio is 1.5 / 0.6 = 2.5. The larger the modal coherence coefficient, the stronger the modal correlation of the leakage signal between sensors.

[0108] The vibration transient impact energy value corresponding to the candidate leak point projection area is extracted synchronously, and the energy correction factor is calculated based on the propagation attenuation model of the leakage shock wave. The vibration transient impact energy value is obtained by integrating the square of the amplitude of the vibration signal within the candidate leak point projection time window. For example, the energy value is obtained by accumulating the square of the vibration signal amplitude within a 0.1 second time window. The propagation attenuation model describes the characteristic that vibration energy decays exponentially with increasing propagation distance. For example, the energy correction factor is inversely proportional to the propagation distance. For every 1 meter increase in propagation distance, the energy value decays to 50% of the original value. The attenuation coefficient is calibrated through a standard leakage experiment. For example, if the measured energy value at a distance of 1 meter is 50% of the initial value, the attenuation coefficient is set to 0.693. The corrected vibration energy value is the exponential function result of the original energy value multiplied by the product of the attenuation coefficient and the propagation distance.

[0109] The modal coherence coefficient and energy correction factor are fused according to preset weights to generate a leakage intensity factor. The preset weights are dynamically assigned based on the signal-to-noise ratio of the pressure signal to the vibration signal. For example, when the pressure signal-to-noise ratio is higher than the vibration signal, the modal coherence coefficient weight is set to 0.7 and the energy correction factor weight is set to 0.3; otherwise, the weights are set to 0.4 and 0.6. The weight allocation table is pre-calibrated through experiments. For example, when the signal-to-noise ratio is 20dB, the weights are set to 0.6:0.4. The leakage intensity factor is calculated by weighted summation. For example, if the modal coherence coefficient is 2.5, the energy correction factor is 0.8, and the weights are 0.6:0.4, then the leakage intensity factor is 2.5 times 0.6 plus 0.8 times 0.4, which is 1.82.

[0110] A two-dimensional detection map is generated based on the distribution density of the projected surface coordinates of the candidate leak points and the leakage intensity factor. The distribution density is calculated using a kernel density estimation algorithm. For example, a Gaussian distribution kernel function is constructed with each candidate leak point as the center, and the kernel function bandwidth is set to 0.05 meters based on the minimum characteristic size of the actuator surface. The leakage intensity factor is used as a density weight in the calculation. For example, if there are three candidate leak points in a certain area with leakage intensity factors of 1.5, 1.8, and 2.0, the weighted density of the area is the sum of the three factors, 5.3. The horizontal and vertical axes of the two-dimensional detection map correspond to the actuator surface coordinates, and the color depth is mapped to different color levels based on the weighted density value. For example, a density of 0-1 corresponds to light yellow, 1-3 corresponds to orange, 3-5 corresponds to red, and 5 or above corresponds to dark red.

[0111] During orthogonal projection, if holes or discontinuous areas exist on the model surface, compensating coordinates are generated by interpolating the coordinates of adjacent projected points. This interpolation method is based on local surface fitting. For example, the coordinates of four projected points on the edge of a hole are selected and fitted using a quadratic surface equation to generate the internal coordinates of the hole. Interpolated coordinates are used only for density calculations and do not contribute to the generation of the leakage intensity factor.

[0112] The calculation of the pressure signal modal space coherence matrix includes noise floor subtraction. The noise floor is determined by the pressure signal modal coherence matrix under leak-free conditions. For example, the main diagonal covariance is 0.1, and the off-diagonal covariance is 0.05. In actual calculations, the noise floor value is subtracted from the measured covariance. For example, if the measured main diagonal covariance is 0.5, the effective covariance after subtracting 0.1 is 0.4.

[0113] The calculation of the energy correction factor takes into account the effects of medium anisotropy. The anisotropic attenuation model adjusts the attenuation coefficient based on the propagation direction. For example, the attenuation coefficient along the shell axis is 0.6, and along the radial axis it is 0.8. The correction factor is the sum of the axial attenuation coefficient and the propagation distance, plus the radial attenuation coefficient and the propagation distance, and then exponentially multiplies the result. The anisotropy coefficient is obtained through multi-directional calibration experiments. For example, the distance at which the energy attenuation reaches 50% is 1 meter along the axial direction and 0.8 meters along the radial direction.

[0114] The fusion weight of the leakage intensity factor is dynamically adjusted based on signal stability. Signal stability is assessed using the coefficient of variation of the pressure and vibration signals. For example, a pressure signal coefficient of variation of 0.1 indicates high stability, and the weight is increased to 0.8. A vibration signal coefficient of variation of 0.3 indicates low stability, and the weight is decreased to 0.2. The coefficient of variation threshold is set based on historical data statistics; for example, a value less than 0.2 indicates high stability.

[0115] After the two-dimensional detection map is generated, isolated noise points are removed through morphological filtering. Morphological filtering uses an opening operation, first erosion and then dilation. For example, erosion removes isolated color blocks smaller than 1 square centimeter, and dilation restores the original outline of the valid area. After filtering, only continuously distributed color blocks are retained in the detection map as valid leakage areas.

[0116] S6. Generate a detection identification map of multiple leakage points based on the coordinate distribution and energy proportion of the leakage area in the detection map through spatial density clustering and leakage intensity classification analysis. The specific implementation can be as follows:

[0117] Based on the coordinate distribution of leak areas in the two-dimensional detection map, a spatial density clustering algorithm is used to divide candidate leak points into spatial clusters. The neighborhood radius is adaptively adjusted based on the energy contribution of the leak area. The algorithm determines the neighborhood radius by calculating the weighted value of the spatial distance and energy contribution between candidate leak points. For example, when the energy contribution of a certain area exceeds 20% of the total energy, the neighborhood radius is set to 0.1 meter; when the energy contribution is between 10% and 20%, the radius is set to 0.05 meter; and when it is below 10%, the radius is set to 0.02 meter. Dynamic adjustment of the neighborhood radius is achieved through a lookup table, such as a table mapping preset energy contribution thresholds to radii, and the radius value is retrieved based on the real-time energy contribution. During the clustering process, if the number of candidate leak points in a neighborhood exceeds a minimum threshold (e.g., 5), it is considered an independent spatial cluster; otherwise, it is marked as a noise point and removed.

[0118] The maximum and mean values ​​of the leakage intensity factor within each spatial cluster are extracted, and the leakage intensity factor is mapped to a preset grading interval based on the dynamic interval partitioning strategy. The maximum and mean values ​​of the leakage intensity factor are obtained by traversing the leakage intensity factors of all points in the spatial cluster. For example, a spatial cluster contains 10 candidate leakage points, and the leakage intensity factors are 1.5, 1.8, 2.0, etc., with a maximum value of 2.0 and a mean of 1.7. The dynamic interval partitioning strategy determines the grading interval range based on the ratio of the maximum value to the mean. For example, when the maximum value / mean value is ≥1.5, the interval is divided into three levels: high, medium, and low; when the ratio is 1.2-1.5, it is divided into two levels: high and low; when it is lower than 1.2, it is considered a single level. The boundary values ​​of the preset grading intervals are set based on historical leakage data statistics. For example, the high-level interval is above the mean + standard deviation, the medium-level interval is from the mean to the mean + standard deviation, and the low-level interval is below the mean.

[0119] Each spatial cluster is assigned a unique identifier based on the leakage intensity grading interval. The size of the identifier is proportional to the total leakage intensity factor within the cluster. The size of the unique identifier is determined by multiplying the base size by the total leakage intensity factor. For example, if the base size is 5 mm and the total leakage intensity factor of a spatial cluster is 15.0, the identifier size is 5 × 15 = 75 mm. The color of the identifier is set according to the grading interval, for example, red for high, orange for medium, and yellow for low. The identifier is circular, with the center coordinates of the weighted centroid of the spatial cluster. For example, the centroid coordinates are calculated by the weighted average of the leakage intensity factors of each point in the cluster.

[0120] Identifiers are superimposed onto the surface of the actuator's 3D structural model to generate a detection map of multiple leak points. This superposition is achieved through coordinate mapping, for example, converting the centroid coordinates in the 2D detection map to corresponding locations on the actuator's 3D structural model. The transparency of the identifiers is dynamically adjusted based on the reliability of the leak intensity factor. For example, when the standard deviation of the leak intensity factor for points within a spatial cluster exceeds 0.5, the transparency is set to 50%; when it is below 0.5, the transparency is set to 100%. The detection map is output in a vector graphics format with zooming and layered display support. For example, users can click on an identifier to view detailed leak intensity data.

[0121] During spatial density clustering, if noise interference areas are detected in the atlas, a noise filtering mechanism is activated. This mechanism works by calculating the energy contribution of a spatial cluster and its correlation with neighboring clusters. For example, if a spatial cluster's energy contribution is less than 1% and its minimum distance from a neighboring cluster exceeds 0.2 meters, it is considered noise and removed. The neighborhood radius of the filtered spatial clusters is recalculated. For example, after noise removal, the neighborhood radius of the remaining clusters is increased by 10% based on their energy contribution.

[0122] The dynamic interval partitioning strategy also includes interval boundary smoothing. This is achieved by introducing a transition interval, such as a ±5% transition band between the high-level and medium-level intervals. Within this transition band, the leakage intensity factor is assigned a color using linear interpolation. This smoothing process prevents sudden changes in intervals that could cause identifier color jumps. For example, a leakage intensity factor of 1.49 (the lower limit of the transition band) would be close to the high-level red, while a leakage intensity factor of 1.51 (the upper limit of the transition band) would be completely red.

[0123] The identifier's graphic size is adaptively scaled based on the display device's resolution. For example, on a 1920×1080 screen, the base size is set to 5 pixels; at a 4K resolution, the base size is adjusted to 10 pixels. The scaling factor is calculated based on the device's pixel density. For example, a scaling factor of 1.0 is used for a pixel density of 300 PPI, and 2.0 for a pixel density of 600 PPI.

[0124] The generation of the detection identification map also includes an overlay of a user interaction layer. This layer displays historical detection data for a leak source through a transparent layer. For example, hovering over an identifier displays a curve showing the leak intensity factor for the last three detections for that cluster. The timestamps of historical data are associated with the detection results and stored, allowing for timeline playback of leak evolution.

[0125] The unique identifier of a spatial cluster maintains positional consistency across multiple inspections. This consistency is achieved by binding the identifier coordinates to key feature points on the actuator's 3D structural model. For example, the center coordinates of the identifier are aligned with bolt holes or welds on the model surface, with a deviation of no more than 0.01 meters. This binding relationship is stored in a hash table, where, for example, the hash value of the key feature point coordinates serves as the identifier's position index.

[0126] Inspection identification images can be exported in formats compatible with industry-standard visualization tools. For example, they can be generated as DWG files for editing in CAD software or exported as PNG images for embedding in reports. Format conversion is accomplished through an intermediate data interface, such as when converting vector graphics to bitmaps, preserving layered metadata for traceability during subsequent analysis.

[0127] Example 2: Figure 2 A schematic structural diagram of an actuator sealing test system according to the present invention is provided. The actuator sealing test system includes the following modules:

[0128] A pressurization acquisition module is used to pressurize the sealed cavity of the actuator to be tested to a preset pressure value and simultaneously acquire the dynamic pressure signal in the sealed cavity and the vibration waveform signal of the shell;

[0129] A signal analysis module is used to perform time-frequency analysis on the dynamic pressure signal to extract the first spectrum peak containing the pressure decay feature, and to perform spectrum analysis on the vibration waveform signal to extract the second spectrum peak related to the leakage impact;

[0130] a feature processing module, configured to calculate a path loss value based on the amplitude gradient of the first spectrum peak and the propagation time difference of the second spectrum peak, and screen sensor groups whose path loss values ​​are less than a loss threshold as candidate leakage nodes;

[0131] The model building module is used to construct a spatial grid model with candidate leakage nodes as vertices and solve the initial three-dimensional coordinate set of the leakage source through the arrival time difference of the vibration waveform signal;

[0132] A map generation module is used to project the initial three-dimensional coordinate set onto the surface of the actuator's three-dimensional structural model and generate a detection map based on the coupled analysis of the pressure signal modal spatial coherence and the vibration transient impact energy;

[0133] The identification generation module is used to generate a detection identification map of multiple leakage points based on the coordinate distribution and energy proportion of the leakage area in the detection map, through spatial density clustering and leakage intensity classification analysis.

[0134] The calculations involved in the embodiments are all dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to actual conditions.

[0135] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0136] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application of the technical solution and the invention constraints. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0137] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0138] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0139] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0140] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for testing the sealing performance of an actuator, characterized in that: The steps include: S1. Pressurize the sealed cavity of the actuator to be tested to a preset pressure value, and simultaneously collect the dynamic pressure signal in the sealed cavity and the vibration waveform signal of the housing; S2. Performing time-frequency analysis on the dynamic pressure signal to extract a first spectrum peak containing pressure attenuation characteristics, and performing spectrum analysis on the vibration waveform signal to extract a second spectrum peak related to leakage impact, including: Perform bandpass filtering on the dynamic pressure signal in the multi-channel signal data set to retain the signal components within the preset frequency band; performing time-frequency analysis on the filtered dynamic pressure signal, extracting a time-frequency spectrum through short-time Fourier transform, and identifying a frequency component in the time-frequency spectrum whose amplitude decay rate exceeds a first threshold as a first spectrum peak; Performing windowing processing on the vibration waveform signal in the multi-channel signal data set, generating a power spectrum density map through fast Fourier transform, and locating the frequency component in the power spectrum density map where the energy mutation amplitude exceeds a second threshold as the second spectrum peak; Performing waveform matching on the first spectrum peak and the preset leakage attenuation feature template, and screening candidate first spectrum peaks whose matching degree exceeds a third threshold; performing correlation calculation on the second spectrum peak and the preset mechanical shock feature template, and screening candidate second spectrum peaks whose correlation coefficient exceeds a fourth threshold; S3, calculating a path loss value based on the amplitude gradient of the first spectrum peak and the propagation time difference of the second spectrum peak, and screening sensor groups with path loss values ​​less than a loss threshold as candidate leakage nodes, including: Calculate the amplitude attenuation coefficient of the received signal of each sensor on the leakage propagation path based on the frequency distribution and amplitude gradient of the candidate first spectrum peak; Calculate the propagation speed of the leakage impulse signal between the sensors based on the arrival time difference of the candidate second spectrum peak and the sensor spacing; Generate a path loss value based on the product of the amplitude attenuation coefficient and the propagation speed; Compare the path loss value with the preset loss threshold, and select the sensor group whose path loss value is less than the loss threshold as the candidate leakage node; The spatial distribution of candidate leakage nodes is topologically verified, and after removing isolated nodes, sensor groups that meet the adjacency density conditions are retained as valid candidate leakage nodes. S4. Construct a spatial grid model with candidate leakage nodes as vertices, and solve the initial three-dimensional coordinate set of the leakage source by the arrival time difference of the vibration waveform signal, including: Construct a spatial grid model with the three-dimensional spatial coordinates of valid candidate leakage nodes as vertices, and the edges between vertices are the actual distances between adjacent nodes; Calculate the propagation time difference vector of the leakage shock wave according to the arrival time difference of the vibration waveform signal between each valid candidate leakage node; Based on the topological relationship of vertices in the spatial grid model and the propagation time difference vector, a set of geometric constraint equations for solving the leakage source location is established; The least squares method is used to iteratively optimize the solution of the geometric constraint equations to generate an initial three-dimensional coordinate set of the leakage source that meets a preset error threshold; S5. Projecting the initial three-dimensional coordinate set onto the surface of the actuator three-dimensional structural model, and generating a detection map based on the coupled analysis of the modal spatial coherence of the pressure signal and the transient impact energy of the vibration; S6. Based on the coordinate distribution and energy proportion of the leakage area in the detection map, a detection identification map of multiple leakage points is generated through spatial density clustering and leakage intensity classification analysis.

2. The actuator sealing test method according to claim 1, characterized in that: Pressurize the sealed cavity of the actuator to be tested to a preset pressure value, and simultaneously collect the dynamic pressure signal in the sealed cavity and the vibration waveform signal of the shell, including: After fixing the actuator to be tested to the test bench, apply increasing pressure in stages until the sealed chamber reaches the preset pressure value; When the pressure in the sealed cavity reaches a preset pressure value, the dynamic pressure signal in the sealed cavity is continuously collected at a preset sampling frequency by the pressure sensor group, and the vibration waveform signal of the shell is collected at a synchronous sampling clock by the vibration sensor array attached to the surface of the shell; The dynamic pressure signal and the vibration waveform signal are aligned in the time domain to generate a multi-channel signal dataset with consistent timestamps.

3. The actuator sealing test method according to claim 1, characterized in that: The initial 3D coordinate set is projected onto the surface of the actuator's 3D structural model. Based on the coupled analysis of the pressure signal modal spatial coherence and the vibration transient impact energy, a detection map is generated, including: Orthogonally project the initial three-dimensional coordinate set of the leakage source along the normal direction of the surface of the three-dimensional structure model of the actuator to generate the surface projection coordinates of the candidate leakage point; Extract the pressure signal modal space coherence matrix corresponding to the projection area of ​​the candidate leak point, and calculate the covariance ratio of the main diagonal elements to the off-diagonal elements in the pressure signal modal space coherence matrix as the modal coherence coefficient; The vibration transient impact energy value corresponding to the projection area of ​​the candidate leakage point is simultaneously extracted, and the energy correction factor is calculated based on the propagation attenuation model of the leakage shock wave; The modal coherence coefficient and the energy correction factor are fused according to preset weights to generate a leakage intensity factor; a two-dimensional detection map is generated based on the distribution density of the surface projection coordinates of the candidate leakage points and the leakage intensity factor.

4. The actuator sealing test method according to claim 3, characterized in that: The color depth of the coordinate points in the two-dimensional detection map represents the leakage intensity level.

5. The actuator sealing test method according to claim 1, characterized in that: Based on the coordinate distribution and energy proportion of the leakage area in the detection map, through spatial density clustering and leakage intensity classification analysis, a detection identification map of multiple leakage points is generated, including: Based on the coordinate distribution of the leakage area in the two-dimensional detection map, the spatial density clustering algorithm is used to divide the spatial clusters of candidate leakage points; Extract the maximum and mean values ​​of the leakage intensity factor within each spatial cluster, and map the leakage intensity factor to a preset classification interval based on a dynamic interval partitioning strategy; A unique identifier is assigned to each spatial cluster according to the leakage intensity classification interval, and the graphic size of the identifier is proportional to the total value of the leakage intensity factor in the cluster; The identifiers are superimposed on the surface of the actuator's three-dimensional structural model to generate a detection identification map of multiple leakage points. The colors of different identifiers in the detection identification map distinguish the leakage intensity levels.

6. The actuator sealing test method according to claim 5, characterized in that: The neighborhood radius of the spatial density clustering algorithm when dividing the spatial clusters of candidate leakage points is adaptively adjusted according to the energy proportion of the leakage area.

7. An actuator sealing test system, used to implement the actuator sealing test method according to any one of claims 1 to 6, characterized in that: Includes the following modules: A pressurization acquisition module is used to pressurize the sealed cavity of the actuator to be tested to a preset pressure value and simultaneously acquire the dynamic pressure signal in the sealed cavity and the vibration waveform signal of the shell; A signal analysis module is used to perform time-frequency analysis on the dynamic pressure signal to extract the first spectrum peak containing the pressure decay feature, and to perform spectrum analysis on the vibration waveform signal to extract the second spectrum peak related to the leakage impact; a feature processing module, configured to calculate a path loss value based on the amplitude gradient of the first spectrum peak and the propagation time difference of the second spectrum peak, and screen sensor groups whose path loss values ​​are less than a loss threshold as candidate leakage nodes; The model building module is used to construct a spatial grid model with candidate leakage nodes as vertices and solve the initial three-dimensional coordinate set of the leakage source through the arrival time difference of the vibration waveform signal; A map generation module is used to project the initial three-dimensional coordinate set onto the surface of the actuator's three-dimensional structural model and generate a detection map based on the coupled analysis of the pressure signal modal spatial coherence and the vibration transient impact energy; The identification generation module is used to generate a detection identification map of multiple leakage points based on the coordinate distribution and energy proportion of the leakage area in the detection map, through spatial density clustering and leakage intensity classification analysis.

Images