An infrasound wave data-based device anomaly detection method and system

Abstract

The embodiment of the application provides a kind of equipment abnormality detection method and system based on infrasonic wave data, belong to equipment state detection technical technical field.The method comprises: collecting infrasonic wave signal and corresponding medium characteristic parameter in the running process of target equipment, establish the medium propagation characteristic table for representing signal propagation path;Based on the medium propagation characteristic table, construct propagation correction model, execute distortion correction to infrasonic wave signal, output correction signal;Based on correction signal, extract time-varying characteristic parameter and construct time-varying characteristic matrix, calculate characteristic drift rate and input drift determination model;Based on drift determination model, execute environment normalization and covariance drift analysis, when drift rate exceeds preset threshold, output equipment abnormality determination result.The present application scheme realizes the distortion correction of infrasonic wave signal in complex propagation medium and multi-feature covariance drift identification, so as to accurately locate equipment abnormal position under environmental interference and determine abnormal type.

Description

Technical Field

[0001] This invention relates to the field of equipment condition monitoring technology, and specifically to a method and system for detecting equipment anomalies based on infrasound data. Background Technology

[0002] In the field of equipment operation status monitoring, acoustic signals are widely used for fault identification and health assessment due to their non-contact nature and high sensitivity. However, existing acoustic detection schemes mainly focus on audible or ultrasonic frequencies, with limited utilization of low-frequency infrasound signals. Infrasound signals, characterized by their long wavelength and strong penetrating power during propagation, can carry deep information about equipment structure and medium coupling, but face significant technical obstacles in practical applications.

[0003] Existing detection methods generally assume that the signal propagation path medium is uniform and the interface characteristics are stable. However, in complex equipment, signals often need to pass through metal casings, supporting components, or multi-medium interfaces, leading to phenomena such as phase reversal, energy attenuation, and multipath superposition during propagation. These propagation distortion effects significantly alter the phase and amplitude characteristics of the signal, causing similar faults to behave inconsistently under different equipment or media conditions, severely impacting the reliability of anomaly identification. Existing solutions typically correct for these issues through simple filtering or amplitude normalization, which struggles to accurately recover the true propagation characteristics.

[0004] Traditional acoustic testing relies heavily on instantaneous signal characteristics for anomaly detection, lacking the ability to effectively identify slow aging or progressive damage to equipment. In long-term operation scenarios, signal changes often occur in the form of minute drifts, influenced by factors such as ambient temperature, humidity, and load variations. Existing algorithms struggle to distinguish between environmental disturbances and genuine structural changes, leading to misjudgments or missed detections.

[0005] In summary, existing acoustic signal-based device detection technologies have significant shortcomings in low-frequency signal propagation distortion compensation and slow-varying anomaly trend identification, and there is still a lack of detection methods that can take into account both multi-media propagation characteristics and time series drift features. Summary of the Invention

[0006] The purpose of this invention is to provide a device anomaly detection method and system based on infrasound data, so as to at least solve the problem that infrasound signals are prone to distortion during multi-media propagation and it is difficult to accurately identify slowly changing anomalies in the prior art.

[0007] To achieve the above objectives, the first aspect of the present invention provides a method for detecting equipment anomalies based on infrasound data. The method includes: acquiring infrasound signals and corresponding medium characteristic parameters during the operation of a target device, and establishing a medium propagation characteristic table to characterize the signal propagation path; constructing a propagation correction model based on the medium propagation characteristic table, performing distortion correction on the infrasound signals to correct the phase and amplitude of the signals, and outputting a corrected signal; extracting time-varying characteristic parameters based on the corrected signal and constructing a time-varying characteristic matrix, calculating the characteristic drift rate and inputting it into a drift judgment model; performing environmental normalization and covariance drift analysis based on the drift judgment model, and outputting an equipment anomaly judgment result when the drift rate exceeds a preset threshold.

[0008] Optionally, the infrasound signal and corresponding medium characteristic parameters during the operation of the target device are collected to establish a medium propagation characteristic table for characterizing the signal propagation path. This includes: synchronously collecting amplitude and phase change data of the corresponding infrasound signal based on multiple infrasound sensors deployed on the outer surface of the target device's casing, support parts, and / or sound energy dissipation path; simultaneously collecting the infrasound signal and recording the corresponding medium characteristic parameters based on the installation location; wherein, the medium characteristic parameters include any one or more of material density, thickness, elastic modulus, and ambient temperature; performing propagation characteristic calculations on the amplitude and phase change data based on the spatial location of each infrasound sensor and the medium characteristic parameters to obtain propagation parameters for characterizing propagation speed and energy attenuation characteristics; and storing each propagation parameter in association with the medium characteristic parameters to form the medium propagation characteristic table.

[0009] Optionally, based on the spatial location of each infrasound sensor and the medium characteristic parameters, propagation characteristic calculations are performed on the amplitude change data and phase change data to obtain propagation parameters characterizing propagation speed and energy attenuation characteristics. This includes: calculating the signal arrival time difference based on the spatial location difference of each infrasound sensor, and combining the material density and thickness in the medium characteristic parameters to solve for the signal propagation speed in different paths; using the propagation speed calculation results as input, analyzing the energy distribution of the amplitude change data on each path to determine the energy attenuation coefficient of the signal during propagation; and pairing and recording the propagation speed and the energy attenuation coefficient with the spatial location of each infrasound sensor to generate propagation parameters corresponding to each path.

[0010] Optionally, a propagation correction model is constructed based on the medium propagation characteristic table to perform distortion correction on the infrasound signal, thereby correcting the phase and amplitude of the signal and outputting a corrected signal. This includes: establishing the propagation correction model using the propagation velocity and energy attenuation coefficient in the medium propagation characteristic table as inputs; identifying the phase reversal interval of the infrasound signal through the Hilbert envelope in the propagation correction model and establishing a propagation path matrix in conjunction with the propagation velocity; correcting the phase distribution of the infrasound signal according to the propagation path matrix and performing amplitude compensation based on the energy attenuation coefficient to output a corrected signal with phase and amplitude correction.

[0011] Optionally, in the propagation correction model, the phase reversal interval of the infrasound signal is identified by the Hilbert envelope, and a propagation path matrix is ​​established in conjunction with the propagation speed. This includes: performing a Hilbert transform on the infrasound signal to extract the instantaneous phase sequence of the signal, and determining the phase reversal interval based on the phase jump amplitude; using the acquisition position corresponding to each phase reversal interval as an index, and combining the propagation speed in the medium propagation characteristic table, calculating the propagation delay of the signal in different paths; and writing the correspondence between the propagation delay of each path and the phase reversal interval into a matrix structure to form the propagation path matrix.

[0012] Optionally, extracting time-varying feature parameters and constructing a time-varying feature matrix based on the correction signal, calculating the feature drift rate and inputting it into the drift determination model includes: performing time-frequency analysis on the correction signal to extract time-varying feature parameters characterizing the operating state of the equipment; wherein, the time-varying feature parameters include signal envelope energy, frequency band center position and phase stability; arranging each time-varying feature parameter in the order of acquisition time to form a time-varying feature matrix, and calculating the change of each feature parameter within adjacent time windows; dividing the change by the corresponding time interval to obtain the feature drift rate, and importing the feature drift rate as input data into the drift determination model.

[0013] Optionally, the drift determination model is used to receive feature drift rates and perform environmental normalization and covariance drift analysis. The construction rules of the drift determination model are as follows: using the corrected time-varying feature matrix as input samples, selecting the feature drift rate as the model training feature, and using the drift rate offset under different environmental conditions as label data; establishing a mapping relationship between the environmental compensation coefficient and the drift rate offset during training, and using the covariance change rate among multiple features as the determination index; setting drift analysis parameters according to the mapping relationship and the covariance change rate, and generating a determination function with environmental normalization and covariance drift analysis functions; and solving the model parameters based on the determination function to obtain the drift determination model.

[0014] Optionally, based on the drift determination model, environmental normalization and covariance drift analysis are performed. When the drift rate exceeds a preset threshold, a device anomaly determination result is output, including: performing environmental normalization processing on the feature drift rates input to the drift determination model to obtain compensated drift rates; calculating the covariance matrix between compensated drift rates in the drift determination model, and identifying a feature set with consistent drift directions based on the covariance change trend; determining whether the drift rates in the feature set all exceed the preset threshold. If the determination condition is met, the device anomaly determination result is generated.

[0015] Optionally, the generation rule for the equipment anomaly determination result is as follows: using the compensated drift rate and covariance matrix output by the drift determination model as input, calculate the spatial distribution of each characteristic drift rate in the equipment structural coordinates; based on the spatial distribution and the propagation path relationship in the medium propagation characteristic table, determine the spatial location coordinates of the abnormal signal source; according to the change pattern and amplitude stability characteristics of the characteristic drift rate, classify the abnormal signal source into anomaly types; output the spatial location coordinates and anomaly type together as the equipment anomaly determination result.

[0016] A second aspect of the present invention provides a device anomaly detection system based on infrasound data. The system includes: a data acquisition unit for acquiring infrasound signals and corresponding medium characteristic parameters during the operation of a target device, and establishing a medium propagation characteristic table to characterize the signal propagation path; a correction unit for constructing a propagation correction model based on the medium propagation characteristic table, performing distortion correction on the infrasound signals to correct the phase and amplitude of the signals, and outputting a correction signal; a processing unit for extracting time-varying characteristic parameters based on the correction signal and constructing a time-varying characteristic matrix, calculating the characteristic drift rate and inputting it into a drift judgment model; and an output unit for performing environmental normalization and covariance drift analysis based on the drift judgment model, and outputting a device anomaly judgment result when the drift rate exceeds a preset threshold.

[0017] Through the above technical solution, this invention achieves the correction of phase and amplitude distortion of subacoustic signals propagating through multiple media by constructing a media propagation characteristic table and a propagation correction model. This makes signals under different structural conditions comparable, improving the stability and accuracy of the detection results. Based on this, by extracting time-varying feature parameters from the corrected signal and establishing a time-varying feature matrix, continuous characterization of equipment operating state changes is achieved. Furthermore, by utilizing a drift judgment model to perform environmental normalization and covariance drift analysis, environmental disturbances and structural changes are effectively distinguished, and abnormal trends can be accurately identified when the drift rate exceeds a preset threshold. Overall, this method achieves high-precision detection and reliable judgment of abnormal states under conditions of slowly changing equipment faults and multi-media propagation distortion.

[0018] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0020] Figure 1 This is a flowchart of a device anomaly detection method based on infrasound data provided in one embodiment of the present invention;

[0021] Figure 2 This is a system structure diagram of a device anomaly detection system based on infrasound data provided in one embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for illustration and explanation of the embodiments of this application and are not intended to limit the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0023] It should be noted that if the embodiments of this application involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0024] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0025] like Figure 1 As shown, an embodiment of the present invention provides a method for high-precision forming of semiconductor carrier tape, the method comprising:

[0026] Step S1: Collect infrasound signals and corresponding medium characteristic parameters during the operation of the target equipment, and establish a medium propagation characteristic table to characterize the signal propagation path.

[0027] Specifically, multiple infrasound sensors deployed on the outer surface of the target device's housing, support components, and / or sound energy dissipation paths synchronously collect amplitude and phase change data of corresponding infrasound signals. While collecting the infrasound signals, corresponding medium characteristic parameters are recorded based on the installation location. These medium characteristic parameters include any one or more of material density, thickness, elastic modulus, and ambient temperature. Based on the spatial location of each infrasound sensor and the medium characteristic parameters, propagation characteristic calculations are performed on the amplitude and phase change data to obtain propagation parameters characterizing propagation speed and energy attenuation characteristics. Each propagation parameter is associated with and stored with the medium characteristic parameters to form the medium propagation characteristic table.

[0028] Furthermore, based on the spatial positions of each infrasound sensor and the medium characteristic parameters, propagation characteristic calculations are performed on the amplitude change data and phase change data to obtain propagation parameters characterizing propagation speed and energy attenuation characteristics. This includes: calculating the signal arrival time difference based on the spatial position difference of each infrasound sensor, and combining the material density and thickness in the medium characteristic parameters to solve for the signal propagation speed in different paths; using the propagation speed calculation results as input, analyzing the energy distribution of the amplitude change data on each path to determine the energy attenuation coefficient of the signal during propagation; and pairing and recording the propagation speed and the energy attenuation coefficient with the spatial positions of each infrasound sensor to generate propagation parameters corresponding to each path.

[0029] In this embodiment of the invention, during the acoustic detection of the equipment's operating status, in order to accurately reflect the propagation law of infrasound signals in the equipment structure and medium, it is necessary to first establish a physically representative propagation characteristic data foundation, also known as the medium propagation characteristic table. This part is actually the underlying support of the entire detection method, which determines the accuracy of subsequent signal correction, feature extraction, and even anomaly detection. Therefore, the integrity and physical consistency of its data source must be guaranteed.

[0030] Specifically, the first step is to determine the strategy for deploying the infrasound sensors. Because infrasound signals have low frequencies and long wavelengths, they are highly sensitive to the overall structure of the equipment, making single-point sampling often insufficient to accurately distinguish differences in propagation paths. Typically, multiple infrasound sensors are evenly distributed on the outer surface of the equipment casing, major supporting components, and along the sound energy dissipation path. The number of sensors depends on the equipment's size and structural complexity, generally ranging from 4 to 16. The sensor spacing should ideally be controlled within 1 / 10 to 1 / 20 of the equipment's characteristic dimensions. This ensures spatial resolution while avoiding signal coupling interference caused by overly dense placement. All sensors should be calibrated before installation to ensure their amplitude-frequency response curves remain linear within the 1Hz to 30Hz range, guaranteeing signal acquisition accuracy in the infrasound frequency band.

[0031] During normal operation, each sensor synchronously acquires infrasound signals at its corresponding location. "Synchronization" is crucial here; synchronization errors directly lead to deviations in propagation speed calculations. Therefore, a unified clock source or GPS time base is typically used to trigger the acquisition module, ensuring that the time error across all channels does not exceed 1 millisecond. Each sensor channel records the instantaneous amplitude and phase changes of the signal. The amplitude reflects the attenuation characteristics of sound energy during propagation, while the phase change reflects the differences in the signal propagation path. A sampling frequency of at least 200Hz is recommended to obtain sufficient time resolution.

[0032] Simultaneously with signal acquisition, it is necessary to record the medium characteristic parameters at each sensor installation location. This data directly determines the accuracy of the propagation characteristic calculation. Medium characteristic parameters generally include one or more of the following: material density, thickness, elastic modulus, and ambient temperature. For metal shell structures, material density and elastic modulus can be found in material standard handbooks or determined by ultrasonic methods; thickness can be obtained from design drawings or on-site thickness gauges; and ambient temperature is recorded in real-time by temperature sensors. These parameters collectively determine the propagation speed and energy loss of sound waves in the medium, and are indispensable inputs for subsequent propagation characteristic calculations.

[0033] The next step is the core step of calculating propagation characteristics. The goal of this step is to extract the propagation velocity and energy attenuation coefficient along the path between each sensor, to characterize the signal transmission characteristics in the device medium. First, based on the spatial position difference ΔL between the sensors and the signal arrival time difference Δt, the propagation velocity v can be calculated using the following formula:

[0034] v=ΔL / Δt

[0035] Δt can be obtained through cross-correlation algorithms or phase difference calculations. Direct peak-based methods should be avoided during calculation to minimize the impact of noise interference. To further improve accuracy, sliding window cross-correlation or Hilbert envelope peak matching methods can be used to extract the arrival time, thereby obtaining more stable propagation delay results.

[0036] After determining the propagation speed, it is necessary to analyze the energy changes of the signal propagating along different paths. Typically, based on the collected amplitude variation data, the energy ratio between the start and end points of each path is calculated. Combined with the path length L, the energy attenuation coefficient α is obtained, the definition of which can be found in the logarithmic attenuation formula.

[0037] α=(1 / L)×ln(A1 / A2)

[0038] Where A1 and A2 are the signal amplitudes at the starting and ending points of the sensor, respectively. The attenuation coefficient calculated in this way can reflect the damping characteristics of the medium and the energy loss characteristics of the structure. If there are contact interfaces or local cracks in the medium, the energy attenuation coefficient will usually increase significantly. Therefore, this parameter can also be used as an important characteristic indicator for anomaly detection.

[0039] After obtaining the propagation speed and energy attenuation coefficient, it is necessary to associate these two key propagation parameters with the corresponding spatial locations of the sensors. Specifically, the spatial coordinates, propagation speed, and energy attenuation coefficient of each pair of sensors can be recorded using path index numbers. The data from all paths together constitute a propagation parameter set and are stored sequentially according to the path number. For ease of subsequent processing, it is recommended to use a unified data structure format, such as a matrix, where each row represents a propagation path, and each column corresponds to the path number, starting coordinates, ending coordinates, propagation speed, attenuation coefficient, and associated medium characteristic parameters, respectively.

[0040] Subsequently, the aforementioned propagation parameter set is integrated with the medium characteristic parameters to establish a complete medium propagation characteristic table. This table not only records the physical properties of each propagation path but also includes propagation response data under the corresponding medium environment. Through this table, a quantitative description of the sound wave propagation patterns in different regions within the equipment can be achieved. Especially in multi-medium structures (such as combinations of metal, supporting rubber layers, and sealing media), the phase changes and energy loss characteristics at material transition interfaces for different paths can be clearly identified.

[0041] Step S2: Construct a propagation correction model based on the medium propagation characteristic table, perform distortion correction on the infrasound signal to correct the phase and amplitude of the signal, and output the corrected signal.

[0042] Specifically, the propagation correction model is established using the propagation speed and energy attenuation coefficient in the medium propagation characteristic table as inputs; in the propagation correction model, the phase reversal interval of the infrasound signal is identified by the Hilbert envelope, and a propagation path matrix is ​​established in combination with the propagation speed; the phase distribution of the infrasound signal is corrected according to the propagation path matrix, and amplitude compensation is performed based on the energy attenuation coefficient to output a corrected signal with amplitude and phase correction.

[0043] Furthermore, in the propagation correction model, the phase reversal intervals of the infrasound signal are identified by the Hilbert envelope, and a propagation path matrix is ​​established in conjunction with the propagation speed. This includes: performing a Hilbert transform on the infrasound signal to extract the instantaneous phase sequence of the signal, and determining the phase reversal intervals based on the phase jump amplitude; using the acquisition position corresponding to each phase reversal interval as an index, and combining the propagation speed in the medium propagation characteristic table, calculating the propagation delay of the signal in different paths; and writing the correspondence between the propagation delay of each path and the phase reversal interval into the matrix structure to form the propagation path matrix.

[0044] In this embodiment of the invention, in practical applications of equipment operation status detection, infrasound signals often need to propagate through the shell, support structure, and multiple layers of media before being received by the sensor. Due to differences in density, thickness, and elastic modulus among the various media layers, the signal inevitably experiences multipath effects, phase reversal, energy attenuation, and nonlinear superposition during propagation. If these propagation distortions are not corrected, subsequent feature extraction and anomaly identification will show significant deviations, leading to incomparable data from different devices or operating conditions. Therefore, after constructing the media propagation feature table, a propagation correction model needs to be established based on this feature table to correct the amplitude and phase of the acquired infrasound signal in order to restore its true propagation characteristics.

[0045] Specifically, the propagation correction model is built using a medium propagation characteristic table as its core input. This table records the propagation velocity *v* and energy attenuation coefficient *α* for each path, reflecting the time delay and energy loss characteristics of signal propagation, respectively. When constructing the model, the propagation velocity of each path is used as the primary basis for time delay correction, while the energy attenuation coefficient is used for subsequent amplitude compensation. To establish the mathematical relationships between propagation paths, a propagation path matrix between each sensor needs to be defined in the model. This matrix describes the propagation path, time delay, and phase response of the signal from the source point to each sampling point.

[0046] Hilbert envelope analysis is a crucial step in establishing a propagation correction model. The Hilbert transform converts real-valued infrasound signals into analytic signals, providing instantaneous phase and amplitude information. By calculating the envelope of the analytic signal after the Hilbert transform, phase transition points during propagation can be identified. For low-frequency infrasound, medium reflection and interface refraction often cause phase reversal within the range of π or 2π. Therefore, a phase reversal identification mechanism needs to be included in the model. Specifically, the instantaneous phase sequence of each signal segment is differentially calculated. When the phase change rate exceeds a preset threshold (e.g., a phase change exceeding π / 2 per sampling period), the region is determined to be a phase reversal interval.

[0047] After identifying all phase reversal intervals, the sensor acquisition positions corresponding to each phase reversal interval are used as indices, and combined with the propagation velocity data in the medium propagation characteristic table, the propagation delay of the signal in different paths is calculated. Propagation Delay It can be calculated using the following formula:

[0048]

[0049] Where L is the path length and v is the propagation speed recorded in the medium propagation characteristic table. The calculated propagation delay is used to construct the delay component in the propagation path matrix, while the phase reversal interval serves as the phase identifier unit in the matrix. In this way, a multi-dimensional matrix structure can be obtained, which uses spatial location as the row index and path number as the column index. The matrix elements record the propagation delay, phase state, and corresponding medium properties of each path.

[0050] After constructing the propagation path matrix, the signal correction stage begins. First, the propagation delay information in the path matrix is ​​used to align the original infrasound signal in the time domain, rearranging signal segments from different paths according to propagation speed and delay to eliminate phase misalignment caused by multipath propagation. Then, based on the phase reversal interval markers, reverse compensation is performed on the signal phase within the reversal interval; that is, π or 2π compensation is performed at the phase jump points to restore the signal phase to the same phase direction as the direct wave, thereby eliminating waveform distortion caused by phase reversal.

[0051] Next, amplitude correction is performed. The energy attenuation coefficient α recorded in the medium propagation characteristic table is used to compensate for the signal amplitude along different paths. Amplitude compensation can be performed according to the following formula:

[0052]

[0053] in The corrected amplitude. The measured original amplitude is given by L, where L is the propagation path length. This compensation process can be understood as a reverse recovery of signal energy loss, ensuring that the amplitude of the signal along each path matches the theoretical energy distribution. This process effectively reduces energy shifts caused by differences in material damping or structural coupling.

[0054] After amplitude and phase correction, the corrected signals are recombined into a corrected signal sequence. During this process, time synchronization must be maintained to ensure that the start and end times of each path signal are consistent, thereby avoiding spurious features or phase drift in subsequent feature extraction. The final output corrected signal, compared to the original signal, has a more continuous phase distribution and more stable amplitude changes, accurately reflecting the true acoustic response characteristics of the equipment under operating conditions.

[0055] Step S3: Extract time-varying feature parameters based on the correction signal and construct a time-varying feature matrix, calculate the feature drift rate and input it into the drift determination model.

[0056] Specifically, time-frequency analysis is performed on the correction signal to extract time-varying characteristic parameters that characterize the operating state of the equipment. The time-varying characteristic parameters include signal envelope energy, frequency band center position, and phase stability. The time-varying characteristic parameters are arranged in the order of acquisition time to form a time-varying characteristic matrix, and the change of each characteristic parameter is calculated within adjacent time windows. The change is divided by the corresponding time interval to obtain the characteristic drift rate, and the characteristic drift rate is used as input data to import into the drift determination model.

[0057] Furthermore, the drift determination model is used to receive feature drift rates and perform environmental normalization and covariance drift analysis. The construction rules of the drift determination model are as follows: using the corrected time-varying feature matrix as input samples, selecting the feature drift rate as the model training feature, and using the drift rate offset under different environmental conditions as label data; establishing a mapping relationship between the environmental compensation coefficient and the drift rate offset during training, and using the covariance change rate among multiple features as the determination index; setting drift analysis parameters according to the mapping relationship and the covariance change rate, and generating a determination function with environmental normalization and covariance drift analysis functions; and solving the model parameters based on the determination function to obtain the drift determination model.

[0058] In this embodiment of the invention, when processing the propagation-corrected infrasound signal, the phase and amplitude of the signal have already been corrected by previous steps. Therefore, the signal at this point can be considered relatively pure dynamic response information. To ensure that the detection results reflect the changing trends of the equipment's operating status, rather than just instantaneous anomalies, it is necessary to further extract a set of parameters reflecting time-varying characteristics from the corrected signal, i.e., the so-called time-varying characteristic parameters. The goal of this part of the processing is to transform the original time-series signal into quantitative indicators that can characterize energy distribution, spectral shift, and phase stability changes, thereby providing a basic input for the subsequent drift determination model.

[0059] Specifically, time-frequency analysis is performed on the correction signal to obtain its local energy distribution characteristics in different time and frequency dimensions. The time-frequency analysis method can employ either Short-Time Fourier Transform (STFT) or Continuous Wavelet Transform (CWT). Given the lower frequency band and smoother time-varying characteristics of infrasound signals, the STFT is preferred. Let the correction signal be... The window function is Then its time-frequency distribution can be expressed as:

[0060]

[0061] By analyzing the time-frequency distribution The analysis can simultaneously obtain the evolution information of the signal on the time axis and the changes in energy distribution on the frequency axis. From this, three typical time-varying characteristic parameters can be extracted: signal envelope energy, frequency band center position, and phase stability.

[0062] Signal envelope energy Characterization equipment in time window The overall vibration intensity within the area can be calculated using the following formula:

[0063]

[0064] Changes in envelope energy can reflect differences in energy release caused by changes in equipment load or operational abnormalities. For example, when internal components experience friction or fluid leakage, the envelope energy curve will show a continuous rise or periodic fluctuations.

[0065] Frequency band center position The location of the spectral barycenter used to characterize the signal energy is defined as follows:

[0066]

[0067] When equipment experiences structural loosening or changes in operating load, the dominant frequency band of the acoustic signal shifts, causing a drift in the center frequency. The direction and magnitude of this drift usually have significant engineering implications and can serve as important indicators of changes in equipment condition.

[0068] The third characteristic parameter is phase stability, which reflects the phase consistency of the device's response signal. It is calculated by first performing a Hilbert transform on the corrected signal to obtain the analytic signal. And from this, the instantaneous phase sequence is extracted. Within the sliding time window, the variance of the phase sequence is calculated. And take its reciprocal as the phase stability index:

[0069]

[0070] in, The width of the time window. To prevent the denominator from being a tiny constant. Higher phase stability indicates stable equipment operation and smooth signal phase changes; however, this value will decrease significantly when the equipment experiences periodic interference, loosening, or vibration resonance.

[0071] After extracting the aforementioned feature parameters, the feature parameter values ​​within each time window are arranged in chronological order of acquisition time to form a time-varying feature matrix. Where m is the number of time windows and n is the feature dimension (3 in this example). To ensure that the values ​​of different features are comparable, the parameters can be normalized to limit their range to [0,1].

[0072] Subsequently, the feature values ​​within adjacent time windows are differentially analyzed to obtain the change of each feature over time:

[0073]

[0074] Then divide the change by the center interval of adjacent time windows. The characteristic drift rate is obtained as follows:

[0075]

[0076] Characteristic drift rate It directly reflects the changing trend of the equipment's condition. When the equipment is operating smoothly, the drift rate of each characteristic is close to zero; if a slow-changing anomaly occurs (such as component wear, structural loosening, etc.), the drift rate will show a continuous shift; while when a sudden anomaly occurs (such as impact or leakage), the drift rate will increase instantaneously. Through continuous calculation, a drift rate sequence can be obtained. , which serves as the input data for the drift determination model.

[0077] The purpose of constructing the drift determination model is to distinguish between drift caused by environmental factors and real-state changes caused by equipment malfunctions. To achieve this, a calibrated time-varying feature matrix is ​​first used as training samples, with the drift rate selected as the feature input, and drift offsets under different environmental conditions used as label data. During the training phase, a mapping relationship between the environmental compensation coefficient and the drift offset is established through least squares or nonlinear fitting; for example, a linear form can be used. ,in b and are environmental compensation parameters. This mapping relationship enables the model to compensate for external disturbances such as changes in temperature, humidity, or load, thereby achieving environmental normalization.

[0078] Next, the model calculates the covariance matrix among multiple features. The consistency of drift direction among features is assessed based on its rate of change over time. The covariance rate of change index is defined as follows:

[0079]

[0080] in, This represents the Frobenius norm. It is the rate of change of covariance when there is correlated drift between features, i.e., multiple features shift simultaneously in the same direction. This will increase significantly. Based on this, the model sets drift analysis parameters and judgment thresholds to determine whether the drift is caused by anomalies.

[0081] Finally, based on the above environmental normalization and covariance rate of change analysis results, the model generates a decision function to comprehensively evaluate the input drift rate vector. When the drift rate exceeds a preset threshold and the feature drift direction is consistent, the model outputs an anomaly identification signal, which serves as the basis for subsequent anomaly type identification and localization.

[0082] Step S4: Perform environmental normalization and covariance drift analysis based on the drift determination model. When the drift rate exceeds a preset threshold, output the device anomaly determination result.

[0083] Specifically, the feature drift rates input to the drift determination model are subjected to environmental normalization processing to obtain compensated drift rates; the covariance matrix between compensated drift rates is calculated in the drift determination model, and a feature set with consistent drift directions is identified based on the covariance change trend; it is determined whether the drift rates in the feature set all exceed a preset threshold. If the determination condition is met, the device anomaly determination result is generated.

[0084] Furthermore, the generation rules for the equipment anomaly determination results are as follows: using the compensated drift rate and covariance matrix output by the drift determination model as input, calculate the spatial distribution of each characteristic drift rate in the equipment structural coordinates; based on the spatial distribution and the propagation path relationship in the medium propagation characteristic table, determine the spatial location coordinates of the abnormal signal source; according to the change pattern and amplitude stability characteristics of the characteristic drift rate, classify the abnormal signal source into anomaly types; output the spatial location coordinates and anomaly types together as the equipment anomaly determination results.

[0085] In this embodiment of the invention, after training and establishing the drift determination model, the feature drift rate needs to be input into the model for dynamic judgment during real-time monitoring of equipment operation to determine whether the equipment has potential abnormal states. The main technical objective of this stage is to decouple the drift rate from environmental factors and identify abnormal drift patterns through the covariance variation trend between multidimensional features, ultimately achieving spatial localization and type identification of the anomaly. The entire process not only requires threshold judgment of the amplitude of signal changes but also spatial reconstruction of the drift distribution pattern to determine the origin and nature of the anomaly.

[0086] Specifically, the feature drift rate input to the drift judgment model is first normalized using environmental normalization. Since the environment in which the device operates is often subject to external disturbances such as fluctuations in temperature, humidity, air pressure, or load, these environmental factors can cause apparent changes in feature drift, but do not necessarily indicate an abnormality in the device itself. To avoid such misjudgments, the drift rate needs to be corrected using the environmental compensation coefficient obtained during the training phase. Let the measured drift rate be... The environmental compensation coefficient is If the offset constant is b, then the compensated drift rate is... It can be calculated in the following linear form:

[0087]

[0088] This normalization process can effectively eliminate non-abnormal drift caused by environmental fluctuations, allowing subsequent analysis to focus on the structural response of the equipment itself.

[0089] After obtaining the compensated drift rate sequence, the drift determination model calculates the covariance matrix among multiple features to reflect the correlation between the drifts of each feature. The elements of the covariance matrix are defined as follows:

[0090]

[0091] in, Let represent the mean of the i-th feature. By calculating the covariance matrix, the coupling relationship between the drifts of different features can be obtained. When the operating state of the equipment changes, the drifts between features usually exhibit some synchronicity or directional consistency. This feature coupling is manifested in the covariance matrix as a concentrated shift of eigenvalues ​​or an amplified change in a specific direction. To facilitate quantitative judgment, the covariance change rate index is introduced. :

[0092]

[0093] in, This represents the Frobenius norm. A higher rate of change in covariance indicates a significant change in the correlation between features over a short period of time, often corresponding to abrupt changes in the internal dynamic characteristics of the equipment, such as loosening of components, formation of leakage channels, or changes in the state of the medium.

[0094] Based on the calculation of the covariance matrix and its rate of change, the model further identifies feature sets with consistent drift directions. The judgment criterion is typically based on the angular relationship between the eigenvector directions of the covariance matrix or the drift rate vectors. When multiple features have highly consistent drift directions and their covariance rate of change exceeds a threshold, these features are considered to be influenced by a common anomaly source. In this case, the mean drift rate is extracted from the feature set, and it is determined whether all of them exceed a preset threshold. If this condition is met, it indicates that there is an anomaly in the device area corresponding to this feature set.

[0095] To further achieve visualized localization and type determination of anomalies, it is necessary to calculate the spatial distribution of each characteristic drift rate in the device structural coordinates based on the compensated drift rate and covariance matrix output by the drift determination model. This spatial mapping process is based on the medium propagation characteristic table established in the previous steps, which records the propagation path, propagation speed, and energy attenuation characteristics between different sensor locations. By matching the drift rate vector with the medium propagation path matrix, the spatial coordinates of the abnormal signal source can be solved. For example, when the drift rates corresponding to multiple sensors show synchronous enhancement and consistent direction in a certain area, the signal source location can be inferred from the propagation time difference and energy attenuation gradient.

[0096] The weighted least squares method can be used in the calculation process, based on the change in drift rate. With path length The relationship between the two is used to solve for the three-dimensional coordinates of the abnormal signal source. The model minimizes the propagation error function:

[0097]

[0098] in, The function represents the weighting factors for each sensor location. This represents the propagation attenuation relationship based on path length. By taking the partial derivative of this error function and solving iteratively, the optimal coordinate estimate of the anomalous signal source can be obtained.

[0099] After determining the location of the abnormal signal source, it is necessary to classify the anomaly type. The classification of anomaly types is mainly based on the variation pattern of the characteristic drift rate and its amplitude stability characteristics. Specifically, if the drift rate suddenly increases and then quickly decreases within a short period, it usually indicates a transient impact or pulse interference type anomaly; if the drift rate continues to rise and the amplitude stability decreases, it indicates that the equipment may have structural loosening or leakage accumulation type anomalies; and if the drift rate oscillates periodically in a certain direction, it is mostly an imbalance of rotating parts or resonance type anomalies. Through this variation pattern identification, the anomaly type can be initially classified without relying on external detection equipment.

[0100] Finally, the model outputs the spatial coordinates of the abnormal signal source along with the corresponding anomaly type, forming the equipment anomaly determination result. This result includes both quantitative indicators of the anomaly (drift rate exceeding the threshold, covariance change rate, etc.) and qualitative information (anomaly category and spatial location). The output format can be structured data, such as... This facilitates subsequent integration with and maintenance databases or digital twin models.

[0101] Figure 2 This is a system structure diagram of a device anomaly detection system based on infrasound data provided in one embodiment of the present invention. Figure 2 As shown, this invention provides a device anomaly detection system based on infrasound data. The system includes: a data acquisition unit for acquiring infrasound signals and corresponding medium characteristic parameters during the operation of the target device, and establishing a medium propagation characteristic table to characterize the signal propagation path; a correction unit for constructing a propagation correction model based on the medium propagation characteristic table, performing distortion correction on the infrasound signal to correct the phase and amplitude of the signal, and outputting a correction signal; a processing unit for extracting time-varying characteristic parameters based on the correction signal and constructing a time-varying characteristic matrix, calculating the characteristic drift rate and inputting it into a drift judgment model; and an output unit for performing environmental normalization and covariance drift analysis based on the drift judgment model, and outputting a device anomaly judgment result when the drift rate exceeds a preset threshold.

[0102] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a microcontroller, chip, or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0103] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details described above. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention. It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not further describe the various possible combinations.

[0104] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the embodiments of the present invention, they should also be regarded as the content disclosed by the embodiments of the present invention.

Claims

1. An infrasound wave data-based device anomaly detection method, characterized by, The method comprises: Collecting infrasound wave signals and corresponding medium characteristic parameters during operation of a target device, and establishing a medium propagation characteristic table for representing a signal propagation path, comprising: Synchronously collecting amplitude variation data and phase variation data of corresponding infrasound wave signals based on a plurality of infrasound wave sensors arranged on the outer surface of the shell, the support part and / or the sound energy dissipation path of the target device; while collecting the infrasound wave signals, recording corresponding medium characteristic parameters based on the installation position; wherein the medium characteristic parameters include any one or more of material density, thickness, elastic modulus and environmental temperature; performing propagation characteristic calculation on the amplitude variation data and the phase variation data according to the spatial positions of the infrasound wave sensors and the medium characteristic parameters, to obtain propagation parameters for representing propagation speed and energy attenuation characteristics; storing each propagation parameter in association with the medium characteristic parameters to form the medium propagation characteristic table; Constructing a propagation correction model based on the medium propagation characteristic table, and performing distortion correction on the infrasound wave signals to correct the phase and amplitude of the signals, and outputting a corrected signal; Extracting time-varying characteristic parameters based on the corrected signal and constructing a time-varying characteristic matrix, calculating a characteristic drift rate and inputting the drift determination model, comprising: Performing time-frequency analysis on the corrected signal to extract time-varying characteristic parameters representing the operating state of the device; wherein the time-varying characteristic parameters include signal envelope energy, frequency band center position and phase stability; arranging each time-varying characteristic parameter in time sequence to form a time-varying characteristic matrix, and calculating the variation of each characteristic parameter in adjacent time windows; dividing the variation by the corresponding time interval to obtain a characteristic drift rate, and inputting the characteristic drift rate as input data into the drift determination model; The drift determination model is used to receive the characteristic drift rate and perform environmental normalization and covariance drift analysis; the construction rule of the drift determination model is: taking the corrected time-varying characteristic matrix as input samples, selecting the characteristic drift rate as model training features, and taking the drift rate offset under different environmental conditions as label data; in the training process, a mapping relationship between the environmental compensation coefficient and the drift rate offset is established, and the covariance change rate among multiple features is taken as a determination index; according to the mapping relationship and the covariance change rate, drift analysis parameters are set to generate a determination function with environmental normalization and covariance drift analysis functions; based on the determination function, model parameter solving is completed to obtain the drift determination model; Performing environmental normalization and covariance drift analysis based on the drift determination model, and outputting a device abnormality determination result when the drift rate exceeds a preset threshold.

2. The infrasound wave data-based equipment abnormality detection method according to claim 1, characterized by, Performing propagation characteristic calculation on the amplitude variation data and the phase variation data according to the spatial positions of the infrasound wave sensors and the medium characteristic parameters, to obtain propagation parameters for representing propagation speed and energy attenuation characteristics, comprising: Calculating the time difference of the signals arriving at different positions of each infrasound wave sensor, and combining the material density and thickness in the medium characteristic parameters to solve the propagation speed of the signals in different paths; Taking the propagation speed calculation result as input, energy distribution of the amplitude variation data on each path is analyzed to determine energy attenuation coefficient of the signal in the propagation process; The propagation speed and the energy attenuation coefficient corresponding to the spatial position of each infrasound sensor are recorded in pairs to generate the propagation parameters corresponding to each path.

3. The infrasound wave data-based equipment abnormality detection method according to claim 1, characterized by, Based on the medium propagation characteristic table, a propagation correction model is constructed to perform distortion correction on the infrasound signal to correct the phase and amplitude of the signal, and output a corrected signal, including: Taking the propagation speed and the energy attenuation coefficient in the medium propagation characteristic table as input, the propagation correction model is established; In the propagation correction model, the phase inversion interval of the infrasound signal is identified by Hilbert envelope, and a propagation path matrix is established in combination with the propagation speed; According to the propagation path matrix, the phase distribution of the infrasound signal is corrected, and amplitude compensation is performed based on the energy attenuation coefficient, and a corrected signal with corrected amplitude and phase is output.

4. The infrasound wave data-based equipment abnormality detection method according to claim 3, characterized by, In the propagation correction model, the phase inversion interval of the infrasound signal is identified by Hilbert envelope, and the propagation path matrix is established in combination with the propagation speed, including: Performing Hilbert transform on the infrasound signal to extract the instantaneous phase sequence of the signal, and determining the phase inversion interval according to the phase jump amplitude; Taking the collection position corresponding to each phase inversion interval as an index, and combining the propagation speed in the medium propagation characteristic table, the propagation time delay of the signal in different paths is calculated; The correspondence between the propagation time delay of each path and the phase inversion interval is written into a matrix structure to form the propagation path matrix.

5. The infrasound wave data-based equipment abnormality detection method according to claim 1, characterized by, Based on the drift judgment model, environment normalization and covariance drift analysis are performed, and when the drift rate exceeds a preset threshold, a device abnormality judgment result is output, including: Performing environment normalization processing on the feature drift rate input to the drift judgment model to obtain a compensated drift rate; In the drift judgment model, a covariance matrix between the compensated drift rates is calculated, and a feature set with consistent feature drift direction is identified according to the covariance change trend; Judge whether the drift rates in the feature set all exceed the preset threshold, if the judgment condition is met, the device abnormality judgment result is generated.

6. The infrasound wave data-based equipment abnormality detection method according to claim 5, characterized by, The generation rule of the device abnormality judgment result is: Taking the compensated drift rate and the covariance matrix output by the drift judgment model as input, the spatial distribution of each feature drift rate in the device structure coordinate is calculated; Based on the spatial distribution and the propagation path relationship in the medium propagation characteristic table, the spatial position coordinates of the abnormal signal source are determined; According to the change mode and amplitude stability characteristics of the feature drift rate, the abnormal type of the abnormal signal source is divided; The spatial position coordinates and the abnormal type are output together as the device abnormality judgment result.

7. An infrasound wave data-based equipment abnormality detection system characterized by comprising: The system is used to execute the device abnormality detection method based on infrasound data in any one of claims 1-6, and the system includes: An acquisition unit is configured to acquire an infrasound signal and corresponding medium characteristic parameter during operation of a target device, and establish a medium propagation characteristic table for representing a signal propagation path; The correction unit is configured to construct a propagation correction model based on the medium propagation characteristic table, perform distortion correction on the infrasound wave signal to correct the phase and amplitude of the signal, and output a corrected signal; The processing unit is configured to extract time-varying characteristic parameters based on the corrected signal and construct a time-varying characteristic matrix, calculate a characteristic drift rate, and input the drift rate into a drift determination model; The output unit is configured to perform environment normalization and covariance drift analysis based on the drift determination model, and output a device anomaly determination result when the drift rate exceeds a preset threshold.

Images