Non-stationary instantaneous impact monitoring method and system

Abstract

The invention discloses a non-stationary instantaneous impact monitoring method and system, and belongs to the technical field of vibration signal analysis and fault diagnosis. The method comprises the steps of firstly obtaining a vibration signal and framing; calculating an envelope signal of each passband and an envelope spectrum of a frequency domain of each passband through multi-passband filtering and Hilbert transform; calculating a relative envelope spectrum probability curve, and identifying a local maximum pulse by calculating a second-order difference of the relative envelope spectrum probability curve within an analysis frequency band of 0-200Hz; and calculating the ratio of each pulse to the average value in the minimum pulse interval on the left side and the right side of the pulse to carry out significance judgment, and finally judging whether non-stationary instantaneous impact damage exists or not in combination with multi-level threshold statistical test. According to the method, high-precision and high-reliability extraction and identification of non-stationary instantaneous impact characteristics can be realized under a strong noise background.

Description

Technical Field

[0001] This application is a divisional application of patent application No. 202512033234.7, filed on December 30, 2025, entitled "A Method and System for Anti-vandalism Monitoring". The present invention relates to the field of vibration signal analysis and fault diagnosis technology, and in particular to a method and system for monitoring non-stationary instantaneous impacts. Background Technology

[0002] In fields such as mechanical equipment condition monitoring, structural health diagnosis, and safety protection, the accurate detection and identification of non-stationary transient impact characteristics in vibration signals is a crucial fundamental technology. These characteristics typically correspond to early failures of mechanical components (such as bearing pitting or gear tooth breakage), transient impacts to the structure, or illegal acts of sabotage (such as hammering or blasting). Early and reliable identification of these characteristics is essential for predictive maintenance, ensuring structural safety, and providing timely warnings of destructive activities. Existing technologies have proposed various methods for detecting non-stationary transient impact characteristics in vibration signals, but each has its limitations. The most direct method is to observe the time-domain waveform of the vibration signal and, by setting an amplitude threshold, consider pulse points exceeding that threshold as impacts. This method is simple to implement, but it suffers from poor anti-interference capability or insufficient robustness. Any random noise spike or accidentally generated large-amplitude signals during normal operation (such as equipment start-up and shutdown) may be misjudged as impacts, leading to an excessively high false alarm rate and making it unreliable in complex noise environments.

[0003] To improve anti-interference capabilities, another common method is to calculate the signal's envelope spectrum. This method first extracts the signal envelope using techniques such as Hilbert transform to amplify the amplitude modulation phenomenon caused by the impact. Then, it performs a Fourier transform on the envelope to obtain the envelope spectrum, and identifies potential periodic impact characteristics by searching for spectral peaks in the envelope spectrum. However, in complex industrial sites or natural environments, background vibrations often contain abundant harmonic components, which generate inherent spectral peaks in the envelope spectrum. These impact-independent peaks severely interfere with the judgment of true impact characteristics and are also prone to false alarms. This method is also not ideal for detecting non-periodic or weakly periodic impacts.

[0004] To further focus on the frequency bands where impacts may occur, some improved schemes introduce bandpass filtering, perform envelope analysis within specific frequency bands, or employ time-frequency analysis methods such as wavelet transform and empirical mode decomposition to extract impact components from the signal. These methods improve frequency focusing capabilities to some extent, but often face problems such as complex parameter selection, high computational cost, and strong algorithmic adaptability. For example, the selection of wavelet basis, the setting of the number of decomposition levels, and mode aliasing in empirical mode decomposition all require extensive prior knowledge and experience, and their adaptability to different types and intensities of impact characteristics is weak. Ultimately, these methods still rely on setting empirical thresholds for judging the extracted components, and their overall robustness is not fundamentally improved.

[0005] In summary, the main shortcomings of existing technical solutions lie in their simplification of non-stationary transient impact detection into a static peak detection problem. They fail to fully consider the complex relationships between the frequency domain distribution characteristics, significance relative to local background noise, and possible temporal occurrence patterns of non-stationary transient impacts generated by actual destructive behavior or faults. A single amplitude threshold cannot distinguish between impact and noise spikes; traditional envelope spectrum analysis struggles to suppress harmonic interference from stationary background vibrations; and complex time-frequency analysis methods suffer from parameter sensitivity and computational efficiency bottlenecks. This results in existing methods failing to achieve both high sensitivity and specificity in detecting non-stationary transient impact characteristics when facing varying environmental noise and various vibration interferences in real-world monitoring scenarios. They either miss weak, real impacts or generate numerous false alarms, failing to meet the application requirements of high-reliability monitoring systems. Therefore, there is an urgent need for an analytical method that can more precisely and robustly identify non-stationary transient impact characteristics from complex background vibrations. Summary of the Invention

[0006] To address the aforementioned problems, this invention provides a method for monitoring non-stationary instantaneous impacts. The invention first acquires and frames the vibration signal; then, it calculates the envelope signal and its frequency domain envelope spectrum for each passband using multi-passband filtering and Hilbert transform; it calculates the relative envelope spectrum probability curve and identifies local maxima pulses within the 0-200Hz analysis frequency band by calculating their second-order differences; it calculates the ratio of each pulse to the average value within the minimum pulse interval to its left and right sides for significance assessment, and combines this with multi-level threshold statistical testing to ultimately determine whether non-stationary instantaneous impact damage exists. This invention can extract and identify non-stationary instantaneous impact characteristics with high accuracy and high reliability from strong noise backgrounds.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is: a non-stationary instantaneous impact monitoring method, comprising the following steps: S1: acquiring vibration signals x ( t ), regarding the vibration data x ( tThe vibration signal of each frame is obtained by performing frame segmentation processing. ; S2: Each frame of vibration signal Signals with multiple passbands are obtained by filtering using multiple parallel bandpass filters. ,in p It is the passband number; S3: Signal for each passband The envelope signal is calculated using the Hilbert transform. The envelope signal is then converted to the frequency domain to obtain the envelope spectrum. ; S4: Envelope spectrum for each passband Calculate its relative envelope spectrum probability curve The relative envelope spectrum probability curve is obtained by selecting 0-200Hz as the analysis frequency band. for: in, ω For frequency; S5: Within the analysis frequency band, the relative envelope spectrum probability curve is calculated. Second-order difference To identify local maximum pulses, when When a local maximum is obtained. and its location information ; S6: Based on the location information Local maxima Sort in ascending order and calculate local maxima. Interval with the minimum pulse on the left span Average value within The ratio R1, and its difference from the minimum pulse interval on the right. span Average value within The ratio R2; S7: If the ratios R1 and R2 are greater than or equal to the threshold... thr_R If the condition is met, count the number of pulses N1 and N2; otherwise, the current frame does not meet the condition. S8: If the number of pulses N1 and N2 is greater than or equal to the threshold... thr_N, Then calculate the number of frames that meet the conditions. M Otherwise, the current frame does not meet the condition; if the number of frames M Greater than or equal to the threshold thr_M If a signal is detected, it is marked that the current frame signal has a non-stationary instantaneous impact; otherwise, it is not marked.

[0008] Preferably, the vibration signal of each frame for: in, t For time; l For the current frame; L Total number of frames; For the current frame l The Hanning window function; Len The frame length.

[0009] Preferably, in step S2, the bandpass filter has a passband size of 2kHz and a passband range of 0-8kHz; the passband number... p The value is 4.

[0010] Preferably, the envelope signal for: in, t For time; l For the current frame; p It is the passband number; v It is the integral variable.

[0011] Preferably, the envelope spectrum for: in, T For the entire time domain; ω For frequency; t For time; l For the current frame; p It is the passband number; Fourier transform; These are the signals obtained in each passband after filtering; j It is the imaginary unit.

[0012] Preferably, the second-order difference for: in, ω Δ represents the frequency; Δ represents the difference operator. l For the current frame; p It is the passband number; This is the relative envelope spectrum probability curve.

[0013] Preferably, the formulas for the ratios R1 and R2 are as follows: in, It is a local maximum; The interval between the local maximum and the left minimum pulse. span The average value within; The interval between the local maximum and the right minimum pulse. span The average value within the range.

[0014] Preferably, the minimum pulse interval span= 3Hz.

[0015] Preferably, the threshold thr_R The value range is (0, 100); threshold thr_N The value range is (0, 100); the threshold value... thr_M The range of values ​​for is (0, ...). L ),in L This represents the total number of frames after the vibration signal is framed.

[0016] A non-stationary instantaneous impact monitoring system includes: a data acquisition module for acquiring vibration signals. x ( t ); The front-end signal processing module is communicatively connected to the data acquisition module and is used to process the acquired vibration signals. x ( t The above-described non-stationary instantaneous impact monitoring method is executed, and a preliminary judgment result indicating the presence of potential non-stationary instantaneous impact characteristics is output. A platform state assessment module, communicatively connected to the front-end signal processing module, receives the preliminary judgment result and performs statistical verification on the received preliminary judgment result in a time series. The statistical verification includes: within a preset time window, counting the number of frames output by the front-end signal processing module marked as having potential non-stationary instantaneous impact characteristics. M Statistical analysis was performed when the number of frames... M Greater than or equal to the threshold thr_M When a sustained, non-stationary instantaneous impact damage is detected, the system determines that such damage has occurred. The image feedback module, which is communicatively connected to the platform status assessment module, is used to collect and upload on-site images according to instructions issued by the platform status assessment module when the sustained, non-stationary instantaneous impact damage is detected.

[0017] By adopting the above technical solution, the present invention has the following beneficial effects.

[0018] (1) This invention constructs a robust non-stationary instantaneous impact feature enhancement channel by combining multi-passband parallel processing with relative envelope spectrum probability curve analysis. This invention frames the vibration signal and passes it in parallel through multiple bandpass filters, achieving focused detection of different center frequency bands that may exist in the impact response, avoiding the obscuring of impact features during broadband analysis. Subsequently, a Hilbert transform is performed on the signal in each passband to obtain the envelope, and the relative probability distribution curve of its envelope spectrum is calculated, transforming the analysis object from the absolute amplitude of the envelope spectrum to the normalized spectral probability. This effectively suppresses the influence of overall signal intensity differences caused by factors such as sensor gain variations and different impact source distances, making the feature curves of impact events of different intensities comparable, significantly improving the stability and consistency of impact feature expression, and laying the foundation for subsequent accurate identification.

[0019] (2) This invention introduces a mathematical method based on second-order differences to accurately locate the local maxima pulses of the relative envelope spectrum probability curve within the 0-200Hz analysis frequency band. This method calculates the second-order difference of the probability curve and utilizes... The present invention uses certain conditions to determine and locate extreme points. Compared with traditional peak-finding methods that search for the first derivative crossing zero or simply compare the amplitude of neighboring regions, this invention provides more rigorous and accurate mathematical criteria, effectively avoiding technical problems such as misjudgment or ambiguous location of extreme points caused by plateaus, spikes, or noise fluctuations in the curve. The local maxima and their frequency location information obtained in this way are also provided. It has sub-resolution accuracy and higher reliability, ensuring that the extracted candidate pulses are true spectral peaks, providing high-quality, high-confidence input data for subsequent significance discrimination.

[0020] (3) This invention uses the pulse significance criterion based on the ratio of the pulse to the average values ​​of its left and right neighboring regions (R1 and R2). For each precisely located local maximum pulse, the ratio of its amplitude to the average value of all values ​​within the left minimum pulse interval span (R1) and the ratio to the corresponding average value on the right (R2) are calculated. A true non-stationary instantaneous impact pulse appears as an "isolated" spike in the envelope spectrum, and its amplitude should be significantly higher than the average level of its left and right neighboring frequency points (i.e., larger R1 and R2 values). In contrast, harmonic peaks or random noise spikes generated by stationary vibrations have relatively lower prominence relative to their neighborhood. By setting a threshold t... hr_R Judging this ratio can effectively filter out real impact pulses with significant local contrast from numerous spectral peaks, greatly suppressing the interference of background harmonics and random noise.

[0021] (4) The present invention determines the significance of a single pulse based on the R1 / R2 criterion; secondly, it counts the number of pulses (N1, N2) that satisfy the significance condition within a single passband and compares them with a threshold. thr_N The comparison was used to confirm the clustering of impact features in the frequency space; finally, the number of frames that met the aforementioned conditions was counted across multiple consecutive frames. M and with threshold thr_M The comparison confirms the duration of the impact event. A single significant pulse may originate from sporadic interference (Level 1 judgment), but it is unlikely to occur simultaneously at multiple frequency points (Level 2 judgment), and even less likely to persist across multiple frames of data (Level 3 judgment). This invention employs a three-level, multi-dimensional decision mechanism, ensuring that the system only makes a final affirmative judgment on events that meet the characteristics of impact behavior in terms of amplitude significance, frequency clustering, and duration. This provides strong resistance to sporadic interference and complex background noise, ensuring the reliability and robustness of the output results.

[0022] (5) This invention solves the technical problems of existing non-stationary instantaneous impact detection methods being susceptible to interference from background harmonics and random noise, and having poor robustness and high false alarm rate due to the single reliance on amplitude threshold for pulse discrimination. It adopts multi-passband parallel filtering, Hilbert transform to obtain the envelope spectrum, calculates the relative envelope spectrum probability curve, and uses second-order difference to accurately locate local maximum pulses. It also introduces a significance judgment criterion based on the ratio of the pulse to the average value of its left and right neighbors, combined with multi-level threshold statistical testing. It can extract and confirm the real non-stationary instantaneous impact features with extremely high accuracy and high reliability from strong noise background, providing core and robust algorithmic support for the identification of destructive behavior. Attached Figure Description

[0023] The following provides a detailed discussion of the manufacture and application of preferred embodiments of the present invention. However, it should be understood that the present invention provides many applicable inventive concepts that can be embodied in various specific environments. The specific embodiments discussed are merely illustrative of specific ways of manufacturing and using the present invention and do not limit the scope of the invention. For those skilled in the art, other drawings can be obtained from these drawings without any creative effort.

[0024] Figure 1 This is a flowchart of the present invention.

[0025] Figure 2 This is a time-domain information, spectrum, and relative envelope spectrum probability curve of the normal signal of this invention.

[0026] Figure 3 This is a time-domain information, spectrum, and relative envelope spectrum probability curve of the actual damage signal of the present invention.

[0027] Figure 4This chart compares the probability of alarms and the probability of missed alarms in normal monitoring and experimental data destruction. Detailed Implementation

[0028] The following provides a detailed discussion of the manufacture and application of preferred embodiments of the present invention. However, it should be understood that the present invention provides many applicable inventive concepts that can be embodied in various specific environments. The specific embodiments discussed are merely illustrative of specific ways of manufacturing and using the invention and do not limit the scope of the invention.

[0029] like Figure 1 The non-stationary instantaneous impact monitoring method shown includes the following steps: Step 1: Acquire vibration signal x ( t ), regarding the vibration data x ( t The vibration signal of each frame is obtained by performing frame segmentation processing. The vibration signal of each frame for: in, t For time; l For the current frame; L Total number of frames; For the current frame l The Hanning window function; Len The frame length.

[0030] Step 2: Extract each frame of vibration signal Signals with multiple passbands are obtained by filtering using multiple parallel bandpass filters. ,in p The passband number is the number of bandwidths. The passband size of the bandpass filter is 2kHz, and the passband range is 0-8kHz; the passband number is... p The envelope signal is 4. for: in, t For time; l For the current frame; p It is the passband number; v It is the integral variable.

[0031] Step 3: For the signal in each passband The envelope signal is calculated using the Hilbert transform. The envelope signal is then converted to the frequency domain to obtain the envelope spectrum. The envelope spectrum for: in, TFor the entire time domain; ω For frequency; t For time; l For the current frame; p It is the passband number; Fourier transform; These are the signals obtained in each passband after filtering; j It is the imaginary unit.

[0032] Step 4: Envelope spectrum for each passband Calculate its relative envelope spectrum probability curve The relative envelope spectrum probability curve is obtained by selecting 0-200Hz as the analysis frequency band. for: in, ω For frequency; Step 5: Identify the local maxima pulses in the curve within the analysis frequency band. This is done by calculating the relative envelope spectrum probability curve. Second-order difference To identify local maximum pulses, when When a local maximum is obtained. and its location information The second-order difference for: in, ω Δ represents the frequency; Δ represents the difference operator. l For the current frame; p It is the passband number; This is the relative envelope spectrum probability curve.

[0033] Step 6: Based on the location information Local maxima Sort in ascending order and calculate local maxima. Interval with the minimum pulse on the left span Average value within The ratio R1, and its difference from the minimum pulse interval on the right. span Average value within The ratio R2; the minimum pulse interval span= 3Hz. The formulas for the ratios R1 and R2 are: in, It is a local maximum; The interval between the local maximum and the left minimum pulse. span The average value within; The interval between the local maximum and the right minimum pulse. span The average value within the range.

[0034] Step 7: If the ratios R1 and R2 are greater than or equal to the threshold... thr_R If the condition is met, the number of pulses N1 and N2 that meet the condition are counted; otherwise, the current frame does not meet the condition. In this embodiment, the threshold... thr_R The value range is (0, 100).

[0035] Step 8: If the number of pulses N1 and N2 is greater than or equal to the threshold... thr_N, Then calculate the number of frames that meet the conditions. M Otherwise, the current frame does not meet the condition; in this embodiment, the threshold thr_N The value range is (0, 100). If the number of frames... M Greater than or equal to the threshold thr_M If the threshold is set, the current frame signal is marked as having a non-stationary instantaneous impact disruption; otherwise, it is not marked. In this embodiment, the threshold... thr_M The range of values ​​for is (0, ...). L ),in L This represents the total number of frames after the vibration signal is framed.

[0036] Furthermore, this embodiment also provides a non-stationary instantaneous impact monitoring system, including: a data acquisition module for acquiring vibration signals. x ( t The front-end signal processing module, which is communicatively connected to the data acquisition module, is used to process the acquired vibration signals. x ( t The above-described non-stationary instantaneous impact monitoring method is executed, and a preliminary judgment result indicating whether potential non-stationary instantaneous impact characteristics exist is output.

[0037] The platform status assessment module, communicatively connected to the front-end signal processing module, is used to receive the preliminary judgment result and perform statistical verification on the preliminary judgment result received in a time series. The statistical verification includes: within a preset time window, counting the number of frames output by the front-end signal processing module that are marked as having potential non-stationary instantaneous impact characteristics. M Statistical analysis was performed when the number of frames... M Greater than or equal to the threshold thr_M When a sustained, non-stationary instantaneous impact damage is detected, the system determines that such damage has occurred. The image feedback module, which is communicatively connected to the platform status assessment module, is used to collect and upload on-site images according to instructions issued by the platform status assessment module when the sustained, non-stationary instantaneous impact damage is detected.

[0038] The following is in conjunction with the appendix Figure 1-4Further explanation is provided. In a specific embodiment of the present invention, to verify the effectiveness of a non-stationary instantaneous impact monitoring method, it is applied to the early fault diagnosis of rotating machinery bearings. Vibration signal x ( t The sampling frequency is 16kHz. This implementation details the entire process from signal processing to decision output, and highlights the collaboration and effectiveness of various technical features through comparative experiments.

[0039] First, regarding vibration signals x ( t Frame division, frame length Len The signal has 4096 points, using a Hanning window. Each frame of signal... The signals in each passband were obtained by passing the signals through four bandpass filters with passbands of 0-2kHz, 2-4kHz, 4-6kHz, and 6-8kHz, respectively. .right The envelope signal is obtained by performing a Hilbert transform. And calculate its envelope spectrum. The role of "multi-passband filtering" here is frequency separation. It physically isolates the high-frequency resonant band (such as 2-4kHz) that may contain fault impulses from the strong low-frequency gear meshing components (concentrated in 0-1kHz), providing an initial, noise-suppressed signal subspace for subsequent analysis.

[0040] Next, the relative envelope spectral probability curve for each passband is calculated. : in, ω The frequency was set as the threshold, and the low-frequency band of 0-200Hz was selected for analysis. The "relative envelope spectrum probability curve" played a crucial normalization role here. It eliminated the direct impact of overall signal amplitude fluctuations caused by load changes on the absolute height of the spectral line, making the same fault impact characteristics under different operating conditions exhibit similar relative heights on the probability curve, thus improving the comparability and stability of the characteristics.

[0041] Within the 0-200Hz analysis frequency band, to accurately locate the fault characteristic frequency and the pulses at its harmonics, the second-order difference of the probability curve is calculated. .when When a local maximum is found at a given frequency point, its location is recorded. and local maxima The probability curve provides a smooth, normalized analytical object, and the second-order difference rule, by mathematically differentiating this object, can accurately find the position of all "peaks" on the curve. Even if there are noise points with similar amplitudes next to the peak, it will not be misjudged. Its positioning accuracy and resistance to small disturbances are significantly better than simple first-order difference or global maximum methods.

[0042] Next, we proceed to the core pulse significance determination. This involves setting the minimum pulse interval. span =3Hz, for each located pulse, calculate its frequency relative to the left side. span Inner average The ratio R1, and the ratio to the right side span Inner average The ratio R². Set a threshold. thr_R =5. In test signals containing strong gear harmonics, there are also spectral peaks at the harmonic frequencies, but the height difference between them and adjacent harmonic peaks is not large (i.e., R1 and R2 values ​​are small). However, the impact pulse at the bearing inner ring fault characteristic frequency (123Hz), due to its non-harmonic characteristics, appears as an "isolated" spike in the spectrum, with an amplitude significantly higher than the background average within a 3Hz range to the left and right, thus generating large R1 and R2 values. Experiments show that using only this ratio judgment can reduce the number of false alarm pulses caused by harmonic interference by 80%. Precise positioning provides an accurate pulse center frequency, thereby ensuring the accuracy of the calculation range of the average values ​​of the left and right neighborhoods, excluding the pulse itself, allowing the ratio judgment to be implemented correctly.

[0043] Then, multi-level statistical tests were performed. Within a single passband, both the statistical R1 and R2 values ​​were greater than [value missing]. thr_R The number of pulses N1 and N2 are given, and N1 ≥ 2 and N2 ≥ 2 (i.e., thr_N =2). This level utilizes the characteristic that fault impulses often occur in pairs or multiples at characteristic frequencies and their harmonics. Finally, in the continuously analyzed frames, the number of frames that satisfy the above conditions is counted. M ,when M ≥5 thr_M When the ratio is 5, a persistent impact is ultimately determined. A random noise spike might be determined by chance through the ratio, but it is extremely difficult to generate multiple pulses that meet the conditions within the same passband of the same frame (N1 / N2 test), let alone for them to appear continuously across multiple frames. M test).

[0044] Test results on an early bearing failure dataset show that the method of this invention achieves a detection rate of 96% for minor impacts. In contrast, traditional envelope spectrum analysis combined with a simple amplitude thresholding method, while achieving a detection rate of 90%, has a false alarm rate as high as 15%; while the false alarm rate of the method of this invention is only 1.2%. This reduction of more than 90% in the false alarm rate is achieved by the ratio judgment mechanism suppressing the interference of stationary background harmonics, while the statistical test filters out occasional random noise from a probabilistic perspective.

[0045] Figure 2 This provides the time-domain information, spectrum, and relative envelope spectrum probability curve of a normal signal. Figure 3The two graphs show the time-domain information, spectrum, and relative envelope spectrum probability curve of the actual disruptive signal. It can be seen from the graphs that the normal signal does not exhibit regular impulses in the time domain and its spectrum is a wideband signal. The disruptive signal, on the other hand, shows pulsed frequency peaks in its spectrum and equally spaced disruptive peaks in its envelope spectrum, which are absent in the normal signal. Algorithmic identification of these characteristics yields the probability of detecting disruption and triggering an alarm, as shown below. Figure 4 As shown, the system can identify 10 destructive data points in the destructive experiment with 100% accuracy and no missed reports. For the 585 data points in the normal monitoring, there are no missed reports, but the detection of 2 destructive data points and the subsequent alarms are false alarms, which are within an acceptable range and greatly reduce the false alarm rate of the system.

[0046] Although the specification has provided a detailed description, it should be understood that various changes, substitutions, and modifications can be made without departing from the spirit and scope of the invention as defined by the appended claims. Furthermore, the specific embodiments described are not intended to limit the scope of the invention, and those skilled in the art will readily understand based on this invention that existing or future-developed processes, machines, manufactures, compositions of matter, means, methods, or steps can perform substantially the same functions or achieve substantially the same results as the embodiments of the invention. Therefore, the appended claims are intended to include such processes, machines, manufactures, compositions of matter, means, methods, or steps within their scope.

Claims

1. A method for monitoring non-stationary instantaneous impacts, characterized in that: Includes the following steps: S1: Acquire vibration signal x ( t ), regarding the vibration data x ( t The vibration signal of each frame is obtained by performing frame segmentation processing. ; S2: Each frame of vibration signal Signals with multiple passbands are obtained by filtering using multiple parallel bandpass filters. ,in p It is the passband number; S3: Signal for each passband The envelope signal is calculated using the Hilbert transform. The envelope signal is then converted to the frequency domain to obtain the envelope spectrum. ; S4: Envelope spectrum for each passband Calculate its relative envelope spectrum probability curve The relative envelope spectrum probability curve is obtained by selecting 0-200Hz as the analysis frequency band. for: in, ω For frequency; S5: Within the analysis frequency band, the relative envelope spectrum probability curve is calculated. Second-order difference To identify local maximum pulses, when When a local maximum is obtained. and its location information ; S6: Based on the location information Local maxima Sort in ascending order and calculate local maxima. Interval with the minimum pulse on the left span Average value within The ratio R1, and its difference from the minimum pulse interval on the right. span Average value within The ratio R2; S7: If the ratios R1 and R2 are greater than or equal to the threshold... thr_R If the condition is met, count the number of pulses N1 and N2; otherwise, the current frame does not meet the condition. S8: If the number of pulses N1 and N2 is greater than or equal to the threshold... thr_N, Then calculate the number of frames that meet the conditions. M Otherwise, the current frame does not meet the condition; if the number of frames M Greater than or equal to the threshold thr_M If a signal is detected, it is marked that the current frame signal has a non-stationary instantaneous impact; otherwise, it is not marked.

2. The non-stationary instantaneous impact monitoring method as described in claim 1, characterized in that: The vibration signal of each frame for: in, t For time; l For the current frame; L Total number of frames; For the current frame l The Hanning window function; Len The frame length.

3. The non-stationary instantaneous impact monitoring method as described in claim 1, characterized in that: In step S2, the passband size of the bandpass filter is 2kHz, and the passband range is 0-8kHz; the passband number... p The value is 4.

4. The non-stationary instantaneous impact monitoring method as described in claim 1, characterized in that: The envelope signal for: in, t For time; l For the current frame; p It is the passband number; v It is the integral variable.

5. The non-stationary instantaneous impact monitoring method as described in claim 1, characterized in that: The envelope spectrum for: in, T For the entire time domain; ω For frequency; t For time; l For the current frame; p It is the passband number; Fourier transform; These are the signals obtained in each passband after filtering; j It is the imaginary unit.

6. The non-stationary instantaneous impact monitoring method as described in claim 1, characterized in that: The second-order difference for: in, ω Δ represents the frequency; Δ represents the difference operator. l For the current frame; p It is the passband number; This is the relative envelope spectrum probability curve.

7. The non-stationary instantaneous impact monitoring method as described in claim 1, characterized in that: The formulas for the ratios R1 and R2 are as follows: in, It is a local maximum; The interval between the local maximum and the left minimum pulse. span The average value within; The interval between the local maximum and the right minimum pulse. span The average value within the range.

8. The non-stationary instantaneous impact monitoring method as described in claim 1, characterized in that: The minimum pulse interval span= 3Hz.

9. The non-stationary instantaneous impact monitoring method as described in claim 1, characterized in that: The threshold thr_R The value range is (0, 100); threshold thr_N The value range of is (0, 100); the threshold value thr_M The range of values ​​for is (0, ...). L ),in L This represents the total number of frames after the vibration signal is framed.

10. A non-stationary instantaneous impact monitoring system, characterized in that: include: The data acquisition module is used to collect vibration signals. x ( t ); The front-end signal processing module is communicatively connected to the data acquisition module and is used to process the acquired vibration signals. x ( t The non-stationary instantaneous impact monitoring method as described in any one of claims 1-9 is executed, and a preliminary judgment result indicating whether there are potential non-stationary instantaneous impact characteristics is output. The platform status assessment module, communicatively connected to the front-end signal processing module, is used to receive the preliminary judgment result and perform statistical verification on the preliminary judgment result received in a time series. The statistical verification includes: within a preset time window, counting the number of frames output by the front-end signal processing module that are marked as having potential non-stationary instantaneous impact characteristics. M Statistical analysis was performed when the number of frames... M Greater than or equal to the threshold thr_M When this occurs, it is determined that a sustained, non-stationary, instantaneous impact has occurred; The image feedback module is communicatively connected to the platform status assessment module and is used to collect and upload on-site images according to the instructions issued by the platform status assessment module when the platform status assessment module determines that the continuous non-stationary instantaneous impact damage has occurred.

Images