Structure shock location method based on kurtosis time series and multi-feature image fusion

Abstract

The application discloses a structure impact positioning method based on kurtosis time series and multi-feature image fusion, and belongs to the technical field of structure health monitoring. The method calculates the kurtosis time series of an impact stress wave signal, extracts three features of a wave arrival time difference, a signal amplitude and a waveform correlation coefficient of the kurtosis time series, respectively uses the wave arrival time difference, the signal amplitude and the waveform correlation coefficient to perform imaging by using a hyperbolic curve principle, a delay accumulation principle and a similarity principle, and finally realizes impact positioning by normalizing and multiplying three images. The impact positioning method does not depend on a narrowband wave model, does not need to use complex signal processing means and interpret a signal mode, extracts multiple features for positioning, has high imaging resolution and positioning precision, and has a good application prospect in impact positioning of a plate-shaped structure.

Description

Technical Field

[0001] This invention belongs to the field of structural health monitoring, specifically involving a structural impact localization method based on kurtosis time series and multi-feature image fusion. Background Technology

[0002] Panel structures are crucial components in engineering equipment, widely used in aerospace, marine engineering, and new energy fields. In-service panel structures inevitably suffer damage from external impacts, affecting their reliability and integrity. Regularly inspecting structures using traditional non-destructive testing methods such as ultrasonic scanning, thermal imaging, and eddy current testing is both uneconomical and carries the risk of missed detections. Impact monitoring technology can monitor impact loads in real-time, providing guidance for the health and safety maintenance of panel structures. Therefore, research on impact monitoring technology is currently an important topic in the field of structural health monitoring.

[0003] The first step in impact monitoring is impact source localization. Piezoelectric, strain, and acceleration sensors can capture stress wave signals generated by the impact. By extracting characteristic information from the stress wave signal through signal processing and combining it with specific algorithms, the location of the impact source can be predicted. Currently, time-difference-based impact location methods, including triangulation, four-point circular arc, and hyperbolic positioning, require precise measurement of the stress wave's arrival time. However, due to boundary reflections and noise, obtaining accurate arrival times is difficult, making accurate impact location identification challenging. Data-driven structural impact location methods require large databases for large engineering equipment, consuming significant computer storage resources and presenting difficulties in engineering applications. Impact location methods based on phase synthesis, time-reversal focusing, multiple signal classification, and wavenumber filtering all require complex signal decomposition techniques to obtain narrowband wave signals. Then, they use amplitude or phase information of specific signal patterns for location. However, the difficulty in interpreting signal patterns and the insufficient utilization of single feature information lead to large impact location errors. Summary of the Invention

[0004] To address the aforementioned issues, this invention proposes a structural impact localization method based on kurtosis time series and multi-feature image fusion, which reduces impact localization error and improves impact localization accuracy.

[0005] The technical solution of the present invention is as follows:

[0006] A structural impact localization method based on kurtosis time series and multi-feature image fusion includes the following steps:

[0007] Step 1: Mark the position coordinates of the sensor and record the impact stress wave signal;

[0008] Step 2: Calculate the kurtosis time series and normalize it;

[0009] Step 3: Calculate the relative time delay of the impact stress wave signals from different sensors;

[0010] Step 4: Multi-feature impact imaging;

[0011] Step 5: Image fusion to predict impact location.

[0012] Further, in step 1, the position coordinates of all sensors receiving impact stress wave signals within the monitoring area on the marked structure are recorded; for the monitoring area of ​​the structure with M piezoelectric sensors, a rectangular coordinate system is established, and the sensor P is... m The position coordinates are (x m ,y m The impact stress wave signal received by the sensor is denoted as S. m (t), m=1,2,…,M.

[0013] Furthermore, the sensors include piezoelectric sensors, strain sensors, and acceleration sensors.

[0014] Further, in step 2, the kurtosis time series of the impact stress wave signal is calculated based on the impact stress wave signal window; specifically, the kurtosis time series of the impact stress wave signal received by each sensor is calculated according to the following formula:

[0015]

[0016] Where L is the total number of sampling points for the impact stress wave signal, and t i Let i be the time corresponding to the i-th sampling point of the signal. The impact stress wave signal received by the m-th sensor within the time window (0 <t<t l The average value within ); k m (t l ) is the impact stress wave signal received by the m-th sensor within the time window (0 <t<t l kurtosis value within )

[0017] The kurtosis time series of the impact stress wave signals received by each sensor are normalized:

[0018]

[0019] Where max(k) m (t) is k m (t) in the time window (0 <t<t L The maximum value within ).

[0020] Furthermore, in step 3, assuming any point within the monitoring area is the location of the impact source, the relative time delay of the impact stress wave signals between different sensors is calculated based on the distance from the sensor to the assumed impact source location and the wave velocity of the impact stress wave; the specific process is as follows:

[0021] The wave velocity C of the impact stress wave is measured using a known impact and a sensor in the same direction:

[0022]

[0023] Where Δl is the distance between sensors in the same direction, and Δt is the time difference of arrival; the time difference of arrival is measured by the first peak of the normalized kurtosis time series of the impact stress wave signal.

[0024] Assuming that any point (x, y) in the structural monitoring area is the location of the impact source, and its distance to the m-th sensor is L. m (x,y):

[0025]

[0026] Calculate the distance difference D between different sensors and the assumed impact source location. mn (x,y):

[0027] D mn (x,y)=L n (x,y)-L m (x,y)m,n=1,2,..,M (5)

[0028] Where n represents the nth sensor;

[0029] Given the assumed location of the impact source, calculate the relative time delay τ of the impact stress wave signal from different sensors. mn (x,y):

[0030]

[0031] Furthermore, in step 4, three features of the kurtosis time series are extracted: time difference of arrival, signal amplitude, and waveform correlation coefficient. These features are then used to perform impact imaging based on different principles. The specific process is as follows:

[0032] The Time of Arrival (TOA) of two different sensors was extracted based on the first peak of the normalized kurtosis time series. m and TOA n The time difference between arrival and arrival is τ' mn (x,y):

[0033] τ' mn (x,y)=TOAn -TOA m m,n=1,2,...,M (7)

[0034] Shock imaging 1 is obtained using the time-of-arrival (TOA) characteristic and the hyperbolic principle. The result of shock imaging 1 is represented as I1:

[0035]

[0036] Where σ is the time delay factor;

[0037] The signal amplitude and characteristics of the normalized kurtosis time series are extracted based on the relative time delay of the shock stress wave signal:

[0038]

[0039] Where, τ m (x,y) represents the relative time delay of the m-th sensor relative to the 1st sensor; T1 is the end time of the window of the normalized kurtosis time series of the shock stress wave signal received by the 1st sensor. The normalized kurtosis time series of the shock stress wave signals received by the 1st to Mth sensors in the time interval [τ] m (x,y),T1+τ m The sum of signal amplitudes within [x,y];

[0040] Impact imaging 2 is obtained by using the signal amplitude and characteristics according to the principle of delay accumulation. The result of impact imaging 2 is represented as I2:

[0041] I2(x,y)=max(amp(x,y)) (10)

[0042] Where max(amp(x,y)) is the normalized kurtosis time series of the impact stress wave signals received by the 1st to Mth sensors in the time interval [τ]. m (x,y),T1+τ m The maximum value of the sum of signal amplitudes within the range (x,y);

[0043] Based on the relative time delay of the shock stress wave signal, the waveform correlation coefficients of the normalized kurtosis time series from different sensors are extracted:

[0044]

[0045] Among them, T m Let K be the termination time of the window for the normalized kurtosis time series of the shock stress wave signal received by the m-th sensor; cov(K) m (t),K n(t) represents the normalized kurtosis time series of the impact stress wave signals received by the m-th and n-th sensors in the time interval [τ]. mn (x,y),T m +τ mn Covariance within [x,y]; and The normalized kurtosis time series of the shock stress wave signals received by the m-th and n-th sensors in the time interval [τ] mn (x,y),T m +τ mn The variance within [x,y];

[0046] Impact imaging 3 is obtained by utilizing waveform correlation coefficient characteristics and based on the waveform similarity principle. The result of impact imaging 3 is represented as I3:

[0047]

[0048] Furthermore, in step 5, the impact imaging results with different features are multiplied and fused; the brightest position in the fused image represents the predicted impact location; the specific process is as follows:

[0049] Normalize the impact imaging results I1, I2, and I3:

[0050]

[0051] Perform cumulative multiplication and fusion on the normalized imaging results:

[0052]

[0053] The brightest position in the fused image represents the predicted impact location.

[0054] The beneficial technical effects of this invention are as follows:

[0055] (1) Impact location is achieved by using the kurtosis time series of the impact stress wave signal. It is not limited to the narrowband wave signal model and does not require signal processing methods to decompose the impact stress wave signal and interpret its pattern.

[0056] (2) Extract various features such as time difference of arrival, signal amplitude and waveform correlation coefficient of kurtosis time series, use various features to perform impact imaging and image fusion to achieve impact positioning, improve the information utilization rate of sparse observation system, and thus improve imaging resolution and impact positioning accuracy.

[0057] (3) This method is different from the data-driven method based on statistical domain features such as kurtosis. It does not require sensor calibration or reference database. The implementation process is very simple, and the positioning algorithm is conducive to the integration of impact monitoring equipment. Attached Figure Description

[0058] Figure 1 This is a flowchart of the structural impact localization method based on kurtosis time series and multi-feature image fusion according to the present invention;

[0059] Figure 2 This is a schematic diagram illustrating the impact positioning principle of the present invention;

[0060] Figure 3 This is a schematic diagram of a typical impact stress wave signal in an embodiment of the present invention;

[0061] Figure 4 This is a schematic diagram of the kurtosis time series of a typical impact stress wave signal in an embodiment of the present invention;

[0062] Figure 5 The following are comparison diagrams of imaging results in embodiments of the present invention: (a) is a schematic diagram of the impact imaging result based on the hyperbolic principle using wave arrival time difference features; (b) is a schematic diagram of the impact imaging result based on the delay accumulation principle using amplitude and features; (c) is a schematic diagram of the impact imaging result based on the similarity principle using waveform similarity features; and (d) is a schematic diagram of the multi-feature image fusion result. Detailed Implementation

[0063] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0064] like Figure 1 As shown, a structural impact localization method based on kurtosis time series and multi-feature image fusion specifically includes the following steps:

[0065] Step 1: Mark the position coordinates of the sensor and record the impact stress wave signal.

[0066] The location coordinates of all sensors (including piezoelectric sensors, strain sensors, and accelerometers) receiving impact stress wave signals within the monitoring area on the marked structure are recorded. For example... Figure 2 As shown, for the structural monitoring area where M piezoelectric sensors are arranged, a rectangular coordinate system is established. In the figure, P1, P2, P3, P4, P5, P6, P7, P8, P9, P1, P1, P1, P1, P2, P1, P2, P1, P2, P1, P2, P2, P3, P4, P1, P2 ... m P n P M These represent sensors at different locations, L1, L... m L n L M These represent the impact stress wave signals received by different sensors. Sensor P is used to represent these signals.m The position coordinates are (x m ,y m The impact stress wave signal received by the sensor is denoted as S. m (t), m=1,2,…,M.

[0067] Step 2: Calculate the kurtosis time series and normalize it.

[0068] The kurtosis time series of the impact stress wave signal is calculated based on the impact stress wave signal window. Specifically, the kurtosis time series of the impact stress wave signal received by each sensor is calculated according to the following formula:

[0069]

[0070] Where L is the total number of sampling points for the impact stress wave signal, and t i Let i be the time corresponding to the i-th sampling point of the signal. The impact stress wave signal received by the m-th sensor within the time window (0 <t<t l The average value within (k). m (t l ) is the impact stress wave signal received by the m-th sensor within the time window (0 <t<t l The kurtosis value within ).

[0071] The kurtosis time series of the impact stress wave signals received by each sensor are normalized:

[0072]

[0073] Where max(k) m (t) is k m (t) in the time window (0 <t<t L The maximum value within ).

[0074] Step 3: Calculate the relative time delay of the impact stress wave signals from different sensors.

[0075] Assuming any point within the monitoring area is the impact source location, the relative time delay of the impact stress wave signals between different sensors is calculated based on the distance from the sensor to the assumed impact source location and the wave velocity of the impact stress wave. The specific process is as follows:

[0076] The wave velocity C of the impact stress wave is measured using a known impact and a sensor in the same direction:

[0077]

[0078] Where Δl is the distance between sensors in the same direction, and Δt is the time difference of arrival. The time difference of arrival is measured by the first peak of the normalized kurtosis time series of the shock stress wave signal.

[0079] Assuming that any point (x, y) in the structural monitoring area is the location of the impact source, and its distance to the m-th sensor is L. m (x,y):

[0080]

[0081] Calculate the distance difference D between different sensors and the assumed impact source location. mn (x,y):

[0082] D mn (x,y)=L n (x,y)-L m (x,y)m,n=1,2,..,M (5)

[0083] Where n represents the nth sensor.

[0084] Given the assumed location of the impact source, calculate the relative time delay τ of the impact stress wave signal from different sensors. mn (x,y):

[0085]

[0086] Step 4: Multi-feature impact imaging.

[0087] Three features of kurtosis time series are extracted: time difference of arrival, signal amplitude, and waveform correlation coefficient. Shock imaging is then performed using different principles based on these features.

[0088] The specific process is as follows:

[0089] The Time of Arrival (TOA) of two different sensors was extracted based on the first peak of the normalized kurtosis time series. m and TOA n The time difference between arrival and arrival is τ' mn (x,y):

[0090] τ' mn (x,y)=TOA n -TOA m m,n=1,2,...,M (7)

[0091] Shock imaging 1 is obtained using the time-of-arrival (TOA) characteristic and the hyperbolic principle. The result of shock imaging 1 is represented as I1:

[0092]

[0093] Where σ is the time delay factor.

[0094] The signal amplitude and characteristics of the normalized kurtosis time series are extracted based on the relative time delay of the shock stress wave signal:

[0095]

[0096] Where, τ m (x,y) represents the relative time delay of the m-th sensor relative to the 1st sensor; T1 is the end time of the window of the normalized kurtosis time series of the impact stress wave signal received by the 1st sensor. The normalized kurtosis time series of the shock stress wave signals received by the 1st to Mth sensors in the time interval [τ] m (x,y),T1+τ m The sum of signal amplitudes within [x,y]

[0097] Impact imaging 2 is obtained by using the signal amplitude and characteristics according to the principle of delay accumulation. The result of impact imaging 2 is represented as I2:

[0098] I2(x,y)=max(amp(x,y)) (10)

[0099] Where max(amp(x,y)) is the normalized kurtosis time series of the impact stress wave signals received by the 1st to Mth sensors in the time interval [τ]. m (x,y),T1+τ m The maximum value of the sum of signal amplitudes within [x,y] .

[0100] Based on the relative time delay of the shock stress wave signal, the waveform correlation coefficients of the normalized kurtosis time series from different sensors are extracted:

[0101]

[0102] Among them, T m Let K be the termination time of the window for the normalized kurtosis time series of the shock stress wave signal received by the m-th sensor; cov(K) m (t),K n (t) represents the normalized kurtosis time series of the impact stress wave signals received by the m-th and n-th sensors in the time interval [τ]. mn (x,y),T m +τ mn Covariance within [x,y]; and The normalized kurtosis time series of the shock stress wave signals received by the m-th and n-th sensors in the time interval [τ] mn(x,y),T m +τ mn The variance within [x,y]

[0103] Impact imaging 3 is obtained by utilizing waveform correlation coefficient characteristics and based on the waveform similarity principle. The result of impact imaging 3 is represented as I3:

[0104]

[0105] Step 5: Image fusion to predict impact location.

[0106] Impact imaging results with different features are multiplied and fused. The brightest position in the fused image represents the predicted impact location. The specific process is as follows:

[0107] Normalize the impact imaging results I1, I2, and I3:

[0108]

[0109] Perform cumulative multiplication and fusion on the normalized imaging results:

[0110]

[0111] The brightest position in the fused image represents the predicted impact location.

[0112] Example

[0113] The experiment was conducted on a glass fiber / epoxy resin composite laminate measuring 600mm×500mm×3mm. The impact was triggered by a hammer. Twelve piezoelectric sensors were arranged on the panel to capture the impact stress wave signal. The position coordinates of the sensors are shown in Table 1, with the lower left corner of the panel as the origin. After capturing the impact stress wave signal, the piezoelectric sensors transmitted it to the signal acquisition system, which had a sampling rate of 200kHz. Figure 3 This is a typical impact stress wave signal obtained by the sensor. The kurtosis time series of the impact stress wave signal calculated according to step 2 is shown below. Figure 4 As shown. By Figure 3 and Figure 4 It can be seen that at the moment of arrival of the impact stress wave, the kurtosis time series exhibits a large peak, then rapidly decays to a very small value and tends to stabilize. Using a known impact and sensors in the same direction, the impact stress wave velocity was measured to be 1050 m / s. Taking an impact at (180 mm, 325 mm) as an example, firstly, the sensor coordinates and impact stress wave signals from each sensor are recorded according to step 1. Then, according to step 2, the kurtosis time series of the impact stress wave signals received by each sensor is calculated and normalized. Next, according to step 3, the relative time delay of the impact stress wave signals from different sensors is calculated. Finally, according to steps 4 and 5, the kurtosis time series of the impact stress wave signals from different sensors is obtained. Figure 5(a), 5(b), 5(c) and 5(d). Figure 5 (a), 5(b), 5(c), and 5(d) are schematic diagrams showing the results of impact imaging based on the hyperbolic principle using time-of-arrival characteristics, impact imaging based on the delay accumulation principle using amplitude and features, impact imaging based on the similarity principle using waveform similarity characteristics, and multi-feature image fusion, respectively. Figure 5 It can be seen that the imaging resolution of multi-feature image fusion is higher than that of single-feature imaging. The brightest position in the multi-feature image fusion represents the predicted impact position, with a positioning error of only 5mm, which meets the engineering requirements for impact positioning.

[0114] Table 1 Sensor location coordinates

[0115]

[0116]

[0117] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A structure shock locating method based on kurtosis time series and multi-feature image fusion, characterized in that, The method comprises the following steps: Step 1, mark the position coordinates of the sensors, and record the impact stress wave signals; Step 2, calculate the kurtosis time series and normalize them; Step 3, calculate the relative time delay of the impact stress wave signals of different sensors; Step 4, multi-feature impact imaging; Step 5, image fusion and prediction of the impact position; In step 4, the wave arrival time difference, signal amplitude and waveform correlation coefficient of the kurtosis time series are extracted, and the wave arrival time difference, signal amplitude and waveform correlation coefficient of the kurtosis time series are used to perform impact imaging by different principles; the specific process is as follows: Extracting the time of arrival of two different sensors from the first peak of the normalized kurtosis time series TOA m and TOA n , the time difference of arrival is : (7) The impact imaging 1 is obtained according to the hyperbolic principle using the time difference of arrival characteristic, and the result of the impact imaging 1 is expressed as I 1: (8) where σ is a time delay factor; is the relative time delay of the impact stress wave signals of different sensors; According to the relative time delay of the impact stress wave signals, the signal amplitude and features of the normalized kurtosis time series are extracted: (9) wherein, is the relative time delay of the i-th sensor with respect to the 1-st sensor; m is the termination time of the window of the normalized kurtosis time series of the impact stress wave signal received by the 1-st sensor; is the termination time of the window of the normalized kurtosis time series of the impact stress wave signal received by the 1-st sensor; is the signal amplitude sum of the normalized kurtosis time series of the impact stress wave signal received by the 1-st to i-th sensors over the time period M is the signal amplitude sum of the normalized kurtosis time series of the impact stress wave signal received by the 1-st to i-th sensors over the time period ​ The impact imaging 2 is obtained according to the delay accumulation principle using the signal amplitude and characteristics, and the result of the impact imaging 2 is represented as I 2: (10) wherein, the maximum value of the signal amplitude and of the normalized kurtosis time series of the impact stress wave signals received by the 1st to the M nth sensor within the time period tmax. According to the relative time delay of the impact stress wave signals, the waveform correlation coefficients of the normalized kurtosis time series of different sensors are extracted: (11) wherein is the termination time of a window of the normalized kurtosis time series of the impact stress wave signal received by the jth sensor; m is the termination time of a window of the normalized kurtosis time series of the impact stress wave signal received by the jth sensor; m n is the covariance of the normalized kurtosis time series of the impact stress wave signals received by the jth and jth sensors over the time period and m is the variance of the normalized kurtosis time series of the impact stress wave signals received by the jth and jth sensors over the time period n is the variance of the normalized kurtosis time series of the impact stress wave signals received by the jth and jth sensors over the time period​​​​​ The waveform correlation coefficient feature is used to obtain the impact image 3 according to the waveform similarity principle, and the result of the impact image 3 is represented as I 3: (12) In step 5, the impact imaging results of different features are multiplied and fused; the highest bright position of the fused image represents the predicted impact position; the specific process is as follows: Impingement imaging results I 1、 I 2 and I 3 were normalized: (13) The normalized imaging results are multiplied and fused: (14) The highest bright position of the fused image represents the predicted impact position.

2. The structure shock locating method based on kurtosis time series and multi-feature image fusion according to claim 1, characterized in that, In step 1, the position coordinates of all sensors receiving the impact stress wave signals in the monitoring area of the structure are marked, and the impact stress wave signals received by all sensors are recorded; for the monitoring area of the structure arranged with M piezoelectric sensors, a rectangular coordinate system is established, and the position coordinates of the sensors are marked as P m x m y m m = 1, 2, …, M ; the impact stress wave signals received by the sensors are recorded as S m t m = 1, 2, …, M .​​​​​​ 3. The structure shock locating method based on kurtosis time series and multi-feature image fusion according to claim 2, characterized in that, The sensors include piezoelectric sensors, strain sensors and acceleration sensors.

4. The structure shock locating method based on kurtosis time series and multi-feature image fusion according to claim 1, characterized in that, In step 2, the kurtosis time series of the impact stress wave signals is calculated according to the impact stress wave signal window; specifically, the kurtosis time series of the impact stress wave signals received by each sensor is calculated according to the following formula: (1) wherein, L is the total number of sampling points of the impact stress wave signal, t i is the time corresponding to the i-th sampling point of the signal, i is the i-th sensor, m t is the average value of the impact stress wave signal received by the i-th sensor within the time window (0 t l is the kurtosis value of the impact stress wave signal received by the i-th sensor within the time window (0 m t t l ​​​​​​ The kurtosis time series of the impact stress wave signals received by each sensor is normalized: (2) wherein is The maximum value within a time window (0 t t L ) is determined.​ 5. The structure shock locating method based on kurtosis time series and multi-feature image fusion according to claim 1, characterized in that, In step 3, assuming that any point in the monitoring area is the impact source position, the relative time delay of the impact stress wave signals between different sensors is calculated according to the distance from the sensor to the assumed impact source position and the wave speed of the impact stress wave; the specific process is as follows: Wave velocity of the impact stress wave is measured with known impact and same direction sensors C : (3) where Δ l is the distance of the sensors in the same direction, Δ t is the time difference of arrival; the time difference of arrival is measured by the first peak of the normalized kurtosis time series of the impact stress wave signal; Assuming any point in the structural monitoring area ( x , y ) represents the location of the impact source, which is the distance to the first m The distance between the sensors is : (4) Calculating the difference in distance of different sensors to the assumed impact source location D mn ( x , y ) (5) wherein, n represents the first n sensor; For a given impact source location, calculate the relative time delays of the impact stress wave signals of different sensors : (6)。

Images