Method for two-dimensional construction and online imaging of dynamic probabilistic features of aircraft structure damage

Abstract

The application discloses a dynamic probability characteristic two-dimensional construction and online imaging method of aircraft structure damage, and belongs to the field of structure health monitoring. Before structure damage monitoring, a two-dimensional characteristic parameter benchmark sample set is obtained based on a guided wave signal of a piezoelectric excitation sensing channel, and a benchmark Gaussian mixture model is established. In the damage monitoring process, a two-dimensional characteristic parameter monitoring sample set is obtained based on the guided wave signal of the piezoelectric excitation sensing channel, and a monitoring Gaussian mixture model is established. The migration distance between the monitoring and benchmark Gaussian mixture models is used to calculate a dynamic probability characteristic signal of the piezoelectric excitation sensing channel. The monitoring area is imaged based on the dynamic probability characteristic signal. The method can effectively inhibit the influence of a time-varying service environment on the uncertainty of the guided wave signal, is more sensitive to monitoring, greatly reduces the calculation amount, improves the monitoring efficiency, realizes reliable and accurate imaging positioning of the structure damage under the influence of the time-varying service environment, and is helpful to realize online monitoring of the damage.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of aircraft structure health monitoring, and particularly relates to a dynamic probability feature two-dimensional construction and online imaging method for aircraft structure damage. BACKGROUND

[0002] Aircraft structure is the key to maintain its aerodynamic shape and load bearing, and the structure health monitoring technology is of great significance to ensure the safety of aircraft service, improve the operation reliability and reduce the maintenance cost. After nearly two decades of development, the aircraft structure health monitoring technology has gradually shifted from early basic theory research to practical engineering application research.

[0003] In the current research of aircraft structure health monitoring methods, the imaging method based on piezoelectric sensor network and guided wave is considered as one of the most promising aircraft structure health monitoring technologies due to its high monitoring sensitivity, large monitoring range, support for online monitoring and offline monitoring, damage monitoring and impact monitoring, and monitoring of metal structures and composite materials. In the imaging method based on piezoelectric sensor network and guided wave, the main methods include path imaging, delay accumulation imaging, multiple signal classification, and spatial wave number filtering imaging. In general, these methods basically rely on the accurate acquisition of damage scattering signals. However, in the actual service environment of aircraft structure, there are often time-varying factors such as temperature and humidity changes, dynamic loads, and changes in structural boundary conditions, which will have a significant impact on guided wave signals, making the amplitude of guided wave signals present uncertain changes, and thus the signal changes caused by the covering damage may occur, which will lead to the above-mentioned methods being able to effectively monitor the damage in the stable environment of the laboratory, but not applicable under the influence of the time-varying service environment of the aircraft. At present, the reliable damage monitoring methods under the influence of time-varying service environment include the imaging method based on environmental factor compensation and the reference-free imaging method. However, the imaging method based on environmental factor compensation can only compensate for a single time-varying factor, and needs to accurately measure the environmental factors during monitoring, but there are often multiple environmental factors coupled in the service environment, and it is often difficult to accurately measure. The reference-free imaging method is only verified by finite element simulation or only on simple structures.

[0004] In recent years, some scholars gradually combine the probabilistic mining modeling method with guided wave array imaging to realize reliable damage monitoring under the influence of time-varying service environment. The Gaussian mixture model is a finite mixture probabilistic mining model, which can approximate the probability distribution of a complex random variable through the weighted combination of multiple Gaussian components without prior knowledge, is suitable for describing the uncertainty distribution of guided wave signals under the influence of time-varying factors, and can suppress the influence of time-varying environment and reliably represent damage by measuring the migration change of the probability distribution characteristics of the model. Therefore, the dynamic probabilistic imaging method combining the Gaussian mixture model and the delay-and-sum imaging of guided wave array imaging has obvious advantages in solving the time-varying problem, has become a new hotspot in the research of aircraft structure damage monitoring, and has good application prospect.

[0005] However, the current dynamic probabilistic imaging method needs to establish the time-invariant characteristic signal of each guided wave channel when accurately imaging and positioning the damage, and there are two main problems: on the one hand, the method of establishing a one-dimensional Gaussian mixture model by only extracting the amplitude of the damage scattering signal as a single feature and measuring the migration change of the model has low sensitivity in the representation of damage under complex service environment and small damage; on the other hand, a reference Gaussian mixture model and a monitoring Gaussian mixture model need to be established for each sampling point in the damage scattering signal of each channel, which has large amount of calculation and low monitoring efficiency, and it is difficult to realize online monitoring of structure damage in actual engineering application. SUMMARY

[0006] In view of the deficiencies of the prior art, the purpose of the present application is to provide a dynamic probabilistic characteristic two-dimensional construction and online imaging method for aircraft structure damage, so as to solve the problems of low sensitivity of the current dynamic probabilistic imaging method under complex time-varying service environment and small damage, and large amount of calculation and low monitoring efficiency leading to difficulty in online monitoring.

[0007] To achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0008] The dynamic probabilistic characteristic two-dimensional construction and online imaging method for aircraft structure damage comprises the following steps: step S1, before starting the monitoring of aircraft structure damage, obtaining a two-dimensional characteristic parameter reference sample set based on the guided wave signal of the piezoelectric excitation sensing channel, and establishing a reference Gaussian mixture model; step S2, during the monitoring of aircraft structure damage, obtaining a two-dimensional characteristic parameter monitoring sample set based on the guided wave signal of the piezoelectric excitation sensing channel, and establishing a monitoring Gaussian mixture model; step S3, calculating the dynamic probabilistic characteristic signal of the piezoelectric excitation sensing channel based on the migration distance between the monitoring Gaussian mixture model and the reference Gaussian mixture model; and step S4, imaging the monitoring area based on the dynamic probabilistic characteristic signal.

[0009] Step S1 above specifically includes the following steps: Step S11, under time-varying service environment, acquire multiple sets of guided wave signals for each piezoelectric excitation sensing channel on the monitored structure; Step S12, based on the guided wave signals acquired in Step S11, extract the reference cross-correlation sequence and reference difference signal envelope for each piezoelectric excitation sensing channel; Step S13, perform discrete sampling on the reference cross-correlation sequence and reference difference signal envelope; Step S14, extract the reference cross-correlation sequence value and reference difference signal envelope amplitude corresponding to each discrete sampling point of each piezoelectric excitation sensing channel, and establish a two-dimensional feature parameter reference sample set for each discrete sampling point; Step S15, based on the two-dimensional feature parameter reference sample set, establish a reference Gaussian mixture model to characterize the uncertainty distribution of the signal characteristics of each sampling point under the influence of time-varying service environment.

[0010] Step S2 above specifically includes the following steps: Step S21, again acquiring multiple sets of guided wave signals for each piezoelectric excitation sensing channel under time-varying service conditions; Step S22, based on the guided wave signals acquired in Step S21, extracting the monitoring cross-correlation sequence and monitoring difference signal envelope for each piezoelectric excitation sensing channel; Step S23, performing discrete sampling on the monitoring cross-correlation sequence and monitoring difference signal envelope; Step S24, extracting the monitoring cross-correlation sequence value and monitoring difference signal envelope amplitude corresponding to each discrete sampling point of each piezoelectric excitation sensing channel, and establishing a two-dimensional feature parameter monitoring sample set for each discrete sampling point; Step S25, based on the two-dimensional feature monitoring sample set, establishing a monitoring Gaussian mixture model to characterize the uncertainty distribution.

[0011] In the aforementioned method for constructing two-dimensional dynamic probabilistic characteristics of aircraft structural damage and for online imaging, multiple sets of guided wave signals from each piezoelectric excitation sensing channel on the monitored structure are acquired. Specifically, this includes acquiring... R+ One set of guided wave signals for each piezoelectric excitation sensing channel, with a signal length of... N , R, N All are natural numbers greater than 1; for each piezoelectric excitation sensing channel, the first group of guided wave signals is used as the reference signal to calculate the remaining... R Any first wavelet in a group of guided wave signals r Group (1≤ r ≤ R Cross-correlation sequence of the signal and the reference signal C r ( m The sequence length is also ). N The expression is:

[0012]

[0013] In the formula, B ( n () is the reference signal. Hr ( n (This refers to the first day before damage monitoring begins) r Guided wave signal.

[0014] In the aforementioned method for constructing two-dimensional dynamic probabilistic features of aircraft structural damage and online imaging, the reference cross-correlation sequence and reference difference signal envelope of each piezoelectric excitation sensing channel are extracted, or the monitoring cross-correlation sequence and monitoring difference signal envelope of each piezoelectric excitation sensing channel are extracted. Specifically, this includes extracting the first... r The difference envelope between the guided wave signal and the reference signal E r ( n ), the length is also N， The expression is:

[0015]

[0016] In the formula, B ( n () is the reference signal. H r ( n (The first day before damage monitoring begins) r Guided wave signal.

[0017] In the aforementioned two-dimensional construction and online imaging method for dynamic probabilistic features of aircraft structural damage, after sampling, the length of the reference cross-correlation sequence and the reference difference signal envelope, or the length of each monitored cross-correlation sequence and the monitored difference signal envelope, is... M ,1≤ M < N Among them, the first m The two-dimensional feature parameter benchmark sample set of sampling points is represented as X( m ) = { x 1, …, x r , …, x R},1≤ m ≤ M The sample set contains a total of R The 2D sample, and the 2D sample r Each sample is represented as this. x r =[ C r ( m ), E r ( m )).

[0018] In the aforementioned two-dimensional construction and online imaging method for dynamic probabilistic characteristics of aircraft structural damage, the Gaussian mixture model Φ(benchmark Gaussian mixture model or monitoring Gaussian mixture model) is used. m The expression for ) is:

[0019]

[0020] wherein, K is the number of Gaussian components in the Gaussian mixture model, μ k ,∑ k , w k is the mean vector, covariance matrix, and weight of the k th Gaussian component, respectively, k = 1, 2, …, K ,Φ k The probability density function of the Gaussian mixture model is expressed as follows:

[0021] .

[0022] The step S3 specifically comprises, for any piezoelectric excitation sensing channel, calculating the migration distance between the monitoring Gaussian mixture model and the reference Gaussian mixture model for each discrete sampling point, and the migration distances of all discrete sampling points together form the dynamic probability characteristic signal of the piezoelectric excitation sensing channel.

[0023] In the method for constructing and imaging the dynamic probability characteristics of the aircraft structure damage, the probability migration distance of the monitoring Gaussian mixture model relative to the reference Gaussian mixture model is calculated according to the following expression: KL m .

[0024]

[0025] wherein, tr is the trace of a matrix, and det is the determinant value of a matrix, μ b m ),∑ b m are the mean matrix and covariance matrix of the reference Gaussian mixture model, μ d m ),∑ d m are the mean matrix and covariance matrix of the monitoring Gaussian mixture model.

[0026] In the step S4, the following steps are specifically included: step S41, interpolating and reconstructing the dynamic probability characteristic signals of the piezoelectric excitation sensing channels after discrete sampling to form complete dynamic probability characteristic signals with the same length, cross-correlation sequence, and difference signal envelope; and step S42, fusing the complete dynamic probability characteristic signals of all piezoelectric excitation sensing channels by using the guided wave array imaging method to image the monitoring region and realize imaging positioning of the structure damage.​​​​​

[0027] In the aforementioned method for constructing two-dimensional dynamic probabilistic features of aircraft structural damage and for online imaging, the interpolation reconstruction in step S41 specifically includes changing the length of the dynamic probabilistic feature signal from... M Expand to N The expression for interpolation reconstruction is:

[0028]

[0029] In the formula, m 1. m 2 represents the sequence number of the discrete sampling point. KL ( m 1) KL ( m 2) Represents the corresponding signal amplitude in the dynamic probability characteristic signal. m 3 indicates that it is intended to be in m 1 and m The sampling points inserted in step 2.

[0030] In the aforementioned method for constructing two-dimensional dynamic probabilistic features of aircraft structural damage and for online imaging, step S42, which involves imaging the monitored area, specifically includes obtaining all... I After interpolating and reconstructing each piezoelectric excitation sensing channel to form a complete dynamic probability characteristic signal, the monitoring area is divided into equally spaced segments. The result is used to calculate the value of any point using the following expression: x , y ) signal energy E ( x , y The signal energy is used as pixel value for image representation, and the point with the largest pixel value in the image is the location of the damage.

[0031]

[0032] In the formula, t off The time offset corresponding to the excitation signal. v The waveguide propagation velocity in the structure. f s The sampling rate of the guided wave signal. KL i For the first i The complete dynamic probabilistic characteristic signal reconstructed by interpolation of each piezoelectric excitation sensing channel, ( x i a , y i a )and( x i s , yi s ) are respectively the first i The coordinates of the exciter and sensor in each piezoelectric excitation sensing channel. i = 1, 2, … , I .

[0033] The beneficial effects of the two-dimensional construction and online imaging method for dynamic probabilistic features of aircraft structural damage in this invention are that it can effectively suppress the uncertainty of guided wave signals caused by the time-varying service environment of the aircraft structure. By extracting the signal features of the difference signal value and the cross-correlation sequence value, a Gaussian mixture model is constructed and its migration change is measured. The resulting dynamic probabilistic feature signal is more sensitive to damage monitoring. Furthermore, by fitting and reconstructing the dynamic probabilistic feature signal through interpolation, the computational load of the method is greatly reduced and the monitoring efficiency is improved. While realizing reliable and accurate imaging and positioning of aircraft structural damage under the influence of time-varying service environment, it also helps to realize online monitoring of damage. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating the two-dimensional construction and online imaging method for dynamic probabilistic features of aircraft structural damage according to the present invention.

[0035] Figure 2 This is a schematic diagram of the specific process of step S1 in this invention.

[0036] Figure 3 This is a schematic diagram of the specific process of step S2 in this invention.

[0037] Figure 4 This is a schematic diagram of the specific process of step S4 in this invention.

[0038] Figure 5a This is a schematic diagram of the porous aluminum alloy wall panel used in the embodiment.

[0039] Figure 5b This is a schematic diagram of the piezoelectric sensor network and the distribution of the excitation sensing channels.

[0040] Figure 6 A schematic diagram illustrating the temperature variation range setting and data acquisition process.

[0041] Figure 7a This is a schematic diagram of 100 differential signals for channels 6-9.

[0042] Figure 7b This is a schematic diagram of 100 sets of cross-correlation sequence signals for channels 6-9.

[0043] Figure 8a This is a schematic diagram of the baseline two-dimensional feature parameter sample set for one sampling point of channel 6-9.

[0044] Figure 8b Reference Gaussian Mixture Model for a sampling point of channel 6-9.

[0045] Figure 9 Time-invariant description for channel 5-9 away from damage and channel 6-9 through damage.

[0046] Figure 10 Time-invariant description for channel 5-9 away from damage and channel 6-9 through damage after interpolation fitting.

[0047] Figure 11 Damage imaging result in the embodiment of the application. DETAILED DESCRIPTION

[0048] In order to make the purpose and technical scheme of the embodiment of the application clearer, the technical scheme of the embodiment of the application will be described clearly and completely below in combination with the drawings of the embodiment of the application. Obviously, the described embodiment is a part of the embodiments of the application, rather than all the embodiments of the application. Based on the described embodiment of the application, all other embodiments obtained by a person of ordinary skill in the art without creative labor belong to the protection scope of the application.

[0049] Figure 1 Flowchart of the method for constructing and imaging the dynamic probability characteristics of aircraft structure damage in two dimensions online according to the application. Referring to Figure 1 the drawing, the method for constructing and imaging the dynamic probability characteristics of aircraft structure damage in two dimensions online according to the application comprises,

[0050] Step S1, before aircraft structure damage monitoring, based on the guided wave signals of the piezoelectric excitation sensing channel, a two-dimensional characteristic parameter reference sample set is obtained, and a reference Gaussian mixture model is established;

[0051] Step S2, during aircraft structure damage monitoring, based on the guided wave signals of the piezoelectric excitation sensing channel, a two-dimensional characteristic parameter monitoring sample set is obtained, and a monitoring Gaussian mixture model is established;

[0052] Step S3, based on the migration distance between the monitoring Gaussian mixture model and the reference Gaussian mixture model, a dynamic probability characteristic signal of the piezoelectric excitation sensing channel is calculated;

[0053] Step S4, based on the dynamic probability characteristic signal, the monitoring area is imaged.

[0054] Figure 2 Specific flowchart of step S1 in the application. As Figure 2 shown in the drawing, in the above step S1, the specific steps include the following.

[0055] Step S11, under the time-varying service environment, a plurality of groups of guided wave signals of each piezoelectric excitation sensing channel on the monitored structure are collected respectively.

[0056] The collection of the plurality of groups of guided wave signals of each piezoelectric excitation sensing channel on the monitored structure specifically includes R+ collecting the guided wave signals of each piezoelectric excitation sensing channel in the first group, and the signal length is N , R, N are natural numbers greater than 1. For each piezoelectric excitation sensing channel, the first group of guided wave signals is taken as the reference signal, and the cross-correlation sequence of the reference signal and any R r group (1≤ r ≤ R ) signal in the remaining C r ( m ) signals is calculated, and the sequence length is also N , and the expression is:

[0057]

[0058] In the formula, B ( n ) is the reference signal, H r ( n ) is the first guided wave signal before starting damage monitoring. r

[0059] Step S12, based on the guided wave signals collected in step S11, the reference cross-correlation sequence and the reference difference signal envelope of each piezoelectric excitation sensing channel are extracted.

[0060] The extraction of the reference cross-correlation sequence and the reference difference signal envelope of each piezoelectric excitation sensing channel specifically includes extracting the difference signal envelope of the first r group of guided wave signals and the reference signal E r ( n ), and the length is also N， , and the expression is:

[0061]

[0062] In the formula, B ( n ) is the reference signal, H r ( n ) is the first guided wave signal before starting damage monitoring. r

[0063] ​​​Step S13 involves discrete sampling of the benchmark cross-correlation sequence and the benchmark difference signal envelope. For both the extracted benchmark cross-correlation sequence and the benchmark difference signal envelope, discrete sampling methods such as interval sampling are used.

[0064] Step S14: Extract the reference cross-correlation sequence value and reference difference signal envelope amplitude corresponding to each discrete sampling point of each piezoelectric excitation sensing channel, and establish a two-dimensional feature parameter reference sample set for each discrete sampling point.

[0065] In step S14 above, after sampling, the lengths of the reference cross-correlation sequence and the reference difference signal envelope are both... M ,1≤ M < N Among them, the first m The two-dimensional feature parameter benchmark sample set of sampling points can be represented as X( m ) = { x 1, …, x r , …, x R},1≤ m ≤ M The sample set contains a total of R The 2D sample, and the 2D sample r Each sample is represented as this. x r =[ C r ( m ), E r ( m )).

[0066] Step S15: Based on the two-dimensional feature parameter benchmark sample set, establish a benchmark Gaussian mixture model to characterize the uncertainty distribution of signal characteristics at each sampling point under the influence of time-varying service environment. The Gaussian mixture model Φ( m The expression for ) is:

[0067]

[0068] In the formula, K Let be the number of Gaussian components in the Gaussian mixture model. μ k Σ k , w k The first k The mean vector, covariance matrix, and weights of each Gaussian component. k = 1, 2, … , K Φ k The probability density function expression is as follows:

[0069]

[0070] Figure 3 is a specific flowchart of step S2 in the present application. As shown in the figure, the above step S2 specifically includes the following steps. Figure 3

[0071] Step S21, under the time-varying service environment, collect multiple groups of guided wave signals of each piezoelectric excitation sensing channel again;

[0072] Step S22, based on the guided wave signals collected in step S21, extract the monitoring cross-correlation sequence and the monitoring difference signal envelope of each piezoelectric excitation sensing channel;

[0073] Step S23, discretely sample the monitoring cross-correlation sequence and the monitoring difference signal envelope;

[0074] Step S24, extract the monitoring cross-correlation sequence value and the monitoring difference signal envelope amplitude corresponding to each discrete sampling point of each piezoelectric excitation sensing channel, and establish a two-dimensional feature parameter monitoring sample set of each discrete sampling point;

[0075] Step S25, based on the two-dimensional feature monitoring sample set, establish a monitoring Gaussian mixture model to represent the uncertainty distribution.

[0076] The calculation method in the above steps S21 to S25 is consistent with that in steps S11 to S15.

[0077] The above step S3 specifically includes, for any one piezoelectric excitation sensing channel, calculating the migration distance between the monitoring Gaussian mixture model and the reference Gaussian mixture model for each discrete sampling point, and the migration distances of all discrete sampling points together constitute the dynamic probability feature signal of the piezoelectric excitation sensing channel. The dynamic probability feature signal of the piezoelectric excitation sensing channel is not affected by the time-varying service environment and is only sensitive to damage.

[0078] The migration distance is preferably measured by KL distance (KL is the abbreviation of Kullback-Leibler), Mahalanobis distance, etc. The probability migration distance of the monitoring Gaussian mixture model relative to the reference Gaussian mixture model is taken as an example. The migration distance calculation expression of KL distance is as follows: KL m

[0079]

[0080] In the formula, tr is the trace of the matrix, det is the determinant value of the matrix, μ b m b m ​​​​​​) is the mean matrix and covariance matrix of the reference Gaussian mixture model, μ d ( m ) is the mean matrix and covariance matrix of the monitoring Gaussian mixture model. d ( m ) is the mean matrix and covariance matrix of the monitoring Gaussian mixture model.

[0081] After the migration distance calculation of all discrete sampling points of each piezoelectric excitation sensing channel is completed, the dynamic probability characteristic signal composed of migration distances with a length of M of each piezoelectric excitation sensing channel can be obtained. M

[0082] Figure 4 is the specific flowchart of step S4 in the application. As shown in Figure 4 , the above step S4 specifically includes the following steps.

[0083] Step S41, the dynamic probability characteristic signal of each piezoelectric excitation sensing channel after discrete sampling is interpolated and reconstructed to form a complete dynamic probability characteristic signal with the same length, cross-correlation sequence and difference signal envelope.

[0084] The interpolation and reconstruction above expands the length of the dynamic probability characteristic signal from M to N . Taking linear interpolation as an example, the expression of interpolation and reconstruction is as follows:

[0085]

[0086] In the formula, m 1, m 2 represent the serial number of the discrete sampling points, KL ( m 1), KL ( m 2) represent the corresponding signal amplitude in the dynamic probability characteristic signal, m 3 represents the sampling point to be inserted in m 1 and m 2.

[0087] Step S42, the complete dynamic probability characteristic signal of all piezoelectric excitation sensing channels is fused by using the guided wave array imaging method to image the monitoring area and realize the imaging positioning of the structural damage.

[0088] The imaging of the monitoring area specifically includes that after the interpolation and reconstruction of all piezoelectric excitation sensing channels to form the complete dynamic probability characteristic signal, the monitoring area is equally divided into I , and the signal energy of any point x , y is calculated according to the following expression: E ​( x , y ), and the signal energy is image characterized as a pixel value, and the point with the maximum pixel value in the image is the damage position,

[0089]

[0090] In the formula, t off is the time offset corresponding to the excitation signal, v is the propagation speed of the guided wave in the structure, f s is the sampling rate of the guided wave signal, KL i is the first i interpolation reconstruction of the complete dynamic probability characteristic signal of the piezoelectric excitation sensing channel, x i a , y i a ) and ( x i s , y i s ) are the coordinates of the exciter and the sensor of the first i piezoelectric excitation sensing channel, i = 1, 2, …, I .

[0091] Hereinafter, the dynamic probability characteristic two-dimensional construction and online imaging method of the aircraft structure damage of the present application will be described in conjunction with a specific embodiment.

[0092] Figure 5a is a schematic diagram of the aluminum alloy porous wall plate used in the embodiment. As shown in Figure 5a , the aviation aluminum alloy porous plate structure with a size of 600mmx600mmx3mm (lengthxwidthxthickness) used in the embodiment is used as the monitored structure, and there are 15 equidistant threaded holes with a diameter of 10mm on the aviation aluminum alloy porous plate, and the mutual distance is 32mm. In order to monitor the damage on the structure, a piezoelectric sensor network is arranged thereon, which contains 16 piezoelectric sensors, which are numbered as piezoelectric sensor 1~piezoelectric sensor 16, and the distance between any two adjacent piezoelectric sensors is 150mm.

[0093] Figure 5b is a schematic diagram of the piezoelectric sensor network and the excitation sensing channel distribution. As shown in Figure 5bAs shown, 16 piezoelectric sensors together form 42 piezoelectric excitation and sensing channels. During implementation, an environmental test chamber is used to simulate the time-varying temperature environment of an aircraft, and the temperature change range and signal acquisition process are as shown in Figure 6 .

[0094] Before carrying out damage monitoring on the above-mentioned porous plate structure of an aircraft aluminum alloy, the monitored structure is in a healthy state. First, at room temperature (20°C), the guided wave signals of the above-mentioned 42 piezoelectric excitation and sensing channels are collected once at a sampling rate of 10 MHz, serving as the reference signals of these piezoelectric excitation and sensing channels, and the signal length is set to 1500 points. Then, the temperature of the environmental test chamber is set to decrease to -50°C, and then increase from -50°C to 80°C. During the process of increasing from -50°C to 80°C, the guided wave signals of all piezoelectric excitation and sensing channels are continuously collected, a total of 100 times. The guided wave signals of each piezoelectric excitation and sensing channel are processed with the reference signal, and 100 sets of difference signal envelopes and cross-correlation sequences of each piezoelectric excitation and sensing channel are obtained. Taking the piezoelectric excitation and sensing of piezoelectric sensor 6 to piezoelectric sensor 9, denoted as channel 6-9, as an example, Figure 7a is a schematic diagram of 100 sets of difference signals of channel 6-9, Figure 7b is a schematic diagram of 100 sets of cross-correlation sequence signals of channel 6-9. As can be seen from the figure, although the structure has no damage at this time, due to the influence of the time-varying temperature environment, there are still obvious differences between the 100 sets of difference signal envelopes and cross-correlation sequences.

[0095] For each of the above-mentioned 42 piezoelectric excitation and sensing channels, linear discrete sampling is performed on all difference signal envelopes and cross-correlation sequences. Still taking channel 6-9 as an example, sampling is performed every 2 points. The difference signal envelopes and cross-correlation sequences after discrete sampling each contain 500 sampling points, the difference signal envelope amplitude and cross-correlation sequence amplitude corresponding to each sampling point are extracted to form two-dimensional feature parameters, a two-dimensional feature parameter sample set containing 100 samples can be obtained, and on this basis, a reference Gaussian mixture model is established to describe the uncertainty distribution of the sample set under the influence of the time-varying temperature environment. Figure 8a and Figure 8b are respectively a two-dimensional feature parameter sample set of a certain sampling point in channel 6-9 and a reference Gaussian mixture model, the number of Gaussian components is 5, and each Gaussian component is described by a set of contour lines.

[0096] The crack damage shown in Figure 5a is applied to the structure, and the damage monitoring of the structure begins. The above-mentioned process is repeated as Figure 6The temperature change control and data acquisition process shown can acquire 100 groups of monitoring signals of each piezoelectric excitation sensing channel, and then respectively processes with the reference signal to obtain 100 groups of difference signal envelopes and cross-correlation sequences of each piezoelectric excitation sensing channel. Similarly, the linear discrete sampling is performed on the above difference signal envelopes and cross-correlation sequences, and the signal length after sampling is 500 points. The two-dimensional feature parameter sample set is extracted for each sampling point, and the monitoring Gaussian mixture model is constructed on the basis.

[0097] The KL distance between the monitoring Gaussian mixture model and the reference Gaussian mixture model of each sampling point is calculated, and then the dynamic probability feature signal composed of the KL distance with a length of 500 points corresponding to each piezoelectric excitation sensing channel is obtained. A small area composed of 4 sensors around the damage is selected for analysis, and there are 6 guided wave excitation sensing channels, wherein the channel 5-9 and the channel 9-10 are relatively far away from the damage position, the channel 5-6, the channel 5-10 and the channel 6-10 are relatively close to the damage position, and the channel 6-9 directly passes through the damage. Figure 9 The dynamic probability feature signals corresponding to the channel 5-9 far away from the damage and the channel 6-9 passing through the damage are respectively shown.

[0098] The linear interpolation method is used to reconstruct the complete dynamic probability feature signal with the same length as the original guided wave signal, and the length is 1500 sampling points. The complete dynamic probability feature signals of the above channel 5-9 and channel 6-9 are shown as Figure 10 It can be seen from Figure 10 that the amplitude of the dynamic probability feature signal of the damage channel remains at a very small extent, the dynamic probability feature signal of the near channel passing through the damage is obviously larger, which conforms to the actual situation, so that the dynamic probability feature signal of each piezoelectric excitation sensing channel can be used for damage imaging positioning.

[0099] After obtaining the complete dynamic probability feature signals of all 6 excitation sensing channels in the small area, the delay accumulation method is used for damage imaging, wherein the wave speed used is 2550 m / s. Figure 11 The focusing processing image of the imaging result in the embodiment is shown, and the position of the damage can be finally determined from the image, the positioning error is 0.9 cm, and the reliable imaging positioning of the aviation structure damage under the influence of temperature time-varying environment is realized.

[0100] The application has many specific application ways, and the above description is only the preferred embodiment of the application. It should be pointed out that, for ordinary skilled in the art, without departing from the principle of the application, some improvements can be made, and these improvements should also be regarded as the protection scope of the application.

Claims

1. A method for two-dimensional construction and online imaging of dynamic probabilistic features of damage in aircraft structures, characterized in that, The method comprises the following steps: Step S1, before aircraft structure damage monitoring, based on the guided wave signals of the piezoelectric excitation sensing channels, a two-dimensional feature parameter reference sample set is obtained, and a reference Gaussian mixture model is established; Step S2, during aircraft structure damage monitoring, based on the guided wave signals of the piezoelectric excitation sensing channels, a two-dimensional feature parameter monitoring sample set is obtained, and a monitoring Gaussian mixture model is established; Step S3, based on the migration distance between the monitoring Gaussian mixture model and the reference Gaussian mixture model, a dynamic probability feature signal of the piezoelectric excitation sensing channel is calculated; for any piezoelectric excitation sensing channel, the migration distance between the monitoring Gaussian mixture model and the reference Gaussian mixture model is calculated for each discrete sampling point, and the migration distances of all discrete sampling points together form the dynamic probability feature signal of the piezoelectric excitation sensing channel; Step S4, based on the dynamic probability feature signal, the monitoring area is imaged; In the above step S1, the following steps are specifically included: Step S11, under a time-varying service environment, a plurality of groups of guided wave signals of each piezoelectric excitation sensing channel on the monitored structure are collected respectively; Step S12, based on the guided wave signals collected in step S11, a reference cross-correlation sequence and a reference difference signal envelope of each piezoelectric excitation sensing channel are extracted; Step S13, the reference cross-correlation sequence and the reference difference signal envelope are discretely sampled; Step S14, the reference cross-correlation sequence value and the reference difference signal envelope amplitude corresponding to each discrete sampling point of each piezoelectric excitation sensing channel are extracted, and a two-dimensional feature parameter reference sample set of each discrete sampling point is established; Step S15, based on the two-dimensional feature parameter reference sample set, a reference Gaussian mixture model is established to represent the uncertainty distribution of the signal characteristics of each sampling point under the influence of the time-varying service environment; extracting a reference cross-correlation sequence and a reference difference signal envelope for each piezoelectrically excited sensing channel, or extracting a monitoring cross-correlation sequence and a monitoring difference signal envelope for each piezoelectrically excited sensing channel, in particular, extracting a first r a difference signal envelope of the group guided wave signal and the reference signal E r ( n ) with a length of N， is given by wherein B n is a reference signal, H r n is the number of days before the start of damage monitoring r guided wave signals;​​ In the above step S2, the following steps are specifically included: Step S21, again under a time-varying service environment, a plurality of groups of guided wave signals of each piezoelectric excitation sensing channel are collected; Step S22, based on the guided wave signals collected in step S21, a monitoring cross-correlation sequence and a monitoring difference signal envelope of each piezoelectric excitation sensing channel are extracted; Step S23, the monitoring cross-correlation sequence and the monitoring difference signal envelope are discretely sampled; Step S24, the monitoring cross-correlation sequence value and the monitoring difference signal envelope amplitude corresponding to each discrete sampling point of each piezoelectric excitation sensing channel are extracted, and a two-dimensional feature parameter monitoring sample set of each discrete sampling point is established; Step S25, based on the two-dimensional feature monitoring sample set, a monitoring Gaussian mixture model is established to represent the uncertainty distribution.

2. The method of claim 1, wherein, Collecting multiple groups of guided wave signals of each piezoelectric excitation sensing channel on the monitored structure, specifically including R+ 1 group of guided wave signals of each piezoelectric excitation sensing channel, the signal length is N , R, N are all natural numbers greater than 1; For each piezoelectric excitation sensing channel, the remaining signals are calculated using the first group of guided wave signals as the reference signal. R Any first wavelet in a group of guided wave signals r Group (1≤ r ≤ R Cross-correlation sequence of the signal and the reference signal C r ( m The sequence length is also ). N The expression is: wherein B n is a reference signal, H r n is the number of days before the onset of injury monitoring, r a guided wave signal.​​ 3. The method of claim 2, wherein, The length of the reference cross-correlation sequence and the reference difference signal envelope, or the length of the monitoring cross-correlation sequence and the monitoring difference signal envelope, after the sampling is completed is M , 1≤ M < N ; wherein the two-dimensional feature parameter reference sample set of the first m sampling point is represented as X m = { x 1, …, x r , …, x R}, 1≤ m ≤ M , the sample set contains a total of R two-dimensional samples, and the first r sample is represented as this x r = [ C r ( m ), E r ( m )].

4. The method of claim 3, wherein, Gaussian mixture model Φ of the reference Gaussian mixture model or the monitoring Gaussian mixture model m The expression of the Gaussian mixture model Φ of the reference Gaussian mixture model or the monitoring Gaussian mixture model is: wherein K is the number of Gaussian components in the Gaussian mixture model, μ k ,∑ k , w k are the mean vector, covariance matrix, and weight of the k th Gaussian component, respectively, k = 1, 2, …, K ,Φ k is expressed as follows: 。 5. The method of claim 1, wherein monitoring a probability migration distance of the Gaussian mixture model relative to the baseline Gaussian mixture model, the migration distance computation expression KL ( m ) is: where tr is the trace of the matrix, det is the determinant value of the matrix, μ b ( m ),∑ b ( m ) are the mean matrix and the covariance matrix of the reference Gaussian mixture model, μ d m ),∑ d m ) are the mean matrix and the covariance matrix of the monitoring Gaussian mixture model.​​ 6. The method of claim 1, wherein In step S4, the following steps are specifically included: Step S41, the discrete sampling dynamic probability feature signals of each piezoelectric excitation sensing channel are interpolated and reconstructed to form complete dynamic probability feature signals with the same length, cross-correlation sequence and difference signal envelope; Step S42, the guided wave array imaging method is used to fuse the complete dynamic probability feature signals of all piezoelectric excitation sensing channels, the monitoring area is imaged, and the imaging positioning of the structure damage is realized.

7. The method of claim 6, wherein, The interpolation reconstruction in step S41 specifically includes lengthening the dynamic probability feature signal from M to N , and the expression of the interpolation reconstruction is: wherein m 1、 m 2 denotes the sequence number of the discrete sampling point, KL ( m 1)、 KL ( m 2) denotes the corresponding signal amplitude in the dynamic probability characteristic signal, m 3 denotes the sampling point to be inserted in m 1 and m 2.

8. The method of claim 6, wherein, Step S42, imaging the monitoring area, specifically includes obtaining all... I After interpolating and reconstructing each piezoelectric excitation sensing channel to form a complete dynamic probability characteristic signal, the monitoring area is divided into equally spaced areas, and the value of any point is calculated according to the following expression ( x , y ) signal energy E ( x , y The signal energy is used as pixel value for image representation, and the point with the largest pixel value in the image is the location of the damage. wherein t off is the time offset corresponding to the excitation signal, v is the guided wave propagation speed in the structure, f s is the guided wave signal sampling rate, KL i is the full dynamic probability feature signal of the interpolation reconstruction of the i x i a y i a x i s y i s i are the coordinates of the exciter and sensor of the i = 1, 2, …, I .​​​​​

Images