An integrated detection method for spacecraft structure collision and leakage based on acoustic emission
Through the detection method based on acoustic emission, the acoustic emission sensor network array and signal processing algorithm are used to solve the problem of rapid and accurate detection of collisions and leakage in spacecraft structures, real-time and long-term detection capabilities are achieved, and the spacecraft health monitoring needs are met. .
Patent Information
- Application Number
- CN202410061344.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-01-16
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-01-16
AI Technical Summary
The prior art cannot quickly and accurately detect collisions and leakages of spacecraft structures, and the existing methods can only conduct single detection, and cannot achieve real-time and long-term stable detection.
Using acoustic emission-based detection method, acoustic emission sensor network array is arranged in a distributed manner, and a feature vector and signal processing algorithms are used to achieve the distinction and judgment of collision and leakage, including establishing a mapping grid, calculating feature vectors, judging signal types and outputs. Positioning results.
Real-time detection of spacecraft structure collisions and leakage is realized, detection speed and accuracy are improved, suitable for harsh on-orbit working conditions, and meet the needs of spacecraft health monitoring.
Smart Images

Figure CN118067319B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of spacecraft structure health detection, and specifically relates to an integrated detection method for spacecraft structure collision and leakage based on acoustic emission. Background Art
[0002] The integrity of the spacecraft structure is the key to the success or failure of the space mission and the key to ensuring the safety of astronauts. At present, with the rapid development of the global aerospace industry and the increase in human space activities, more and more space debris remain in space, seriously threatening operational safety and the safety of astronauts. If the collision between space debris and spacecraft and the leakage caused by collision and other reasons cannot be detected and repaired in time, it may cause huge economic losses and even catastrophic consequences such as casualties. In order to minimize the losses, it is very important to quickly sense and locate the collision and leakage at the first time, which is of great significance to the long-term safe operation of spacecraft in orbit.
[0003] For locating collision sources, existing technologies include infrared imaging, fiber Bragg grating, resistance film, etc.; for locating leakage sources, existing technologies include infrared thermal imaging, fiber Bragg grating, helium mass spectrometry, multi-parameter leak detection technology, etc. However, the above methods have the defects of complex detection devices or unstable and slow detection, and can only perform single detection of collision or leakage.
[0004] Compared with the above detection methods, the acoustic emission method is simpler and more convenient, has more application scenarios, is faster in detection, and has higher sensitivity. It has great advantages in positioning accuracy, environmental applicability, system integration, etc. Moreover, the acoustic emission method can detect both collision and leakage situations by performing different algorithm processing on the acoustic signal, and can achieve long-term and stable real-time detection, which is more suitable for harsh working conditions such as in-orbit spacecraft.
[0005] Therefore, this patent proposes an integrated detection method for spacecraft structure collision and leakage based on acoustic emission, which can meet the needs of real-time detection of both collision and leakage during spacecraft operation. Summary of the invention
[0006] The purpose of the present invention is to overcome the deficiencies of the prior art and to provide an integrated detection method for spacecraft structure collision and leakage based on acoustic emission.
[0007] The present invention solves the technical problem by the following technical solutions:
[0008] A spacecraft structure collision and leakage integrated detection method based on acoustic emission, characterized in that the steps of the method are:
[0009] 1) System construction: The detection system used in the method includes a data acquisition card, a computer, a plurality of acoustic emission sensors and a preamplifier. The plurality of acoustic emission sensors are pre-distributedly coupled and arranged at selected positions on the surface of the spacecraft structure to be tested to receive acoustic signals generated by collision and leakage events. The acoustic emission sensors in each row and each column are arranged at a fixed interval to form an acoustic emission sensor network array;
[0010] The acoustic emission sensors uniformly distributed on the surface of the spacecraft structure to be tested are divided into units, with four acoustic emission sensors forming a small unit, and the spacecraft structure to be tested can be divided into several small units. In each small unit, a rectangular coordinate system is established with the geometric center position of the unit as the origin, and the small unit is divided into four quadrants, so that each of the four quadrants of each small unit has an acoustic emission sensor, and they are named No. 1, 2, 3, and 4 sensors of the unit respectively; the several acoustic emission sensors are sequentially connected to the several preamplifiers to amplify the weak signals received by the acoustic emission sensors; the preamplifiers are sequentially connected to the data acquisition card and the computer to collect the signal data received by the acoustic emission sensors and perform subsequent data processing;
[0011] 2) Establish a mapping grid: Take a small unit as the minimum unit for analysis and processing, divide the grid inside the small unit, generate several grid nodes, and establish the corresponding feature vector for each grid node. Since the collision signal and the leakage signal have different characteristics, two feature vectors need to be established for each grid node:
[0012] a. For the collision signal, calculate the theoretical time difference characteristic matrix of each node. The acoustic signal propagates along the shortest path in the plate. If the collision point is located at (x p ,y p ), the i-th sensor is located at (x i ,y i ), the straight-line distance of the collision signal propagation path is:
[0013]
[0014] The path difference Δl between every two sensor signals is obtained by subtraction, and a total of six sets of theoretical path difference values can be obtained;
[0015] Since the S0 mode in the Lamb wave has the fastest propagation speed and is not easily affected by other modes in the time domain, the arrival time of the S0 mode is used as the arrival time of the signal. It is believed that the propagation speed of the S0 mode Lamb wave in the low frequency band changes little and is approximately a constant. Therefore, the theoretical path difference is divided by the speed of sound to obtain the theoretical time difference:
[0016] Δt i =Δl i / v(i=1,2,...,6)
[0017] The six sets of theoretical time differences form a vector T, which is only related to the node position:
[0018] T=[Δt1,Δt2,...,Δt6]
[0019] This vector is the collision feature vector of the grid node;
[0020] b. For the leakage signal, one of the four acoustic emission sensors is used as a reference sensor, and the other three acoustic emission sensors are used as measurement sensors. Excitation is applied to each point to obtain a characteristic ratio vector;
[0021] A signal generator is used to generate an excitation signal at the grid node to simulate the continuous sound signal generated during leakage. The type of the simulated signal is a chirp signal with a signal frequency of 50kHz to 300kHz, a signal excitation period of 10ms, and an excitation time of 1s. The frequency domain eigenvalues of the signals collected from each grid are extracted. The sensor collects the signal. Since the A0 mode has the largest energy in the signal, the signal A0 mode is selected as the eigenmode for subsequent processing. Through Fourier transform, the frequency domain information of each grid point signal corresponding to each channel signal is obtained. The signal collected by sensor i can be expressed as:
[0022]
[0023] Where: α i Represents the propagation attenuation coefficient, which is related to boundary reflection, Lamb wave dispersion and attenuation;
[0024] P i represents the frequency response function of the sensor numbered i;
[0025] A(f) represents the signal amplitude as a function of signal frequency;
[0026] φ i Represents the phase of the signal. When extracting energy information in the frequency domain, there is no need to consider the phase function of the signal.
[0027] The signals received by the three measuring sensors are compared with the signal received by the selected reference sensor to obtain the ratio V:
[0028]
[0029] Since the path that the signal propagates to the sensor is constant for each grid point when it is generated, and the sensor position and model used in the measurement are fixed, the attenuation coefficient ratio and frequency response coefficient ratio of the two channel signals are fixed values, and the characteristic ratio between different signals is only related to the signal propagation path in the board;
[0030] The received signal is divided into multiple frequency bands in the frequency domain. The sum of the energy of each frequency band can represent the eigenvalue Q ij , where i represents the i-th sensor, j represents the j-th frequency segment, and the characteristic ratio R j It can be expressed as:
[0031]
[0032] The characteristic ratio vector V can be expressed as:
[0033] V=[R1,R2,...,R n ]
[0034] This vector is the leakage feature vector of the grid node;
[0035] 3) Determine the signal type: Based on the different characteristics of the collision signal and the leakage signal, the energy comparison method is used to distinguish and judge the two situations:
[0036] When the sensor numbered i receives an unknown acoustic emission signal S i (t), use a multi-scale sliding window to compare the two sides, find the local maximum of the signal, and determine the time t when the signal peak occurs m , select the energy of the peak area as the energy reference value E m , assuming that the time required to sample 5000 points of the signal is t0, then define:
[0037]
[0038] Select the energy value E of the time domain signal area after a certain length of time t , which is defined as:
[0039]
[0040] Compare the energy value of the peak area with the energy value of the time domain signal area after a certain period of time. If there is E t ≤0.7E m , if the energy difference between the two is large and meets the characteristics of a collision event, it is judged as a collision event; otherwise, it is judged as a leakage event;
[0041] 4) Output positioning results: According to the judgment results, the signal is processed in different ways to achieve the final positioning and output the positioning results:
[0042] a. If the result of the judgment is a collision event, the four acoustic emission signals received are first processed. The acoustic emission signal propagates in the thin plate structure in the form of Lamb waves. Lamb waves have multi-modal and dispersion characteristics, and the wave speeds of each mode are different. According to the Lamb wave dispersion curve and the characteristics of the signal, the signal is band-pass filtered to filter out the noise signal and select the frequency band with more obvious Lamb wave characteristics in the retained signal to facilitate subsequent signal processing. G f represents the transfer function of the bandpass filter, then the sensor numbered i receives the unknown acoustic emission signal S i (t) The signal after preliminary filtering can be expressed as:
[0043] x i (t) = S i (t)G f
[0044] Then, the generalized weighted correlation delay estimation algorithm is used to extract the time difference, and the filtered sensor signal x i (t) Perform Fourier transform to obtain the frequency domain data X of the signal i (f) The four sensors perform cross-correlation calculations between each other, that is, conjugate the spectrum of one signal and multiply it with the spectrum of another signal to obtain two segments of signal x i (t) and x j (t) the cross-correlation signal X ij (f):
[0045] X ij (f) = X i (f)X j * (f)
[0046] Select the appropriate weighting function H(f) according to the environment to perform weighting and calculate the weighted cross power spectrum R of the two signal data segments ij (τ):
[0047]
[0048] Then, the weighted cross power spectrum is inversely Fourier transformed to obtain time domain data, and the peak position of the time domain data is detected. The delay estimation value can be calculated based on the position of the cross correlation peak, and the time difference between the signals reaching the two sensors can be obtained; one calculation can be performed between each two sensor signals, so a total of six groups of actual arrival time differences can be obtained, forming the actual time difference vector T';
[0049] Compare the actual time difference vector with the feature vectors of each grid node. Assume that the error between the collision feature vector of the grid node at the i-th row and j-th column position and the actual time difference vector of the unknown sound source signal is the smallest, then this node is considered the best matching position:
[0050] (x,y)=min{|T'-T ij |}| (x,y)
[0051] The node with the highest vector matching and the smallest difference is the collision source, and the coordinates (x,y) corresponding to this node are the collision source position, achieving the final positioning;
[0052] b. If the judgment result is a leakage event, the sensor collects the signal of the unknown leakage source;
[0053] First, preprocess the signal by the Empirical Wavelet Transform (EWT) method. This method is an adaptive decomposition method for non-stationary signals, which can adaptively select frequency bands and overcome the mode mixing problem caused by the discontinuity of the signal time-frequency scale;
[0054] First, perform Fourier transform on the leakage signal, normalize it to the range of 2π, divide the normalized Fourier spectrum of the signal into N continuous intervals, and the boundaries between each segment are ω n (n = 0,1,...,N), where ω0 = 0, ω N = π, search for the local maxima of the spectrum to determine the remaining N - 1 boundaries. Assume that the algorithm finds M local maxima points in the spectrum, sort them in descending order. When M≥N, retain the first N - 1 local maxima as the boundary values. When M<N, retain all the local maxima and correct the value of N to make N = M, and then
[0055] Define a transition section centered on the boundary ω n with a width of 2τ n The width of which is defined as:
[0056]
[0057] Construct a wavelet filter bank for filtering in each small interval [ω n-1 ,ω n (n = 0,1,...,N). The empirical wavelet function ψ n (ω) and the empirical scaling function can be expressed as:
[0058]
[0059]
[0060] where: β(x) = x 4(35-84x+70x 2 -20x 3 ), a wavelet filter bank is constructed from the above formula to decompose and modally separate the signal. The A0 modal signal obtained after EWT processing can be expressed as:
[0061]
[0062] Then, the frequency domain feature ratio vector of the actual continuous sound signal is constructed by the same method as in step 2) b., and the actual feature ratio vector V' is constructed;
[0063] Match the feature ratio vector of the unknown leakage source with the feature vector of each node, and evaluate the matching degree of the signal by the minimum variance method. If the frequency domain feature ratio vector of the grid node at the i-th row and j-th column has the smallest error with the frequency domain feature ratio vector of the unknown sound source signal, then the node is considered to be the best matching position:
[0064] (x,y)=min{|V'-V ij |}| (x,y)
[0065] The coordinates (x, y) corresponding to the node are the location of the leakage source, achieving the final positioning.
[0066] The advantages and beneficial effects of the present invention are:
[0067] The integrated detection method for spacecraft structure collision and leakage based on acoustic emission of the present invention can accurately determine whether the burst signal received by the sensor is a collision signal or a leakage signal. Two positioning detection functions are realized by processing data with two different algorithms through a set of measuring devices. In addition, the method has no continuity requirements for the spacecraft structure and can realize measurement on complex test surfaces with discontinuous geometric structures. It can meet the two positioning requirements of spacecraft structure health monitoring, and the measurement is convenient and has high measurement accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0068] Figure 1 It is a schematic diagram of the detection system of the present invention;
[0069] Figure 2 It is a time domain diagram of a typical collision signal and leakage signal of the present invention;
[0070] Figure 3 It is a flow chart of the present invention. DETAILED DESCRIPTION
[0071] The present invention is further described in detail below through specific examples. The following examples are only illustrative and not restrictive, and the protection scope of the present invention cannot be limited thereto.
[0072] like Figure 3As shown, the present invention proposes an integrated detection method for spacecraft structure collision and leakage based on acoustic emission, and its innovation lies in: the steps of the method are:
[0073] 1) System construction: The detection system used in the method includes an acoustic emission sensor, a preamplifier, a data acquisition card and a computer. The schematic diagram of the detection system is shown in Figure 1 shown.
[0074] The acoustic emission sensors are pre-distributed and coupled at certain selected positions on the surface of the structure to be tested to receive the acoustic signals generated by collision and leakage events, covering the entire cabin. The acoustic emission sensors use coupling agents to ensure close contact between the piezoelectric sensor and the metal of the tested part. The acoustic emission sensors in each row and column are arranged at a fixed interval to form an acoustic emission sensor network array. The number and position can be changed according to the requirements of the specific structure. A certain number of acoustic emission sensors evenly distributed on the surface of the structure to be tested are divided into units. With four acoustic emission sensors as a small unit, the structure to be tested can be divided into several small units. The units are flexibly divided according to the geometric characteristics of the structure to be tested and the distribution positions of the acoustic emission sensors. Some acoustic emission sensors at the intersection can belong to different small units at the same time. In each small unit, a rectangular coordinate system is established with the geometric center position of the unit as the origin, and the small unit is divided into four quadrants, so that each of the four quadrants of each small unit has an acoustic emission sensor, and they are named as sensors 1, 2, 3, and 4 of the unit respectively.
[0075] Each acoustic emission sensor is connected to a preamplifier in turn to amplify the weak signal received by the acoustic emission sensor and reduce the noise signal to improve the signal-to-noise ratio of the acoustic emission signal. The preamplifier is connected to the acquisition card and computer in turn to collect the signal data received by the acoustic emission sensor and perform subsequent data processing.
[0076] 2) Establish a mapping grid: Take a small unit as the smallest unit for analysis and processing, divide the grid inside the small unit, generate several grid nodes, and establish the corresponding feature vector for each grid node. Since the collision signal and the leakage signal have different characteristics, two feature vectors need to be established for each grid node.
[0077] a. For the collision signal, use sensors 1, 2, 3, and 4 to calculate the theoretical time difference characteristic matrix of each node. The acoustic signal propagates along the shortest path in the plate. If the collision point is located at (x p ,y p ), the i-th sensor is located at (x i ,y i ), the straight-line distance of the collision signal propagation path is:
[0078]
[0079] The path difference Δl between every two sensor signals is obtained by subtraction, and a total of six groups of theoretical path difference values can be obtained.
[0080] Since the S0 mode of Lamb waves has the fastest propagation speed and is not easily affected by other modes in the time domain, the arrival time of the S0 mode is used as the arrival time of the signal. It is believed that the propagation speed of the S0 mode Lamb wave in the low frequency band changes little and is approximately a constant, so the theoretical path difference is divided by the speed of sound to obtain the theoretical time difference:
[0081] Δt i =Δl i / v(i=1,2,...,6)
[0082] The six sets of theoretical time differences form a vector T, which is only related to the node position:
[0083] T=[Δt1,Δt2,...,Δt6]
[0084] This vector is the collision feature vector of the mesh node.
[0085] b. For the leakage signal, one of the four acoustic emission sensors is used as the reference sensor, and the other three acoustic emission sensors are used as measurement sensors. Excitation is applied to each point to obtain the characteristic ratio vector.
[0086] A signal generator is used to generate an excitation signal at the grid node to simulate the continuous sound signal generated during leakage. The type of the simulated signal is a chirp signal with a signal frequency of 50kHz to 300kHz, a signal excitation period of 10ms, and an excitation time of 1s. The frequency domain eigenvalues of the signals collected from each grid are extracted. The sensor collects the signal. Since the A0 mode has the largest energy in the signal, the signal A0 mode is selected as the characteristic mode for subsequent processing. Through Fourier transform, the frequency domain information of each grid point signal corresponding to each channel signal is obtained. The signal collected by sensor i can be expressed as:
[0087]
[0088] Among them, α i It represents the propagation attenuation coefficient, which is related to boundary reflection, Lamb wave dispersion and attenuation. i represents the frequency response function of sensor number i, A(f) represents the signal amplitude of the signal frequency function, φ i Represents the phase of the signal. When extracting energy information in the frequency domain, there is no need to consider the phase function of the signal.
[0089] The signal received by each measuring sensor is compared with the signal received by the reference sensor to obtain the ratio V:
[0090]
[0091] Since the path that the signal propagates to the sensor is constant for each grid point when it is generated, and the sensor position and model used during measurement are fixed, the attenuation coefficient ratio and frequency response coefficient ratio of the two channel signals are fixed values, and the characteristic ratio between different signals is only related to the signal propagation path in the board.
[0092] The received signal is divided into multiple frequency bands in the frequency domain. The sum of the energy of each frequency band can represent the eigenvalue Q ij , where i represents the i-th sensor, j represents the j-th frequency segment, and the characteristic ratio R j It can be expressed as:
[0093]
[0094] The characteristic ratio vector V can be expressed as:
[0095] V=[R1,R2,...,R n ]
[0096] This vector is the leakage eigenvector of the grid node.
[0097] 3) Determine the signal type: The signal generated by the collision is a burst-type sound signal with a clear arrival time, while the signal generated by the leakage is a continuous sound signal with little change in amplitude within a period of time after the collision. The two have certain differences in the time domain. The typical time domain diagrams of collision and leakage signals are as follows: Figure 2 shown.
[0098] According to the different characteristics of collision signal and leakage signal, the energy comparison method is used to distinguish and judge the two situations. i (t), use a multi-scale sliding window to compare the two sides, find the local maximum of the signal, and determine the time t when the signal peak occurs m The energy in the peak area is selected as the energy reference value E m , assuming that the time required to sample 5000 points of the signal is t0, then define:
[0099]
[0100] Then select the energy value E of the time domain signal area after a certain length of time t , which is defined as:
[0101]
[0102] Compare the energy value of the peak area with the energy value of the time domain signal area after a certain period of time. If there is E t ≤0.7E m If the energy difference between the two is large and meets the characteristics of a collision event, it is judged as a collision event. Otherwise, it is judged as a leakage event.
[0103] 4) Output positioning results: Based on the judgment results, the signal is processed by different methods to achieve the final positioning and output the positioning results.
[0104] a. If the result of the judgment is a collision event, the signals of sensors 1, 2, 3, and 4 are processed first. The acoustic emission signal propagates in the thin plate structure in the form of Lamb waves. Lamb waves have multi-modal and dispersion characteristics, and the wave speeds of each mode are different. When the propagation distance of the Lamb wave increases, the Lamb wave dispersion phenomenon will become obvious. The wave speeds of the Lamb wave modes overlap in the high-frequency band, which is not conducive to acoustic positioning. Since there is noise in the outside world, there are low-frequency signals in the signal before the Lamb wave arrives. Therefore, the collision signal is band-pass filtered with a frequency band of 150kHz to 250kHz, which can not only filter out noise signals and high-frequency Lamb waves, but also retain the frequency bands with more obvious Lamb wave characteristics in the signal.
[0105] Then the generalized weighted correlation delay estimation algorithm is used to extract the time difference. i (t) Perform Fourier transform to obtain the frequency domain data X of the signal i (f) The four sensors perform cross-correlation calculations between each other, that is, conjugate the spectrum of one signal and multiply it with the spectrum of another signal to obtain two segments of signal x i (t) and x j (t) the cross-correlation signal X ij (f):
[0106] X ij (f) = X i (f)X j * (f)
[0107] Select the appropriate weighting function H(f) according to the environment to perform weighting and calculate the weighted cross power spectrum R of the two signal data segments ij (τ):
[0108]
[0109] Then, perform an inverse Fourier transform on the weighted cross-power spectrum to obtain time-domain data. Detect the peak position of the time-domain data. According to the position where the cross-correlation peak is located, the time-delay estimation value can be calculated, and the time difference between the signals reaching the two sensors can be obtained. A calculation can be performed for each pair of sensor signals. Therefore, a total of six groups of actual arrival time differences can be obtained, forming an actual time difference vector T'.
[0110] Compare the actual time difference vector with the eigenvector of each grid node. The smaller the deviation, the greater the possibility of a collision event occurring at that point. Assume that the error between the collision eigenvector of the grid node at the i-th row and j-th column position and the actual time difference vector of the unknown sound source signal is the smallest. Then, this node is considered the best matching position:
[0111] (x,y) = min{|T' - T ij |}| (x,y)
[0112] The node with the highest vector matching and the smallest difference is the collision source, and the corresponding coordinates (x,y) of this node are the collision source position, realizing the final positioning.
[0113] b. If the judgment result is a leakage event, the sensor collects the signal of the unknown leakage source. First, preprocess the signal by the empirical wavelet transform (EWT) method. This method is an adaptive decomposition method for non-stationary signals, which can adaptively select frequency bands and overcome the mode mixing problem caused by the discontinuous time-frequency scale of the signal. First, perform a Fourier transform on the leakage signal and normalize it to the range of 2π. Divide the normalized Fourier spectrum of the signal into N continuous intervals, and the boundaries between each segment are ω n (n = 0,1,...,N), where ω0 = 0, ω N = π, search for the local maxima of the spectrum to determine the remaining N - 1 boundaries. Assume that the algorithm finds M local maxima points in the spectrum and sorts them in descending order. When M ≥ N, retain the first N - 1 local maxima as the boundary values. When M < N, retain all the local maxima and correct the value of N to make N = M. Then
[0114] To facilitate the construction of the filter in the subsequent process, a transition segment centered on the boundary ω n with a width of 2τ n is defined, and its width is defined as:
[0115]
[0116] Construct a wavelet filter bank for filtering in each small interval [ω n-1 , ω n (n = 0,1,...,N). The empirical wavelet function ψ n (ω) and the empirical scaling function It can be expressed as:
[0117]
[0118]
[0119] Where β(x) = x 4 (35-84x+70x 2 -20x 3 ).
[0120] The wavelet filter bank is constructed from the above formula to decompose and separate the signal. It can effectively identify the location of the characteristic information in the Fourier spectrum of the signal, adaptively extract the different frequency components of the signal, and simplify the difficulty of extracting the frequency domain features of the signal. The A0 modal signal obtained after EWT processing can be expressed as:
[0121]
[0122] Then, the frequency domain feature ratio vector of the actual continuous sound signal is constructed by the same method as in step 2) b., thereby constructing the actual feature ratio vector V'.
[0123] Match the feature ratio vector of the unknown leakage source with the feature vector of each node, and evaluate the matching degree of the signal by the minimum variance method. If the frequency domain feature ratio vector of the grid node at the i-th row and j-th column has the smallest error with the frequency domain feature ratio vector of the unknown sound source signal, then the node is considered to be the best matching position:
[0124] (x,y)=min{|V'-V ij |}| (x,y)
[0125] The coordinates (x, y) corresponding to the node are the location of the leakage source, achieving the final positioning.
[0126] Although the embodiments and drawings of the present invention are disclosed for illustrative purposes, those skilled in the art will appreciate that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.
Claims
1. A spacecraft structure collision and leakage integrated detection method based on acoustic emission, characterized by: The steps of the method are: 1) System construction: The detection system used in the method includes a data acquisition card, a computer, a plurality of acoustic emission sensors and a preamplifier. The plurality of acoustic emission sensors are pre-distributedly coupled and arranged at selected positions on the surface of the spacecraft structure to be tested to receive acoustic signals generated by collision and leakage events. The acoustic emission sensors in each row and each column are arranged at a fixed interval to form an acoustic emission sensor network array; The acoustic emission sensors uniformly distributed on the surface of the spacecraft structure to be tested are divided into units, with four acoustic emission sensors forming a small unit, and the spacecraft structure to be tested is divided into several small units. In each small unit, a rectangular coordinate system is established with the geometric center position of the unit as the origin, and the small unit is divided into four quadrants, so that each of the four quadrants of each small unit has an acoustic emission sensor, and they are named No. 1, 2, 3, and 4 sensors of the unit respectively; the several acoustic emission sensors are sequentially connected to the several preamplifiers to amplify the weak signals received by the acoustic emission sensors; the preamplifiers are sequentially connected to the data acquisition card and the computer to collect the signal data received by the acoustic emission sensors and perform subsequent data processing; 2) Establish a mapping grid: Take a small unit as the minimum unit for analysis and processing, divide the grid inside the small unit, generate several grid nodes, and establish the corresponding feature vector for each grid node. Since the collision signal and the leakage signal have different characteristics, two feature vectors need to be established for each grid node: a. For the collision signal, calculate the theoretical time difference characteristic matrix of each node. The acoustic signal propagates along the shortest path in the plate. If the collision point is located at (x p ,y p ), the i-th sensor is located at (x i ,y i ), the straight-line distance of the collision signal propagation path is: The path difference Δl between every two sensor signals is obtained by subtraction, and a total of six sets of theoretical path difference values can be obtained; Since the S0 mode in the Lamb wave has the fastest propagation speed and is not easily affected by other modes in the time domain, the arrival time of the S0 mode is used as the arrival time of the signal. It is believed that the propagation speed of the S0 mode Lamb wave in the low frequency band changes little and is approximately a constant. Therefore, the theoretical path difference is divided by the speed of sound to obtain the theoretical time difference: Δt i =Δl i / v(i=1,2,...,6) The six sets of theoretical time differences form a vector T, which is only related to the node position: T=[Δt1,Δt2,...,Δt6] This vector is the collision feature vector of the grid node; b. For the leakage signal, one of the four acoustic emission sensors is used as a reference sensor, and the remaining three acoustic emission sensors are used as measurement sensors. Excitation is applied to each point to obtain a characteristic ratio vector; A signal generator is used to generate an excitation signal at the grid node to simulate the continuous sound signal generated during leakage. The type of the simulated signal is a chirp signal with a signal frequency of 50kHz to 300kHz, a signal excitation period of 10ms, and an excitation time of 1s. The frequency domain eigenvalues of the signals collected from each grid are extracted. The sensor collects the signal. Since the A0 mode has the largest energy in the signal, the signal A0 mode is selected as the eigenmode for subsequent processing. Through Fourier transform, the frequency domain information of each grid point signal corresponding to each channel signal is obtained. The signal collected by sensor i can be expressed as: Where: α i Represents the propagation attenuation coefficient, which is related to boundary reflection, Lamb wave dispersion and attenuation; P i represents the frequency response function of the sensor numbered i; A(f) represents the signal amplitude as a function of signal frequency; φ i Represents the phase of the signal. When extracting energy information in the frequency domain, there is no need to consider the phase function of the signal. The signals received by the three measuring sensors are compared with the signal received by the selected reference sensor to obtain the ratio V: Since the path that the signal propagates to the sensor is constant for each grid point when it is generated, and the sensor position and model used in the measurement are fixed, the attenuation coefficient ratio and frequency response coefficient ratio of the two channel signals are fixed values, and the characteristic ratio between different signals is only related to the signal propagation path in the board; The received signal is divided into multiple frequency bands in the frequency domain. The sum of the energy of each frequency band can represent the eigenvalue Q ij , where i represents the i-th sensor, j represents the j-th frequency segment, and the characteristic ratio R j It can be expressed as: The characteristic ratio vector V can be expressed as: V=[R1,R2,...,R n ] This vector is the leakage feature vector of the grid node; 3) Determine the signal type: Based on the different characteristics of the collision signal and the leakage signal, the energy comparison method is used to distinguish and judge the two situations: When the sensor numbered i receives an unknown acoustic emission signal S i (t), use a multi-scale sliding window to compare the two sides, find the local maximum of the signal, and determine the time t when the signal peak occurs m , select the energy of the peak area as the energy reference value E m , assuming that the time required to sample 5000 points of the signal is t0, then define: Select the energy value E of the time domain signal area after a certain length of time t , which is defined as: Compare the energy value of the peak area with the energy value of the time domain signal area after a certain period of time. If there is E t ≤0.7E m , if the energy difference between the two is large and meets the characteristics of a collision event, it is judged as a collision event; otherwise, it is judged as a leakage event; 4) Output positioning results: According to the judgment results, the signal is processed in different ways to achieve the final positioning and output the positioning results: a. If the result of the judgment is a collision event, the four acoustic emission signals received are first processed. The acoustic emission signal propagates in the thin plate structure in the form of Lamb waves. Lamb waves have multi-modal and dispersion characteristics, and the wave speeds of each mode are different. According to the Lamb wave dispersion curve and the characteristics of the signal, the signal is band-pass filtered to filter out the noise signal and select the frequency band with more obvious Lamb wave characteristics in the retained signal to facilitate subsequent signal processing. G f represents the transfer function of the bandpass filter, then the sensor numbered i receives the unknown acoustic emission signal S i (t) The signal after preliminary filtering can be expressed as: x i (t)=S i (t)G f Then, the generalized weighted correlation delay estimation algorithm is used to extract the time difference, and the filtered sensor signal x i (t) Perform Fourier transform to obtain the frequency domain data X of the signal i (f) The four sensors perform cross-correlation calculations between each other, that is, conjugate the spectrum of one signal and multiply it with the spectrum of another signal to obtain two segments of signal x i (t) and x j (t) the cross-correlation signal X ij (f): X ij (f)=X i (f)X j * (f) Select the appropriate weighting function H(f) according to the environment to perform weighting and calculate the weighted cross power spectrum R of the two signal data segments ij (τ): Then, the weighted cross power spectrum is inversely Fourier transformed to obtain time domain data, and the peak position of the time domain data is detected. The delay estimation value can be calculated based on the position of the cross correlation peak, and the time difference between the signals reaching the two sensors can be obtained; a calculation is performed once between each two sensor signals, so a total of six groups of actual arrival time differences can be obtained, forming an actual time difference vector T'; The actual time difference vector is compared with the feature vector of each grid node. If the collision feature vector of the grid node at the i-th row and j-th column has the smallest error with the actual time difference vector of the unknown sound source signal, then the node is considered to be the best matching position: (x,y)=min{|T'-T ij |}| (x,y) The node with the highest vector matching and the smallest difference is the collision source, and the coordinates (x, y) corresponding to the node are the location of the collision source, thus achieving the final positioning; b. If the judgment result is a leakage event, the sensor collects the signal of the unknown leakage source; Firstly, the signal is preprocessed by the empirical wavelet transform (EWT) method, which is an adaptive decomposition method for non-stationary signals. It can adaptively select frequency bands and overcome the modal aliasing problem caused by the discontinuity of signal time-frequency scale. First, perform a Fourier transform on the leakage signal and normalize it to the range of 2π. Divide the normalized Fourier spectrum of the signal into N consecutive intervals, and the boundaries between each segment are ω n (n = 0, 1,..., N), where ω0 = 0, ω N = π. Search for the local maxima of the spectrum to determine the remaining N - 1 boundaries. Assume that the algorithm finds M maxima points in the spectrum and sort them in descending order. When M ≥ N, retain the first N - 1 maxima as the boundary values. When M < N, retain all the maxima and correct the value of N to make N = M, and then Defined by the boundary ω n Centered at 2τ n The width of the transition segment is defined as: In each small interval [ω n-1 ,ω n ](n=0,1,...,N) constructs a wavelet filter bank for filtering, and the empirical wavelet function ψ n (ω) and the empirical scaling function It can be expressed as: Where: β(x) = x 4 (35-84x+70x 2 -20x 3 ), a wavelet filter bank is constructed from the above formula to decompose and modally separate the signal. The A0 modal signal obtained after EWT processing can be expressed as: Then, the frequency domain feature ratio vector of the actual continuous sound signal is constructed by the same method as in step 2) b., and the actual feature ratio vector V' is constructed; Match the feature ratio vector of the unknown leakage source with the feature vector of each node, and evaluate the matching degree of the signal by the minimum variance method. If the frequency domain feature ratio vector of the grid node at the i-th row and j-th column has the smallest error with the frequency domain feature ratio vector of the unknown sound source signal, then the node is considered to be the best matching position: (x,y)=min{|V'-V ij |}| (x,y) The coordinates (x, y) corresponding to the node are the location of the leakage source, achieving the final positioning.
Citation Information
Patent Citations
Pipeline impact and leakage locating method based on distributed FBG (Fiber Bragg Grating) sensing network
CN104595727A
Method for detecting structure health of spacecraft based on acoustic sensor
CN104597126A
Spacecraft discontinuous structure leakage positioning system and method based on frequency domain characteristic spectrum mapping
CN111307947A
Method for monitoring space debris impact on spacecraft sealed cabin of special-shaped structure
CN111645886A
Automobile collision fuel oil leakage detection method based on image recognition
CN113514193A
