CN114061848B - Method for identifying leak hole of reinforced sealing structure of spacecraft - Google Patents

Abstract

The invention discloses a method for identifying leakage holes of a reinforced sealing structure of a spacecraft, which comprises the following steps: s1, performing a simulated leakage experiment, extracting characteristic values, and performing learning training to form an identification model library; s2, collecting leakage sound signals, performing 30kHz-500kHz band-pass filtering pretreatment on the leakage sound signals, and obtaining a frequency band f with less reinforcing rib attenuation _d ‑f _u The method comprises the steps of carrying out a first treatment on the surface of the S3, positioning the leakage holes; s4, compensating according to the number of the bars passing signals; s5, extracting leakage sound signal characteristics, and comparing the leakage sound signal characteristics with an identification model library to obtain leakage hole characteristics. According to the invention, the spacecraft leakage identification of EMD-WPD feature fusion is realized by applying the reliefF algorithm, and the accuracy is greatly improved by utilizing the fusion signal processing method to identify different leak holes compared with the prior pure spectrum signal leak holes, and the identification of the shape, the size and other characteristics of the leak holes is realized.

Description

A method for identifying leaks in reinforced sealing structures of spacecraft

Technical Field

The invention relates to the technical field of spacecraft leakage detection, and in particular to a method for identifying leakage holes in a spacecraft reinforced sealing structure.

Background Art

With the development of aerospace technology and the increasing frequency of human space activities, the number of space debris has increased significantly. Once the debris collides with the spacecraft, it will cause the spacecraft sealing structure to leak, seriously affecting the spacecraft's on-orbit operation. Timely and accurate detection of leaks and identification of the location, size, shape and other characteristics of the leak hole can provide support for subsequent spacecraft plugging and repair and astronaut emergency escape. Currently, the commonly used spacecraft leak detection technologies include pressure change leak detection, infrared thermal imaging leak detection, helium mass spectrometry leak detection, acoustic leak detection, etc. Pressure change leak detection can only determine whether there is a leak, but cannot determine the leak location; infrared thermal imaging leak detection has poor sensitivity and can only perform qualitative analysis; helium mass spectrometry leak detection can detect leaks with extremely low leak rates, but cannot quickly locate the leak. Once the search range is too large, it is more difficult to find the leak; acoustic leak detection is an emerging leak detection technology that can quickly determine the leak location, but it is difficult to achieve quantitative analysis of the leak, and structures such as reinforcement ribs affect the propagation of sound waves and thus affect the accuracy of acoustic leak detection. This patent proposes a method for identifying leaks in a spacecraft reinforced sealing structure, which realizes the judgment of leaks and the identification of leak size and shape through database training and learning methods.

Summary of the invention

The purpose of the present invention is to solve the above-mentioned problem and to propose a method for identifying leaks in a spacecraft reinforced sealing structure.

In order to achieve the above object, the present invention adopts the following technical solutions:

A method for identifying leaks in a spacecraft reinforced sealing structure comprises the following steps:

S1. Establish an identification model library: Conduct simulated leakage experiments on leak holes of different shapes and sizes in advance, collect leakage sound signals, extract characteristic values for learning and training to form an identification model library;

S2. Collection of leakage sound signals: Collect leakage sound signals. To remove background noise, generally 30kHz-500kHz filtering pre-processing is performed first. The attenuation of the sound signal before and after passing through the reinforcement is observed, and the sound signal frequency band f _{d -fu} with weaker attenuation is selected;

y(f)是不同频率下信号幅值，传感器S₂、S₃的计算方法相同，以此建立方程组：S3. Leak hole location: Establish a coordinate system, use the data of three sensors (numbered S ₁ , S ₂ , S ₃ ), and use FIR filters to obtain the f _{d -fu} band energy signals of these sensors. Take the S ₁ sensor as an example,

y(f) is the signal amplitude at different frequencies. The calculation method of sensors _S2 and _S3 is the same, and the equation group is established:

(x ₁ ,y ₁ )(x ₂ ,y ₂ )(x ₃ ,y ₃ ) are the known coordinates of the three sensors. The linear iteration algorithm can be used to solve the equations to obtain the leak coordinates (x,y);

S4. Signal cross-rib compensation: After obtaining the coordinates of the leak hole, the original data of the signal of the sensor closest to the leak hole is used. The number of reinforcing ribs α passed by the leak hole and the sensor in the straight-line propagation process can be known. For the signal cross-rib compensation, the attenuation coefficient of each reinforcing rib in the frequency band f _{d -fu} is β(f), so the frequency band of y(f) is multiplied by the compensation coefficient α*β(f) (α is the number of cross-ribs, β(f) is the attenuation coefficient);

S5. Leakage hole identification: Extract the leakage sound signal characteristics and compare them with the identification model library to obtain the leakage hole characteristics.

Preferably, the method for determining the leakage identification model library by ground learning training in step S1 comprises the following steps:

圆孔泄漏、

圆孔泄漏、

圆孔泄漏、1mm×1mm方孔泄漏、三边1mm三角孔泄漏、0.5mm×2mm长方孔泄漏的七种信号，每次提取信号都同时用两个相同的传感器(A₁传感器和A₂传感器)，各组声学信号各提取3s，信号提取时，传感器距离漏孔中心10cm；A1. Extract the following leakage signals in the simulated spacecraft internal and external pressure and vacuum: extract the distribution without leakage,

Round hole leakage,

Round hole leakage,

There are seven types of signals: circular hole leakage, 1mm×1mm square hole leakage, three-sided 1mm triangle hole leakage, and 0.5mm×2mm rectangular hole leakage. Two identical sensors ( _A1 sensor and _A2 sensor) are used simultaneously for each signal extraction. Each group of acoustic signals is extracted for 3s. When extracting signals, the sensor is 10cm away from the center of the leak hole.

A2. Cut each group of extracted signals into 200 groups (0.015s each), where _A1 sensor data is used as the test group data to test the accuracy of the classification model, and _A2 sensor data is used as the training group data of the classification algorithm model;

A3. Perform digital filtering on all raw signal data. The effective frequency band of the selected acoustic sensor includes 30kHz-500kHz. Considering the influence of background noise, it is generally necessary to filter out the components below 30kHz. Therefore, a bandpass FIR filter is used for filtering to obtain the signal components between 30kHz-500kHz.

A4. These data are processed by EMD algorithm and WPD algorithm respectively. The interpolation function of EMD adopts cubic spline interpolation, and the wavelet mother function of WPD adopts dmey wavelet. The two algorithms obtain IMF_1-IMF_n ₁ and subband signal 1-n ₂ respectively. The middle 4 groups of IMF obtained by EMD and the first 4 groups of subband signals obtained by WPD are taken. 9 eigenvalues are extracted from these 8 groups of signals respectively. EMD and WPD methods can obtain 36 features respectively, for a total of 72 features. These 9 eigenvalues are time domain kurtosis factor, skewness factor and waveform factor, as well as frequency domain kurtosis factor, skewness factor, waveform factor, peak frequency, spectrum bandwidth and bandwidth centroid frequency, for a total of 9 feature parameters. The algorithms for the 9 feature parameters are as follows:

(x_n为时域信号值)；[1]Time domain kurtosis factor:

(x _n is the time domain signal value);

(x_n为信号时域值，

为信号时域均值)；[2] Time domain skewness factor:

(x _n is the signal time domain value,

is the time domain mean of the signal);

(x_rms为信号时域均方根值，x_avr为信号时域均值)；[3]Time domain waveform factor:

(x _rms is the root mean square value of the signal in time domain, x _avr is the mean value of the signal in time domain);

(y_n为频域值)；[4] Frequency domain kurtosis factor:

(y _n is the frequency domain value);

(y_n为信号频域值，

为信号频域值均值)；[5] Frequency domain skewness factor:

(y _n is the signal frequency domain value,

is the mean value of the signal frequency domain);

(y_rms为频域均方根值，y_avr为频域均值)；[6] Frequency domain waveform factor:

(y _rms is the root mean square value in the frequency domain, y _avr is the mean value in the frequency domain);

[7] Peak frequency: f _max = max(y(f))| _f ; (y(f) is the signal frequency domain function, the value of y(f) is called the frequency domain amplitude, and f _max represents the frequency point corresponding to the maximum amplitude in the frequency domain);

[8] Spectral bandwidth: f _dB = f _up - f _down (f _up is the maximum frequency point corresponding to the frequency domain amplitude y(f) being 0.3 times the maximum amplitude, f _down is the minimum frequency point corresponding to the frequency domain amplitude y(f) being 0.3 times the maximum amplitude);

[9]Bandwidth centroid frequency:

A5. The ReliefF feature evaluation algorithm is used to evaluate the 36 features in each of these two groups to obtain their respective weight values. The size of these weight values is equivalent to the distinguishing ability of the features themselves. The higher the weight value, the higher the distinguishing ability of the feature for different leaks. Application of the ReliefF algorithm: randomly select a sample R from a certain leak feature set A, and find k nearest neighbor sample sets H from samples of the same type as R, and find k nearest neighbor sample sets M from samples of different types from R, and then update the weight of each leak feature according to the following two formulas (1)(2):

The above formula gives the weights of 36×2 features corresponding to each leak;

A6. Sum and average the weights of the same features of each leak, so as to obtain the new weights of the 36 features of EMD and WPD, and then sort the new weights of the 36 features of the two groups, first select the top 2 features of each group and bring them into SVM classification training. The SVM kernel function uses the radial basis kernel function (RBF). The trained leak identification model library is tested with the test group data to see how accurate the discrimination is. If the accuracy is not enough, the number of high-weight features can be gradually increased. The experimental results show that the accuracy will increase with the increase of the number of features, but the training volume will undoubtedly increase. Therefore, it is necessary to determine the number of features to be brought in according to the required requirements.

A7. Finally determine the features selected for the simulation test and train the leak identification model library.

Preferably, in step S2, f _d and _fu are determined by using the same leak hole at the same distance, respectively, to collect signals and compare spectra when passing through the ribs and not passing through the ribs. The lower frequency limit of the part where the energy of the passing spectrum is greater than 80% of the energy of the not passing spectrum is f _d , and the upper frequency limit is _fu . At the same time, the attenuation β(f) of the frequency band f _{d -fu} can also be obtained.

In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

1. In this application, by applying the feature evaluation method of ReliefF, the screening and mixed application of the various feature values obtained by the two signal processing methods of EMD and WPD are realized. Compared with the previous method of obtaining feature values based on pure spectrum signals for identification, this EMD-WPD feature fusion recognition method based on ReliefF not only improves the correct recognition rate, but also effectively controls the number of features introduced and reduces unnecessary training.

2. In this application, the reinforcement ribs in the spacecraft structure have a great influence on the propagation of acoustic signals and the identification of leak holes. By applying this method, the number of ribs is calculated by realizing the location of the leak hole, and the signal is compensated according to the rib attenuation function, thereby overcoming the interference of the spacecraft reinforcement ribs on the identification of leak holes.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 shows a schematic diagram of a spacecraft reinforced structure sensor and a leak hole layout according to an embodiment of the present invention;

FIG2 shows a flow chart of leak detection and leak hole feature identification according to an embodiment of the present invention;

FIG. 3 shows a flow chart of establishing a leak hole identification model according to an embodiment of the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

Please refer to Figures 1-3, the present invention provides a technical solution:

A method for identifying leaks in a spacecraft reinforced sealing structure comprises the following steps:

S1. Conduct simulated ground training to determine the leak hole identification model library, and put the leak hole identification model library into the actual on-orbit operation of the spacecraft;

S2. Observe the energy change of the frequency band 30kHz-500kHz before and after the reinforcement, determine the reinforcement attenuation frequency band _fd - _fu , and obtain the reinforcement attenuation coefficient β(f);

y(f)是不同频率下信号幅值，传感器S₂、S₃的计算方法相同，以此建立方程组：S3. To locate the leak, first establish a coordinate system. The coordinates of each sensor are known. Since the f _d - _fu frequency band is less affected by the reinforcement ribs, the f _d - _fu frequency band signal is used for positioning. The positioning principle is: use at least 3 sensor data and use FIR filters to intercept the f _d - _fu frequency band signals of these sensors. Take the S ₁ sensor as an example.

y(f) is the signal amplitude at different frequencies. The calculation method of sensors _S2 and _S3 is the same, and the equation group is established:

(x ₁ ,y ₁ )(x ₂ ,y ₂ )(x ₃ ,y ₃ ) are the known coordinates of the three sensors. The linear iteration algorithm can be used to solve the equations to obtain the leak coordinates (x,y);

S4. After obtaining the coordinates of the leak hole, the original data of the signal of the sensor closest to the leak hole is used. The number of reinforcing ribs α passed by the leak hole and the sensor in the straight-line propagation process can be known. The signal is compensated for passing through the ribs. The attenuation coefficient of each reinforcing rib in the frequency band f _{d -fu} is β(f), so the frequency band of y(f) is multiplied by the compensation coefficient α*β(f) (α is the number of passing ribs, β(f) is the attenuation coefficient);

S5. Extract the leakage sound signal features and compare them with the identification model library to obtain the leakage hole features.

Specifically, as shown in FIG3 , the method for simulating ground training to determine the leak hole identification model library in step S1 includes the following steps:

圆孔泄漏、

圆孔泄漏、

圆孔泄漏、1mm×1mm方孔泄漏、三边1mm三角孔泄漏、0.5mm×2mm长方孔泄漏的七种信号。每次提取信号都同时用两个相同的传感器(A₁传感器和A₂传感器)，各组声学信号各提取3s，信号提取时，传感器距离漏孔中心10cm；A1. Extract the following leakage signals in the simulated spacecraft internal and external pressure and vacuum: extract the distribution without leakage,

Round hole leakage,

Round hole leakage,

There are seven types of signals: circular hole leakage, 1mm×1mm square hole leakage, three-sided 1mm triangular hole leakage, and 0.5mm×2mm rectangular hole leakage. Two identical sensors ( _A1 sensor and _A2 sensor) are used simultaneously for each signal extraction. Each group of acoustic signals is extracted for 3s. When extracting the signal, the sensor is 10cm away from the center of the leak hole.

A2. Cut each group of extracted signals into 200 groups (0.015s each), where _A1 sensor data is used as the test group data to test the accuracy of the classification model, and _A2 sensor data is used as the training group data of the classification algorithm model;

A3. Perform digital filtering on all raw signal data. The effective frequency band of the selected acoustic sensor includes 30kHz-500kHz. Considering the influence of background noise, a bandpass FIR filter is used for filtering to obtain the signal components between 30kHz-500kHz;

A4. These data are processed by EMD algorithm and WPD algorithm respectively. The interpolation function of EMD adopts cubic spline interpolation, and the wavelet mother function of WPD adopts dmey wavelet. The two algorithms obtain IMF_1-IMF_n ₁ and subband signal 1-n ₂ respectively. The middle 4 groups of IMF obtained by EMD and the first 4 groups of subband signals obtained by WPD are taken. 9 eigenvalues are extracted from these 8 groups of signals respectively. EMD and WPD methods can obtain 36 features respectively, for a total of 72 features. These 9 eigenvalues are time domain kurtosis factor, skewness factor and waveform factor, as well as frequency domain kurtosis factor, skewness factor, waveform factor, peak frequency, spectrum bandwidth and bandwidth centroid frequency, for a total of 9 feature parameters. The algorithms for the 9 feature parameters are as follows:

(x_n为时域信号值)；[1]Time domain kurtosis factor:

(x _n is the time domain signal value);

(x_n为信号时域值，

为信号时域均值)；[2] Time domain skewness factor:

(x _n is the signal time domain value,

is the time domain mean of the signal);

(x_rms为信号时域均方根值，x_avr为信号时域均值)；[3]Time domain waveform factor:

(x _rms is the root mean square value of the signal in time domain, x _avr is the mean value of the signal in time domain);

(y_n为频域值)；[4] Frequency domain kurtosis factor:

(y _n is the frequency domain value);

(y_n为信号频域值，

为信号频域值均值)；[5] Frequency domain skewness factor:

(y _n is the signal frequency domain value,

is the mean value of the signal frequency domain);

(y_rms为频域均方根值，y_avr为频域均值)；[6] Frequency domain waveform factor:

(y _rms is the root mean square value in the frequency domain, y _avr is the mean value in the frequency domain);

[7] Peak frequency: f _max = max(y(f))| _f ; (y(f) is the signal frequency domain function, the value of y(f) is called the frequency domain amplitude, and f _max represents the frequency point corresponding to the maximum amplitude in the frequency domain);

[8] Spectral bandwidth: f _dB = f _up - f _down (f _up is the maximum frequency point corresponding to the frequency domain amplitude y(f) being 0.3 times the maximum amplitude, f _down is the minimum frequency point corresponding to the frequency domain amplitude y(f) being 0.3 times the maximum amplitude);

[9]Bandwidth centroid frequency:

A5. The ReliefF feature evaluation algorithm is used to evaluate the 36 features in each of these two groups to obtain their respective weight values. The size of these weight values is equivalent to the distinguishing ability of the features themselves. The higher the weight value, the higher the distinguishing ability of the feature for different leaks. Application of the ReliefF algorithm: randomly select a sample R from a certain leak feature set A, and find k nearest neighbor sample sets H from samples of the same type as R, and find k nearest neighbor sample sets M from samples of different types from R, and then update the weight of each leak feature according to the following two formulas (3) and (4):

The above formula gives the weights of 36×2 features corresponding to each leak;

A6. Sum and average the weights of the same features of each leak, so as to obtain the new weights of the 36 features of EMD and WPD, and then sort the new weights of the 36 features of the two groups, first select the top 2 features of each group and bring them into SVM classification training. The SVM kernel function uses the radial basis kernel function (RBF). The trained leak identification model library is tested with the test group data to see how accurate the discrimination is. If the accuracy is not enough, the number of high-weight features can be gradually increased. The experimental results show that the accuracy will increase with the increase of the number of features, but the training volume will undoubtedly increase. Therefore, it is necessary to determine the number of features to be brought in according to the required requirements.

A7. Finally determine the features selected for the simulation test and the trained leak identification model library.

Specifically, as shown in FIG1 , to determine f _d and _fu in step S2, it is only necessary to use the same leak hole at the same distance, respectively, to collect signals and make spectrum comparisons when passing through the ribs and not passing through the ribs. The lower frequency limit of the part where the energy of the passing spectrum is greater than 80% of the energy of the not passing spectrum is f _d , and the upper frequency limit is _fu . At the same time, the attenuation β(f) of the frequency band f _{d -fu} can also be obtained.

To sum up, the leak identification method of the reinforced sealing structure of a spacecraft provided in this embodiment can effectively improve the ability to identify the occurrence of leaks and the characteristics of leaks, while also controlling the number of features that need to be brought into training. Compared with the previous method of obtaining characteristic values based on simple spectral signals for identification, this EMD-WPD feature fusion identification method based on ReliefF not only improves the correct recognition rate, but also realizes the identification of leak shape, size and other characteristics. It is an effective method for identifying the occurrence of spacecraft leaks.

The above description of the embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims (2)

1. The method for identifying the leak hole of the reinforced sealing structure of the spacecraft is characterized by comprising the following steps of:

s1, establishing an identification model library: leakage simulation experiments of leakage holes with different shapes and sizes are performed in advance, leakage sound signals are collected, characteristic values are extracted, and learning and training are performed to form an identification model;

the method for establishing the identification model library in the step S1 comprises the following steps:

A1. the following leakage signal extraction is performed in the internal and external air pressure and vacuum degree of the simulated spacecraft: no leakage of distribution extraction,

Leakage of round hole>

Leakage of round hole>

Seven signals of round hole leakage, 1mm multiplied by 1mm square hole leakage, three-side 1mm triangular hole leakage and 0.5mm multiplied by 2mm rectangular hole leakage are simultaneously extracted by two identical A signals each time ₁ Sensor and A ₂ The sensors are used for extracting each group of acoustic signals for 3s, and the distance between the sensors and the center of the leak hole is 10cm during signal extraction;

A2. cutting each extracted signal into 200 groups of 0.015s each, wherein A ₁ The sensor is used as test group data for testing the accuracy of the classification model, A ₂ The sensor data is used as training set data of a classification algorithm model;

A3. the method comprises the steps of carrying out digital filtering on all original signal data, wherein an effective frequency band of an acoustic sensor is selected to comprise 30kHz-500kHz, and filtering is carried out by a band-pass FIR filter to obtain signal components between 30kHz and 500kHz in consideration of background noise influence and the general need of filtering components below 30 kHz;

A4. processing the data by an EMD algorithm and a WPD algorithm respectively, wherein an interpolation function of the EMD adopts cubic spline interpolation, a wavelet mother function of the WPD adopts dmey wavelet, and the two algorithms respectively obtain IMF_1-IMF_n ₁ Hezi (Hezi)With signals 1-n ₂ Taking the intermediate 4 groups of IMF obtained by EMD and the first 4 groups of subband signals obtained by WPD, extracting 9 characteristic values by using the 8 groups of signals respectively, and enabling the two methods of EMD and WPD to obtain 36 characteristics respectively, wherein the 9 characteristic values are respectively a time domain kurtosis factor, a skewness factor and a waveform factor, and meanwhile, the frequency domain kurtosis factor, the skewness factor, the waveform factor, a peak frequency, a spectrum bandwidth and a bandwidth centroid frequency, and the algorithm of 9 characteristic parameters is as follows:

[1]time domain kurtosis factor:

(x _n is a time domain signal value);

[2]time domain skewness factor:

(x _n for signal time domain values, +.>

Is the time domain mean value of the signal);

[3]time domain form factor:

(x _rms is the signal time domain root mean square value, x _avr Is the time domain mean value of the signal);

[4]frequency domain kurtosis factor:

(y _n is a frequency domain value);

[5]frequency domain skewness factor:

(y _n for signal frequency domain values, +.>

Is the mean value of the signal frequency domain values);

[6]frequency domain form factor:

(y _rms is the root mean square value of the frequency domain, y _avr Is the frequency domain mean value);

[7]peak frequency: f (f) _max ＝max(y(f))| _f The method comprises the steps of carrying out a first treatment on the surface of the (y (f) is a signal frequency domain function, the value of y (f) is called frequency domain amplitude, f _max Representing a frequency point corresponding to the maximum amplitude of the frequency domain);

[8]spectrum bandwidth: f (f) _dB ＝f _up -f _down (f _up For the frequency domain amplitude y (f) is at the maximum frequency point corresponding to the maximum amplitude of 0.3 times, f _down The frequency domain amplitude y (f) is at the lowest frequency point corresponding to the maximum amplitude of 0.3 times);

[9]bandwidth centroid frequency:

A5. the two groups of 36 features are respectively applied to a ReliefF feature evaluation algorithm to obtain respective weight values, the weight values are equivalent to the distinguishing capability of the features, and the higher the weight values are, the higher the distinguishing capability of the features to different leakage holes is, so that the application of the ReliefF algorithm is: randomly selecting one sample R from a certain leak leakage characteristic set A, searching k nearest neighbor sample sets H from samples of the same type of R, searching k nearest neighbor sample sets M from samples of different types of R, and updating the weight of each leakage characteristic according to the following two formulas (1) (2):

obtaining the weight of 36 multiplied by 2 features corresponding to each leak hole from the above steps;

A6. the weights of the same features of each leak hole are all summed and calculated to be average, so that new weights of 36 features of EMD and WPD are obtained, then new weights of 36 features of two groups are ordered, features of 2 before each weight in the two groups are selected to be brought into SVM classification training, a Radial Basis Function (RBF) is adopted as a kernel function of the SVM, a leak hole identification model library after training is tested by test group data, how the leak hole identification model library is distinguished in accuracy is seen, if the accuracy is insufficient, the number of features with high weights can be gradually increased, and experimental results show that the number of the features with high weight is increased, but the training amount is definitely increased, so that the number of the features with high weight is finally determined according to the required requirements;

A7. finally, determining the characteristics selected in the next simulation test and an identification model library of the trained SVM;

s2, collecting leakage sound signals: collecting leakage sound signals, filtering and preprocessing at 30kHz-500kHz to remove background noise, observing attenuation condition of sound signals before and after passing through reinforcing ribs, and selecting sound signal frequency band f with weaker attenuation _d -f _u ；

S3, positioning the leak holes: establishing a coordinate system by adopting S ₁ ，S ₂ ，S ₃ 3 sensor data, and FIR filter is used to obtain these sensors f _d -f _u Band energy signal, S ₁ The sensor is exemplified by a sensor such as a sensor,

y (f) is the signal amplitude at different frequencies, sensor S ₂ 、S ₃ The same calculation method is used for establishing an equation set:

(x ₁ ,y ₁ )(x ₂ ,y ₂ )(x ₃ ,y ₃ ) Knowing coordinates for three sensors, using a linear iterative algorithm, the equation set can be solved for leak coordinates (x, y);

S4.and (3) signal bar-crossing compensation: after the coordinates of the leak hole are obtained, the original data of the signal of the sensor closest to the leak hole is applied, the number alpha of the reinforcing ribs passing through in the linear propagation process between the leak hole and the sensor can be known, the signal is compensated by the reinforcing ribs, and each reinforcing rib pair f _d -f _u The attenuation coefficient of this band range is β (f), so multiplying this band of y (f) by the compensation coefficient α×β (f), α being the number of over-tendons, β (f) being the attenuation coefficient;

s5, identifying leakage holes: and extracting leakage sound signal characteristics, and comparing the leakage sound signal characteristics with an identification model library to obtain leakage hole characteristics.

2. The method for identifying leakage holes of reinforcement seal structure of spacecraft according to claim 1, wherein f in the step S2 _d And f _u The frequency lower limit of the part of the over-reinforcement spectrum, which is larger than 80% of the over-reinforcement spectrum energy, is f only by using the same drain hole to collect signals and make spectrum comparison under the condition of over-reinforcement and over-reinforcement respectively at the same distance _d The upper frequency limit is f _u At the same time, f can also be obtained _d -f _u Attenuation β (f) of this band.

Images