CN103542261B - Pipeline leak acoustic emission signal processing method based on compressive sensing and mask signal method HHT - Google Patents

Abstract

The present invention relates to a kind of pipeline leakage acoustic emission signals processing method based on compressed sensing and mask signal method HHT.Mainly can apply in the leakage acoustic emission signal detection of pipeline, boiler, main processing steps: the first step: obtain acoustic emission primary signal, utilize digital filter to filter the high-frequency noise beyond leakage acoustic emission signal frequency domain；Second step: introduce compressive sensing theory and acoustic emission signal is compressed sampling；3rd step: utilize OMP algorithm that compressed signal is carried out Accurate Reconstruction；4th step: use the EMD of mask signal method to decompose acoustic emission signal, the component of different frequency in signal is separated from high frequency to low frequency one by one；5th step: each acoustic emission signal frequency component is done Hilbert conversion, determines the beginning and ending time of acoustic emission signal.

Description

Acoustic Emission Signal Processing of Pipeline Leakage Based on Compressed Sensing and Mask Signal Method HHT method

technical field

The invention relates to a pipeline leakage acoustic emission signal processing method based on compressed sensing and mask signal method HHT.

Background technique

The signal acquisition method based on the traditional Nyqusit sampling theorem requires that the sampling frequency be at least twice the highest frequency of the signal. The frequency range of the acoustic emission signal is several thousand Hz to several hundred thousand Hz. In practice, a multi-channel data high-speed acquisition system is generally used, and the sampling frequency is 5 to 10 times the highest frequency of the signal. When monitoring the acoustic emission signal in real time, it will inevitably generate The huge amount of data makes the cost of hardware implementation relatively high, which puts huge pressure on subsequent data transmission, storage and processing, which is very unfavorable for the long-term monitoring of leakage signals.

The acoustic emission signal is a non-stationary transient signal containing noise. The acoustic emission signal is decomposed from the non-stationary transient acoustic emission signal containing noise, and determining the start and end position of the acoustic emission signal is the feature extraction and identification of the acoustic emission signal. key. HHT analysis is usually available.

HHT (Hilbert-Huang Transform, Hilbert-Huang Transform) consists of two parts: Empirical Mode Decomposition (EMD) and Hilbert Transform, among which EMD is the core part. EMD is an effective processing method for non-stationary signals. It starts from the characteristic time scale and decomposes each frequency component in the signal according to the frequency. However, when there are multiple similar frequencies in the signal to be decomposed, the traditional EMD decomposition method will have mode aliasing. The EEMD method proposed later continuously adds white noise during the signal decomposition process, so that the additional white noise is evenly distributed in the entire time-frequency space, and then the signal is tested independently, and the noise will be eliminated by using the overall mean value of enough tests. EEMD can solve the modal aliasing problem in EMD to a certain extent, but it still cannot achieve the ideal detection effect when the signal amplitude-to-frequency ratio does not meet the requirements.

Contents of the invention

The object of the present invention aims at the problems existing in the above traditional signal processing methods, and proposes a processing method that applies compressive sensing theory and EMD with mask signal added to pipeline leakage acoustic emission signal detection technology.

The technical solution of the present invention is: a pipeline leakage acoustic emission signal processing method based on compressed sensing and mask signal method HHT, which is characterized in that: the compressive sensing theory is introduced into the acoustic emission signal detection technology, and then the mask signal is added The EMD decomposition method realizes the decomposition of the acoustic emission signal, and performs Hilbert transformation on the components to determine the start and end time of the leakage. The specific processing steps are as follows:

Step 1: Obtain the original acoustic emission signal, and use a digital filter to filter out high-frequency noise beyond the frequency domain of the leaked acoustic emission signal;

The second step: apply the compressive sensing theory to compress the acoustic emission emission signal obtained in the first step, and use an observation matrix whose length is much smaller than the original signal to perceive the original acoustic emission signal, and obtain a set of observation values;

The third step: using the OMP (Orthogonal Matching Pursuit, Orthogonal Matching Pursuit) algorithm to reconstruct the compressed signal obtained in the second step to obtain the reconstructed acoustic emission signal;

Step 4: Decompose the acoustic emission signal by EMD using the mask signal method, and separate the components of different frequencies in the reconstructed acoustic emission signal in the third step from high frequency to low frequency;

Step 5: Perform Hilbert transformation on the frequency components of the acoustic emission signals separated in the fourth step to determine the start and end times of the acoustic emission signals.

The first step and the second step are known measurement matrices, the length of the acoustic emission signal X is N, and the degree of sparsity is K. Following the theory of compressed sensing, X is projected onto , to get a set of measured values Y whose length is much smaller than N, the signal X can be reconstructed according to the inverse operation of the measured values and the measured matrix, and the sparse approximation can be obtained .

The second step makes signal sampling no longer limited by the Nyquist sampling theorem, improves data compression efficiency, and reduces data collection, transmission and storage costs.

The third step adopts the OMP algorithm, which essentially completes the optimization of the norm of the first step and realizes the fast and accurate reconstruction of the compressed acoustic emission signal.

In the fourth step, a mask signal is added to the acoustic emission signal, and HHT (Hilbert-Huang Transform) is performed on the acoustic emission signal containing the mask signal.

The mask signal added to the acoustic emission signal in the fourth step is, that is, the frequency value of the mask signal is the sum of two adjacent highest frequencies, and the amplitude is the amplitude of the highest frequency.

In the fourth step, when the acoustic emission signal contains N adjacent frequency components, it is necessary to continuously add N-1 mask signals in the decomposition process until the last single frequency component remains.

In the fifth step, Hilbert transform is performed on the decomposed acoustic emission signal components, and the start and end times of its occurrence can be determined.

A pipeline leakage acoustic emission signal processing method based on compressed sensing and mask signal method HHT, that is, compressive sensing theory is applied to leakage acoustic emission signal detection technology.

The invention has the advantages of: (1) the signal sampling is no longer limited by the Nyquist sampling theorem, the data acquisition cost is reduced, and the data compression efficiency is improved; (2) the problem of large data transmission and storage is solved; (3) the decomposition is effectively suppressed The modal aliasing phenomenon in the process makes the decomposition result more accurate and effective, and facilitates the feature extraction of the acoustic emission signal; (4) it is a new type of efficient acoustic emission signal compression and processing method.

Description of drawings

Fig. 1 is a flow chart of processing acoustic emission signals in the patent of the present invention.

Fig. 2 is a flow chart of signal sampling compression by compressed sensing in the patent of the present invention.

Fig. 3 is a flowchart of the patented OMP algorithm of the present invention.

Fig. 4 is a flow chart of mask signal decomposition of the patent of the present invention taking IMF1 as an example.

Fig. 5 is an example of signal reconstruction based on CS in the present invention.

Fig. 6 is an example of EMD decomposition based on mask signal in the present invention.

detailed description

The purpose of the present invention is to solve the problem of ultra-high frequency and large data transmission and storage in the acoustic emission signal detection technology, and obtain accurate decomposition results to provide the possibility for feature extraction, classification and positioning. That is to introduce the compressed sensing theory into the acoustic emission signal detection technology, and then use the EMD decomposition method of adding the mask signal to realize the decomposition of the acoustic emission signal, and perform Hilbert transformation on the components to determine the start and end time of the leakage, compressive sensing theory; and A signal analysis method adding the Hilbert-Huang transform of the masked signal.

The specific treatment plan is:

Step 1: Obtain the original acoustic emission signal and filter it. Since the leakage acoustic emission signal has a certain frequency domain, a Butterworth low-pass filter can be designed to filter out high-frequency noise beyond the frequency domain of the leakage acoustic emission signal.

The second step: introduce the theory of compressive sensing. Analyze signals for sparsity and non-correlation. The acoustic emission signal is compressed and sampled, and an observation matrix whose length is much smaller than the original signal is used to perceive the original signal to obtain a set of observation values;

Step 3: Reconstruct the compressed signal. OMP (Orthogonal Matching Pursuit, Orthogonal Matching Pursuit) algorithm is an improved algorithm based on the MP algorithm. Use the OMP algorithm to accurately reconstruct the compressed signal to obtain the reconstructed acoustic emission signal;

Step 4: Decompose and reconstruct the acoustic emission signal. EMD decomposes the acoustic emission signal by adding a mask signal method, and separates the components of different frequencies in the signal from high frequency to low frequency.

Step 5: Hilbert transform is performed on the frequency components of each acoustic emission signal to determine the start and end time of the acoustic emission signal.

The overall process of leakage acoustic emission signal processing based on compressive sensing theory and mask signal method empirical mode decomposition is shown in Figure 1. In the actual propagation process of the leakage acoustic emission signal, it is inevitable to be interfered by various external sources, so that the collected original signal is mixed with a lot of noise. Since the frequency of the acoustic emission signal has a certain range of frequency domain, a filter can be used to filter out the part with too high frequency in the signal. The invention designs a Butterworth filter to filter out the high-frequency noise in the original signal. Then, based on the compressed sensing theory, the filtered signal is compressed and sampled to facilitate signal transmission and storage. At the receiving end of the signal, the accurate reconstruction of the signal is achieved by using less observations combined with the OMP (Orthogonal Matching Pursuit, Orthogonal Matching Pursuit) algorithm. Finally, the fast Fourier transform is performed on the reconstructed signal to analyze the amplitude-frequency characteristics of the signal, which is convenient for constructing the mask signal. The mask signal is continuously added to the reconstructed signal, and the superimposed signal is decomposed by EMD to obtain a decomposed single frequency component. Finally, the Hilbert transform is performed on the frequency components of each acoustic emission signal to determine the start and end time of the leakage acoustic emission signal.

The processing principle and process of the present invention will be further described below in conjunction with the accompanying drawings.

Figure 2 shows the signal compression process based on compressed sensing. For a signal X, transform it to domain, its representation on this domain is: . This is because the signals used for compressed sensing must be sparse, and many natural signals are not sparse, and these signals have sparse representations in a specific domain, The sparsity of is K. With the known observation matrix to perceive , to obtain a set of measured values Y whose length is much smaller than the length of the original signal, when the measured value Y is transmitted and stored, the processing cost can be saved. The signal can be reconstructed by inverse transforming the measured value at the receiving end of the signal.

Figure 3 is a flowchart of the OMP algorithm. In the known observation matrix , in the case of the observation value Y and the sparsity K, the signal can be reconstructed using the OMP algorithm to obtain the approximate value of the signal X . When initializing, let the residual , set the empty index set and iteration count. First find out the subscript corresponding to the maximum value of the residual and the column inner product of the observation matrix, after finding the subscript, update the index set, record and reconstruct the atom, and then obtain an approximation value of the signal by the least square method, and then update the residual value, The iteration terminates until the number of iterations is greater than or equal to the sparsity, otherwise it keeps looping to find the approximation value.

Figure 4 shows the EMD decomposition steps of the mask signal method taking IMF1 as an example. Before decomposition, it is necessary to perform fast Fourier transform on the signal X, and preliminarily determine the mask signal to be constructed according to its amplitude-frequency characteristics. The specific decomposition process is as follows:

(1) For the analyzed signal , to construct the mask signal . Wherein, the frequency value of the mask signal is the sum of two adjacent highest frequencies, and the amplitude is the amplitude of the highest frequency.

(2 pairs Carry out EMD decomposition, take the first IMF, denote as ,also EMD decomposition takes its first IMF and records it as .

(3) Calculate the mean , resulting in IMF1 decomposed as a signal.

When the signal to be decomposed contains N similar frequency components, N-1 mask signals need to be added during the decomposition process, and the above three steps are repeated until the last single frequency component remains.

According to the above theoretical basis, Figure 5 is an example of signal reconstruction based on CS. It can be seen from the figure that the error between the reconstructed signal and the original signal is very small, and the accurate reconstruction of the signal is realized.

Figure 6 shows an example of EMD decomposition based on mask signals. It can be seen from the decomposition results that the original signal is modulated by three signals of different frequencies, and the signals of different frequencies can be clearly distinguished, and the modal aliasing phenomenon is well suppressed.

The invention breaks the limitation of Nyquist sampling theorem in traditional signal collection, reduces the cost of data collection, and realizes accurate reconstruction of acoustic emission signals through a reconstruction algorithm. The EMD decomposition with mask signal can effectively suppress the modal aliasing phenomenon in the decomposition process, making the decomposition results more accurate and effective.

Claims (9)

1. a pipeline leakage acoustic emission signals processing method based on compressed sensing and mask signal method HHT, it is characterised in that: Compressive sensing theory is incorporated in acoustic emission signal detection technique, then utilizes the EMD decomposition method adding mask signal real The now decomposition to acoustic emission signal, and component is done Hilbert conversion, determine the beginning and ending time that leakage occurs, specifically process step Rapid as follows:

The first step: obtain acoustic emission primary signal, utilizes digital filter to filter the high frequency beyond leakage acoustic emission signal frequency domain Noise；

Second step: applied compression perception theory is compressed sampling to acoustic emission signal X that the first step obtains, remote by a length Less than the calculation matrix perception acoustic emission primary signal of primary signal, obtain a group observations；

3rd step: utilize OMP (Orthogonal Matching Pursuit, orthogonal matching pursuit) algorithm that second step is obtained Compressed signal X ＇ be reconstructed, obtain reconstruct acoustic emission signal；

4th step: use the EMD of mask signal method to decompose acoustic emission signal, by different frequency in the 3rd step reconstruct acoustic emission signal Component separate successively from high frequency to low frequency；

5th step: each acoustic emission signal frequency component separating the 4th step does Hilbert conversion, determines rising of acoustic emission signal The only time.

Pipeline leakage acoustic emission signals based on compressed sensing and mask signal method HHT the most according to claim 1 processes Method, it is characterised in that: the first step and second step known calculation matrix Φ, the acoustic emission signal X length that the first step obtains For N, degree of rarefication is K, it then follows compressive sensing theory, projects on calculation matrix Φ by acoustic emission signal X that the first step obtains, To one group of length much smaller than the measured value Y of N, do inverse operation according to measured value and calculation matrix and just can reconstruct what second step obtained Compressed signal X ＇, obtains sparse bayesian learning

At pipeline leakage acoustic emission signals based on compressed sensing and mask signal method HHT the most according to claim 1 and 2 Reason method, it is characterised in that: described second step makes signal sampling no longer be limited by Nyquist sampling thheorem, improves data pressure Contracting efficiency, reduces data acquisition, transmission and carrying cost.

Pipeline leakage acoustic emission signals based on compressed sensing and mask signal method HHT the most according to claim 1 processes Method, it is characterised in that: the 3rd step uses OMP algorithm, substantially completes the optimization of first step norm, it is achieved to second step The compressed signal quick Accurate Reconstruction of X ＇ obtained.

Pipeline leakage acoustic emission signals based on compressed sensing and mask signal method HHT the most according to claim 1 processes Method, it is characterised in that: in the 4th step acoustic emission signal, add mask signal, the acoustic emission signal containing mask signal is entered Row HHT Hilbert-Huang transform.

The most according to claim 1 or 5 at pipeline leakage acoustic emission signals based on compressed sensing and mask signal method HHT Reason method, it is characterised in that: the mask signal added in the 4th step acoustic emission signal is { S₁(t)……S_n-1(t) }, i.e. The frequency values of mask signal is adjacent two highest frequency sums, and amplitude is the amplitude of highest frequency.

Pipeline leakage acoustic emission signals based on compressed sensing and mask signal method HHT the most according to claim 1 processes Method, it is characterised in that: the 4th step, when acoustic emission signal comprises N number of near by frequency component, needs in catabolic process constantly Add N-1 mask signal, until last single frequency component remaining.

Pipeline leakage acoustic emission signals based on compressed sensing and mask signal method HHT the most according to claim 1 processes Method, it is characterised in that: the acoustic emission signal component decompositing the 5th step does Hilbert conversion, it may be determined that it occurs Beginning and ending time.

9. according to the pipe leakage acoustic emission based on compressed sensing and mask signal method HHT described in any one of claim 1-8 Signal processing method is applied to pipeline leakage testing.