Frequency signal real-time monitoring method based on spatial reconstruction and principal component analysis

Abstract

The invention discloses a frequency signal real-time monitoring method based on spatial reconstruction and principal component analysis. The method comprises the following steps: acquiring a multi-dimensional power spectral density matrix by adopting spatial reconstruction and power spectral density analysis aiming at acquired time domain signals, then performing principal component analysis and standardization processing on the power spectral density matrix, and drawing a time-frequency diagram of signal characteristics according to screened principal components; according to the invention, the interference of noise signals in the flow field can be obviously reduced, the change condition of characteristic signals along with time can be monitored in the measurement process, and the method has a good practical application prospect.

Description

Technical Field

[0001] The invention relates to a frequency signal real-time monitoring method based on space reconstruction and principal component analysis, and belongs to the field of dynamic testing and parameter identification. Background Art

[0002] When measuring dynamic parameters and analyzing the frequency of a flow field, there are multiple signal interferences in the flow field, making it difficult to obtain accurate frequency characteristics in the flow field. When the frequency characteristics of the disturbance source are mixed with the background noise of the flow field, as the flow velocity increases, the noise source in the flow field gradually increases, and the characteristic energy of the noise signal is too high, which often disrupts the analysis of the signal in the entire frequency band, making it difficult to quickly extract the main useful signal.

[0003] Traditional fast Fourier transform (FFT) and power spectral density analysis (PSD) often obtain a large number of frequency domain signals, in which characteristic frequencies are often submerged and difficult to be accurately extracted. Among general noise reduction methods, wavelet transform, empirical mode decomposition, and HHT transform often obtain time domain information under many frequency signal combinations, which requires a large amount of calculation, especially when the frequency band occupied by the noise signal is very wide, it is not enough to separate the main characteristic frequencies. In addition, when extracting the frequency domain features of the signal, it is often necessary to extract and analyze a large amount of time domain information after the experiment is completed. The signal analysis is not timely and cannot meet the real-time extraction of frequency information. Summary of the invention

[0004] In the case of noise interference, the present invention provides a frequency dynamic monitoring method based on spatial reconstruction and principal component analysis. The collected time domain signal is normalized and spatial reconstruction and power spectrum density analysis are used to obtain a multidimensional pre-multiplied power spectrum density matrix. The pre-multiplied power spectrum density matrix is ​​then subjected to principal component analysis and screening to obtain the main frequency characteristics, and the distribution of the main frequencies over time is plotted. The pre-multiplied power spectrum density matrix is ​​averaged and standardized to obtain a time-frequency diagram of the process signal characteristics.

[0005] In order to solve the problems of the prior art, the technical solution adopted by the present invention is:

[0006] A frequency dynamic monitoring method based on spatial reconstruction and principal component analysis comprises the following steps:

[0007] Step 1, given the length t and dimension d of the read data, read the real-time data p(t) of the collected time domain signal;

[0008] Step 2, constructing a multidimensional array matrix A(d, t) for the real-time data of the time domain signal read in step 1;

[0009] Step 3, standardize the multidimensional array matrix A(d,t);

[0010] Step 4: Calculate the power spectrum density of each row vector of the standardized multidimensional array matrix to obtain the power density matrix P at different frequencies. d (d,t);

[0011] Step 5: Construct the pre-multiplied power spectral density matrix P dr (d,t);

[0012] Step 6: Pre-multiply the power spectral density matrix P dr (d, t) Conduct principal component analysis, screen according to the contribution rate, set the contribution rate threshold according to the actual situation, screen and reconstruct its components, and obtain the main frequency;

[0013] Step 7: Pre-multiply the power spectral density matrix P dr (d, t) are averaged and standardized;

[0014] Step 8, draw the time-frequency diagram of the main frequency characteristics and analyze the time domain and frequency domain characteristics of the signal;

[0015] Step 9: With the real-time acquisition of the signal, the multidimensional array is updated, and steps 2 to 8 are repeated to obtain the dynamic change of the frequency domain characteristic signal over time, thereby realizing the real-time monitoring of the characteristic signal.

[0016] As an improvement, the power spectrum density calculation method in step 4 is a periodogram method, an autocorrelation method or a parameter model spectrum estimation method.

[0017] A further improvement is that the calculation formula of the power spectrum density in step 4 is as follows:

[0018]

[0019]

[0020] Where T is the period, G T (f) is the energy signal g T The Fourier transform of (t), P d (f) is the power spectral density, in W / Hz.

[0021] As an improvement, the contribution rate in step 6 is between 0 and 1, which can be selected according to actual conditions.

[0022] Beneficial effects:

[0023] Compared with the prior art, the frequency dynamic monitoring method based on spatial reconstruction and principal component analysis of the present invention performs normalization processing on the collected time domain signal, and uses spatial reconstruction and power spectrum density analysis to obtain a multidimensional pre-multiplied power spectrum density matrix, and then performs principal component analysis and screening on the pre-multiplied power spectrum density matrix to obtain the main frequency characteristics, draw the distribution of the main frequency over time, average and standardize the pre-multiplied power spectrum density matrix, and obtain the time-frequency diagram of the process signal characteristics. Therefore, the present invention can reduce the interference of noise signals in the flow field, and can monitor the changes of characteristic signals over time during the measurement process, and can realize the real-time dynamic extraction of the main frequency characteristics, and obtain the time-frequency diagram in a timely and accurate manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The present invention is a flow chart of a frequency dynamic monitoring method based on space reconstruction and principal component analysis.

[0025] Figure 2 is the premultiplied power spectral density frequency spectrum. DETAILED DESCRIPTION

[0026] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and experimental scenarios.

[0027] The test scenario selected in the present invention is the film cooling flow field under a flat plate environment. The mainstream flow velocity of the low-speed wind tunnel is about 60m / s. The cold air is mixed with the mainstream through the film holes. The degree of mixing can be expressed by the blowing ratio (the ratio of the cold air outlet dense flow to the mainstream dense flow). Different blowing ratios can be obtained by changing the cold air flow rate. The vortex flow caused by the mixing of the mainstream and the cold air will produce periodic disturbances to the pressure of the flow field, and the disturbance frequency changes with the change of the blowing ratio. In the flat film cooling flow field, the wall pressure pulsation signal behind the film hole is monitored. After data processing, the change of the flow field frequency signal at the monitoring position in the flow field can be obtained. The frequency signal can characterize the state of the current flow field. Accurately identifying the characteristic signal can provide strong support for the development of monitoring flow.

[0028] A frequency dynamic monitoring method based on spatial reconstruction and principal component analysis comprises the following steps:

[0029] Step 1, according to the requirements of time and frequency resolution, select the length t and dimension d of the time domain signal, and read a certain time domain signal p(t) of the wall pressure pulsation. The longer the data, the larger the dimension, and the slower the calculation time will be. It can be adjusted according to actual conditions. In this embodiment, the acquisition frequency is 40kHz, the signal length t=2500, and d=16.

[0030] Step 2: spatially reconstruct the real-time data p(t) of the time domain signal read in step 1 to construct a multidimensional matrix A(d, t).

[0031] Step 3, standardize the multidimensional array matrix A(d,t).

[0032] Step 4: Calculate the power spectrum density of each row vector of the standardized multidimensional array matrix to obtain the power density matrix P at different frequencies. d (d, t), the formula for power spectral density is as follows:

[0033]

[0034]

[0035] Where T is the period, G T (f) is the energy signal g T The Fourier transform of (t), P d (f) is the power spectrum density, in W / Hz. The methods for calculating the power spectrum density of a signal include the periodogram method, the autocorrelation method, and the parameter model spectrum estimation method.

[0036] Step 5: Construct the pre-multiplied power spectral density matrix P dr (d,t).

[0037] Step 6: Pre-multiply the power spectral density matrix P dr (d, t) Conduct principal component analysis and perform screening based on the contribution rate (valued from 0 to 1). The contribution rate threshold can be selected based on actual conditions. In this embodiment, features with a contribution rate greater than 0.8 are selected for data screening and reconstruction, and the main frequency is obtained;

[0038] Step 7: Pre-multiply the power spectral density matrix P dr (d, t) are averaged and standardized;

[0039] Step 8, draw the time-frequency diagram of the main frequency characteristics and analyze the time domain and frequency domain characteristics of the signal;

[0040] Step 9: With the real-time acquisition of the signal, the multidimensional array is updated, and steps 2 to 8 are repeated to obtain the dynamic change of the frequency domain characteristic signal over time, thereby realizing the real-time monitoring of the characteristic signal.

[0041] The present invention and its embodiments are described schematically above, and the description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. Therefore, if a person skilled in the art is inspired by it and designs a structural method and an embodiment similar to the technical solution without creativity without departing from the purpose of the invention, they shall all fall within the protection scope of the present invention.

Claims

1. A frequency dynamic monitoring method based on spatial reconstruction and principal component analysis, characterized in that: The specific steps are: Step 1, given the length t and dimension d of the read data, read the real-time data p(t) of the collected time domain signal; Step 2, constructing a multidimensional array matrix A(d, t) for the real-time data of the time domain signal read in step 1; Step 3, standardize the multidimensional array matrix A(d,t); Step 4: Calculate the power spectrum density of each row vector of the standardized multidimensional array matrix to obtain the power density matrix P at different frequencies. d (d,t); Step 5: Construct the pre-multiplied power spectral density matrix P dr (d,t); Step 6: Pre-multiply the power spectral density matrix P dr (d, t) Conduct principal component analysis, screen according to the contribution rate, set the contribution rate threshold according to the actual situation, screen and reconstruct its components, and then obtain the main frequency; Step 7: Pre-multiply the power spectral density matrix P dr (d, t) are averaged and standardized; Step 8, draw the time-frequency diagram of the main frequency characteristics and analyze the time domain and frequency domain characteristics of the signal; Step 9: With the real-time acquisition of the signal, the multidimensional array is updated, and steps 2 to 8 are repeated to obtain the dynamic change of the frequency domain characteristic signal over time, thereby realizing the real-time monitoring of the characteristic signal.

2. The frequency dynamic monitoring method based on spatial reconstruction and principal component analysis according to claim 1 is characterized in that: The power spectrum density in step 4 is calculated by a periodogram method, an autocorrelation method or a parameter model spectrum estimation method.

3. The frequency dynamic monitoring method based on spatial reconstruction and principal component analysis according to claim 2 is characterized in that: The calculation formula of the power spectrum density in step 4 is as follows: Where T is the period, G T (f) is the energy signal g T The Fourier transform of (t), P d (f) is the power spectral density, in W / Hz.

4. The frequency dynamic monitoring method based on spatial reconstruction and principal component analysis according to claim 1 is characterized in that: The contribution rate in step 6 ranges from 0 to 1.

Images