Frequency domain signal feature extraction method based on premultiplication power spectral density

Abstract

The invention discloses a frequency domain signal feature extraction method based on premultiplication power spectral density. The method comprises the following steps: carrying out power spectral density analysis on time domain signals collected in a flow field, and carrying out standardization processing on pre-multiplication power spectral density dimensionless parameters formed based on flow field information, frequency and power spectral density to obtain main characteristic signals in the flow field; according to the method, the interference of noise signals in a flow field can be obviously reduced, the frequency domain processing time is shortened, and accurate and rapid extraction of characteristic signals is realized.

Description

Technical Field

[0001] The invention relates to a method for extracting frequency domain signal features based on premultiplied power spectrum density. Background Art

[0002] When measuring dynamic parameters of a flow field, when there is interference in the flow field, the frequency characteristics of the disturbance source are often mixed in the background noise of the flow field. As the flow velocity increases, the noise in the flow field gradually increases, making it difficult to quickly extract the main frequency parameters. Traditional fast Fourier transform (FFT) and power spectral density analysis (PSD) often obtain a large number of frequency domain signals, in which the characteristic frequencies are often submerged and difficult to be accurately extracted. In addition, when performing spectrum calculations, the more data there is, the longer the calculation time. Summary of the invention

[0003] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a method for extracting frequency domain signal features based on premultiplied power spectral density, perform power spectral density estimation on the obtained time domain signal, and then perform dimensionless and standardized processing on the obtained premultiplied power spectral density, so as to improve the efficiency and accuracy of frequency feature extraction.

[0004] In order to solve the problems of the prior art, the technical solution adopted by the present invention is:

[0005] A frequency domain signal feature extraction method based on premultiplied power spectral density comprises the following steps:

[0006] Step 1, read the time domain signal p(t) collected in the flow field, and reconstruct the one-dimensional time domain signal p(t) into a multi-dimensional array;

[0007] Step 2, standardizing the multi-dimensional reconstructed signal;

[0008] Step 3, performing power spectrum density calculation on the multidimensional array calculated in step 2 to obtain a multidimensional matrix of power density at different frequencies;

[0009] Step 4: Based on the power spectrum density, construct a dimensionless parameter pair, i.e., dimensionless frequency f * and the dimensionless premultiplied power density

[0010] Step 5: Pre-multiply the dimensionless power density obtained After averaging and standardization, we get

[0011] Step 6, by plotting the dimensionless parameter against f * and The dimensionless frequency and the corresponding dimensionless premultiplied power spectral density of the main characteristic signals in the flow field are extracted from the graph, and the frequency and power spectral density of the main characteristic signals are further obtained.

[0012] As an improvement, the specific steps of performing standardization processing on the multi-dimensionally reconstructed signal in step 2 are:

[0013] Step 2.1, obtain the average value of the collected time domain signal and standard deviation μ;

[0014] Step 2.2, according to the standardization formula, the standardized time domain signal p is obtained n (t);

[0015] Step 2.3, obtain the normalized time domain signal p n The root mean square of (t)p rms , providing conditions for dimensionless processing.

[0016] As an improvement, the power spectrum density calculation method in step 3 is a periodogram method, an autocorrelation method or a parameter model spectrum estimation method, and the parameters are set specifically according to the requirements for frequency domain resolution.

[0017] A further improvement is that the calculation formula of the power spectrum density in step 3 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 g (f) is the power spectral density, in W / Hz.

[0021] As an improvement, the dimensionless parameter f described in step 4 is * and The calculation formula is as follows:

[0022] f * =2πfA

[0023] Where f is the frequency and the dimension of A is T 1 , is to construct the dimensionless parameter f * The required parameter combination is determined according to the source of the original signal, f * is the dimensionless frequency;

[0024]

[0025] Where f is the frequency, Pg is the power density function, and the dimension of B is ML 2 T -4 , is to construct the dimensionless parameter The required parameter combination is determined according to the source of the original signal. is the dimensionless premultiplied power density. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A flow chart of a signal feature extraction method based on premultiplied power spectral density according to the present invention;

[0027] Figure 2 The FFT processing results and PSD processing results of the original time domain signal read in by the present invention and the time domain signal used for comparison, (a) is the original time domain signal read in, (b) is the FFT spectrum distribution of the time domain signal, and (c) is the power spectrum density distribution of the time domain signal;

[0028] Figure 3 The dimensionless premultiplied power spectral density results before and after normalization obtained after processing in each step of the present invention, (a) is the dimensionless premultiplied power spectral distribution, and (b) is the dimensionless premultiplied power spectral distribution after normalization. DETAILED DESCRIPTION

[0029] The specific implementation of the present invention is further described below in conjunction with the drawings and examples.

[0030] A method for extracting frequency domain signal features based on premultiplied power spectral density is proposed. First, the wall pressure pulsation signal behind the film hole in the flat film cooling flow field is read under the condition of a blowing ratio of 1.6. The vortex flow caused by the mixing of mainstream and cold air will produce periodic disturbances on the pressure of the flow field, and the disturbance frequency changes with the change of the blowing ratio.

[0031] The specific steps are as follows:

[0032] Step 1: Read the time domain signal p(t) collected in the flow field. The signal graph is shown in Figure 2 (a), reconstruct the one-dimensional time domain signal into a multidimensional array;

[0033] Step 2: Standardize the multi-dimensional reconstructed signal. The specific steps are as follows:

[0034] Step 2.1, obtain the average value of the collected time domain signal and standard deviation μ;

[0035] Step 2.2, according to the standardization formula, the standardized time domain signal p is obtained n (t);

[0036] Step 2.3, obtain the time domain signal pn The root mean square of (t)p rms ;

[0037] Step 3, calculate the power spectrum density of the multidimensional array calculated in step 2 to obtain a multidimensional matrix of power density at different frequencies. The formula of power spectrum density is as follows:

[0038]

[0039]

[0040] Where T is the period, G T (f) is the energy signal g T The Fourier transform of (t), P g (f) is the power spectrum density, in W / Hz. The power spectrum density of the signal can be calculated by the periodogram method, the autocorrelation method, the parameter model spectrum estimation method, etc. This embodiment uses the Burg method in the autocorrelation method for estimation, and the window length is slightly smaller than the length of the array;

[0041] Step 4: Construct a dimensionless parameter pair f based on frequency and power spectrum density * and The specific calculation formula is as follows:

[0042] Dimensionless frequency f * =2πfA,

[0043] Where f is the frequency and the dimension of A is T 1 , is to construct the dimensionless parameter f * The required parameter combination, in this embodiment δ is the displacement boundary layer thickness, U is the mainstream velocity of the flow field;

[0044] Dimensionless power density Where f is the frequency, P g is the power density function, and the dimension of B is ML 2 T -4 , is to construct the dimensionless parameter The required parameter combination, in this embodiment in is the mean square of the pressure signal per unit mass, and δ is the displacement boundary layer thickness.

[0045] Step 5: average and normalize the dimensionless premultiplied power density to obtain

[0046] Step 6, by plotting the dimensionless parameter against f * and Graphics such as Figure 3As shown in (b), the dimensionless frequency and the corresponding dimensionless power spectral density of the main characteristic signal in the flow field are extracted, and the frequency and power spectral density of the main characteristic signal are further obtained.

[0047] In this embodiment, Figure 3 The point with the highest power density in (b) is f * =0.548, corresponding to frequency f=7450 Hz. According to the mechanism of mixing of cold air and mainstream when the blowing ratio is 1.6, the shedding vortex frequency under this blowing ratio obtained by experiment is 7500 Hz (the test method can be referred to in the literature: Wu, Q., Zhang, H., Yin, Z., Gao, A., Liu, Y., Xu, Y., and Chen, H. (September 10, 2024). "Understanding Unsteady Vortex Structure and Shedding Frequency of Cylindrical Hole Film Cooling: Insights From Experimental and Numerical Approaches." ASME. J. Turbomach. February 2025; 147(2): 021003. https: / / doi.org / 10.1115 / 1.4066273). Figure 2 (b) and Figure 2 In (c), the characteristic signal can be observed near the frequency of 7500 Hz, but the selection is somewhat difficult. In comparison, the method of the present invention has a higher accuracy in extracting the existing main frequencies, and the frequency selection is more intuitive.

[0048] In summary, the present invention performs power spectral density analysis on time domain signal features and obtains dimensionless parameter pairs, which can improve the accuracy and speed of frequency feature extraction, especially in the presence of noise interference, and can achieve fast and accurate extraction of frequency signal features, which has good application prospects.

[0049] 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 domain signal feature extraction method based on premultiplied power spectral density, characterized in that: The following steps are involved: Step 1, read the time domain signal p(t) collected in the flow field, and reconstruct the one-dimensional time domain signal p(t) into a multi-dimensional array; Step 2, standardizing the multi-dimensional reconstructed signal; Step 3, performing power spectrum density calculation on the multidimensional array calculated in step 2 to obtain a multidimensional matrix of power density at different frequencies; Step 4: Based on the power spectrum density, construct a dimensionless parameter pair, i.e., dimensionless frequency f * and the dimensionless premultiplied power density Step 5: Pre-multiply the dimensionless power density obtained After averaging and standardization, we get Step 6, by plotting the dimensionless parameter against f * and The dimensionless frequency and the corresponding dimensionless premultiplied power spectral density of the main characteristic signals in the flow field are extracted from the graph, and the frequency and power spectral density of the main characteristic signals are further obtained.

2. The method for extracting frequency domain signal features based on premultiplied power spectral density according to claim 1, characterized in that: The specific steps of performing standardization processing on the multi-dimensional reconstructed signal in step 2 are: Step 2.1, obtain the average value of the collected time domain signal and standard deviation μ; Step 2.2, according to the standardization formula, the standardized time domain signal p is obtained n (t); Step 2.3, obtain the time domain signal p n The root mean square of (t)p rms , providing conditions for dimensionless processing.

3. The method for extracting frequency domain signal features based on premultiplied power spectral density according to claim 1, characterized in that: The power spectrum density in step 3 is calculated by a periodogram method, an autocorrelation method or a parameter model spectrum estimation method.

4. The method for extracting frequency domain signal features based on premultiplied power spectral density according to claim 3, characterized in that: The calculation formula of the power spectral density described in step 3 is as follows: Where T is the period, G T (f) is the energy signal g T The Fourier transform of (t), P g (f) is the power spectral density, in W / Hz.

5. The method for extracting frequency domain signal features based on premultiplied power spectral density according to claim 1, characterized in that: The dimensionless parameter f described in step 4 * and The calculation formula is as follows: f * =2πfA Where f is the frequency and the dimension of A is T 1 , is to construct the dimensionless parameter f * The required parameter combination is determined according to the source of the original signal, f * is the dimensionless frequency; Where f is the frequency, P g is the power density function, and the dimension of B is ML 2 T -4 , is to construct the dimensionless parameter The required parameter combination is determined according to the source of the original signal. is the dimensionless premultiplied power density.

Images