Flip chip ultrasonic excitation vibration signal denoising method and system

Abstract

The invention relates to the technical field of flip chips, in particular to a flip chip ultrasonic excitation vibration signal denoising method and system, and the method comprises the steps: obtaining an ultrasonic excitation vibration signal of a flip chip; empirical Fourier decomposition is improved based on Welch transformation and automatic multi-scale peak detection, and the vibration signals are decomposed; screening decomposition components of the vibration signals based on an information entropy and kurtosis joint index; and performing dimensionality reduction compression on the signal component by using principal component analysis, extracting main frequency characteristics to create a reference signal of a frequency domain adaptive filter, and performing noise reduction processing on a chip vibration signal. According to the method, the modal characteristics of the chip can be effectively extracted while the ultrasonic excitation vibration signal noise of the flip chip is removed, excellent denoising performance is achieved, the noise reduction effect on the ultrasonic excitation vibration signal of the flip chip is improved, and therefore the accuracy and reliability of flip chip defect detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of flip chip, and in particular to a method and system for denoising ultrasonic excitation vibration signals of flip chips. Background Art

[0002] Flip chip (FC) is an advanced integrated circuit packaging technology. Compared with the traditional wire bonding packaging technology, the flip chip turns the chip upside down so that its face is downward and directly connected to the packaging substrate or circuit board. Due to its excellent characteristics such as high frequency, low latency, and low crosstalk, it has become the mainstream technology in the field of microelectronic packaging. With the continuous improvement of packaging density, the bump size and pitch of flip chips are gradually reduced and hidden inside the package, which makes it very difficult to detect solder joint defects inside the chip. Solder joint defects will surely seriously affect the electrical, thermal, and mechanical properties of the entire packaged product.

[0003] Vibration detection technology converts a set excitation signal into ultrasonic waves through an ultrasonic transducer and acts on the chip to be detected, and then uses a laser Doppler vibrometer to detect the vibration signal on the chip surface. Through further analysis of the signal, the defect detection of the solder joints of the flip chip is realized. However, in actual detection, the ultrasonic excitation vibration signal of the flip chip will be affected by factors such as environmental noise and detection system noise, resulting in interference with the effective information in the signal, and further reducing the accuracy of defect detection of the flip chip, and it is impossible to comprehensively analyze the faults of the flip chip. Therefore, denoising the ultrasonic excitation vibration signal of the flip chip is a prerequisite for comprehensively analyzing the chip faults. Summary of the Invention

[0004] Therefore, the technical problem to be solved by the present invention is to overcome the problem that the ultrasonic excitation vibration signal of the flip chip in the prior art contains noise, resulting in a reduction in the accuracy of defect detection of the flip chip.

[0005] To solve the above technical problem, the present invention provides a method for denoising ultrasonic excitation vibration signals of flip chips, including: Obtain the ultrasonic excitation vibration signal of the flip chip; Decompose the ultrasonic excitation vibration signal of the flip chip to obtain a number of IMF components; Calculate the normalized kurtosis and normalized information entropy of each IMF component, and screen the IMF components based on the normalized comprehensive index jointly based on the normalized kurtosis and normalized information entropy, remove the noise components, and obtain the initial signal components; Use the principal component analysis method to perform dimensionality reduction and compression on all IMF components, extract the maximum frequency component features of the fitting signal corresponding to each IMF component, and construct the reference signal of the frequency-domain adaptive filter based on all the maximum frequency component features; use the frequency-domain adaptive filter to further reduce the noise of the reconstructed excitation signal of the initial signal component to obtain the target noise-reduced signal.

[0006] Preferably, decompose the ultrasonic excitation vibration signal of the flip chip to obtain several IMF components, including: Use the Welch transform to divide the ultrasonic excitation vibration signal of the flip chip into multiple overlapping short time periods, and calculate the power spectrum estimate of the ultrasonic excitation vibration signal of the flip chip; Obtain the boundary coordinates for dividing the power spectrum estimate through the automatic multi-scale peak detection algorithm; Construct a zero-phase filter bank according to the boundary coordinates, and perform zero-phase filtering on the ultrasonic excitation vibration signal of the flip chip to obtain several IMF components.

[0007] Preferably, the formula for calculating the power spectrum estimate of the ultrasonic excitation vibration signal of the flip chip is: ; where is the power spectrum estimate of the ultrasonic excitation vibration signal of the flip chip, is the power spectrum estimate of the i-th overlapping short time period, , represents the number of overlapping short time periods, represents the total length of the ultrasonic excitation vibration signal of the flip chip, represents the length of each overlapping short time period; is the normalization factor, which is used to ensure that the obtained power spectrum estimate conforms to the asymptotically unbiased estimate; represents the window function, represents the i-th overlapping short time period, represents the frequency, represents the time, represents the imaginary unit.

[0008] Preferably, the obtaining of the boundary coordinates for dividing the power spectrum estimate through the automatic multi-scale peak detection algorithm includes: Use a multi-scale window to detect the local maximum of the power spectrum estimate, and the formula is: ; where is the size of the local window, is the local maximum of the power spectrum estimate; Calculate the peak index of the local maximum at each scale, and the formula is: ; Among them, is the peak index of the local maximum, which is used to evaluate the peak degree of the local maximum at this scale; Screen out the peak index with the largest peak at each scale, and use the local maximum corresponding to this peak as the boundary coordinates. Preferably, the calculation formula for the normalized kurtosis of the IMF component is:

[0009] ; ; Among them, is the IMF component, is the normalized kurtosis, is the expected value of the IMF component, is the standard deviation of the IMF component, represents the expectation operation, represents the normalization operation; The calculation formula for the normalized information entropy of the IMF component is: ; Among them, is the normalized information entropy, is the time; The calculation formula for the normalized comprehensive index is: ; Among them, is the normalized comprehensive index.

[0010] Preferably, screening the IMF components using the normalized comprehensive index jointly based on the normalized kurtosis and the normalized information entropy includes: when the normalized comprehensive index of the IMF component is less than the sensitivity coefficient, then this IMF component is a noise component.

[0011] Preferably, using the principal component analysis method to perform dimensionality reduction and compression on all IMF components and extract the maximum frequency component features includes: Obtaining the signal feature matrix of each IMF component through piecewise reconstruction and performing standardization processing on the signal feature matrix; Obtaining the correlation matrix based on the standardized signal feature matrix; Calculating the eigenvalues and corresponding eigenvectors of the correlation matrix; Selecting the eigenvector with the largest eigenvalue and reconstructing it into a fitting signal; Converting the fitting signal to the frequency domain and extracting the maximum frequency component features of the fitting signal corresponding to each IMF vector.

[0012] Preferably, the maximum frequency component feature includes frequency and amplitude, and the calculation formula is: ; Wherein, represents the fitting signal of the m-th IMF component, and respectively represent the frequency and amplitude of the maximum frequency component feature of the fitting signal ; represents frequency, represents time, represents the imaginary unit.

[0013] Preferably, the reference signal for constructing the frequency-domain adaptive filter according to all the maximum frequency component features is: ; Wherein, represents the reference signal of the frequency-domain adaptive filter, is the total number of IMF components.

[0014] The present invention also provides a flip-chip ultrasonic excitation vibration signal denoising system, including: A signal acquisition module for acquiring the ultrasonic excitation vibration signal of the flip-chip; A modal decomposition module for decomposing the ultrasonic excitation vibration signal of the flip-chip to obtain a number of IMF components; A preliminary denoising module for calculating the normalized kurtosis and normalized information entropy of each IMF component, and screening the IMF components based on the normalized comprehensive index jointly composed of the normalized kurtosis and normalized information entropy to remove the noise components and obtain the initial signal components; A secondary denoising module for using the principal component analysis method to perform dimensionality reduction and compression on all IMF components, extracting the maximum frequency component features of the fitting signals corresponding to each IMF component, and constructing the reference signal of the frequency-domain adaptive filter according to all the maximum frequency component features; further denoising the reconstructed excitation signal of the initial signal components by using the frequency-domain adaptive filter to obtain the target denoising signal.

[0015] The above technical solutions of the present invention have the following beneficial effects compared with the prior art: A denoising method for ultrasonic excitation vibration signals of flip chips according to the present invention uses a comprehensive index combining kurtosis and information entropy to screen IMF components, realizes accurate discrimination between noise components and effective components, preliminarily denoises the ultrasonic excitation vibration signals of flip chips, and then further uses the principal component analysis method to reconstruct the IMF components, establish a reference signal for an adaptive filter, and perform secondary denoising on the ultrasonic excitation vibration signals of flip chips to remove the aliasing noise in the signals and obtain the target denoised signal. The present invention improves the denoising effect of the ultrasonic excitation vibration signals of flip chips, thereby improving the accuracy and reliability of flip chip defect detection.

[0016] Moreover, in order to solve the problems of difficult and inaccurate identification of modal components in the ultrasonic excitation vibration signals of flip chips by traditional modal decomposition methods, the present invention improves the empirical Fourier decomposition technology. By using the power spectrum estimation obtained by Welch transformation instead of the frequency spectrum, the spectral characteristics of non-linear vibration can be identified and distinguished from noise. Furthermore, an automatic multi-scale peak detection algorithm is used instead of a single-scale peak detection algorithm to achieve more reasonable power spectrum segmentation, improving the modal decomposition ability of the ultrasonic excitation vibration signals of flip chips. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to make the content of the present invention easier to be clearly understood, the following further describes the present invention in detail according to specific embodiments of the present invention in conjunction with the drawings, wherein: Figure 1 is a flowchart of a denoising method for ultrasonic excitation vibration signals of a flip chip according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The following further describes the present invention in conjunction with the drawings and specific embodiments, so that those skilled in the art can better understand the present invention and be able to implement it, but the embodiments cited are not intended to limit the present invention.

[0019] Referring to Figure 1 as shown, the present invention provides a denoising method for ultrasonic excitation vibration signals of flip chips, including: S1. Use an air-coupled ultrasonic transducer to emit ultrasonic excitation with a periodically changing frequency, cause the flip chip to vibrate, and then use a laser Doppler vibrometer to obtain the ultrasonic excitation vibration signal of the flip chip; S2. Improve the Improved Empirical Fourier Decomposition (IEFD) technology based on the Welch transform and the Automatic multiscale-based peak detection (AMPD) algorithm to decompose the ultrasonic excitation vibration signal of the flip chip, and obtain several IMF components; S3. Calculate the normalized kurtosis and normalized information entropy of each IMF component, and screen the IMF components based on the normalized comprehensive index combined with the normalized kurtosis and normalized information entropy to remove the noise components and obtain the initial signal components; S4. Use the Principal Components Analysis (PCA) to perform dimensionality reduction and compression on all IMF components, extract the maximum frequency component characteristics of the fitting signal corresponding to each IMF component, and construct a reference signal for the Frequency-Domain Adaptive Filter (FDAF) according to all the maximum frequency component characteristics; use the frequency-domain adaptive filter to further denoise the reconstructed excitation signal of the initial signal components to obtain the target denoised signal.

[0020] The specific content of S1 includes: S101. The air-coupled ultrasonic transducer creates a swept-frequency sine wave signal: ; where is time, is the period, is the swept-frequency sine wave signal of a single period, and are respectively the upper and lower bounds of the frequency change of the swept-frequency sine signal.

[0021] S102. Through the signal generator, signal amplifier and ultrasonic transducer, convert the swept-frequency sine wave signal into sound pressure and transmit it to the flip chip; the conversion formula is: ; where is the converted sound pressure, is the peak value of the signal generator, is the amplification factor of the signal amplifier, is the sensitivity of the ultrasonic transducer.

[0022] S103. The flip chip receives the sound pressure excitation and generates forced vibration: ; where is the time-domain response function of the chip system, is the vibration response of the flip chip.

[0023] S104. Measure the vibration velocity signal on the surface of the flip chip using a laser Doppler vibrometer, which is the ultrasonic excitation vibration signal of the flip chip. The formula is: ; where, is the ultrasonic excitation vibration signal of the flip chip, is the change in the echo frequency of the laser Doppler vibrometer, is the wavelength.

[0024] Preferably, after obtaining the ultrasonic excitation vibration signal of the flip chip in S1, preprocess the ultrasonic excitation vibration signal of the flip chip, including signal truncation, low-pass and high-pass filtering, first-order differentiation, first-order integration, and resampling operations.

[0025] To solve the problems of difficulty and inaccurate identification of modal components in the ultrasonic excitation vibration signal of the flip chip by traditional modal decomposition methods, this embodiment improves the empirical Fourier decomposition technique. The specific steps of S2 are as follows: S201. Use the Welch transform to divide the ultrasonic excitation vibration signal of the flip chip into multiple overlapping short time segments. The length of each overlapping short time segment is usually shorter than the length of the entire signal. The purpose of overlapping is to reduce the variance in the spectral estimation and improve the stability of the spectral estimation.

[0026] Calculate the power spectral estimate of the ultrasonic excitation vibration signal of the flip chip. The formula is: ; where, is the power spectral estimate of the ultrasonic excitation vibration signal of the flip chip, is the power spectral estimate of the i-th overlapping short time segment, , represents the number of overlapping short time segments, represents the total length of the ultrasonic excitation vibration signal of the flip chip, represents the length of each overlapping short time segment; is the normalization factor, which is used to ensure that the obtained power spectral estimate meets the asymptotic unbiased estimate; represents the window function, represents the i-th overlapping short time segment, represents the frequency, represents the time, represents the imaginary unit.

[0027] The power spectrum estimate obtained using the Welch transform is used instead of the spectrum because spectral spurs can significantly reduce the accuracy of frequency band segmentation, while the power spectrum can identify the spectral characteristics of non-linear vibrations and distinguish them from noise.

[0028] S202. When the characteristic frequencies in the signal are distributed in a relatively wide frequency band and are greatly affected by noise, it is easy to cause the characteristic frequencies selected during empirical Fourier decomposition to be concentrated in a relatively narrow interval. Therefore, in this embodiment, the automatic multi-scale peak detection algorithm is used instead of the single-scale peak detection algorithm, and multi-scale sliding windows are used to globally extract local maxima from the power spectrum to achieve more reasonable frequency spectrum segmentation. Specifically, it includes: Obtaining the boundary coordinates for segmenting the power spectrum estimate through the automatic multi-scale peak detection algorithm, including: Using multi-scale windows to detect the local maxima of the power spectrum estimate, with the formula: ; where is the size of the local window, is the local maximum of the power spectrum estimate; Calculating the peak index of the local maximum at each scale, with the formula: ; where is the peak index of the local maximum for evaluating the peak degree of the local maximum at this scale; Selecting the peak with the largest peak index at each scale, and taking the local maximum corresponding to this peak as the boundary coordinate.

[0029] S203. Constructing a zero-phase filter bank based on the boundary coordinates, and performing zero-phase filtering on the ultrasonic excitation vibration signal of the flip chip to obtain several IMF components, with the formula: ; where represents the ultrasonic excitation vibration signal of the flip chip, represents the reverse sequence of the ultrasonic excitation vibration signal of the flip chip, represents the zero-phase filter bank, represents the IMF component, is the total number of IMF components.

[0030] In this embodiment, the power spectrum estimation obtained by Welch transform is used to replace the frequency spectrum, which can identify the frequency spectrum characteristics of nonlinear vibration and distinguish them from noise. Furthermore, the automatic multi-scale peak detection algorithm is used to replace the single-scale peak detection algorithm to achieve more reasonable power spectrum segmentation, improving the modal decomposition ability of the ultrasonic excitation vibration signal of the flip chip.

[0031] Specifically, S3 includes: S301. Calculate the normalized kurtosis of each IMF component. The formula is: ; where is the IMF component, is the normalized kurtosis, is the expected value of the IMF component, is the standard deviation of the IMF component, represents the expectation operation, represents the normalization operation.

[0032] S302. Calculate the normalized information entropy of each IMF component. The formula is: ; where is the normalized information entropy, is the time.

[0033] S303. Calculate the normalized comprehensive index of the combination of the normalized kurtosis and the normalized information entropy. The formula is: ; where is the normalized comprehensive index.

[0034] S304. Screen the IMF components based on the normalized comprehensive index. When is less than the sensitivity coefficient , the IMF component is a noise component. Remove the noise component to obtain the initial signal component.

[0035] Specifically, S4 includes: S401. Use the principal component analysis method to perform dimensionality reduction and compression on all IMF components, including: Taking the mth IMF component as an example, the signal feature matrix of is obtained by segmented reconstruction, and the signal feature matrix is standardized. The formula is: ; ; where denotes the signal feature matrix, denotes the reconstruction operation, denotes the m-th IMF component, denotes the th segment; denotes the normalized signal feature matrix, is the expected value of the IMF component, is the standard deviation of the IMF component; According to the normalized signal feature matrix obtain the correlation matrix , the formula is: ; wherein, denotes the correlation matrix, denotes the transpose of the normalized signal feature matrix, denotes the number of points of the m-th IMF component; The eigenvalues of the calculated correlation matrix are , and the corresponding eigenvectors are ; Select the eigenvector with the largest eigenvalue and reconstruct it into a fitting signal .

[0036] S402. Transform the fitting signal into the frequency domain and extract the maximum frequency component feature of each IMF vector corresponding to the fitting signal; the maximum frequency component feature includes frequency and amplitude, and the specific calculation formula is: ; wherein, and respectively represent the frequency and amplitude of the maximum frequency component feature of the fitting signal .

[0037] S403. Construct the reference signal of the frequency domain adaptive filter according to the maximum frequency component features of all IMF components, and the formula is: ; wherein, denotes the reference signal of the frequency domain adaptive filter, is the total number of IMF components.

[0038] S404. Reconstruct the initial signal component to obtain the reconstructed excitation signal ; Use the reconstructed chip excitation signal as the input signal of the adaptive filter for further noise reduction to obtain the target noise reduction, and the formula is: ; wherein, is the target noise-reduced signal output by the adaptive filter, is the filter order, are the filter coefficients.

[0039] In summary, for the flip-chip ultrasonic excitation vibration signal denoising method described in the present invention, a comprehensive index combining kurtosis and information entropy is used to screen IMF components, so as to accurately distinguish noise components and effective components, preliminarily denoise the flip-chip ultrasonic excitation vibration signal, and then further use the principal component analysis method to reconstruct the IMF components to establish a reference signal for the adaptive filter, and perform secondary denoising on the flip-chip ultrasonic excitation vibration signal to remove the aliasing noise in the signal and obtain the target denoised signal. The present invention improves the denoising effect of the flip-chip ultrasonic excitation vibration signal, thereby improving the accuracy and reliability of flip-chip defect detection.

[0040] Based on the above flip-chip ultrasonic excitation vibration signal denoising method, this embodiment also provides a flip-chip ultrasonic excitation vibration signal denoising system, including: A signal acquisition module, configured to acquire the flip-chip ultrasonic excitation vibration signal; A modal decomposition module, configured to decompose the flip-chip ultrasonic excitation vibration signal to obtain a plurality of IMF components; A preliminary denoising module, configured to calculate the normalized kurtosis and normalized information entropy of each IMF component, and screen the IMF components based on the normalized comprehensive index combining the normalized kurtosis and the normalized information entropy, and remove the noise components to obtain the initial signal components; A secondary denoising module, configured to use the principal component analysis method to perform dimensionality reduction and compression on all IMF components, extract the maximum frequency component characteristics of the fitting signal corresponding to each IMF component, and construct a reference signal for the frequency-domain adaptive filter according to all the maximum frequency component characteristics; use the frequency-domain adaptive filter to further denoise the reconstructed excitation signal of the initial signal components to obtain the target noise-reduced signal.

[0041] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0042] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce a means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or a means for implementing the functions specified in multiple blocks

[0043] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including an instruction means that implements the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or a means for implementing the functions specified in multiple blocks

[0044] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or a means for implementing the functions specified in multiple blocks

[0045] Obviously, the above embodiments are only examples for clear illustration and are not limitations on the implementation manners. For those of ordinary skill in the art, other different forms of changes or variations can be made based on the above description. It is not necessary and impossible to exhaustively list all implementation manners here. And the obvious changes or variations derived therefrom are still within the protection scope of the present invention

Claims

1. A method for denoising a flip chip ultrasonic excitation vibration signal, characterized in that: include: Acquiring ultrasonic excitation vibration signals of the flip chip; The ultrasonic excitation vibration signal of the flip chip is decomposed to obtain several IMF components; Calculate the normalized kurtosis and normalized information entropy of each IMF component, and screen the IMF components based on the normalized comprehensive index of the normalized kurtosis and normalized information entropy, remove the noise components, and obtain the initial signal components; The principal component analysis method is used to reduce the dimension of all IMF components, and the maximum frequency component characteristics of the fitting signal corresponding to each IMF component are extracted. The reference signal of the frequency domain adaptive filter is constructed according to all the maximum frequency component characteristics. The reconstructed excitation signal of the initial signal component is further denoised using the frequency domain adaptive filter to obtain the target denoised signal.

2. A flip chip ultrasonic excitation vibration signal denoising method according to claim 1, characterized in that: The ultrasonic excitation vibration signal of the flip chip is decomposed to obtain several IMF components, including: The ultrasonic excitation vibration signal of the flip chip is divided into multiple overlapping short time periods by using Welch transform, and the power spectrum estimation of the ultrasonic excitation vibration signal of the flip chip is calculated; The boundary coordinates for segmentation power spectrum estimation are obtained through an automatic multi-scale peak detection algorithm; A zero-phase filter bank is constructed according to the boundary coordinates, and the ultrasonic excitation vibration signal of the flip chip is subjected to zero-phase filtering to obtain several IMF components.

3. A flip chip ultrasonic excitation vibration signal denoising method according to claim 2, characterized in that: The power spectrum estimation of the ultrasonic excitation vibration signal of the flip chip is calculated by the formula: ； in, Power spectrum estimation of ultrasonic excitation vibration signal of flip chip, is the power spectrum estimate of the i-th overlapping short period, , represents the number of overlapping short periods, represents the total length of the ultrasonic excitation vibration signal of the flip chip, Indicates the length of each overlapping short period; is a normalization factor used to ensure that the obtained power spectrum estimate conforms to the asymptotically unbiased estimate; represents the window function, represents the i-th overlapping short period, Indicates frequency, Indicates time, Represents an imaginary unit.

4. A flip chip ultrasonic excitation vibration signal denoising method according to claim 2, characterized in that: The step of obtaining boundary coordinates for segmenting power spectrum estimation by an automatic multi-scale peak detection algorithm includes: Use a multi-scale window to detect the local maximum of the power spectrum estimate. The formula is: ； in, is the size of the local window, is the local maximum of the power spectrum estimate; Calculate the local maximum at each scale The peak index is: ； in, is a local maximum The peak index is used to evaluate the local maximum the degree of peak value on this scale; Filter out the peak indicators at each scale The largest peak value is selected, and the local maximum value corresponding to the peak value is used as the boundary coordinate.

5. The method for denoising a flip chip ultrasonically excited vibration signal according to claim 1, characterized in that: The calculation formula of the normalized kurtosis of the IMF component is: ； in, is the IMF component, is the normalized kurtosis, is the expected value of the IMF component, is the standard deviation of the IMF component, represents the expectation operation, Represents a normalization operation; The calculation formula of the normalized information entropy of the IMF component is: ； in, is the normalized information entropy, For time; The calculation formula of the normalized comprehensive index is: ； in, It is a normalized comprehensive indicator.

6. A flip chip ultrasonic excitation vibration signal denoising method according to claim 1, characterized in that: The method of screening the IMF component based on the normalized comprehensive index combined with the normalized kurtosis and the normalized information entropy includes: when the normalized comprehensive index of the IMF component is less than the sensitivity coefficient, the IMF component is a noise component.

7. A flip chip ultrasonic excitation vibration signal denoising method according to claim 1, characterized in that: The principal component analysis method is used to reduce the dimension of all IMF components and extract the maximum frequency component features, including: The signal characteristic matrix of each IMF component is obtained by piecewise reconstruction, and the signal characteristic matrix is ​​standardized; The correlation matrix is ​​obtained according to the standardized signal feature matrix; Calculate the eigenvalues ​​and corresponding eigenvectors of the correlation matrix; Select the eigenvector with the largest eigenvalue and reconstruct it into a fitting signal; The fitting signal is converted into the frequency domain, and the maximum frequency component characteristics of the fitting signal corresponding to each IMF vector are extracted.

8. A flip chip ultrasonic excitation vibration signal denoising method according to claim 7, characterized in that: The maximum frequency component characteristics include frequency and amplitude, and the calculation formula is: ； in, represents the fitted signal of the mth IMF component, and Represent the fitting signals The frequency and amplitude of the maximum frequency component feature; Indicates frequency, Indicates time, Represents an imaginary unit.

9. A flip chip ultrasonic excitation vibration signal denoising method according to claim 8, characterized in that: The reference signal of the frequency domain adaptive filter is constructed according to the characteristics of all maximum frequency components, and the formula is: ； in, represents the reference signal of the frequency domain adaptive filter, is the total number of IMF components.

10. A flip chip ultrasonic excitation vibration signal denoising system, characterized in that: include: A signal acquisition module, used to acquire the ultrasonic excitation vibration signal of the flip chip; A modal decomposition module is used to decompose the ultrasonic excitation vibration signal of the flip chip to obtain several IMF components; The initial noise reduction module is used to calculate the normalized kurtosis and normalized information entropy of each IMF component, and screen the IMF components based on the normalized comprehensive index of the normalized kurtosis and normalized information entropy, remove the noise components, and obtain the initial signal components; The secondary denoising module is used to use the principal component analysis method to perform dimensionality reduction compression on all IMF components, extract the maximum frequency component characteristics of the fitting signal corresponding to each IMF component, and construct the reference signal of the frequency domain adaptive filter according to all the maximum frequency component characteristics; the reconstructed excitation signal of the initial signal component is further denoised using the frequency domain adaptive filter to obtain the target denoised signal.

Images