GNSS-PPP structural dynamic deformation monitoring signal filtering algorithm based on improved Butterworth-CEEMDAN

By improving the Butterworth-CEEMDAN algorithm, combined with Butterworth low-pass filter and CEEMDAN decomposition, the problem of high low-frequency noise in GNSS-PP is solved, and accurate monitoring of dynamic deformation of the structure is achieved and monitoring accuracy is improved.

CN117665876BActive Publication Date: 2024-11-26TIANJIN UNIVERSITY OF SCIENCE & TECHNOLOGY ASSET MANAGEMENT CO LTD +1
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202311661014.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-12-05
Publication Date
2024-11-26
Estimated Expiration
2043-12-05

AI Technical Summary

Technical Problem

GNSS-PPP has high low-frequency noise in dynamic structural deformation monitoring, which is difficult to effectively weaken. The existing filter transition band is long, resulting in signal distortion, making it difficult to accurately capture dynamic structural deformation.

Method used

Using the improved Butterworth-CEEMDAN algorithm, the low-pass filter signal is obtained through the Butterworth low-pass filter, combined with the maximum correlation coefficient alignment and CEEMDAN decomposition, filter the real signal dominant components, and reconstruct the secondary filter signal to weaken low-frequency noise and high Frequent white noise.

Benefits of technology

Effectively eliminate the time-delay effect of Butterworth low-pass filter, minimize useful signal loss, improve GNSS-PPP monitoring accuracy, weaken low-frequency noise and high-frequency white noise components, and improve the accuracy of dynamic deformation monitoring of structures.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117665876B_ABST
    Figure CN117665876B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of GNSS positioning technology, and specifically to a GNSS-PPP structural dynamic deformation monitoring signal filtering algorithm based on an improved Butterworth-CEEMDAN, wherein a Butterworth low-pass filter is used to filter the GNSS-PPP monitoring original signal to obtain a low-pass filtered signal; the original signal and the low-pass filtered signal are aligned and subtracted based on the maximum correlation coefficient to obtain a high-frequency dynamic deformation component of the structure, that is, a primary filtered signal; the primary filtered signal is decomposed by CEEMDAN to obtain multiple IMF components and one residual component; the real signal dominant component is screened according to the correlation coefficient and energy coefficient threshold, and the real signal dominant IMF component is further determined and screened based on the power spectrum density function as a supplementary condition; the real signal dominant IMF component is reconstructed to obtain a secondary filtered signal. The combined filtering algorithm of the present invention eliminates the time lag effect of the traditional Butterworth filter filtering, can effectively separate the background noise component in the GNDD-PPP monitoring signal, and minimizes the loss of useful signals.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of GNSS positioning technology, and in particular to a GNSS-PPP structure dynamic deformation monitoring signal filtering algorithm based on improved Butterworth-CEEMDAN. Background Art

[0002] Real-time dynamic deformation monitoring plays an important role in the operation and maintenance of large civil engineering structures. It ensures safe operation by capturing the operating status of the structure in real time and detects abnormal deformation early, thus playing a role in structural disaster prevention and mitigation. In this context, real-time dynamic differential technology based on the Global Navigation Satellite System (GNSS) provides an efficient means for this monitoring work.

[0003] GNSS real-time dynamic differential technology, namely Real Time Kinematic (RTK), is a widely used method. However, due to the requirements of signal differential analysis and the challenges of base station deployment, GNSS-RTK has monitoring bottlenecks in some scenarios. In view of this, Precise Point Positioning (PPP) can be used as a supplement or alternative to GNSS-RTK and demonstrate its application value in real-time dynamic deformation monitoring of structures.

[0004] It is worth noting that although the spectrum distribution of low-frequency noise and white noise of GNSS-PPP is similar to that of GNSS-RTK, the low-frequency noise of GNSS-PPP is usually larger than that of GNSS-RTK because its signal also contains multipath errors, white noise inside the instrument, and other interference factors such as tropospheric delay and ionospheric delay.

[0005] In order to effectively attenuate these low-frequency noises, researchers turned to passband filtering technology. Butterworth filters and Chebyshev filters are the most commonly used passband filters. They can filter and reduce noise within a specific frequency band according to the design parameters. In particular, both filters have excellent performance in reducing low-frequency multipath errors in GNSS-RTK monitoring signals. The advantage of the Butterworth filter is that the frequency response within its passband is very flat and has no ripples, which can maximize the retention of useful signals and minimize background noise. However, its disadvantage is that its transition band is long, which may cause signal distortion.

[0006] In fact, the ideal filter should attenuate directly to 0 in the stop band and have no transition band. But this is difficult to achieve in practical applications. In addition, the quasi-static deformation of the structure and the low-frequency noise of GNSS-PPP are similarly distributed in the frequency domain, making it difficult to effectively separate the two. For this reason, in view of the current background, further research and adjustment of the relevant filtering algorithm are needed to ensure that the filter does not lose the accurate capture of the dynamic deformation of the structure while removing noise. Summary of the invention

[0007] Based on the above objectives, the present invention provides a GNSS-PPP structure dynamic deformation monitoring signal filtering algorithm based on improved Butterworth-CEEMDAN.

[0008] The GNSS-PPP structure dynamic deformation monitoring signal filtering algorithm based on the improved Butterworth-CEEMDAN includes the following steps:

[0009] S1: Use Butterworth low-pass filter to filter the GNSS-PPP monitoring raw signal to obtain the low-pass filtered signal;

[0010] S2: Align and subtract the original signal and the low-pass filtered signal based on the maximum correlation coefficient to obtain the high-frequency dynamic deformation component of the structure, which is the primary filtered signal;

[0011] S3: Perform CEEMDAN decomposition on the primary filtered signal to obtain multiple IMF components and one residual component;

[0012] S4: screening the real signal dominant component according to the correlation coefficient and the energy coefficient threshold, and judging whether the screened component is the real signal dominant IMF component based on the supplementary conditions;

[0013] S5: Reconstruct the IMF component to obtain a secondary filtered signal.

[0014] Furthermore, the S1 specifically includes:

[0015] S11: Determine the technical indicators of the filter, passband cutoff frequency ω p , stopband cutoff frequency ω s 、Passband maximum attenuation α p and stopband minimum attenuation α s , and their mutual relationship is:

[0016] α p =-10lg|H a (jω p )| 2 ;

[0017] α s =-10lg|H a (jωs )| 2 ;

[0018] S12: Let the constant be ε, which is used to adjust the maximum attenuation allowed in the passband, and then determine the passband cutoff frequency. When ω=ω p When , the maximum attenuation in the channel occurs, and ε is expressed as:

[0019]

[0020] S13: Assume that the minimum filter order is m, where m depends on the minimum stopband attenuation α s , when ω=ω s When , the minimum attenuation in the channel occurs, and m is expressed as:

[0021]

[0022] Furthermore, the constant ε is determined by the maximum attenuation α allowed in the passband. p Determined, and the constant ε is less than 3dB.

[0023] Furthermore, the S2 specifically includes:

[0024] S21: Calculate the original signal X(t) and the low-pass filtered signal x 0 (t+a), and determine the time difference a corresponding to the maximum value of the correlation coefficient;

[0025] S22: x 0 Subtract the translation time difference a from X(t) to obtain a filtered signal x 1 (t);

[0026] S23: Further, the time difference a corresponding to the maximum value of the correlation coefficient is calculated as follows:

[0027]

[0028] Where X(t) is the original signal, and x(t+a) represents the original signal X(t) delayed by a units in time.

[0029] Furthermore, the S3 specifically includes:

[0030] S31: Add adaptive white noise ε to the original signal x(t) 0 E 1 ω j (t), and then perform empirical mode decomposition (EMD) to obtain the preliminary IMF components expressed as f 1 j (t), and then f 1 j(t) is averaged to obtain the first-order IMF component

[0031] S32: Original signal x(t) and first-order IMF component Subtract and get the new signal to be decomposed r 1 (t);

[0032] S33: R 1 (t) Add white noise ε 1 E 1 ω i (t) Continue to decompose and obtain the second-order IMF component By analogy, we get the i-th signal to be decomposed r i (t) and the i+1th order IMF component;

[0033] S34: cyclically calculating the signal to be decomposed and the IMF component until the decomposition is completed and the final residual component is obtained.

[0034] Furthermore, the first-order IMF component The formula is:

[0035]

[0036] The new signal to be decomposed r 1 (t) The formula is:

[0037] Second-order IMF components It is expressed as:

[0038] Where: E 1 (·) represents an empirical mode decomposition (EMD), ω j (t) is the white noise added for the jth time, ε 1 is the standard deviation of added white noise;

[0039] The i-th signal to be decomposed r i (t) is expressed as:

[0040] i+1th order IMF component It is expressed as:

[0041] The final residual component is expressed as:

[0042] The original signal x(t) is expressed as:

[0043] Among them, the 1st to kth IMF components are dominated by high-frequency white noise, the (k+1)th to pth IMF components are dominated by real signals, and the (p+1)th to nth IMF components and the residual component r(t) are dominated by low-frequency multipath errors.

[0044] Further S4 specifically includes:

[0045] S41: The correlation between the real IMF component and the original signal is high, while the correlation between the white noise-dominated IMF component and the original signal is low. Therefore, the correlation coefficient can be introduced to screen and identify the real IMF component, which can be expressed as: Where x(t) and y(t) represent two signals respectively, and N represents the signal length.

[0046] Furthermore, the threshold for distinguishing the noise-dominated IMF component based on the correlation coefficient can be expressed as: In the formula, ρ i is the correlation coefficient between the ith IMF component and the original signal, max(ρ i ) is the maximum value of the IMF component and the original correlation coefficient, and n is the number of IMF components;

[0047] S42: Considering that in actual monitoring applications, only the correlation coefficient between the IMF component and the original noisy signal can be calculated, when the original monitoring signal contains a large amount of noise components, the false IMF component will also show a strong correlation with the original monitoring signal. In view of this, an energy coefficient threshold is proposed to further filter the IMF component, which can be expressed as: Where D i is the energy density, is the average period.

[0048] Furthermore, the expression of energy density is: In the formula, f i (t) is the amplitude of the ith IMF component, and the expression of the average period is: In the formula, O i is the number of extreme values ​​of the ith IMF component.

[0049] Furthermore, when the energy coefficient of a certain item increases geometrically compared to the average value of the energy coefficients of the previous items, it can be determined that this item is non-constant, and the previous constants are the white noise characteristic components. The energy coefficient threshold can be expressed as:

[0050] S43: The correlation coefficient and energy coefficient thresholds are used to select the dominant components of the real signal as IMF(k+1)~IMFp, and then the power spectrum density (PSD) function is used as a supplementary condition to determine whether IMFk and IMF(p+1) are the dominant IMF components of the real signal.

[0051] Beneficial effects of the present invention:

[0052] The present invention proposes an improved Butterworth low-pass filtering algorithm based on maximum correlation coefficient alignment, which can eliminate the time lag effect of the Butterworth low-pass filter, effectively extract the low-pass filtering signal, and avoid the loss of useful signals to the greatest extent.

[0053] The present invention adopts the improved Butterworth algorithm and the CEEMDAN algorithm for joint filtering, which has a good application effect on the processing of GNSS-PPP structure dynamic deformation monitoring data, and can effectively weaken the low-frequency noise component and the high-frequency white noise component in the monitoring signal, thereby improving the GNSS-PPP monitoring accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 A schematic diagram of an algorithm application flow of an embodiment of the present invention;

[0056] Figure 2 A simulation signal diagram of an embodiment of the present invention;

[0057] Figure 3 is a translation signal diagram of an embodiment of the present invention;

[0058] Figure 4 This is a GNSS-PPP monitoring signal diagram of an embodiment of the present invention;

[0059] Figure 5 This is a GNSS-PPP low-pass filtering signal diagram of an embodiment of the present invention;

[0060] Figure 6 This is a diagram of high-frequency signal components of GNSS-PPP in the east-west direction according to an embodiment of the present invention;

[0061] Figure 7 The east-west direction IMF component diagram and the corresponding PSD function diagram of an embodiment of the present invention;

[0062] Figure 8 It is a dynamic deformation component diagram of the structure in the east-west direction according to an embodiment of the present invention;

[0063] Fig. 9 It is a PSD function diagram of the dynamic deformation component of the structure in the east-west direction according to an embodiment of the present invention. DETAILED DESCRIPTION

[0064] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0065] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0066] like Figure 1-Figure 9 As shown in the figure, the GNSS-PPP structural dynamic deformation monitoring signal filtering algorithm based on the improved Butterworth-CEEMDAN is firstly, the GNSS-PPP monitoring signal is filtered by a Butterworth low-pass filter to obtain a low-pass filtered signal; then, the original signal and the low-pass filtered signal are aligned based on the maximum correlation coefficient and then subtracted to obtain the high-frequency dynamic deformation component of the structure; finally, the high-frequency dynamic deformation component of the structure is decomposed by CEEMDAN, the white noise-dominated IMF component is eliminated based on the triple threshold method, and the real IMF component is reconstructed to obtain the secondary filtering signal, which specifically includes the following steps:

[0067] 1. Design a Butterworth low-pass filter to perform primary filtering.

[0068] First, determine the technical indicators of the filter: passband cutoff frequency ω p , stopband cutoff frequency ω s 、Passband maximum attenuation α p and stopband minimum attenuation α s , and their mutual relationship is:

[0069] α p =-10lg|H a (jω p )| 2 ;

[0070] α s =-10lg|H a (jω s )| 2 ;

[0071] The function of the constant ε is to adjust the maximum attenuation allowed in the passband to less than 3dB, thereby determining the cutoff frequency. ε is determined by the maximum attenuation allowed in the passband α p Determine, when ω=ω p When , the maximum attenuation in the channel occurs. ε can be expressed as:

[0072]

[0073] The minimum filter order m depends on the minimum attenuation α in the stop band. s , when ω=ω s When , the minimum attenuation in the channel occurs. m can be further expressed as:

[0074]

[0075] After the original signal X(t) is processed by the Butterworth low-pass filter, the low-frequency signal component x is obtained. 0 (t).

[0076] 2. Alignment based on correlation coefficients to eliminate the time lag effect of the Butterworth low-pass filter.

[0077] Calculate X(t) and x 0 (t+a), and determine the time difference a corresponding to the maximum value of the correlation coefficient. The calculation formula is:

[0078]

[0079] Finally, x 0 (t) is shifted by a and subtracted from X(t) to obtain a filtered signal x 1 (t).

[0080] 3. Carry out secondary filtering based on CEEMDAN algorithm.

[0081] Add adaptive white noise ε to the original signal x(t) 0 E 1 ω j (t), and then perform EMD decomposition to obtain the first-order IMF component f 1 j (t). Then for f 1 j (t) is averaged to obtain the first-order IMF component:

[0082]

[0083] Original signal x(t) and IMF components Subtract and get the new signal to be decomposed r 1 (t):

[0084]

[0085] R 1 (t) Add white noise ε 1 E 1 ω i (t) Continue to decompose and obtain the second-order IMF component:

[0086]

[0087] Where: E 1 (·) represents an EMD decomposition, ω j (t) is the white noise added for the jth time, ε 1 is the standard deviation of added white noise;

[0088] And so on, we get the i-th signal to be decomposed:

[0089]

[0090] The i+1th order IMF component is:

[0091]

[0092] Repeat steps 4) to 5) until the decomposition is complete and the final residual component is obtained:

[0093]

[0094] x(t) can be further expressed as:

[0095]

[0096] Where: the 1st to kth IMF components are dominated by high-frequency white noise, the (k+1)th to pth IMF components are dominated by real signals, and the (p+1)th to nth IMF components and the residual component r(t) are dominated by low-frequency multipath errors.

[0097] 4. Screen the dominant IMF components of the real signal based on the correlation coefficient, energy coefficient threshold and PSD function.

[0098] The correlation between the real IMF component and the original signal is high, while the correlation between the white noise-dominated IMF component and the original signal is low. Therefore, the correlation coefficient can be introduced to screen and identify the real IMF component, which can be expressed as: In the formula, x(t) and y(t) represent two signals respectively, and N represents the signal length;

[0099] The threshold for distinguishing the noise-dominated IMF component based on the correlation coefficient can be expressed as: In the formula, ρi is the correlation coefficient between the ith IMF component and the original signal, max(ρ i ) is the maximum value of the IMF component and the original correlation coefficient, and n is the number of IMF components.

[0100] Considering that only the correlation coefficient between the IMF component and the original noisy signal can be calculated in actual monitoring applications, if there is a strong correlation between the noisy IMF component and the original noisy signal, it may lead to inaccurate judgment based solely on the correlation coefficient. In view of this, the energy coefficient threshold is proposed to further screen the IMF component, which can be expressed as: Where D i is the energy density, is the average period;

[0101] The expression of energy density is: In the formula, f i (t) is the amplitude of the ith IMF component, and the expression of the average period is: In the formula, O i is the number of extreme values ​​of the ith IMF component;

[0102] When the energy coefficient of a certain item increases geometrically compared to the average value of the energy coefficients of the previous items, it can be determined that this item is non-constant, and the previous constants are the white noise characteristic components. The energy coefficient threshold can be expressed as:

[0103]

[0104] The correlation coefficient and energy coefficient thresholds are used to screen the dominant components of the real signal as IMF(k+1)~IMFp, and then the power spectral density (PSD) function is used as a supplementary condition to determine whether IMFk and IMF(p+1) are the dominant IMF components of the real signal;

[0105] The real signal dominant components IMF(k+1)~IMFp are screened according to the correlation coefficient and energy coefficient threshold, and the PSD function is used as a supplementary condition to determine whether IMF k and IMF(p+1) are the real signal dominant IMF components. Finally, the IMF components are reconstructed to obtain the secondary filtered signal.

[0106] The present invention is divided into the application of simulated signals and the application of measured signals for verification respectively.

[0107] Simulation signal application:

[0108] The fundamental frequency of vibration of large civil engineering structures is relatively low. For example, the natural fundamental frequency of the sky tower is about 0.16 Hz. Considering the large amplitude of low-frequency noise of GNSS-PPP, this paper uses a 0.03 Hz sine signal with an amplitude of 3 cm to simulate low-frequency noise, and uses superimposed sine signals of 0.10 Hz and 0.20 Hz to simulate structural deformation components, with amplitudes of 2 cm and 1 cm respectively. The simulation signal is expressed as:

[0109] y 1 =3·sin(2π·0.03·t);

[0110] y 2 =2·sin(2π·0.10·t);

[0111] y 3 = sin(2π·0.20·t);

[0112] y=y 1 +y 2 +y 3 ;

[0113] In addition, Gaussian white noise is added to the simulation signal y, and the noisy signal can be expressed as:

[0114] Y=y+y noise ;

[0115] The signal sampling interval is set to 1s, the signal length is 600s, and the simulation signal is as follows Figure 2 (From left to right, from top to bottom: sinusoidal signal y 1 , sinusoidal signal y 2 , sinusoidal signal y 3 , white noise signal y noise , noise signal Y) is shown.

[0116] First, a Butterworth low-pass filter is used for filtering. Based on past experience and signal simulation, an 8th-order Butterworth low-pass filter with a cutoff frequency of 0.5Hz is determined. The theoretical process of filter design is as follows:

[0117] First, determine the technical indicators of the filter: passband cutoff frequency ω p , stopband cutoff frequency ω s 、Passband maximum attenuation α p and stopband minimum attenuation α s , and their mutual relationship is:

[0118] α p =-10lg|H a (jω p )| 2 ;

[0119] αs =-10lg|H a (jω s )| 2 ;

[0120] The function of the constant ε is to adjust the maximum attenuation allowed in the passband to less than 3dB, thereby determining the cutoff frequency. ε is determined by the maximum attenuation allowed in the passband α p Determine, when ω=ω p When , the maximum attenuation in the channel occurs, and ε can be expressed as:

[0121]

[0122] The minimum filter order m depends on the minimum attenuation α in the stop band. s , when ω=ω s When , the minimum attenuation in the channel occurs, and m can be further expressed as:

[0123]

[0124] The original noisy signal Y(t) is processed by the Butterworth low-pass filter to obtain the low-pass filtered signal Y 0 (t); calculate Y(t) and Y 0 (t+a), and determine the time difference a corresponding to the maximum value of the correlation coefficient. The calculation formula is:

[0125]

[0126] After calculation, when a=12s, the sinusoidal signal y 1 With the low-pass filtered signal Y 0 The correlation coefficient between the two is the largest, which is 0.9933. 1 The translation signal obtained by translation for 12 seconds is as follows Figure 3 As shown, the translation signal and the sinusoidal signal y 1 ( Figure 2 The results show that the improved Butterworth low-pass filter can effectively extract the low-frequency signal components, and the correlation coefficient-based translation alignment can effectively eliminate the influence of signal time lag effect.

[0127] The high-frequency signal component can be expressed as:

[0128] Y 1 =YY 0 ;

[0129] Next, the high-frequency signal components are decomposed by CEEMDAN to weaken the white noise interference. The specific decomposition principle is as follows:

[0130] Add adaptive white noise ε to the original signal x(t)0 E 1 ω j (t), and then perform EMD decomposition to obtain the first-order IMF component f 1 j (t). Then for f 1 j (t) is averaged to obtain the first-order IMF component:

[0131]

[0132] Original signal x(t) and IMF components Subtract and get the new signal to be decomposed r 1 (t):

[0133]

[0134] R 1 (t) Add white noise ε 1 E 1 ω i (t) Continue to decompose and obtain the second-order IMF component:

[0135]

[0136] Where: E 1 (·) represents an EMD decomposition, ω j (t) is the white noise added for the jth time, ε 1 is the standard deviation of the added white noise.

[0137] And so on, we get the i-th signal to be decomposed:

[0138]

[0139] The i+1th order IMF component is:

[0140]

[0141] Repeat the above two steps until the decomposition is completed and the final residual component is obtained:

[0142]

[0143] This formula can be further expressed as:

[0144]

[0145] Where: the 1st to kth IMF components are dominated by high-frequency white noise, the (k+1)th to pth IMF components are dominated by real signals, and the (p+1)th to nth IMF components and the residual component r(t) are dominated by low-frequency multipath errors.

[0146] The calculation process of the above algorithm is not described here. Based on the improved Butterworth-CEEMDAN, the secondary filtering signal is obtained, and the signal-to-noise ratio (SNR), standard deviation (STD) and correlation coefficient are used as indicators to quantitatively evaluate the filtering and denoising results of different algorithms. The results are shown in Table 1. Compared with the noisy signal Y, the SNR and correlation coefficient of the primary filtering signal obtained by the improved Butterworth low-pass filter based on the correlation coefficient alignment are significantly increased, and the STD is significantly reduced; on this basis, the SNR value of the secondary filtering signal obtained by the improved Butterworth-CEEMDAN algorithm is further increased, the STD value is further reduced, and the correlation coefficient with the original sinusoidal superposition signal is the largest, which confirms the filtering and denoising effect of the improved Butterworth-CEEMDAN algorithm.

[0147] Table 1 Summary of signal filtering and noise reduction effects of different algorithms

[0148]

[0149]

[0150] Measured signal application:

[0151] This experiment aims to verify the feasibility of GNSS-PPP monitoring of the dynamic deformation of the Tianta Tower and the effectiveness of the joint filtering algorithm. The monitoring station is arranged on the south side of the outdoor circular observation deck at a height of 284 meters. During the experiment, the observation deck was closed to the public to avoid interference from adverse factors such as the flow of people. The GNSS receiver is supported by a pole and fixed to the inner railing of the observation deck with the help of tape. Generally speaking, the dynamic response of super-high-rise structures under environmental excitation is mainly excited by wind loads. During the experiment, the wind speed was relatively stable, varying between 1.8m / s and 6.0m / s. The GNSS-PPP monitoring signal is as follows: Figure 4 As shown (the starting point is at the top for the south-north direction, and the starting point is at the bottom for the east-west direction).

[0152] First, a Butterworth low-pass filter is used for filtering, and an 8th-order Butterworth low-pass filter with a cutoff frequency of 0.5Hz is designed. The theoretical process of filter design is as follows:

[0153] First, determine the technical indicators of the filter: passband cutoff frequency ω p , stopband cutoff frequency ω s 、Passband maximum attenuation α p and stopband minimum attenuation α s , and their mutual relationship is:

[0154] α p =-10lg|H a (jω p )| 2;

[0155] α s =-10lg|H a (jω s )| 2 ;

[0156] The function of the constant ε is to adjust the maximum attenuation allowed in the passband to less than 3dB, thereby determining the cutoff frequency. ε is determined by the maximum attenuation allowed in the passband α p Determine that when ω=ω p When , the maximum attenuation in the channel occurs, and ε can be expressed as:

[0157]

[0158] The minimum filter order m depends on the minimum attenuation α in the stop band. s , when ω=ω s When , the minimum attenuation in the channel occurs. m can be further expressed as:

[0159]

[0160] After the original signal X(t) is processed by the Butterworth low-pass filter, the low-frequency signal component x is obtained. 0 (t). Calculate X(t) and x 0 (t+a), and determine the time difference a corresponding to the maximum value of the correlation coefficient. The calculation formula is:

[0161]

[0162] Finally, x 0 (t) is translated by a and is subtracted from X(t) to obtain a low-pass filtered signal;

[0163] The GNSSS-PPP low-pass filtered signal is Figure 5 As shown in the figure (the starting point is at the top for the south-north direction, and the starting point is at the bottom for the east-west direction), it can be seen that the amplitudes of the low-pass filtered signals in the south-north and east-west directions are basically the same, which is ±5cm, and have certain similarities locally, with a correlation coefficient of -0.4120. The low-pass filtered signal contains the GNSS-PPP low-frequency noise component and the quasi-static deformation component of the structure. Although the monitoring accuracy is low, it can still reflect the quasi-static deformation trend and abnormal deformation of the structure to a certain extent, thereby providing a reference for the operation status and safety assessment of the tower.

[0164] In order to eliminate the time lag effect of the Butterworth low-pass filter, combined with the simulation signal verification results, the GNSS-PPP low-pass filtered signal is shifted forward by 12s. At this time, the correlation coefficient with the GNSS-PPP original monitoring signal is the largest. The original monitoring signal is subtracted from the shifted low-pass filtered signal to obtain the GNSS-PPP high-frequency signal component (primary filtered signal).

[0165] Considering that the data with a duration of 12000s has a large amount of calculation, and the data length has no significant effect on the dynamic deformation analysis results, in order to improve the calculation efficiency, the data with a duration of 3600s is intercepted for decomposition and analysis. The intercepted east-west GNSS-PPP high-frequency signal components are as follows: Figure 6 As shown, it can be seen that the displacement amplitudes of the high-frequency signal components in the two directions are basically the same. Compared with the original signal, the amplitude of the high-frequency signal component is significantly reduced due to the removal of the low-frequency trend term, but the instrument white noise component still exists. Next, the high-frequency signal component is decomposed by CEEMDAN to weaken the white noise interference. The specific decomposition principle is as follows:

[0166] Add adaptive white noise ε to the original signal x(t) 0 E 1 ω j (t), and then perform EMD decomposition to obtain the first-order IMF component f 1 j (t), and then f 1 j (t) is averaged to obtain the first-order IMF component:

[0167]

[0168] Original signal x(t) and IMF components Subtract and get the new signal to be decomposed r 1 (t):

[0169]

[0170] R 1 (t) Add white noise ε 1 E 1 ω i (t) Continue to decompose and obtain the second-order IMF component:

[0171]

[0172] Where: E 1 (·) represents an EMD decomposition, ω j (t) is the white noise added for the jth time, ε 1 is the standard deviation of the added white noise.

[0173] And so on, we get the i-th signal to be decomposed:

[0174]

[0175] The i+1th order IMF component is:

[0176]

[0177] Repeat the above two steps until the decomposition is completed and the final residual component is obtained:

[0178]

[0179] This formula can be further expressed as:

[0180]

[0181] Where: the 1st to kth IMF components are dominated by high-frequency white noise, the (k+1)th to pth IMF components are dominated by real signals, and the (p+1)th to nth IMF components and the residual component r(t) are dominated by low-frequency multipath errors.

[0182] Taking the east-west direction as an example, the CEEMDAN decomposition is performed on the GNSS-PPP primary filtering signal to obtain 10 IMF components. The IMF components and the corresponding PSD functions are shown in Figure 7 As shown, it can be seen that the IMF 1 component presents obvious white noise characteristics. The white noise is widely and evenly distributed in the frequency domain. The PSD function corresponding to the IMF 2 component can observe an obvious peak, while the peaks of the other components are not obvious.

[0183] The correlation coefficient and energy coefficient of each IMF component and the high-frequency signal component are calculated. The results are shown in Table 2. The correlation coefficient threshold δ of the IMF component E-W =max(C i ) / 10=0.0679, based on which IMF 1~4 are judged as real components; the energy coefficient threshold is set to 1, and IMF 2 is judged as the real deformation component of the structure according to the energy coefficient threshold. It can be seen from the PSD function that the IMF1 component mainly presents white noise characteristics, and there are no obvious peaks in IMF 1 and IMF 5 components. Based on the triple threshold method of correlation coefficient, energy coefficient and PSD function, IMF 2~4 components are reconstructed as structural dynamic deformation components, as shown in Figure 8 shown.

[0184] Table 2 Correlation coefficients and energy coefficient thresholds between IMF components and signals before decomposition

[0185]

[0186] Note: 1) ** indicates significant correlation at the 0.01 level (two-tailed); 2) Quantities that meet the threshold conditions are marked in bold.

[0187] Fig. 9 is the PSD function of the dynamic deformation component of the structure in the east-west direction. After the improved Butterworth-CEEMDAN decomposition, the energy of the low-frequency noise component and the high-frequency white noise is significantly weakened, while the energy loss of the corresponding structural vibration fundamental frequency is small, indicating that the dynamic deformation component and the natural frequency of the structure can be effectively extracted. In addition, an obvious peak can be identified from the figure, and the corresponding frequency is 0.1586Hz. This result is consistent with the structural vibration fundamental frequency obtained based on the GNSS-RTK synchronous monitoring value, and is also consistent with the field measurement results of Chen Zhipeng et al. The above research results show that the dynamic deformation of the sky tower in the horizontal direction can be effectively monitored based on the GNSS-PPP technology to extract the natural frequency.

[0188] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention is limited to these examples. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity.

[0189] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A GNSS-PPP structural dynamic deformation monitoring signal filtering algorithm based on the improved Butterworth-CEEMDAN, characterized by: The following steps are involved: S1: Use Butterworth low-pass filter to filter the original signal of GNSS-PPP monitoring to obtain low-pass filtered signal; S2: Align and subtract the original signal and the low-pass filtered signal based on the maximum correlation coefficient to obtain the high-frequency dynamic deformation component of the structure, which is the primary filtered signal; S3: Perform CEEMDAN decomposition on the primary filtered signal to obtain multiple IMF components and one residual component; S4: Screening the dominant component of the real signal according to the correlation coefficient and the energy coefficient threshold, and judging whether the screened component is the dominant IMF component of the real signal based on the supplementary condition of the power spectrum density PSD function; S5: Reconstruct the dominant IMF component of the real signal and obtain the secondary filtered signal.

2. The GNSS-PPP structure dynamic deformation monitoring signal filtering algorithm based on improved Butterworth-CEEMDAN according to claim 1 is characterized in that: The S1 specifically includes: S11: Determine the technical indicators of the filter, passband cutoff frequency ω p , stopband cutoff frequency ω s 、Passband maximum attenuation α p and stopband minimum attenuation α s , and their mutual relationship is: α p =-10lg|H a (jω p )| 2 ; α s =-10lg|H a (jω s )| 2 ; S12: Let the constant be ε, which is used to adjust the maximum attenuation allowed in the passband, and then determine the passband cutoff frequency. When ω=ω p When , the maximum attenuation in the channel occurs, and ε is expressed as: S13: Assume that the minimum filter order is m, where m depends on the minimum stopband attenuation α s , when ω=ω s When , the minimum attenuation in the channel occurs, and m is expressed as: S14: The original signal X(t) is processed by a Butterworth low-pass filter to obtain a low-pass filtered signal x0(t).

3. The GNSS-PPP structure dynamic deformation monitoring signal filtering algorithm based on improved Butterworth-CEEMDAN according to claim 2 is characterized in that: The constant ε is determined by the maximum attenuation allowed in the passband α p Determined, and the constant ε is less than 3dB.

4. The GNSS-PPP structure dynamic deformation monitoring signal filtering algorithm based on improved Butterworth-CEEMDAN according to claim 3 is characterized in that: The S2 specifically includes: S21: Calculate the correlation coefficient between the original signal X(t) and the low-pass filtered signal x0(t+a), and determine the time difference a corresponding to the maximum value of the correlation coefficient; S22: Subtract the time difference a of x0(t) from X(t) to obtain a filtered signal x1(t); S23: Further, the time difference a corresponding to the maximum value of the correlation coefficient is calculated as follows: Where X(t) is the original signal, and x(t+a) represents the original signal X(t) delayed by a units in time.

5. The GNSS-PPP structure dynamic deformation monitoring signal filtering algorithm based on improved Butterworth-CEEMDAN according to claim 4 is characterized in that: The S3 specifically includes: S21: Add adaptive white noise ε0E1ω to the original signal x(t) j (t), and then perform empirical mode decomposition to obtain the preliminary IMF component expressed as f1 j (t), and then f1 j (t) is averaged to obtain the first-order IMF component S22: Original signal x(t) and first-order IMF component Make a difference and get the new signal to be decomposed r1(t); S23: Add white noise ε1E1ω to r1(t) i (t) Continue to decompose and obtain the second-order IMF component By analogy, we get the i-th signal to be decomposed r i (t) and the i+1th order IMF component; S24: cyclically calculating the signal to be decomposed and the IMF component until the decomposition is completed and the final residual component is obtained.

6. The GNSS-PPP structure dynamic deformation monitoring signal filtering algorithm based on improved Butterworth-CEEMDAN according to claim 5 is characterized in that: The first-order IMF component The formula is: The formula of the new signal to be decomposed r1(t) is: Second-order IMF components It is expressed as: Where: E1(·) represents an empirical mode decomposition (EMD), ω j (t) is the jth added white noise, ε1 is the standard deviation of the added white noise; The i-th signal to be decomposed r i (t) is expressed as: i+1th order IMF component It is expressed as: The final residual component is expressed as: The signal x(t) is expressed as: Among them, the 1st to kth IMF components are dominated by high-frequency white noise, the (k+1)th to pth IMF components are dominated by real signals, and the (p+1)th to nth IMF components and the residual component r(t) are dominated by low-frequency multipath errors.

7. The GNSS-PPP structure dynamic deformation monitoring signal filtering algorithm based on improved Butterworth-CEEMDAN according to claim 6 is characterized in that: The correlation coefficient and energy coefficient thresholds screen the real signal dominant components as IMF(k+1) to IMFp, and then use the power spectrum density PSD function as a supplementary condition to determine whether IMFk and IMF(p+1) are the real signal dominant IMF components.

8. The GNSS-PPP structure dynamic deformation monitoring signal filtering algorithm based on improved Butterworth-CEEMDAN according to claim 7 is characterized in that: The correlation coefficient in S4 for screening and discriminating the true IMF component is expressed as: In the formula, x(t) and y(t) represent two signals respectively, and N represents the signal length; The threshold for distinguishing the noise-dominated IMF component based on the correlation coefficient is expressed as: In the formula, ρ i is the correlation coefficient between the ith IMF component and the original signal, max(ρ i ) is the maximum value of the IMF component and the original correlation coefficient, and n is the number of IMF components; The energy coefficient threshold screening IMF component in S4 is expressed as: Where D i is the energy density, is the average period; The expression of energy density is: In the formula, f i (t) is the amplitude of the ith IMF component, and the expression of the average period is: In the formula, O i is the number of extreme values ​​of the ith IMF component; When the energy coefficient of a certain item increases geometrically compared to the average value of the energy coefficient of the previous item, it can be determined that this item is non-constant. The previous constant is the characteristic component of white noise. The energy coefficient threshold is expressed as: Based on the correlation coefficient and energy coefficient threshold, the dominant components of the real signal are screened as IMF(k+1)~IMFp, and then the power spectral density PSD function is used as a supplementary condition to determine whether IMFk and IMF(p+1) are the dominant IMF components of the real signal.

Citation Information

Patent Citations

  • Ultrasonic signal processing method based on CEEMDAN combined wavelet packet threshold

    CN114264724A

  • CEEMDAN decomposition and periodic term extraction method, system, equipment and medium

    CN115964614A

  • Aircraft water entry impact simulation nonlinear data processing method and application

    CN116522087A

  • Systems and methods for supporting a location uncertainty for a scheduled location

    US20220360943A1