CN-122019976-A - Double-constraint second-order blind identification bridge dynamic deflection signal noise reduction method

CN122019976ACN 122019976 ACN122019976 ACN 122019976ACN-122019976-A

Abstract

The invention discloses a double-constraint second-order blind identification bridge dynamic deflection signal noise reduction method which comprises the steps of collecting dynamic deflection observation signals of bridge target points, constructing observation signal matrixes by at least three observation signals of adjacent target points, carrying out centering treatment on the observation signal matrixes, obtaining whitened signals of which a covariance matrix is a unit matrix through whitening transformation, selecting a group of non-zero time delay values, calculating delay covariance matrixes of the whitened signals under each time delay, carrying out joint approximate diagonalization on all the delay covariance matrixes to obtain a separation matrix and initial separation source signals, optimizing the separation matrix based on sparsity constraint to obtain a mixed matrix, carrying out main frequency optimization on the initial separation source signals, and carrying out reconstruction based on the optimized mixed matrix and the optimized main frequency source signals to obtain the bridge dynamic deflection signals after noise reduction. The mixed matrix optimization based on sparsity constraint enables the reconstruction result to be closer to the real bridge dynamic deformation.

Inventors

WU HAIQIAN
CAI SIYAO
WANG RUNJIE
LIU XIANGLEI
LU ZHAO
KANG JUNYU
YU BAIHUI
JIA YUNXIANG
Ayi Chuwak
WANG DAPENG

Assignees

北京建筑大学

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (7)

1. A double-constraint second-order blind identification bridge dynamic deflection signal noise reduction method is characterized by comprising the following steps: Collecting dynamic deflection observation signals of bridge target points, and constructing an observation signal matrix from the observation signals of at least three adjacent target points; Performing centering treatment on the observation signal matrix, and obtaining a whitened signal with a covariance matrix as a unit matrix through whitening transformation; selecting a group of non-zero time delay values, calculating a time delay covariance matrix of the whitened signal under each time delay, and carrying out joint approximate diagonalization on all the time delay covariance matrices to obtain a separation matrix and an initial separation source signal; optimizing the separation matrix based on sparsity constraint to obtain a mixed matrix; and reconstructing based on the optimized mixing matrix and the optimized main frequency source signal to obtain a bridge dynamic deflection signal after noise reduction.
2. The dual-constraint second-order blind identification bridge dynamic deflection signal noise reduction method of claim 1 is characterized in that optimizing the separation matrix based on sparsity constraint to obtain a mixed matrix specifically comprises extracting absolute value vectors of each row of the separation matrix, identifying and marking elements which have the weakest contribution to output signals row by row, and zeroing the elements to obtain the optimized mixed matrix.
3. The dual-constraint second-order blind identification bridge dynamic deflection signal noise reduction method according to claim 1, wherein the method is characterized by comprising the following steps of: the main frequency optimization of the initial separation source signal specifically comprises the following steps: performing power spectrum density estimation on an original target signal in an observation signal matrix, and determining a main frequency of a bridge dynamic deflection signal; and carrying out power spectrum density estimation on the component with the weakest contribution degree in the initial separation source signal, screening and retaining the signal component containing the main frequency, and obtaining the optimized source signal.
4. The double-constraint second-order blind identification bridge dynamic deflection signal noise reduction method is characterized in that a dynamic deflection observation signal of a collected bridge target point is sampled by adopting a ground-based synthetic aperture radar, and the sampling frequency is not lower than 100Hz.
5. The dual-constraint second-order blind identification bridge dynamic deflection signal noise reduction method of claim 1 is characterized in that the selection range of the non-zero time delay value is 1-50 sampling periods, joint approximate diagonalization is achieved through orthogonal matrix transformation, and the transformed time delay covariance matrix approximates to a diagonal matrix.
6. The double-constraint second-order blind identification bridge dynamic deflection signal noise reduction method according to claim 2 is characterized in that the element with the weakest contribution is determined by comparing the absolute values of elements in each row of a separation matrix, and the element with the smallest absolute value in each row is selected as an element to be zeroed.
7. The method for noise reduction of double-constraint second-order blind identification bridge dynamic deflection signals is characterized in that when signal components containing main frequencies are screened, signal components with peak amplitudes corresponding to the main frequencies in power spectral density not lower than 5% of the main peak values are reserved.

Description

Double-constraint second-order blind identification bridge dynamic deflection signal noise reduction method Technical Field The invention relates to the field of signal noise reduction, in particular to a double-constraint second-order blind identification bridge dynamic deflection signal noise reduction method. Background Currently, dynamic deflection monitoring of most urban bridges depends on contact sensors such as accelerometers, strain gauges and the like. Although the methods can realize high-precision measurement, the method has the inherent defects of low space coverage, high layout cost, difficult maintenance and the like. For this purpose, noncontact measurement techniques such as Global Navigation Satellite System (GNSS), terrestrial Laser Scanning (TLS), and satellite-borne synthetic aperture radar (In-SAR) are increasingly used. However, the GNSS sampling rate is usually lower than 20Hz, high-frequency dynamic response is difficult to capture, TLS measurement accuracy only reaches the centimeter level and cannot meet the micro-deformation monitoring requirement, in-SAR is limited by satellite revisit period, and real-time monitoring is difficult to realize. The foundation synthetic aperture radar (GB-SAR) is used as an emerging non-contact measurement technology, has the advantages of high precision, high sampling rate, all-weather operation, integral monitoring and the like, can acquire the dynamic deformation field of the bridge in real time, effectively overcomes the defects of the traditional technology in the aspects of monitoring range, precision and environmental adaptability, and provides a new solution for bridge health monitoring. However, when GB-SAR monitors bridge dynamic deflection, the dynamic load born by the bridge is generally unknown, so that the useful components in the monitored signal lack prior information and are difficult to directly extract. Meanwhile, in urban environments, traffic flow, pedestrians, ground vibration and electromagnetic interference can introduce complex multi-scale noise, and the noise and nonlinear and non-stationary deflection signals of a bridge are seriously aliased in time domains and frequency domains, so that effective signals are very difficult to extract. Aiming at the noise reduction of nonlinear and non-stationary signals, the traditional single time domain and frequency domain noise reduction method has obvious limitations. Time domain methods rely on signal stationarity assumptions that tend to cause distortion or noise residuals of the useful signal under time-varying loads. The frequency domain method assumes that signals are separated from noise frequency bands, is difficult to process spectrum aliasing components, and is easy to damage transient useful signals or retain background noise. Although the conventional time-frequency analysis method (such as wavelet transformation, empirical mode decomposition and the like) can provide time-frequency localization analysis, the conventional time-frequency analysis method is still generally limited by the problems of mode aliasing, end-point effect, parameter selection sensitivity and the like, so that noise reduction is incomplete or signals are distorted. Blind Source Separation (BSS) technology, in particular Second Order Blind Identification (SOBI) algorithm, provides a new idea for the above-mentioned problems by utilizing the time correlation of signals to perform source separation. However, the traditional SOBI algorithm is based on the assumption that source signals are linearly independent, has poor separation effect when processing highly nonlinear bridge deflection signals, often has residual noise leakage and signal distortion, and is difficult to meet the requirement of high-precision micro-deformation monitoring. Other weak residues of source signals often exist in source signals separated by the traditional SOBI algorithm, so that cross interference exists among the signals. If the first component is directly retained for reconstruction, signal quality may be affected because interference is not effectively suppressed. Therefore, a double-constraint second-order blind identification bridge dynamic deflection signal noise reduction method is needed. Disclosure of Invention The invention aims to overcome the defects of the prior art and provides a double-constraint second-order blind identification bridge dynamic deflection signal noise reduction method based on a mixed matrix weight-dominant frequency double constraint improved SOBI algorithm, so as to realize more accurate and more robust noise reduction treatment on GB-SAR bridge monitoring signals. In order to achieve the above purpose, the invention is implemented according to the following technical scheme: the invention provides a double-constraint second-order blind identification bridge dynamic deflection signal noise reduction method, which comprises the following steps: Collecting dynamic deflection observation signa