CN-121980138-A - Bridge monitoring data noise reduction method and system

CN121980138ACN 121980138 ACN121980138 ACN 121980138ACN-121980138-A

Abstract

The invention discloses a method and a system for noise reduction of bridge monitoring data, which relate to the technical field of bridge data monitoring and comprise the following steps of collecting the monitoring data to be noise reduced of a target bridge through a structural health monitoring system; and inputting the monitoring data to be noise reduced into a pre-trained noise reduction model, and carrying out noise reduction on the monitoring data to be noise reduced. The noise reduction model provided by the invention has triple capacities of local detail perception, global structure understanding and time sequence dependent modeling, and the robustness under a high noise environment is obviously improved. The first self-attention module acts on the local feature map to strengthen the attention to local dependence, and the second self-attention module acts on the high-level feature map to refine global semantic information. So that the model can focus on both local and global features. BiLSTM learn the long-term dependence of the sequence data from the forward direction and the backward direction, and realize end-to-end mapping from the noise-containing signal to the noise-reducing signal.

Inventors

CAO YANMEI
ZHANG YUXIAN
ZHAO QINGHAO
WU JING

Assignees

北京交通大学

Dates

Publication Date: 20260505
Application Date: 20260126

Claims (5)

1. The noise reduction method for the bridge monitoring data is characterized by comprising the following steps of: collecting monitoring data to be noise reduced of a target bridge through a structural health monitoring system; Inputting the monitoring data to be noise reduced into a pre-trained noise reduction model, and noise reducing the monitoring data to be noise reduced, wherein the noise reduction model comprises a first convolution layer, a first self-attention module, a first average pooling layer, a second convolution layer, a second self-attention module, a second average pooling layer, biLSTM layers and a full-connection layer; the method comprises the steps of carrying out local feature extraction on monitored data to be denoised through a first convolution layer to obtain a local feature map, distributing different attention weights to different time points in the local feature map through a first self-attention module, carrying out downsampling on the local feature map after attention weight distribution through a first average pooling layer, combining the downsampled local feature map to obtain a high-level feature map through a second convolution layer, distributing different attention weights to different time points in the high-level feature map through a second self-attention module, carrying out downsampling on the high-level feature map after attention weight distribution through a second average pooling layer to obtain a feature sequence, carrying out contextual information of a BiLSTM-layer learning feature sequence, and mapping the contextual information through a full-connection layer to obtain the monitored data after denoised.
2. A method of noise reduction of bridge monitoring data according to claim 1, wherein the pre-training of the noise reduction model comprises the steps of: collecting real acceleration monitoring data of a target bridge, and injecting Gaussian white noise with different degrees into the real acceleration monitoring data to obtain a plurality of groups of noise-containing signals; And pre-training the noise reduction model through the real acceleration monitoring data and the corresponding noise-containing signals.
3. The method for noise reduction of bridge monitoring data according to claim 1, wherein the noise-containing signal is specifically as follows: ; Wherein Signal noisy is a noisy Signal, signal clean is real acceleration monitoring data, noise is a standard normal distribution random vector with zero mean and unit standard deviation, var is variance Ep is the noise level.
4. The method for noise reduction of bridge monitoring data according to claim 1, wherein the monitoring data is acceleration data.
5. A bridge monitor data noise reduction system, comprising: the acquisition module is used for acquiring the monitoring data to be noise reduced of the target bridge through the structural health monitoring system; The system comprises an input module, a pre-training noise reduction module and a noise reduction module, wherein the input module is used for inputting the monitoring data to be noise reduced to the pre-training noise reduction module, and the noise reduction module is used for noise reduction of the monitoring data to be noise reduced, and comprises a first convolution layer, a first self-attention module, a first average pooling layer, a second convolution layer, a second self-attention module, a second average pooling layer, biLSTM layers and a full-connection layer; The processing module is used for carrying out local feature extraction on the monitoring data to be noise reduced through the first convolution layer to obtain a local feature map, the first self-attention module is used for distributing different attention weights to different time points in the local feature map, the first averaging pooling layer is used for carrying out downsampling on the local feature map after attention weight distribution, the second convolution layer is used for combining the local feature map after downsampling to obtain an advanced feature map, the second self-attention module is used for distributing different attention weights to different time points in the advanced feature map, the second averaging pooling layer is used for carrying out downsampling on the advanced feature map after attention weight distribution to obtain a feature sequence, the BiLSTM layers of context information of learning feature sequences are used for mapping the context information, and the monitoring data after noise reduction is obtained.

Description

Bridge monitoring data noise reduction method and system Technical Field The invention relates to the technical field of bridge data monitoring, in particular to a method and a system for noise reduction of bridge monitoring data. Background The safety and reliability of the bridge structure directly determine the smoothness of urban traffic and the safety of resident trip. The SHM system is used for monitoring the structure, the state, the performance and the like of the bridge in real time through a plurality of sensor arrays by applying various sensing technologies and combining the knowledge of the related fields such as building structures, civil engineering, computer science, communication technology and the like, collecting and analyzing the related data of the bridge, analyzing and predicting the health condition, and providing data support and technical support for the maintenance and operation management of the bridge. More and more SHM systems are installed on various structural facilities such as dams, large bridges, high-rise buildings and the like, the dynamic response of the structure is measured permanently or periodically, the running condition of the structure is known in real time, and the evaluation of the structural state and the structural damage is carried out according to the abnormal state early warning of the SHM systems. The field test conditions of civil engineering structures are generally difficult to control, and the dynamic response of the structures measured by the SHM system is vulnerable to strong noise contamination from various sources, such as environmental noise, measurement noise, instrument noise, and the like. In this case, strong noise may mask variations in the structural response caused by minor damage, resulting in an inability to effectively and accurately identify structural modes or detect damage. Therefore, it is necessary to perform signal noise reduction to improve the quality and usability of the measurement signal. Many scholars have conducted related studies on the noise problem of signals and proposed some signal noise reduction methods. In existing studies, noise that is aliased into the vibration signal by 10% -20% is often considered a high noise level. In fact, under some poor test conditions, measurement noise can easily exceed 20% noise levels. According to the difference of signal identification domains, the conventional signal noise reduction method can be mainly divided into a time domain method, a frequency domain method and a time-frequency domain method. Most of the traditional signal noise reduction methods depend on priori knowledge of signals or noise and artificial selection of parameters in the noise reduction process, and certain difficulties exist in the process of monitoring signal processing in which monitoring data of signal noise distribution cannot be obtained in the SHM field and monitoring signal processing in which noise frequency domain distribution is the same. In recent years, with the development of artificial intelligence, deep learning (DEEP LEARNING, DL) has been widely studied in the field of signal processing by virtue of its strong feature extraction and data processing capabilities. In the field of SHM, the DL technology is mainly applied to the aspects of structural damage recognition and the like, the application research in the problem of SHM vibration data noise reduction is few, and the existing research is mainly focused on realizing noise reduction treatment of various signals based on CNN. Although CNN is superior to traditional noise reduction methods in terms of processing noise interference and extracting weak signal features, the local receptive field of the CNN model convolution kernel is good at capturing the local features of the signal, but it is difficult to effectively model the long-range dependency relationship existing in the signal, ignoring the dependency features and the data global features between the local information, and causing the performance of the model to be reduced. Disclosure of Invention Based on the defects in the prior art, the invention provides a method and a system for noise reduction of bridge monitoring data, and solves the existing problems. The invention adopts the following technical scheme: in a first aspect, the present invention provides a method for noise reduction of bridge monitoring data, comprising the following steps: collecting monitoring data to be noise reduced of a target bridge through a structural health monitoring system; Inputting the monitoring data to be noise reduced into a pre-trained noise reduction model, and noise reducing the monitoring data to be noise reduced, wherein the noise reduction model comprises a first convolution layer, a first self-attention module, a first average pooling layer, a second convolution layer, a second self-attention module, a second average pooling layer, biLSTM layers and a full-connection layer; the method comprise