CN-121723272-B - Long-term monitoring method and system for building settlement

CN121723272BCN 121723272 BCN121723272 BCN 121723272BCN-121723272-B

Abstract

The invention relates to the technical field of building safety monitoring, and discloses a long-term building settlement monitoring method and a monitoring system. The method comprises the steps of collecting historical settlement data of all buildings in a building community and resampling the historical settlement data into a standard time sequence, extracting local trend characteristics of all the sequences, constructing a characteristic similarity network which takes the buildings as nodes and the trend matching degree as the side weight, automatically dividing building clusters with similar settlement behaviors through community discovery, identifying settlement mutation points shared in all communities, fusing and screening to obtain global significance time nodes, training a time sequence prediction model by using settlement amount data corresponding to the nodes, and performing settlement prediction on a target building. The method identifies the commonality sedimentation rule through group behavior analysis, overcomes the defect that single-point analysis is easy to be interfered by noise, and improves the accuracy and the robustness of long-term prediction.

Inventors

ZHU LICHAO
XIANG TAO
SU XU
WANG HUIDONG
ZHANG XUPENG
HE MINGGANG
SUN CHAO
ZHANG LIANG
CAI QIANG
MA WEILI

Assignees

山东省水利勘测设计院有限公司

Dates

Publication Date: 20260508
Application Date: 20260225

Claims (8)

1. A method for long-term monitoring of settlement of a building, the method comprising the steps of: Collecting a historical sedimentation data set of each building in a building community, wherein the historical sedimentation data set consists of a time stamp and a corresponding sedimentation value; resampling the historical sedimentation data set of each building at equal intervals to generate a standard time sequence, wherein a time axis of the standard time sequence is provided with uniformly distributed time nodes; Calculating local trend characteristics of the standard time sequence of each building, and extracting sedimentation change rates near each time node by the local trend characteristics in a sliding window mode; constructing a feature similarity network among the buildings based on the local trend features of all the buildings, wherein nodes of the feature similarity network represent the buildings, and the side weights represent the trend feature matching degree; Community discovery is carried out according to the characteristic similarity network, and a building is divided into a plurality of communities with similar sedimentation behaviors; Analyzing a standard time sequence of a building in the community aiming at each community, and identifying a settlement mutation point set shared by the communities; fusing the settlement mutation point sets of all communities, and screening out global significance time nodes; training a time sequence prediction model by utilizing sedimentation values corresponding to the global significance time nodes; inputting settlement data of the building to be monitored into a time sequence prediction model to obtain a settlement amount prediction result; The identifying a set of sedimentation mutation points shared by communities comprises: calculating a first-order differential sequence for a standard time sequence of each building in the community; detecting extreme points in each first-order differential sequence, and taking a time node corresponding to the extreme points as a candidate mutation point; counting the occurrence frequency of all buildings in the community on each candidate mutation point, and incorporating the candidate mutation points with the frequency exceeding a threshold value into a settlement mutation point set shared by the community; The screening out the global significance time node comprises the following steps: calculating the weight of a time node in a settlement mutation point set of each community, wherein the weight is normalized based on the number of buildings in the community; combining the settlement mutation point sets of all communities, and carrying out weight superposition on repeated time nodes; The first k time nodes with the highest weight sum are selected as global significance time nodes, wherein k is a preset integer.
2. A method of long-term monitoring of settlement of buildings according to claim 1, wherein said equally-spaced resampling of historical settlement data sets for each building to generate a standard time series comprises: setting a resampling time interval, and aligning the time stamp in the historical sedimentation data set according to the resampling time interval; For each time node, calculating the missing sedimentation quantity value by adopting a linear interpolation method, so that the sedimentation quantity value of each building is defined on the same time node; and (3) carrying out standardization treatment on the interpolated sedimentation value, removing dimension influence, and generating a standard time sequence.
3. A method of long-term monitoring of building settlement according to claim 1, wherein said calculating local trend features of the standard time series for each building comprises: Defining the size of a sliding window, and intercepting a subsequence of sedimentation values in the window by taking each time node as the center; performing linear fitting on each subsequence to obtain the slope of a fitting straight line as the local change rate of the time node; and arranging the local change rates of all the time nodes in time sequence to form a local trend characteristic sequence.
4. A method of long term monitoring of building settlement according to claim 1, wherein said constructing a network of feature similarities between buildings comprises: calculating a dynamic time warping distance between the local trend feature sequences of each pair of buildings; converting the dynamic time warping distance into similarity weight, wherein the similarity weight and the distance value are in negative correlation; And constructing a fully-connected feature similarity network by taking the building as a node and the similarity weight as an edge value.
5. The method of claim 1, wherein the community discovery based on a feature similarity network comprises: Dividing the feature similarity network by using a modularity optimization algorithm to maximize the edge weight in the community; And iteratively adjusting community division until the modularity index converges to obtain a stable building cluster division result.
6. The method for long-term monitoring of building settlement according to claim 1, wherein the training time series prediction model comprises: organizing the sedimentation magnitude of each building on a global saliency time node as a feature vector; taking the characteristic vector as input and the final settlement of the building as output to construct a deep neural network structure; parameters of the deep neural network are optimized using a back propagation algorithm until the loss function stabilizes.
7. The method for long-term monitoring of settlement of a building according to claim 1, wherein the obtaining of the settlement amount prediction result comprises: extracting the sedimentation value of a building to be monitored on a global significance time node to form an input feature vector; Inputting the input feature vector into a trained time sequence prediction model; and outputting the model to obtain a settlement amount prediction result of the building to be monitored.
8. A long-term monitoring system of building settlement, the system comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method of any one of claims 1 to 7 when the computer program is executed.

Description

Long-term monitoring method and system for building settlement Technical Field The invention relates to the technical field of building safety monitoring, in particular to a long-term monitoring method and a long-term monitoring system for building settlement. Background In the long-term monitoring of settlement of building communities, conventional techniques rely primarily on independent analysis of historical settlement data of monomer buildings. The method directly models and predicts a single time sequence, ignores the building community as a whole, and has inherent correlation between member settlement behaviors. Building settlement is the result of its structural characteristics in combination with external environmental factors. The limitations of isolated analysis are particularly pronounced when the monitoring range extends from monomer to population. Static grouping is simply carried out according to geographic adjacency or design drawings, and the internal rule of the dynamic sedimentation process is difficult to accurately reflect. Geographically adjacent buildings may produce distinct settlement responses to the same external factors due to differences in foundation form, loading, or subsurface conditions. The coarse granularity grouping based on the static attribute causes mismatching of community dividing results and real sedimentation dynamics behaviors, and cannot provide effective support for regional sedimentation mechanism analysis. Identifying key event points from the sedimentation data is an important pre-step in constructing the predictive model. The prior art generally performs mutation point detection independently over the time sequence of a single building. This approach is highly susceptible to strong interference by building individual specific factors. Abnormal fluctuations in the data formed by these individual events can be misinterpreted as meaningful sedimentation mutations. The resulting set of mutations is doped with a large amount of non-universal, sporadic local information. The model is difficult to focus on the reality and global sedimentation modes which affect the whole area and are caused by common driving factors, and the accuracy and generalization capability of the model in long-term prediction are severely restricted. Disclosure of Invention The invention aims to provide a long-term monitoring method and a long-term monitoring system for building settlement, which are used for solving the problems in the background technology. To achieve the above object, the present invention provides a method for long-term monitoring of settlement of a building, the method comprising: Collecting a historical sedimentation data set of each building in a building community, wherein the historical sedimentation data set consists of a time stamp and a corresponding sedimentation value; resampling the historical sedimentation data set of each building at equal intervals to generate a standard time sequence, wherein a time axis of the standard time sequence is provided with uniformly distributed time nodes; Calculating local trend characteristics of the standard time sequence of each building, and extracting sedimentation change rates near each time node by the local trend characteristics in a sliding window mode; constructing a feature similarity network among the buildings based on the local trend features of all the buildings, wherein nodes of the feature similarity network represent the buildings, and the side weights represent the trend feature matching degree; Community discovery is carried out according to the characteristic similarity network, and a building is divided into a plurality of communities with similar sedimentation behaviors; Analyzing a standard time sequence of a building in the community aiming at each community, and identifying a settlement mutation point set shared by the communities; fusing the settlement mutation point sets of all communities, and screening out global significance time nodes; training a time sequence prediction model by utilizing sedimentation values corresponding to the global significance time nodes; and inputting the settlement data of the building to be monitored into a time sequence prediction model to obtain a settlement amount prediction result. Preferably, the resampling the historical settlement data set of each building at equal intervals to generate a standard time sequence includes: setting a resampling time interval, and aligning the time stamp in the historical sedimentation data set according to the resampling time interval; For each time node, calculating the missing sedimentation quantity value by adopting a linear interpolation method, so that the sedimentation quantity value of each building is defined on the same time node; and (3) carrying out standardization treatment on the interpolated sedimentation value, removing dimension influence, and generating a standard time sequence. Preferably, the calculating the local