A method for early warning and monitoring deformation of upper bridge structure during construction of pile foundation underpinning structure in tunnel

Abstract

The present invention provides a deformation early warning monitoring method for an upper bridge structure during the construction of a pile foundation underpinning structure in a tunnel, which is applicable to the field of tunnel construction monitoring; the early warning monitoring method adopts a GRU-SVM combined method, including the arrangement of monitoring points, the acquisition of monitoring data, the preliminary arrangement of monitoring data, the analysis of monitoring data and the realization of deformation early warning; the deformation early warning monitoring method for an upper bridge structure during the construction of a pile foundation underpinning structure in a tunnel proposed by the present invention has high accuracy and can be widely used in the field of tunnel construction monitoring.

Description

Technical Field

[0001] The invention relates to a method for early warning and monitoring deformation of an upper bridge structure during construction of a pile foundation underpinning structure in a tunnel, and is applicable to the field of tunnel construction monitoring. Background Art

[0002] With the continuous development of urban transportation facilities and the increasingly dense transportation network, coupled with the influence of topography and geology, line selection and design and other factors, tunnels often pass under bridge pile foundations. Due to the limited ground working surface and limited construction space in the tunnel, the method of in-tunnel arch underpinning is often used to ensure construction safety and the stability of existing structures. The traditional in-tunnel arch underpinning system has some safety hazards and technical challenges, such as the problem of pile bottom leakage in single-layer underpinning structures, and the problem of unreliable inter-layer bonding and high construction risks in multi-layer underpinning structures.

[0003] During the construction process, deformation early warning monitoring of the upper bridge structure is the key to ensuring construction safety and the stability of the existing structure. Monitoring methods usually include establishing a three-dimensional numerical simulation model to monitor the settlement changes and laws of the pile foundation, the distribution of friction resistance, etc. in real time. In addition, static monitoring methods, dynamic monitoring methods (such as measuring robots), global positioning system (GPS) monitoring methods, etc. can also be used to achieve real-time monitoring and early warning of bridge deformation. At present, with the popularization of machine learning methods, many scholars use time series data analysis methods to analyze monitoring data, and obtain prediction values ​​by analyzing the distribution trend of data to achieve the effect of early warning. However, it should be clear that the original monitoring data itself is the data observed by people or machines, which itself has certain errors, so directly using the original data to obtain the prediction value will also have certain errors. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that there are certain errors in the current analysis of original monitoring data using a time series data analysis method without performing error correction preprocessing on the original data, and directly using the original data for analysis and processing. A deformation early warning monitoring method for the upper bridge structure during the construction of a pile foundation underpinning structure in a tunnel is proposed. The method adopts a GRU-SVM combined method to obtain a predicted value and then realize deformation early warning work. It has high accuracy and can be widely used in the field of tunnel construction monitoring.

[0005] The purpose of the present invention can be achieved by adopting the following technical solutions:

[0006] S101, monitoring point arrangement;

[0007] The arrangement of the monitoring points includes arranging the monitoring points according to the situation that the tunnel is passing through the bridge pile foundation. The monitoring points are divided into: displacement monitoring points passing through the bridge pile foundation, bridge displacement monitoring points, and bridge stress monitoring points. The arrangement time of the monitoring points should be completed when the horizontal distance between the tunnel face and the bridge pile foundation is 200 meters. The number of the monitoring points is i;

[0008] S102, monitoring data acquisition;

[0009] The acquisition of the monitoring data includes carrying out monitoring work according to the monitoring point i and obtaining the monitoring data set C i {a1, a2, a3, ..., a n}, the starting point of the monitoring work is when the horizontal distance between the tunnel face and the bridge pile foundation is 50d, d is the maximum diameter of the bridge pile foundation, and the end of the monitoring work is when the monitoring data of the monitoring point meets the end of monitoring standards;

[0010] S103, preliminary collation of monitoring data;

[0011] The preliminary arrangement of the monitoring data includes interpolating missing values ​​of the monitoring data Ci to obtain a complete data set Di containing uncertainty estimation, wherein the interpolation method of the missing values ​​adopts a Bayesian method, and the model of the Bayesian method adopts a linear model;

[0012] The preliminary arrangement of the data also includes normalizing the data set Di to obtain a normalized data set E i ={x1, x2, x3, ..., x n}, the normalization method adopts the maximum and minimum method;

[0013] S104, analysis of monitoring data;

[0014] The analysis of the monitoring data includes the following steps: i Perform data detection and correction to obtain the predicted value b n+1 , the steps of data detection and correction are:

[0015] a) Model construction, building a gated recurrent unit model,

[0016] b) The gated recurrent unit model is trained using a sliding window method, where the window starts at t0 and ends at t0+(L-1), where L is the window length, and the segment {x t0 ,......,x t0+(L-1)}, with {x t0+L} is the result mark for iterative training, the time window slides backward in time order, one position at a time, and training is performed in sequence until the entire model converges to meet the preset requirements, and finally the trained gated recurrent unit model M is obtained. GRU ,

[0017] c) Obtaining the predicted value, based on the trained gated recurrent unit model M GRU , input the data set Ei = {x1, x2, x3, ..., x n}, the corresponding predicted value data set is F i ={y L ,......,y n ,y n+1}, subtract the corresponding value of the dataset Ei from the predicted value of the corresponding predicted value dataset to obtain the difference dataset Hi = {d L , ..., d n}, where d j =y j -x j , and based on the value data set Hi = {d L , ..., d n}, using expert judgment method or support vector machine splitting model M SVM The evaluation result is obtained, and the evaluation result is 0 or 1, where 0 represents non-abnormal and 1 represents abnormal. Finally, the expert evaluation result is organized into a data set Gi={(0OR1) L ,...(0OR1) n},

[0018] d) According to the data set G i ={(0OR1) L ,...(0OR1) n}, construct a new data set Mi = {b L ,......,b n}, the construction principle of the new data set Mi is: when the evaluation result is 0, the value of the data set Ei is used; when the evaluation result is 1, the predicted value data set F is used i Data, finally construct a new data set Mi = {b L ,......,b n}, the segment {b n-(L-1) ,......,b n} Input to the trained gated recurrent unit model M GRU Get the predicted value b n+1 ,

[0019] e) constructing a support vector machine classification model, and training the support vector machine classification model using training data, wherein the number of input and output neurons of the support vector machine is 1, and the input data of the training data is the interpolation data set Hi={d L’ ...d n’}, the output value corresponding to the training data is Gi={(0OR1) L ,...(0OR1) n}, and finally obtain the support vector machine classification model M after training. SVM ,

[0020] f) For the same type of monitoring points, the expert judgment method is required for the first monitoring point analysis, and the support vector machine splitting model M is used for the subsequent point analysis SVM , the expert evaluation method is no longer used to obtain the evaluation results;

[0021] S105, implementation of deformation warning;

[0022] The implementation of the deformation warning includes analyzing whether the predicted value bn+1 exceeds the warning standard according to the obtained predicted value bn+1 and the established warning standard, and issuing a warning if it exceeds the warning standard, thereby achieving the purpose of deformation warning;

[0023] Furthermore, in the above S103, the preliminary arrangement of the monitoring data also includes using the Bayesian method to interpolate missing values. When the monitoring point i is a monitoring point for displacement of a bridge pile foundation or a bridge displacement monitoring point, the linear model formula used by the Bayesian method is formula (1):

[0024] y=β0+β1*x+ε (1)

[0025] In the formula, β0 is the intercept term, β1 is the regression coefficient, and ε is the error term, which is assumed to obey a normal distribution with a mean of 0;

[0026] Furthermore, in the above S103, the preliminary arrangement of the monitoring data also includes using the Bayesian method to interpolate missing values. When the monitoring point i is a bridge stress monitoring point, the linear model formula used by the Bayesian method is formula (2):

[0027] y=α0+α1·x+α2·x 2 +η (2)

[0028] In the formula, α0 is the intercept term, α1 and α2 are regression coefficients, and η is the error term;

[0029] Furthermore, in the above S103, the preliminary arrangement of the monitoring data also includes the maximum and minimum method as a maximum and minimum method with a sliding window, and the steps of the maximum and minimum method with a sliding window are: a) determining the window size as W=10~20, b) traversing each data point and applying a sliding window to each data point, c) calculating the maximum and minimum values ​​in each window data, and d) using the minimum and maximum values ​​in the window to normalize the current data point.

[0030] The beneficial effects of the present invention are:

[0031] The beneficial effects of the present invention are: solving the problem of certain errors in analyzing monitoring data using a time series data analysis method without analyzing and processing the original data, and proposing a deformation early warning monitoring method for the upper bridge structure during the construction of a pile foundation underpinning structure in a tunnel. The method adopts a GRU-SVM combined method to obtain a prediction value and then realize deformation early warning work, has high accuracy, and can be widely used in the field of tunnel construction monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 : A flow chart of a method for early warning and monitoring deformation of an upper bridge structure during construction of a pile foundation underpinning structure in a tunnel according to the present invention. DETAILED DESCRIPTION

[0033] The specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings; it should be understood that the specific embodiments given here are only used to illustrate and explain the present invention and cannot be used to limit the present invention.

[0034] The following is a specific embodiment of a method for early warning and monitoring deformation of an upper bridge structure during construction of a pile foundation underpinning structure in a tunnel.

[0035] like Figure 1 As shown, it is a flow chart of a method for early warning and monitoring deformation of an upper bridge structure during construction of a pile foundation underpinning structure in a tunnel according to the present invention.

[0036] S101, monitoring point arrangement;

[0037] The arrangement of the monitoring points includes arranging the monitoring points according to the situation that the tunnel is passing through the bridge pile foundation. The monitoring points are divided into: displacement monitoring points passing through the bridge pile foundation, bridge displacement monitoring points, and bridge stress monitoring points. The arrangement time of the monitoring points should be completed when the horizontal distance between the tunnel face and the bridge pile foundation is 200 meters. The number of the monitoring points is i;

[0038] S102, monitoring data acquisition;

[0039] The acquisition of the monitoring data includes carrying out monitoring work according to the monitoring point i and obtaining the monitoring data set C i {a1, a2, a3, ..., a n}, the starting point of the monitoring work is when the horizontal distance between the tunnel face and the bridge pile foundation is 50d, d is the maximum diameter of the bridge pile foundation, and the end of the monitoring work is when the monitoring data of the monitoring point meets the end of monitoring standards;

[0040] S103, preliminary collation of monitoring data;

[0041] The preliminary arrangement of the monitoring data includes interpolating missing values ​​of the monitoring data Ci to obtain a complete data set Di containing uncertainty estimation, wherein the interpolation method of the missing values ​​adopts a Bayesian method, and the model of the Bayesian method adopts a linear model;

[0042] The preliminary arrangement of the data also includes normalizing the data set Di to obtain a normalized data set E i ={x1, x2, x3, ..., x n}, the normalization method adopts the maximum and minimum method;

[0043] Furthermore, in the above S103, the preliminary arrangement of the monitoring data also includes using the Bayesian method to interpolate missing values. When the monitoring point i is a monitoring point for displacement of a bridge pile foundation or a bridge displacement monitoring point, the linear model formula used by the Bayesian method is formula (1):

[0044] y=β0+β1*x+ε (1)

[0045] In the formula, β0 is the intercept term, β1 is the regression coefficient, and ε is the error term, which is assumed to obey a normal distribution with a mean of 0;

[0046] Furthermore, in the above S103, the preliminary arrangement of the monitoring data also includes using the Bayesian method to interpolate missing values. When the monitoring point i is a bridge stress monitoring point, the linear model formula used by the Bayesian method is formula (2):

[0047] y=α0+α1·x+α2·x 2 +η (2)

[0048] In the formula, α0 is the intercept term, α1 and α2 are regression coefficients, and η is the error term;

[0049] Furthermore, in the above S103, the preliminary arrangement of the monitoring data also includes the maximum and minimum method as a maximum and minimum method with a sliding window, and the steps of the maximum and minimum method with a sliding window are: a) determining the window size as W=10~20, b) traversing each data point and applying a sliding window to each data point, c) calculating the maximum and minimum values ​​in each window data, and d) using the minimum and maximum values ​​in the window to normalize the current data point.

[0050] S104, analysis of monitoring data;

[0051] The analysis of the monitoring data includes the following steps: i Perform data detection and correction to obtain the predicted value b n+1 , the steps of data detection and correction are:

[0052] a) Model construction, building a gated recurrent unit model,

[0053] b) The gated recurrent unit model is trained using a sliding window method, where the window starts at t0 and ends at t0+(L-1), where L is the window length, and the segment {x t0 ,......,x t0+(L-1)}, with {x t0+L} is the result mark for iterative training, the time window slides backward in time order, one position at a time, and training is performed in sequence until the entire model converges to meet the preset requirements, and finally the trained gated recurrent unit model M is obtained. GRU ,

[0054] c) Obtaining the predicted value, based on the trained gated recurrent unit model M GRU , input the data set Ei = {x1, x2, x3, ..., x n}, the corresponding predicted value data set is F i ={y L ,......,y n ,y n+1}, subtract the corresponding value of the dataset Ei from the predicted value of the corresponding predicted value dataset to obtain the difference dataset Hi = {d L， ..., d n}, where d j =y j -x j , and based on the value data set Hi = {d L , ..., d n}, using expert judgment method or support vector machine splitting model M SVMThe evaluation result is obtained, and the evaluation result is 0 or 1, where 0 represents non-abnormal and 1 represents abnormal. Finally, the expert evaluation result is organized into a data set Gi={(0OR1) L ,...(0OR1) n},

[0055] d) According to the data set G i ={(0OR1) L ,...(0OR1) n}, construct a new data set Mi = {b L ,......,b n}, the construction principle of the new data set Mi is: when the evaluation result is 0, the value of the data set Ei is used; when the evaluation result is 1, the predicted value data set F is used i Data, finally construct a new data set Mi = {b L ,......,b n}, the segment {b n-(L-1) ,......,b n} Input to the trained gated recurrent unit model M GRU Get the predicted value b n+1 ,

[0056] e) constructing a support vector machine classification model, and training the support vector machine classification model using training data, wherein the number of input and output neurons of the support vector machine is 1, and the input data of the training data is the interpolation data set Hi={d L’ ...d n’}, the output value corresponding to the training data is Gi={(0OR1) L ,...(0OR1) n}, and finally obtain the support vector machine classification model M after training. SVM ,

[0057] f) For the same type of monitoring points, the expert judgment method is required for the first monitoring point analysis, and the support vector machine splitting model M is used for the subsequent point analysis SVM , the expert evaluation method is no longer used to obtain the evaluation results;

[0058] S105, implementation of deformation warning;

[0059] The implementation of the deformation warning includes analyzing whether the predicted value bn+1 exceeds the warning standard according to the obtained predicted value bn+1 and the established warning standard, and issuing a warning if it exceeds the warning standard, thereby achieving the purpose of deformation warning.

[0060] In the above embodiment, the present invention discloses a deformation early warning monitoring method for the upper bridge structure during the construction of the pile foundation underpinning structure in the tunnel, including the arrangement of monitoring points, the acquisition of monitoring data, the preliminary organization of monitoring data, the analysis of monitoring data and the realization of deformation early warning; the deformation early warning monitoring method for the upper bridge structure during the construction of the pile foundation underpinning structure in the tunnel proposed by the present invention has high accuracy and can be widely used in the field of tunnel construction monitoring.

[0061] The above description is a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for early warning and monitoring deformation of upper bridge structure during construction of pile foundation underpinning structure in a tunnel, characterized in that: The method comprises the following steps: 1) Layout of monitoring points; 2) Acquisition of monitoring data; 3) Preliminary collation of monitoring data; 4) Analysis of monitoring data; 5) Implementation of deformation warning; The arrangement of the monitoring points includes arranging the monitoring points according to the situation that the tunnel is passing through the bridge pile foundation. The monitoring points are divided into: displacement monitoring points passing through the bridge pile foundation, bridge displacement monitoring points, and bridge stress monitoring points. The arrangement time of the monitoring points should be completed when the horizontal distance between the tunnel face and the bridge pile foundation is 200 meters. The number of the monitoring points is i; The acquisition of the monitoring data includes carrying out monitoring work according to the monitoring point i and obtaining the monitoring data set C i {a1, a2, a3, ..., a n }, the starting point of the monitoring work is when the horizontal distance between the tunnel face and the bridge pile foundation is 50d, d is the maximum pile diameter of the bridge pile foundation, and the end of the monitoring work is when the monitoring data of the monitoring point meets the end of monitoring standards; The preliminary arrangement of the monitoring data includes interpolating missing values ​​of the monitoring data Ci to obtain a complete data set Di containing uncertainty estimation, wherein the interpolation method of the missing values ​​adopts a Bayesian method, and the model of the Bayesian method adopts a linear model; The preliminary arrangement of the data also includes normalizing the data set Di to obtain a normalized data set E i ={x1, x2, x3, ..., x n }, the normalization method adopts the maximum and minimum method; The analysis of the monitoring data includes the following steps: i Perform data detection and correction to obtain the predicted value b n+1 , the steps of data detection and correction are: a) Model construction, building a gated recurrent unit model, b) The gated recurrent unit model is trained using a sliding window method, where the window starts at t0 and ends at t0+(L-1), where L is the window length, and the segment {x t0 ,......,x t0+（L-1） }, with {x t0+L } is the result mark for iterative training, the time window slides backward in time order, one position at a time, and training is performed in sequence until the entire model converges to meet the preset requirements, and finally the trained gated recurrent unit model M is obtained. GRU , c) Obtaining the predicted value, based on the trained gated recurrent unit model M GRU , input the data set Ei={x1, x2, x3, ..., x n }, the corresponding predicted value data set is F i ={y L ,......,y n ,y n+1 }, subtract the corresponding value of the dataset Ei from the predicted value of the corresponding predicted value dataset to obtain the difference dataset Hi={d L， ..., d n }, where d j =y j -x j , and based on the difference data set Hi={d L， ..., d n }, using expert judgment method or support vector machine classification model M SVM The evaluation result is obtained, and the evaluation result is 0 or 1, where 0 represents non-abnormal and 1 represents abnormal. Finally, the results are organized into a data set Gi={(0OR1) L ,...(0OR1) n }, d) According to the dataset G i ={(0OR1) L ,...(0OR1) n }, construct a new data set Mi={b L ,......,b n }, the construction principle of the new data set Mi is: when the evaluation result is 0, the value of the data set Ei is used; when the evaluation result is 1, the predicted value data set F is used i Data, finally construct a new data set Mi={b L ,......,b n }, the segment {b n-(L-1) ,......,b n } Input to the trained gated recurrent unit model M GRU Get the predicted value b n+1 , e) constructing a support vector machine classification model, and training the support vector machine classification model using training data, wherein the number of input and output neurons of the support vector machine is 1, and the input data of the training data is the difference data set Hi={d L’ ...d n’ }, the output value corresponding to the training data is Gi={(0OR1) L ,...(0OR1) n }, and finally obtain the support vector machine classification model M after training. SVM , f) For the same type of monitoring points, the expert judgment method is required for the first monitoring point analysis, and the support vector machine classification model M is used for the subsequent point analysis SVM , the expert evaluation method is no longer used to obtain the evaluation results; The implementation of the deformation warning includes: n+1 And the established early warning standards, analyze the predicted value b n+1 Whether it exceeds the warning standard, if it exceeds it, a warning will be issued to achieve the purpose of deformation warning.

2. The method for early warning and monitoring deformation of the upper bridge structure during the construction of the pile foundation underpinning structure in a tunnel according to claim 1 is characterized in that: The preliminary arrangement of the monitoring data also includes the following steps: when the monitoring point i is a monitoring point for displacement of a bridge pile foundation or a bridge displacement monitoring point, the linear model formula used by the Bayesian method is (1): （1） In the formula, β0 is the intercept term, β1 is the regression coefficient, and ε is the error term, which is assumed to obey a normal distribution with a mean of 0.

3. The method for early warning and monitoring deformation of the upper bridge structure during the construction of the pile foundation underpinning structure in a tunnel according to claim 1 is characterized in that: The preliminary arrangement of the monitoring data also includes using the Bayesian method to interpolate missing values. When the monitoring point i is a bridge stress monitoring point, the linear model formula used by the Bayesian method is formula (2): （2） Where α0 is the intercept term, α1 and α2 are regression coefficients, and η is the error term.

4. The method for early warning and monitoring deformation of the upper bridge structure during the construction of the pile foundation underpinning structure in a tunnel according to claim 1 is characterized in that: The preliminary arrangement of the data also includes the maximum and minimum method using a maximum and minimum method with a sliding window, the steps of the maximum and minimum method with a sliding window are: a) determining the window size as W=10~20, b) traversing each data point and applying a sliding window to each data point, c) calculating the maximum and minimum values ​​in each window data, and d) using the minimum and maximum values ​​in the window to normalize the current data point.

Images