A steel cofferdam monitoring method based on multi-source data fusion and a medium

Abstract

The application provides a steel cofferdam monitoring method based on multi-source data fusion, comprising: formulating a steel cofferdam monitoring scheme, determining each monitoring item, corresponding monitoring sensors and corresponding layout points; collecting monitoring data according to corresponding monitoring frequencies and storing the monitoring data in time sequence; for monitoring item data of a certain component of a steel cofferdam, extracting all monitoring item data of the region, and performing data preprocessing; performing correlation analysis on the monitoring data, and dividing main dependent variables and characteristic variables; performing regression analysis on the divided main dependent variables and characteristic variables by using an elastic network regression algorithm, establishing a linear regression equation between the main dependent variables and the characteristic variables, and comprehensively evaluating the safety state of the steel cofferdam through the main dependent variables and the characteristic variables. The method can reasonably evaluate the safety state of the steel cofferdam and has good reliability and stability. The application also provides a computer readable medium.

Description

Technical Field

[0001] This invention relates to the field of steel cofferdam monitoring technology, and in particular to a steel cofferdam monitoring method and medium based on multi-source data fusion. Background Technology

[0002] During bridge foundation construction, environmental factors often necessitate the construction of temporary pit retaining structures to provide a safe and stable construction platform. Steel cofferdams, due to their diverse forms, flexible structures, simple construction, easy support, and mature technology, are widely used in bridge foundation construction. Common types of steel cofferdams include sheet pile cofferdams, pipe pile cofferdams, steel casing cofferdams, and caisson cofferdams. As temporary retaining structures, steel cofferdams often experience significant internal forces and deformations at support nodes and waler midpoints due to uneven soil (water) pressure inside and outside the pit, vibration and impact from nearby construction machinery, and groundwater. Therefore, after the steel cofferdam is constructed, monitoring devices must be installed at the pile tops, supports, and walers to monitor the corresponding internal forces and displacements, preventing construction safety accidents caused by sudden instability of the cofferdam structure.

[0003] Currently, steel cofferdam monitoring schemes include measurements such as vertical displacement of pile tops, horizontal displacement of pile tops, deep horizontal displacement of piles, axial force of supports, and groundwater levels inside and outside the foundation pit. For land-based cofferdams, ground settlement around the foundation pit must also be included. These monitoring items involve various instruments, and the data acquisition frequencies and types differ. Judging the safety status of a steel cofferdam structure solely from data from a single instrument is incomplete. Therefore, establishing an effective multi-source heterogeneous data fusion system and fully leveraging the value of monitoring data are crucial for the comprehensive evaluation of the safety status of steel cofferdams and subsequent analytical decision-making. Summary of the Invention

[0004] The purpose of this invention is to provide a scheme for monitoring steel cofferdams based on multi-source data fusion, addressing the shortcomings of the aforementioned background technology.

[0005] To achieve the above objectives, this invention provides a method for monitoring steel cofferdams based on multi-source data fusion, comprising the following steps:

[0006] S1. Develop a monitoring plan for the steel cofferdam, and determine each monitoring item, the corresponding monitoring sensor, and the corresponding deployment point;

[0007] S2, collect monitoring data according to the corresponding monitoring frequency and store it in time series;

[0008] S3: For monitoring data of a certain area of ​​a certain component of a steel cofferdam, extract all monitoring data of that area and perform data preprocessing.

[0009] S4. Conduct correlation analysis on the monitoring data and classify the main dependent variables and characteristic variables;

[0010] S5. For the divided main dependent variables and characteristic variables, the elastic network regression algorithm is used to perform regression analysis to establish a linear regression equation between the main dependent variables and characteristic variables. The safety status of the steel cofferdam is evaluated by comprehensively evaluating the main dependent variables and characteristic variables.

[0011] Furthermore, the preprocessing steps in S2 include outlier removal, missing value imputation, and time synchronization, ultimately yielding preprocessed data:

[0012]

[0013] Furthermore, for outlier removal, the 3σ principle is used to determine the outlier nature of the data, with the specific formula as follows:

[0014]

[0015] Where, x ij This represents the data for the j-th monitoring item on day i. p=1 indicates that x ij The monitoring data is abnormal.

[0016] For missing values, the average value of the monitored items is used instead.

[0017] For time synchronization, cubic spline interpolation is used.

[0018] Furthermore, the monitoring data in S3 is represented as X = (X1, X2, ..., X...). j ), where j represents the number of monitored items.

[0019] Furthermore, S4 includes the following sub-steps:

[0020] S41, perform partial correlation analysis on the pairwise data of the monitoring items;

[0021] S42, using the partial correlation coefficient matrix R as the evaluation matrix, calculate the information entropy weights of each variable;

[0022] S43, based on the information entropy weight of variables, divides the main dependent variable and characteristic variables.

[0023] Furthermore, the nth-order partial correlation coefficient in S41 is recursively calculated from the (n-1)th-order partial correlation coefficient, and its calculation formula is as follows:

[0024]

[0025] In the formula, This indicates that in controlling x3 to x j Under the indicator variable, the partial correlation coefficient between variables x1 and x2; This indicates that in controlling x3 to xj-1 Under the indicator variable, the partial correlation coefficient between variables x1 and x2; This indicates that in controlling x3 to x j-1 Under the indicator variable, variables x1 and x j The partial correlation coefficient; This indicates that, under the control of the index variable, variables x2 and x... j The partial correlation coefficient;

[0026] After calculating the partial correlation coefficient between the two variables, a t-test is performed. The formula for the t-test is as follows:

[0027]

[0028] In the formula, r is the corresponding partial correlation coefficient, n is the data sample size, and k is the number of control variables. When t > t 0.05 When (nk-2), the partial correlation coefficient passes the test;

[0029] Calculate the partial correlation coefficients between each variable individually to obtain the partial correlation coefficient matrix R:

[0030]

[0031] In S42, the partial correlation coefficient matrix is ​​normalized:

[0032]

[0033] In the formula, z ij R represents the normalized data values. ij This represents the partial correlation coefficient;

[0034] Calculate the information entropy and information utility value of variables:

[0035]

[0036] In the formula, e j Let d represent the information entropy of the j-th variable. j =1-e j d j This represents the utility value of the j-th piece of information;

[0037] Calculate the information weights of variables:

[0038]

[0039] In S43, the information entropy weights of each variable are arranged from largest to smallest. The variable with the largest information entropy weight is regarded as the main dependent variable, and the remaining variables are regarded as characteristic variables.

[0040] Furthermore, S5 specifically includes the following sub-steps:

[0041] S51, data standardization;

[0042] S52, establish the loss function using the elastic network regression algorithm;

[0043] S53, Solving the weight coefficient matrix w based on the coordinate descent method;

[0044] S54, optimized parameters;

[0045] S55 establishes a regression model based on optimal parameters to achieve fusion analysis of the main dependent variable and characteristic variables.

[0046] Furthermore, in S51, deviation standardization is used to process the analyzed data, ensuring that all variable data fall within the range of 0 to 1. The calculation formula is as follows:

[0047]

[0048] In the formula: z ij This represents the standardized data; x ij This represents the raw analysis data;

[0049] In S52, the size of the loss term is controlled by two parameters, λ and ρ, and the specific function form is as follows:

[0050]

[0051] In the formula, w represents the weight matrix, λ represents the first regularization term loss coefficient, and ρ represents the second regularization term loss coefficient; λ and ρ are pre-input into the network as fixed values;

[0052] In S53, input parameters λ and ρ, initialize the weight coefficient matrix w as a zero vector matrix, treat one weight coefficient as a variable, and fix the values ​​of the other weight coefficients to find the optimal solution of the loss function under the current conditions; iterate through all weight coefficients and repeat the above calculation process to obtain the optimal weight coefficient matrix w under the current values ​​of λ and ρ; the process of solving w is expressed as follows:

[0053]

[0054] in, This represents the weight coefficient in the k-th iteration;

[0055] In S54, the mean squared error (MSE) of the regression equation is used as the objective, and the parameters λ and ρ are optimized. The formula for calculating MSE is:

[0056]

[0057] Among them, Represents the regression value, y iLet n represent the true value and n be the sample size; λ+ρ=1, with λ as the baseline value, the initial value is 0, the step size is d, and the final value is 1. Change the input values ​​of λ and ρ, and iteratively solve for the input values ​​of λ and ρ that correspond to the minimum MSE.

[0058] In S55, a regression model is established based on the solved optimal λ and ρ values ​​to achieve the fusion analysis of the main dependent variable and characteristic variables.

[0059] The present invention also provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the steel cofferdam monitoring method based on multi-source data fusion as described above.

[0060] The above-described solution of the present invention has the following beneficial effects:

[0061] The steel cofferdam monitoring method and medium based on multi-source data fusion provided by this invention overcomes the influence of multiple linear correlations between indicator variables by performing correlation analysis on the data through partial correlation analysis. At the same time, it divides the main variables and characteristic variables by calculating the information entropy weights of the indicator variables based on the partial correlation coefficient matrix as the evaluation matrix, which can effectively highlight the dispersion and significance of the main variables and characteristic variables. Compared with other correlation analysis methods, it has objectivity, reliability and universality. The regression model is established by using elastic network regression analysis. When solving the weight coefficients, it combines the characteristics of the penalty function of ridge regression analysis and LASSO regression analysis. It avoids the disadvantage of ridge regression analysis that may select all characteristic variables and fail to achieve the screening of characteristic variables, and overcomes the limitation of LASSO regression analysis that gives some variables a weight coefficient of 0, which is a "hard selection". It has good reliability and stability.

[0062] Other beneficial effects of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0063] Figure 1 This is a flowchart of the steps of the present invention;

[0064] Figure 2 This is a schematic diagram of the elevation layout of various measuring points of the steel cofferdam in an embodiment of the present invention;

[0065] Figure 3 This is a schematic diagram of the planar arrangement of various measuring points of the steel cofferdam in the embodiment of the invention;

[0066] Figure 4 This is a flowchart of the coordinate descent method solution in S5 of the present invention;

[0067] Figure 5 This is a schematic diagram of iterative optimization of a certain weight coefficient in an embodiment of the present invention;

[0068] Figure 6This is a residual plot of the true value and the regression in an embodiment of the present invention. Detailed Implementation

[0069] The following specific examples illustrate the implementation of this disclosure. Those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0070] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0071] It should also be noted that the illustrations provided in the following embodiments are merely schematic representations of the basic concept of this disclosure. The illustrations only show components relevant to this disclosure and are not drawn according to the actual number, shape, and size of components in implementation. In actual implementation, the type, quantity, and proportion of each component can be arbitrarily changed, and the component layout may be more complex. Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0072] like Figure 1 As shown, an embodiment of the present invention provides a method for monitoring steel cofferdams based on multi-source data fusion, comprising the following steps:

[0073] S1. Develop a monitoring plan for the steel cofferdam, determining each monitoring item, its corresponding monitoring sensors, and their deployment locations. According to relevant regulations and specifications, the monitoring items for the steel cofferdam include vertical displacement of the pile top, horizontal displacement of the pile top, deep horizontal displacement of the pile body, axial force of the support, deflection of the support, temperature of the support, and water level outside the pit (or surface settlement outside the pit). Specifically, vertical and horizontal displacement of the pile top are monitored using surface piles and a level, providing data support for monitoring the displacement at the top of the steel cofferdam. Deep horizontal displacement of the pile body is monitored using inclinometers and a clinometer, providing data support for monitoring the overall horizontal displacement of the steel cofferdam. Axial force of the support is monitored using steel wire strain gauges, providing data support for monitoring the internal forces of the support. Deflection of the support is monitored using a level, providing data support for monitoring the deformation of the support. Temperature of the support is monitored using a thermometer, providing data support for monitoring the overall stability of the support. Additionally, the data collection frequency for each monitoring item needs to be determined.

[0074] S2, Collect monitoring data. Based on the monitoring plan established in S1, collect monitoring data according to the corresponding monitoring frequency and store it in time series.

[0075] S3, Monitoring Data Preprocessing. For monitoring data of a specific area within a component of the steel cofferdam, extract all monitoring data for that area, X = (X1, X2, ..., X...). j ), where j represents the number of monitoring items, and the preprocessing steps include outlier removal, missing value supplementation, and time synchronization.

[0076] For outlier removal, the complex environment of the construction site can easily lead to outliers in the collected data. Since the monitoring data follows a normal distribution, the 3σ principle can be used to determine data anomalies. The specific formula is as follows:

[0077]

[0078] Where, x ij This represents the data for the j-th monitoring item on the i-th day. If p = 1, then it means x ij The monitoring data is abnormal.

[0079] For missing value supplementation, if some data is missing due to anomaly removal or missing records, the average value of the monitored item is used as a substitute.

[0080] For time synchronization, due to the difference in data collection frequency of different monitoring projects, the data between the monitoring projects cannot be matched. Therefore, cubic spline interpolation is used for synchronization.

[0081] The final preprocessed data is obtained:

[0082]

[0083] S4, Analyze data correlation and extract features. Perform correlation analysis on the monitoring data and divide it into main dependent variables and characteristic variables. This may include the following sub-steps:

[0084] S41, perform partial correlation analysis on each pair of monitoring project data. Partial correlation analysis explores the association between two variables while controlling for the influence of other variables. The nth-order partial correlation coefficient is recursively calculated from the (n-1)th-order partial correlation coefficient, and its calculation formula is:

[0085]

[0086] In the formula, This indicates that in controlling x3 to x j Under the indicator variable, the partial correlation coefficient between variables x1 and x2; This indicates that in controlling x3 to x j-1 Under the indicator variable, the partial correlation coefficient between variables x1 and x2; This indicates that in controlling x3 to x j-1 Under the indicator variable, variables x1 and x j The partial correlation coefficient; This indicates that, under the control of the index variable, variables x2 and x... j The partial correlation coefficient.

[0087] After calculating the partial correlation coefficient between the two variables, a t-test is performed. The formula for the t-test is as follows:

[0088]

[0089] In the formula, r is the corresponding partial correlation coefficient, n is the data sample size, and k is the number of control variables. When t > t 0.05 When (nk-2), the partial correlation coefficient passes the test.

[0090] Calculate the partial correlation coefficients between each variable individually to obtain the partial correlation coefficient matrix R:

[0091]

[0092] S42, using the partial correlation coefficient matrix R as the evaluation matrix, calculate the information entropy weights of each variable. The partial correlation coefficient effectively controls the influence of other variables and reflects the correlation between the two variables; therefore, it is used as the evaluation matrix to calculate the information entropy weights of each variable. First, normalize the partial correlation coefficient matrix:

[0093]

[0094] In the formula, z ij R represents the normalized data values. ijThis represents the partial correlation coefficient.

[0095] Calculate the information entropy and information utility value of variables:

[0096]

[0097] In the formula, e j Let d represent the information entropy of the j-th variable. j =1-e j d j This represents the utility value of the j-th piece of information.

[0098] Calculate the information weights of variables:

[0099]

[0100] S43. Based on the information entropy weights of variables, distinguish between the dependent variable and the characteristic variables. Arrange the variables according to their information entropy weights from largest to smallest. The larger the entropy weight, the smaller the dispersion of the variable's index with other variables, and the more significant the correlation. Therefore, the variable with the largest information entropy weight is regarded as the dependent variable, and the remaining variables are regarded as characteristic variables.

[0101] S5, Elastic Network Regression Analysis Based on Coordinate Descent. For the identified primary and characteristic variables, elastic network regression is used to establish a linear regression equation between them. Since S4 only used partial correlation analysis to identify primary and characteristic variables, it did not address the potential for multicollinearity among variables. Therefore, elastic network regression analysis is used to screen out variables with high linear correlation among the characteristic variables, reducing the impact of inaccurate regression estimates caused by high correlation among characteristic variables.

[0102] Specifically, S5 can include the following sub-steps:

[0103] S51, Data Standardization. To eliminate the influence of data dimensions, deviation standardization is used to process the analyzed data, ensuring that all variable data fall within the range of 0 to 1. The calculation formula is as follows:

[0104]

[0105] In the formula: z ij This represents the standardized data; x ij This represents the original analysis data.

[0106] S52, Establish the loss function. The Elastic Regression algorithm combines the advantages of two existing machine learning regression algorithms: Ridge Regression and LASSO Regression. For loss function regularization, the magnitude of the loss term is controlled by two parameters, λ and ρ. The specific function form is as follows:

[0107]

[0108] In the formula, w represents the weight matrix, λ represents the first regularization term loss coefficient, and ρ represents the second regularization term loss coefficient. λ and ρ are pre-input as fixed values ​​into the network.

[0109] S53, Solving for the weight coefficient matrix w using the coordinate descent method. Input parameters λ and ρ, initialize the weight coefficient matrix w as a zero vector matrix, treat one weight coefficient as a variable, and fix the values ​​of the other weight coefficients to find the optimal solution of the loss function under the current conditions. Iterate through all weight coefficients, repeating the above calculation process to obtain the optimal weight coefficient matrix w under the current values ​​of λ and ρ. The process of solving for w can be represented as follows:

[0110]

[0111] in, This represents the weight coefficient for the k-th iteration and the m-th weight coefficient.

[0112] S54, Optimize parameters λ and ρ. Using the mean squared error (MSE) of the regression equation as the objective, optimize parameters λ and ρ. The MSE calculation formula is:

[0113]

[0114] in, Represents the regression value, y i Let λ represent the true value and n be the sample size. Generally, λ + ρ = 1 is assumed. λ is taken as the baseline value, initially set to 0, with a step size of d and a final value of 1. The input values ​​of λ and ρ are changed, and the input values ​​of λ and ρ corresponding to the minimum MSE are iteratively solved.

[0115] S55. Based on the solved optimal λ and ρ values, a regression model is established to realize the fusion analysis of the main dependent variable and the characteristic variable, that is, to realize the fusion of various indicator data within the dataset, so as to reasonably evaluate the safety status of the steel cofferdam through the main dependent variable and the characteristic variable.

[0116] The following specific examples further illustrate the effectiveness of this solution:

[0117] The experimental data were obtained from 20 days of real-time monitoring of the steel cofferdam of a railway bridge pier under construction, such as... Figure 2 , Figure 3 As shown, for Figure 2 The monitoring data of steel cofferdams in a certain area were analyzed by this method to achieve data fusion. The monitoring items extracted in this area include: horizontal displacement of pile top, vertical displacement of pile top, and horizontal displacement of pile body (four locations).

[0118] According to S3, the monitoring project data in this area is preprocessed, and after operations such as outlier removal, missing value imputation, and time synchronization, a standard data matrix X is obtained. Then, S4 is used for data correlation analysis and variable segmentation. First, the partial correlation coefficients between each pair of characteristic variables are calculated:

[0119] Table 1 Partial Correlation Coefficients

[0120]

[0121]

[0122] In the table, x1 is the horizontal displacement of the pile top, x2 is the vertical displacement of the pile top, and x3 to x6 are the horizontal displacements of the pile body.

[0123] Using the partial correlation coefficients calculated above as the evaluation matrix, the information entropy weights of each feature variable are calculated based on the entropy weight method:

[0124] Table 2 Information Entropy Weight Table for Feature Variables

[0125]

[0126] Arranged from largest to smallest based on the information entropy weights of the calculated variables:

[0127] x1 > x5 > x4 > x3 > x2 > x6

[0128] It can be seen that the horizontal displacement variable at the pile top has the largest information entropy weight, indicating that this variable has the smallest dispersion from other variables. The higher the significance, the more likely this variable is to be considered the main dependent variable, and the remaining variables are considered characteristic variables. This completes the variable classification.

[0129] Let y denote the monitoring data sequence of pile top level as the main dependent variable, and let X = (x1, x2, ..., x3) denote the monitoring data sequence of characteristic variables. j-1 Based on elastic network regression analysis, a regression model is established, and a loss function is constructed:

[0130]

[0131] The process of solving for w is as follows:

[0132]

[0133] The algorithm used is the coordinate descent method, and its process is as follows: Figure 4 As shown, the initial weight coefficient matrix is ​​w = (0,0,...,0). T Next, one regression weight coefficient is fixed as a variable, while the rest are constants of their current value. The value of the fixed regression weight coefficient is solved using the coordinate descent method. This process is iterated until all weight coefficients have been optimized. A schematic diagram of the optimization of a single weight coefficient is shown below. Figure 5 As shown in the table below, the regression weight coefficients for each feature variable are finally obtained:

[0134] Table 3 Regression weight coefficients of feature variables

[0135]

[0136] In the table, x2 represents the vertical displacement of the pile top, and x3 to x6 represent the horizontal displacement of the pile body.

[0137] When λ = 0.8, the MSE value corresponding to the regression function is minimized, and the established regression equation is: y = 0.3795x1 - 0.1091x3 + 0.2075x5. The residuals between the regression value obtained by solving this equation and the true value are as follows: Figure 6 As shown, the maximum residual value is 0.031, and the residual rate is 2.34%, which is less than the 5% required for engineering accuracy.

[0138] R-squared value for evaluating the goodness of fit of regression equations square Coefficient, calculation formula:

[0139] R square =SSR / SST

[0140] in Calculated R square =0.9366.

[0141] Therefore, this regression model has superior accuracy and can comprehensively evaluate the safety status of steel cofferdams through pile top horizontal displacement, pile top vertical displacement, and pile body horizontal displacement.

[0142] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0143] In summary, the method provided in this embodiment uses partial correlation analysis to perform correlation analysis on the data, overcoming the influence of multiple linear correlations among indicator variables. Furthermore, by calculating the information entropy weights of the indicator variables based on the partial correlation coefficient matrix as the evaluation matrix, it effectively highlights the dispersion and significance of the main and characteristic variables. Compared with other correlation analysis methods, it possesses objectivity, reliability, and universality. The method employs elastic network regression analysis to establish a regression model, combining the characteristics of ridge regression analysis and LASSO regression analysis in calculating the weight coefficients. This avoids the disadvantage of ridge regression analysis potentially selecting all characteristic variables and failing to achieve characteristic variable selection, while overcoming the limitation of LASSO regression analysis giving some variables a "hard selection" weight coefficient of 0, thus exhibiting good reliability and stability.

[0144] Based on the same inventive concept, this embodiment also provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned steel cofferdam monitoring method based on multi-source data fusion.

[0145] Computer-readable media include, but are not limited to, any type of disk (including floppy disks, hard disks, optical disks, CD-ROMs, and magneto-optical disks), ROM, RAM, EPROM (Erasable Programmable Read-Only Memory), EEPROM, flash memory, magnetic cards, or optical cards. In other words, readable media includes any medium by which a device (such as a computer) stores or transmits information in a readable form.

[0146] The computer-readable storage medium provided in this embodiment has the same inventive concept and the same beneficial effects as the aforementioned method, and will not be described again here.

[0147] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0148] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for monitoring steel cofferdams based on multi-source data fusion, characterized in that, Includes the following steps: S1. Develop a monitoring plan for the steel cofferdam, and determine each monitoring item, the corresponding monitoring sensor, and the corresponding deployment point; S2, collect monitoring data according to the corresponding monitoring frequency and store it in time series; S3: For monitoring data of a certain area of ​​a certain component of a steel cofferdam, extract all monitoring data of that area and perform data preprocessing. S4, perform correlation analysis on the monitoring data and classify the main dependent variable and characteristic variables; including: S41, perform partial correlation analysis on each pair of monitoring item data; S42, analyze the partial correlation coefficient matrix. R As an evaluation matrix, calculate the information entropy weights of each variable; S43, based on the information entropy weights of the variables, divide the main dependent variable and the characteristic variable; S5. For the divided main dependent variables and characteristic variables, the elastic network regression algorithm is used to perform regression analysis to establish a linear regression equation between the main dependent variables and characteristic variables. The safety status of the steel cofferdam is evaluated by comprehensively evaluating the main dependent variables and characteristic variables.

2. The method for monitoring steel cofferdams based on multi-source data fusion according to claim 1, characterized in that, The preprocessing steps in S2 include outlier removal, missing value imputation, and time synchronization, ultimately yielding the preprocessed data. 。 3. The method for monitoring steel cofferdams based on multi-source data fusion according to claim 2, characterized in that, For outlier removal, the 3σ principle is used to determine the outlier nature of the data. The specific formula is as follows: in, Indicates the first The first monitoring item Data from the day, , , illustrate The monitoring data is abnormal. For missing values, the average value of the monitored items is used instead. For time synchronization, cubic spline interpolation is used.

4. The method for monitoring steel cofferdams based on multi-source data fusion according to claim 3, characterized in that, The monitoring project data in S3 is represented as follows: ,in j This indicates the number of monitored items.

5. The method for monitoring steel cofferdams based on multi-source data fusion according to claim 1, characterized in that, In S41, the nth-order partial correlation coefficient is recursively calculated from the (n-1)th-order partial correlation coefficient, and its calculation formula is as follows: In the formula, Indicates control Under indicator variables, variables and The partial correlation coefficient; Indicates control Under indicator variables, variables and The partial correlation coefficient; Indicates control Under the indicator variable, the variable and The partial correlation coefficient; This indicates that, under the control index variable, the variable and The partial correlation coefficient; After calculating the partial correlation coefficient between the two variables, proceed with... t The verification is performed using the following formula: In the formula, r This represents the corresponding partial correlation coefficient. n It refers to the data sample size. k To control the number of variables, when At that time, the partial correlation coefficient passed the test; Calculate the partial correlation coefficients between each variable individually to obtain the partial correlation coefficient matrix R: ； In S42, the partial correlation coefficient matrix is ​​normalized: In the formula, This represents the normalized data value. This represents the partial correlation coefficient; Calculate the information entropy and information utility value of variables: In the formula, Indicates the first j Information entropy of each variable , Indicates the first j Individual information utility value; Calculate the information weights of variables: ； In S43, the information entropy weights of each variable are arranged from largest to smallest. The variable with the largest information entropy weight is regarded as the main dependent variable, and the remaining variables are regarded as characteristic variables.

6. The method for monitoring steel cofferdams based on multi-source data fusion according to claim 5, characterized in that, S5 specifically includes the following sub-steps: S51, data standardization; S52, establish the loss function using the elastic network regression algorithm; S53, Solving the weight coefficient matrix based on the coordinate descent method ; S54, optimized parameters; S55 establishes a regression model based on optimal parameters to achieve fusion analysis of the main dependent variable and characteristic variables.

7. A method for monitoring steel cofferdams based on multi-source data fusion according to claim 6, characterized in that, In S51, deviation standardization is used to process the analyzed data, ensuring that all variable data fall within the range of 0 to 1. The calculation formula is as follows: In the formula: This represents the standardized data; This represents the raw analysis data; S52 uses two parameters and To control the size of the loss term, the specific function form is: In the formula, Represents the weight matrix. This represents the loss coefficient of the first regularization term. This represents the loss coefficient of the second regularization term; and As a fixed value, it is pre-input into the network; Input parameters in S53 and Initialize the weight coefficient matrix Given a zero vector matrix, treat one of the weight coefficients as a variable and keep the other weight coefficients fixed, find the optimal solution of the loss function under the current conditions; Iterate through all weight coefficients in sequence, repeat the above calculation process, and obtain the current result. and Under the given conditions, the optimal weight coefficient matrix Solve The process is represented as follows: in, Indicates the first The next iteration, the... Each weighting coefficient; In S54, the mean square error (MSE) of the regression equation is used as the objective to optimize the parameters. and The formula for calculating MSE is: in, Represents the regression value. Represents the actual value. For sample size; ,by The baseline value is 0, and the step size is [value missing]. The termination value is 1, change and Given the input values, iteratively solve for the value with the minimum MSE. and The input value; S55 is based on the optimal solution. and A regression model is established to achieve the fusion analysis of the main dependent variable and characteristic variables.

8. A computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements a steel cofferdam monitoring method based on multi-source data fusion as described in any one of claims 1-7.

Images