Weighted analysis method for multi-attribute data in advanced geological forecast

Abstract

The invention discloses a weighted analysis method for multi-attribute data in advanced geological forecast, belongs to the technical field of detection data optimization, and can solve the problems that redundant and mutually exclusive data are difficult to eliminate and missing and abnormal data processing effects are poor when the multi-attribute data are processed by an existing method. The method comprises the following steps: S1, normalizing various attribute data obtained by each grid point in a target area in advance geological forecast to obtain normalized data of each attribute in each grid point; s2, determining a single attribute weight coefficient of each attribute in each grid point according to the standardized data of each attribute in all the grid points, and determining a comprehensive attribute weight coefficient of each grid point according to the single attribute weight coefficient; and S3, determining update data of each attribute of each grid point by using the normalized data and the single attribute weight coefficient of each attribute in each grid point and the comprehensive attribute weight coefficient of each grid point. The method is used for multi-attribute data analysis.

Description

Technical Field

[0001] The invention relates to a weighted analysis method of multi-attribute data in advanced geological forecasting, belonging to the technical field of detection data optimization. Background Art

[0002] In the advance prediction of tunnel geological conditions, existing technologies are usually based on the analysis method of seismic wave data, using a series of seismic wave parameters (such as longitudinal wave velocity, transverse wave velocity, stress, water content, etc.) to predict the geological conditions ahead. These technologies provide a certain reference for tunnel construction, but in practical applications, due to the complexity of data, variability of geological conditions and other reasons, the accuracy and applicability of predictions are relatively limited. At present, commonly used data processing methods include linear regression, empirical analysis, multi-attribute weighting, interpolation, etc. However, these existing data processing methods still have many shortcomings when processing multidimensional and multi-attribute data. Specifically reflected in the following aspects:

[0003] (1) Redundant and mutually exclusive data are difficult to eliminate: In tunnel geological prediction, multiple parameters such as seismic wave velocity, stress, and water content often have complex correlations, and the relationships between different attributes are not independent. Due to the similar weights between attributes, information overlap often occurs, resulting in redundant data. In addition, there are mutually exclusive relationships between some attributes, which requires the model to find a balance point in contradictory information during data analysis, resulting in a decrease in the accuracy of data analysis. The existence of redundant and mutually exclusive data not only increases the difficulty of data processing, but also directly affects the reliability of geological prediction results, posing a hidden danger to tunnel construction safety.

[0004] (2) Poor processing of missing and abnormal data: In the process of tunnel geological survey, due to the complexity of geological conditions and incomplete field data collection, missing data and abnormal situations often occur. Although traditional interpolation methods can fill in missing data, they can usually only process simple linear or two-dimensional spatial data, and it is difficult to fully deal with the common three-dimensional spatial data missing or abnormal problems in tunnel construction. When the data distribution presents nonlinear and uneven characteristics, the accuracy and robustness of traditional interpolation methods are obviously insufficient, resulting in low quality of filled data, affecting the reliability of subsequent analysis and the accuracy of prediction. Insufficient data processing will not only affect the accurate identification of geological conditions, but also increase construction risks.

[0005] (3) Limited generalization ability of the model: Current multi-attribute weight models usually rely on empirical weighting, that is, weights are assigned to each parameter based on experience or human judgment. However, this empirical weighting method is difficult to adapt to different geological conditions, resulting in poor performance of the model in new or different geological environments and insufficient generalization ability. For example, a weight model established for a specific geological area may not accurately reflect the actual geological structure when faced with other geological conditions, affecting the stability and accuracy of the forecast results. The lack of model generalization ability directly restricts its promotion and application in diverse geological environments, making it difficult to meet the actual needs of complex geological predictions.

[0006] (4) Potential patterns are difficult to reveal: Existing geological prediction methods usually use simple data processing and analysis methods, and fail to deeply explore the potential patterns hidden in multidimensional data. Due to the lack of analysis of the deep structure of the data, it is difficult for the model to identify and reveal the complex correlations between different geological parameters. This limitation limits the accuracy of the prediction model in identifying geological hazards, especially potential risks, and makes it difficult to effectively predict potential geological risks, which may threaten the safety of tunnel construction. The lack of potential pattern disclosure makes the existing methods limited in the identification and early warning of geological hazards. Summary of the invention

[0007] The present invention provides a weighted analysis method for multi-attribute data in advanced geological forecasting, which can solve the problems of existing data processing methods in processing multi-dimensional and multi-attribute data, such as difficulty in eliminating redundant and mutually exclusive data, poor processing effect of missing and abnormal data, and limited generalization ability of the model.

[0008] The present invention provides a weighted analysis method for multi-attribute data in advanced geological prediction, the method comprising:

[0009] S1. Normalizing the various attribute data obtained in the advanced geological prediction for each grid point in the target area to obtain the normalized data of each attribute in each grid point;

[0010] S2. Determine the single attribute weight coefficient of each attribute in each grid point according to the normalized data of each attribute in all grid points, and determine the comprehensive attribute weight coefficient of each grid point according to the single attribute weight coefficient;

[0011] S3. Determine the updated data of each attribute of each grid point using the normalized data and single attribute weight coefficient of each attribute in each grid point, as well as the comprehensive attribute weight coefficient of each grid point.

[0012] Optionally, the S1 specifically includes:

[0013] S11, determining the points to be interpolated among all grid points in the target area under each attribute, and performing data filling processing on the points to be interpolated to obtain multiple attribute data of all grid points after processing;

[0014] S12. Normalize the same attribute data of all grid points respectively to obtain normalized data of each attribute in each grid point.

[0015] Optionally, the data filling process adopts a cubic spline interpolation method.

[0016] Optionally, the step of filling data on the points to be interpolated in S11 is specifically as follows:

[0017] The cubic spline interpolation method is used to interpolate the X-axis direction, Y-axis direction and Z-axis direction of the interpolation point in sequence.

[0018] Optionally, the S12 is specifically:

[0019] Divide the various attribute data of each grid point into benefit-type attribute data and cost-type attribute data;

[0020] The benefit-type attribute data and cost-type attribute data of all grid points are normalized respectively to obtain the normalized data of each attribute in each grid point.

[0021] Optionally, the determining of the single attribute weight coefficient of each attribute in each grid point in S2 specifically includes:

[0022] Obtain the sum of the differences between the normalized data of each attribute of each grid point and the normalized data of the same attribute in each of the remaining grid points as the first sum, obtain the square value of the first sum corresponding to each attribute of each grid point, and obtain the square root of the sum of the square values ​​corresponding to all attributes of each grid point as the second sum;

[0023] The sum of the absolute values ​​of the differences between the normalized data of each attribute of each grid point and the normalized data of the same attribute in each of the remaining grid points is obtained as the third sum, and the ratio of the third sum to the second sum is obtained as the single attribute weight coefficient of each attribute in each grid point.

[0024] Optionally, the step of determining the comprehensive attribute weight coefficient of each grid point according to the single attribute weight coefficient in S2 is specifically:

[0025] The sum of the products of the normalized data of all attributes of each grid point and its single attribute weight coefficient is obtained as the comprehensive attribute weight coefficient of each grid point.

[0026] Optionally, the S3 specifically includes:

[0027] The normalized data of each attribute in each grid point is adjusted by using the comprehensive attribute weight coefficient of each grid point to obtain the attenuation value of each attribute in each grid point;

[0028] Determine the weighted adjustment value of each attribute in each grid point according to the single attribute weight coefficient of each attribute in each grid point and the comprehensive attribute weight coefficient of each grid point;

[0029] Determine the minimum value adjustment of each attribute at each grid point according to the single attribute weight coefficient of each attribute at each grid point and the comprehensive attribute weight coefficient of each grid point;

[0030] The sum of the attenuation value, the weighted adjustment value, and the minimum adjustment value of each attribute in each grid point is obtained as update data of each attribute of each grid point.

[0031] Optionally, various attribute data of each grid point include longitudinal wave velocity, transverse wave velocity, relative stress, water content probability, dynamic Young's modulus, Poisson's ratio, longitudinal and transverse wave velocity ratio and hazard level.

[0032] Optionally, the benefit-type attribute data includes longitudinal wave velocity, transverse wave velocity and dynamic Young's modulus;

[0033] The cost-type attribute data include relative stress, water content probability, Poisson's ratio, longitudinal and transverse wave velocity ratio and hazard level.

[0034] The beneficial effects that the present invention can produce include:

[0035] The weighted analysis method of multi-attribute data in advanced geological prediction provided by the present invention normalizes the data, then calculates the single attribute weight coefficient of each attribute after normalization, and then obtains the comprehensive attribute weight coefficient to update the attribute data. The present application can effectively eliminate redundant and mutually exclusive information in the data processing process, making the data analysis results more accurate and stable.

[0036] The weighted analysis method of multi-attribute data in advanced geological forecasting provided by the present invention can more accurately supplement data in the case of nonlinear and non-uniform distribution of data by adopting cubic spline interpolation when processing missing or abnormal data in three-dimensional space. Compared with traditional linear or simple multidimensional interpolation methods, the cubic spline interpolation method of the present invention can more accurately reflect the real relationship between data in complex geological conditions, reduce interpolation errors and cumulative deviations, and thus improve the quality and reliability of supplementary data.

[0037] The weighted analysis method of multi-attribute data in advanced geological forecasting provided by the present invention normalizes geological data of different dimensions, thereby eliminating scale differences between data, allowing values ​​of different attributes to be more accurately integrated and compared, and significantly improving the accuracy of data analysis.

[0038] The weighted analysis method of multi-attribute data in advanced geological prediction provided by the present invention calculates the single attribute weight coefficient of each attribute, and then obtains the comprehensive attribute weight coefficient to establish an adaptive weight model. This adaptive weight model overcomes the shortcomings of the prior art that relies on experience weighting, so that the model can better adapt to different geological conditions, thereby having stronger generalization ability and prediction stability.

[0039] The weighted analysis method of multi-attribute data in advanced geological prediction provided by the present invention can effectively eliminate redundant and mutually exclusive information in the data processing process through comprehensive analysis of multi-attribute characteristics, making the data analysis results more accurate and stable, effectively eliminating redundant information and mutually exclusive relationships, and improving the accuracy of data analysis

[0040] The weighted analysis method of multi-attribute data in advanced geological prediction provided by the present invention reveals the potential patterns in the data through the integrated processing of multiple attributes, helps the model to identify the potential risks in geological hazards, and especially improves the accuracy of hazard identification under complex geological conditions, making the advanced geological prediction more reliable. The present invention improves the processing accuracy of geological parameters through data filling, normalization processing, weight coefficient calculation and data updating, so that the prediction model can more accurately identify potential geological risks, thereby effectively improving the prediction accuracy of geological risks during tunnel construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of a weighted analysis method for multi-attribute data in advanced geological prediction provided by an embodiment of the present invention;

[0042] Figure 2 A schematic diagram of multi-attribute data in advanced geological forecasting provided by an embodiment of the present invention;

[0043] Figure 3 A schematic diagram of comprehensive attribute weight coefficients of all grid points provided in an embodiment of the present invention;

[0044] Figure 4 A schematic diagram of multi-attribute update data of all grid points provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention is described in detail below in conjunction with embodiments, but the present invention is not limited to these embodiments.

[0046] Seismic wave technology predicts the geological conditions of tunnels in advance, generating multi-attribute and multi-dimensional data parameters including longitudinal wave velocity, transverse wave velocity, relative stress, water content probability, dynamic Young's modulus, Poisson's ratio, longitudinal and transverse wave velocity ratio, and hazard level. However, the complex correlation between these parameters is often accompanied by a large amount of redundant information and sometimes there is a mutually exclusive relationship, which significantly increases the difficulty of data processing and geological interpretation. In response to this problem, the present invention uses cubic spline interpolation to fill in missing or abnormal data in three-dimensional space, normalize multi-attribute data, obtain single attribute weight coefficients for each attribute data in advance geological prediction, and then obtain comprehensive attribute weight coefficients, and update various attribute data, so as to effectively integrate multiple related features, significantly improve the accuracy of data analysis and the generalization ability of prediction models, reveal potential hidden patterns in data, thereby enhancing the prediction accuracy of the model and improving the accuracy of geological hazard identification.

[0047] Specifically, the embodiment of the present invention provides a weighted analysis method for multi-attribute data in advanced geological prediction, such as Figures 1 to 4 As shown, the method includes:

[0048] S1. The various attribute data obtained in the advanced geological forecast for each grid point in the target area are respectively normalized to obtain the normalized data of each attribute in each grid point.

[0049] Among them, the various attribute data of each grid point include longitudinal wave velocity, transverse wave velocity, relative stress, water content probability, dynamic Young's modulus, Poisson's ratio, longitudinal and transverse wave velocity ratio and hazard level.

[0050] The present invention takes the advance prediction of underground caverns as an example, defines the X-axis direction as the direction along the cavern axis, the Y-axis direction as from the left wall to the right wall, the Z-axis direction as from the top arch to the bottom plate, and the coordinate origin is located at the center of the tunnel face. First, obtain the large-scale advance geological prediction results data of the entire space of the underground cavern, which is usually the longitudinal wave velocity. Shear wave velocity Relative stress Δσ TSP (x i ,y j ,z k ), water content probability Dynamic Young's modulus E TSP (x i ,y j ,z k ), Poisson's ratio μ TSP (x i ,y j ,z k ), longitudinal and transverse wave velocity ratio Hazard Level C TSP (x i ,yj ,z k )These 8 types of attribute data.

[0051] The above S1 specifically includes:

[0052] S11, determining the points to be interpolated among all the grid points in the target area under each attribute, and performing data filling processing on the points to be interpolated to obtain multiple attribute data of all the processed grid points.

[0053] In the embodiment of the present invention, the method used for data filling processing is cubic spline interpolation. The above-mentioned points to be interpolated are grid points where attribute values ​​are missing or abnormal.

[0054] The data filling of the interpolation point in S11 is specifically: interpolating the X-axis direction, the Y-axis direction and the Z-axis direction of the interpolation point in sequence by using the cubic spline interpolation method.

[0055] Perform dimension-by-dimensional interpolation in space, assuming that the spatial coordinate position (x m ,y n ,z p ) is the attribute value V(x) of the grid point at m ,y n ,z p ) is missing or has an exception (for the convenience of algorithm description, V represents The points at the surrounding coordinate positions are all valid data and need to be filled in by interpolation. First, under fixed y and z values, the data is interpolated along the X axis to obtain an intermediate interpolation result. Then, under fixed x and z values, the intermediate interpolation result is interpolated along the Y axis. Finally, under fixed x and y values, the result is interpolated along the Z axis.

[0056] (1) First, interpolate along the X-axis. Assume that the position along the X-axis (x m-1 ,y n ,z p )、(x m+1 ,y n ,z p )、(x m+2 ,y n ,z p ) respectively have reasonable attribute data V1=V(x m-1 ,y n ,z p )、V2=V(x m+1 ,y n ,z p )、V3=V(x m+2 ,y n ,z p ) and at position (xm ,y n ,z p ) has missing values ​​or outliers V(x) m ,y n ,z p ), and obtain V(x) by using the cubic spline interpolation algorithm. m ,y n ,z p ).

[0057] Definition: h0 = x m+1 -x m-1 , h1=x m+2 -x m+1 , then the interpolation coefficient is obtained through the spline interpolation condition, according to the function value: a0 = V1; according to the first-order derivative continuity condition and the interpolation condition, we get: Through the second-order derivative continuity condition and boundary conditions, we get: According to the continuity condition and the solution of the second-order derivative, we get: According to formula (1), select the coordinate index q closest to the point to be interpolated:

[0058]

[0059] The x-direction spline interpolation attribute value is obtained according to the following formula (2):

[0060] V(x m ,y n ,z p )=a0+a1(x m -x q )+a2(x m -x q ) 2 +a3(x m -x q ) 3 (2)

[0061] (2) Update the Y-axis data in the same way as the X-axis. Assume that the position along the Y-axis (x m ,y n-1 ,z p )、(x m ,y n+1 ,z p )、(x m ,y n+2 ,z p ) respectively have reasonable attribute data V1=V(x m ,y n-1 ,z p )、V2=V(x m ,y n+1,z p )、V3=V(x m ,y n+2 ,z p ), and at position (x m ,y n ,z p ) has missing values ​​or outliers V(x) m ,y n ,z p ), update V(x) by cubic spline interpolation algorithm m ,y n ,z p ).

[0062] (3) Update the Z-axis data in the same way as the X-axis and Y-axis data. Assume that the position along the Y-axis (x m ,y n ,z p-1 )、(x m ,y n ,z p+1 )、(x m ,y n ,z p+2 ) respectively have reasonable attribute data V1=V(x m ,y n ,z p-1 )、V2=V(x m ,y n ,z p+1 )、V3=V(x m ,y n ,z p+2 ) and at position (x m ,y n ,z p ) has missing values ​​or outliers V(x) m ,y n ,z p ), update V(x) through the cubic spline interpolation algorithm m ,y n ,z p ).

[0063] S12. Normalize the same attribute data of all grid points respectively to obtain normalized data of each attribute in each grid point.

[0064] Specifically include:

[0065] (1) Divide the various attribute data of each grid point into benefit-type attribute data and cost-type attribute data.

[0066] According to the data attributes of advanced geological prediction and seismic wave CT, the data can be divided into benefit-type and cost-type attribute data. Benefit-type attributes indicate that the larger the attribute value, the better the attribute, while cost-type attributes indicate that the smaller the attribute value, the better the attribute. The benefit-type attribute data includes longitudinal wave velocity Shear wave velocity and dynamic Young's modulus E TSP (x i ,y j ,z k ); Cost attribute data includes relative stress Δσ TSP (x i ,y j ,z k ), water content probability Poisson's ratio μ TSP (x i ,y j ,z k ), longitudinal and transverse wave velocity ratio and hazard level C TSP (x i ,y j ,z k ).

[0067] (2) The benefit-based attribute data and cost-based attribute data of all grid points are normalized separately to obtain the normalized data of each attribute in each grid point.

[0068] For benefit-type attribute data, normalization is performed according to formula (3):

[0069]

[0070] For cost-type attribute data, normalization is performed according to formula (4):

[0071]

[0072] According to the above calculation formula, the normalized attribute data of various types are obtained.

[0073] S2. Determine the single attribute weight coefficient of each attribute in each grid point according to the normalized data of each attribute in all grid points, and determine the comprehensive attribute weight coefficient of each grid point according to the single attribute weight coefficient.

[0074] The above single attribute weight coefficient of each attribute in each grid point is determined, specifically including:

[0075] Obtain the sum of the differences between the normalized data of each attribute of each grid point and the normalized data of the same attribute in each of the remaining grid points as the first sum, obtain the square value of the first sum corresponding to each attribute of each grid point, and obtain the square root of the sum of the square values ​​corresponding to all attributes of each grid point as the second sum;

[0076] The sum of the absolute values ​​of the differences between the normalized data of each attribute of each grid point and the normalized data of the same attribute in each of the remaining grid points is obtained as the third sum, and the ratio of the third sum to the second sum is obtained as the single attribute weight coefficient of each attribute in each grid point.

[0077] Specifically, in three-dimensional space, the attribute deviation between two points is defined as:

[0078]

[0079] The subscript A represents different attributes, and in this invention, it represents 8 attributes. A (x i ,y j ,z k ) should be selected to maximize the total deviation of all evaluation indicators for all decision-making options. Based on this, the objective function is constructed and combined with the single attribute weight coefficient W A (x i ,y j ,z k ) The sum of the characteristics is 1, and the single attribute weight coefficient W is solved A (x i ,y j ,z k ) is converted to solve the following optimization problem:

[0080]

[0081] By constructing the Lagrangian function through equation (6) and setting the partial derivative of the objective function to 0, we can obtain the following formula (7) to obtain the single attribute weight coefficient of each attribute of each grid point:

[0082]

[0083] The above method determines the comprehensive attribute weight coefficient of each grid point based on the single attribute weight coefficient, which is:

[0084] The sum of the products of the normalized data of all attributes of each grid point and its single attribute weight coefficient is obtained as the comprehensive attribute weight coefficient of each grid point.

[0085] The comprehensive attribute weight coefficient at each grid point can be obtained by formula (8):

[0086]

[0087] S3. Determine the updated data of each attribute of each grid point using the normalized data and single attribute weight coefficient of each attribute in each grid point, as well as the comprehensive attribute weight coefficient of each grid point.

[0088] Specifically include:

[0089] (1) The normalized data of each attribute in each grid point is adjusted using the comprehensive attribute weight coefficient of each grid point to obtain the attenuation value of each attribute in each grid point.

[0090] The attenuation value of each attribute at each grid point can be obtained by A V (x i ,y j ,z k )(1-W A (x i ,y j ,z k ) 2 ) is obtained, which means that according to W A (x i ,y j ,z k ) Adjust the normalized data A V (x i ,y j ,z k ), mainly to weaken the attribute value of a specific area, because W A (x i ,y j ,z k ) is a weighting function with a value between 0 and 1, so (1-W A (x i ,y j ,z k ) 2 ) is also between 0 and 1. A (x i ,y j ,z k )=0, the adjustment factor is 1, then A V (x i ,y j ,z k ) remains unchanged. A (x i ,y j ,z k ) approaches 1, the adjustment factor approaches 0, which means A V (x i ,y j ,z k) will be weakened, or even approach 0.

[0091] (2) Determine the weighted adjustment value of each attribute at each grid point based on the single attribute weight coefficient of each attribute at each grid point and the comprehensive attribute weight coefficient of each grid point.

[0092] Specifically, the normalized data of each attribute in each grid point are normalized and weighted adjusted using the single attribute weight coefficient of each attribute in each grid point and the comprehensive attribute weight coefficient of each grid point to obtain the weighted adjustment value of each attribute in each grid point.

[0093] The weighted adjustment value of each attribute at each grid point can be obtained by The function of this formula is to obtain D(x i ,y j ,z k ) and A V (x i ,y j ,z k ) changes in the range of A V (x i ,y j ,z k ) for weighted adjustment. It can dynamically adjust A according to the changes in certain position characteristics. V (x i ,y j ,z k ) value. Normalization factor A V (x i ,y j ,z k ) varies with D(x i ,y j ,z k The purpose is to make a relative adjustment between the two quantities to maintain the consistency of the numerical scale. A (x i ,y j ,z k ) 2 The role of is to further weight the adjustment so that D(x i ,y j ,z k The effect of the change of ) on the correction result depends on W A (x i ,y j ,z k ). If W A (x i ,y j ,z k) is larger, the contribution of the weighted adjustment value will be greater; conversely, if W A (x i ,y j ,z k ) is small, the contribution of the weighted adjustment value will be suppressed.

[0094] (3) Determine the minimum adjustment value of each attribute at each grid point based on the single attribute weight coefficient of each attribute at each grid point and the comprehensive attribute weight coefficient of each grid point.

[0095] Specifically, the single attribute weight coefficient of each attribute in each grid point and the comprehensive attribute weight coefficient of each grid point are used to standardize and weighted-adjust the minimum value in the normalized data of the corresponding attributes in all grid points to obtain the minimum value adjustment value of each attribute in each grid point.

[0096] The minimum adjustment value for each attribute at each grid point can be The function of this formula is to ensure that A V (x i ,y j ,z k ) is appropriately adjusted during the correction process, and the magnitude of the adjustment depends on D(x i ,y j ,z k ) and W A (x i ,y j ,z k ). Among them, the normalization factor Used to standardize D(x i ,y j ,z k ) range of change. It will A V (x i ,y j ,z k ) is the minimum value of D(x i ,y j ,z k ) to scale the range of change.

[0097] (4) Obtain the sum of the attenuation value, the weighted adjustment value, and the minimum adjustment value of each attribute in each grid point as the update data of each attribute of each grid point.

[0098] Update various attribute data according to formula (9):

[0099]

[0100] The purpose of formula (9) is to V (x i,y j ,z k ) is dynamically adjusted to make it more in line with regional characteristics, especially through D(x i ,y j ,z k ) and W A (x i ,y j ,z k ) to reflect the impact of spatial changes.

[0101] Compared with the prior art, the present invention is mainly different in the following aspects:

[0102] (1) Cubic spline interpolation method is used to fill in missing data and outliers in three-dimensional space.

[0103] The present invention uses cubic spline interpolation when processing missing or abnormal data in three-dimensional space, which can more accurately supplement data in the case of nonlinear and non-uniform data distribution. Compared with traditional linear or simple multidimensional interpolation methods, cubic spline interpolation can more accurately reflect the real relationship between data in complex geological conditions, reduce error accumulation, and make data processing more reliable.

[0104] (2) Normalization of multi-attribute data.

[0105] The present invention normalizes the multi-attribute data of underground cavern advance prediction, thereby unifying the dimension and range of the data, reducing the scale differences between different attributes, and avoiding errors caused by inconsistent dimensions during data analysis. Compared with the method of directly processing non-normalized data in the prior art, this processing step significantly improves the accuracy of data fusion and analysis.

[0106] (3) Calculation of single attribute weight coefficients and acquisition of comprehensive attribute weight coefficients.

[0107] The present invention calculates the single attribute weight coefficients of eight attribute data, further obtains the comprehensive attribute weight coefficient, and establishes an adaptive weight model that can better adapt to different geological conditions. This weight allocation method effectively reduces the error and subjectivity caused by empirical weighting, and enables the model to have better generalization ability and stability in different geological environments.

[0108] (4) Integrate multi-attribute features and eliminate redundant and mutually exclusive information.

[0109] By integrating and processing multiple attribute features, the present invention effectively eliminates redundant information and mutually exclusive relationships in the data, and avoids analysis errors caused by information overlap and contradictions. Compared with the limited effect of information redundancy processing in traditional methods, the present invention significantly improves the accuracy of data analysis and improves the prediction effect of the model.

[0110] (5) Reveal potential hidden patterns.

[0111] The present invention can deeply explore potential hidden patterns in the data through comprehensive processing and analysis of multi-attribute data, and enhance the accuracy of the model in identifying and predicting geological hazards. This feature is particularly suitable for complex geological conditions in tunnel construction. Compared with the existing method of only processing shallow data, the present invention has higher accuracy in identifying geological hazards.

[0112] (6) Improve the generalization ability of the model.

[0113] Since the present invention adopts a multi-attribute comprehensive analysis based on weight coefficients, the model can adapt to different geological conditions and has stronger generalization ability. This adaptive advantage makes up for the problem of poor generalization ability in the prior art, making the prediction model more stable under complex geological conditions.

[0114] The present invention provides a specific example, which is as follows:

[0115] The underground powerhouse cavern group of a hydropower station is located in the mountain on the left bank. The underground powerhouse cavern group has a relatively developed structure, mostly interlayer faults, and most of them are located below the groundwater level. In order to ensure the safe construction of the underground powerhouse cavern group, grasp the deformation law and development trend of the surrounding rock, and dynamically adjust the excavation and support of the surrounding rock during the construction period, the TGS360Pro complete set of equipment was used at 0+053m on the right side of the main transformer room to conduct geological prediction of the rock structure morphology and surrounding rock classification 100m in front of the face (0+053m on the right side of the plant to 0+047m on the left side of the plant).

[0116] For the convenience of comparison, it is displayed in two-dimensional sections. The advanced geological prediction data is processed by the supporting software (i.e., multiple attribute data) as shown in the figure below. Figure 2 As shown, Figure 2 (a) to (h) are the eight main parameters (relative stress change Δσ, water content probability (water), longitudinal wave velocity V p , shear wave velocity, ratio of longitudinal and transverse wave velocity V p / V s , Poisson's ratio Poisson, dynamic Young's modulus E, hazard level category) related results, various parameter data points are mixed, and it is difficult to distinguish the potential geological risks they indicate. These data show a large overlap and redundancy in the image, and some parameters are even mutually exclusive, resulting in unclear boundaries between various geological hazards and failure to significantly display the stratification characteristics of the rock formations. On the other hand, the inherent similarities and differences in the data are difficult to identify, and it is difficult to accurately identify the geological structure and potential hazard types from each parameter.

[0117] The data of the eight attributes are extracted according to the grid size of 0.5m×0.5m. A total of (100÷0.5)×(30÷0.5)=12000 grid points are extracted for each attribute. According to formula (7), the single attribute weight coefficients of the eight attributes of relative stress change, water content probability, longitudinal wave velocity, transverse wave velocity, longitudinal and transverse wave velocity ratio, Poisson's ratio, dynamic Young's modulus, and hazard level are calculated as (0.268, 0.241, 0.375, 0.362, 0.154, 0.217, 0.353, 0.635). The comprehensive attribute value calculated by formula (8) is as follows after gridding: Figure 3 shown.

[0118] According to formula (9), the data after updating each attribute is as follows: Figure 4 As shown, Figure 4 (a) to (h) are the relevant results of the eight main attributes (relative stress change, water content probability, longitudinal wave velocity, transverse wave velocity, longitudinal and transverse wave velocity ratio, Poisson's ratio, dynamic Young's modulus, and hazard level). The organization and representation of the updated data of each attribute have changed significantly. The originally chaotic data now presents an orderly structure. The results of multi-attribute decision analysis can clearly assign data points to their respective groups, and each group represents a unique geological phenomenon or type of hidden danger. The similarities and differences between the results data have been strengthened, and the boundaries of rock formations and the scope of geological hazards have been clearly revealed.

[0119] The above are only a few embodiments of the present application and do not constitute any form of limitation to the present application. Although the present application is disclosed as above with preferred embodiments, it is not intended to limit the present application. Any technician familiar with the profession, without departing from the scope of the technical solution of the present application, using the technical content disclosed above to make slight changes or modifications are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A weighted analysis method for multi-attribute data in advanced geological prediction, characterized in that: The method comprises: S1. Normalizing the various attribute data obtained in the advanced geological prediction for each grid point in the target area to obtain the normalized data of each attribute in each grid point; S2. Determine the single attribute weight coefficient of each attribute in each grid point according to the normalized data of each attribute in all grid points, and determine the comprehensive attribute weight coefficient of each grid point according to the single attribute weight coefficient; S3. Determine the updated data of each attribute of each grid point using the normalized data and single attribute weight coefficient of each attribute in each grid point, as well as the comprehensive attribute weight coefficient of each grid point.

2. The method according to claim 1, characterized in that: The S1 specifically includes: S11, determining the points to be interpolated among all grid points in the target area under each attribute, and performing data filling processing on the points to be interpolated to obtain multiple attribute data of all grid points after processing; S12. Normalize the same attribute data of all grid points respectively to obtain normalized data of each attribute in each grid point.

3. The method according to claim 2, characterized in that The method used in the data filling process is cubic spline interpolation.

4. The method according to claim 3, characterized in that The step S11 of filling the data of the to-be-interpolated points is specifically as follows: The cubic spline interpolation method is used to interpolate the X-axis direction, Y-axis direction and Z-axis direction of the interpolation point in sequence.

5. The method according to claim 2, characterized in that: The S12 is specifically: Divide the various attribute data of each grid point into benefit-type attribute data and cost-type attribute data; The benefit-type attribute data and cost-type attribute data of all grid points are normalized respectively to obtain the normalized data of each attribute in each grid point.

6. The method according to claim 1, characterized in that Determining the single attribute weight coefficient of each attribute in each grid point in S2 specifically includes: Obtain the sum of the differences between the normalized data of each attribute of each grid point and the normalized data of the same attribute in each of the remaining grid points as the first sum, obtain the square value of the first sum corresponding to each attribute of each grid point, and obtain the square root of the sum of the square values ​​corresponding to all attributes of each grid point as the second sum; The sum of the absolute values ​​of the differences between the normalized data of each attribute of each grid point and the normalized data of the same attribute in each of the remaining grid points is obtained as the third sum, and the ratio of the third sum to the second sum is obtained as the single attribute weight coefficient of each attribute in each grid point.

7. The method according to claim 1, characterized in that Determining the comprehensive attribute weight coefficient of each grid point according to the single attribute weight coefficient in S2 is specifically: The sum of the products of the normalized data of all attributes of each grid point and its single attribute weight coefficient is obtained as the comprehensive attribute weight coefficient of each grid point.

8. The method according to claim 1, characterized in that The S3 specifically includes: The normalized data of each attribute in each grid point is adjusted by using the comprehensive attribute weight coefficient of each grid point to obtain the attenuation value of each attribute in each grid point; Determine the weighted adjustment value of each attribute in each grid point according to the single attribute weight coefficient of each attribute in each grid point and the comprehensive attribute weight coefficient of each grid point; Determine the minimum value adjustment of each attribute at each grid point according to the single attribute weight coefficient of each attribute at each grid point and the comprehensive attribute weight coefficient of each grid point; The sum of the attenuation value, the weighted adjustment value, and the minimum adjustment value of each attribute in each grid point is obtained as update data of each attribute of each grid point.

9. The method according to claim 1, characterized in that: The various attribute data for each grid point include longitudinal wave velocity, shear wave velocity, relative stress, water content probability, dynamic Young's modulus, Poisson's ratio, longitudinal and shear wave velocity ratio and hazard level.

10. The method according to claim 5, characterized in that The benefit-type attribute data include longitudinal wave velocity, transverse wave velocity and dynamic Young's modulus; The cost-type attribute data include relative stress, water content probability, Poisson's ratio, longitudinal and transverse wave velocity ratio and hazard level.

Images