Method and system for predicting deformation of surrounding rock of water diversion tunnel based on extreme random forest

Abstract

The present invention discloses a method and system for predicting deformation of surrounding rock of diversion tunnel based on extreme random forest, and the system includes a data acquisition module and an interval prediction module. First, the data acquisition module acquires the deformation history data of surrounding rock of diversion tunnel, and constructs a complete feature data set based on the deformation history data of surrounding rock of diversion tunnel. Then, the interval prediction module constructs a training set based on the complete feature data set, trains the extreme random forest model, and evaluates the nonlinear mapping relationship between the deformation influencing factors of surrounding rock of diversion tunnel and the deformation of surrounding rock of diversion tunnel through the optimal extreme random forest model obtained by training, so as to realize the interval prediction of deformation of surrounding rock of diversion tunnel. Compared with the traditional point prediction method, this method integrates the prediction results of multiple sub-random trees to construct a probability interval, so as to obtain a relatively stable evaluation interval prediction result, thereby improving the reliability of the interval prediction result of surrounding rock deformation of diversion tunnel.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing systems or methods specifically suitable for administrative, commercial, financial, management, supervision or prediction purposes, and in particular to a method and system for predicting deformation of surrounding rock of a water diversion tunnel based on extreme random forest. Background Art

[0002] The distribution of water resources in my country is extremely unbalanced. The construction of water conservancy projects, including water diversion tunnels, is an important infrastructure for coordinating water resources. As one of the important water conservancy projects for transporting water resources in complex mountainous areas, water diversion tunnels play a key role in shortening the distance of water transport in mountainous areas and improving economic benefits. Therefore, the safety of water diversion tunnels is related to the safety of people's production and life.

[0003] The deformation of the surrounding rock of the water diversion tunnel accompanies the entire process of its excavation and use, and is an important indicator for judging the health of the water diversion tunnel structure. The deformation of the surrounding rock of the water diversion tunnel can be predicted by using parameters with physical significance and engineering environment background as input for deformation point prediction. Most traditional surrounding rock prediction methods are based on point-to-point prediction based on existing monitoring curves. This method can provide high-precision prediction values, but it is difficult to provide uncertainty or degree of error in the prediction results. Interval prediction can give a range or confidence interval of the prediction results, provide uncertainty estimates about the prediction results, and help decision makers more accurately assess risks and formulate corresponding decision strategies. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention proposes a method and system for predicting deformation of surrounding rock of diversion tunnel based on extreme random forest, which can perform interval prediction on deformation of surrounding rock of diversion tunnel, thereby improving the reliability of prediction results. The specific technical scheme is as follows:

[0005] In a first aspect, a method for predicting deformation interval of surrounding rock of a diversion tunnel based on an extreme random forest model is provided. In a first implementable manner of the first aspect, the method includes:

[0006] Acquire the deformation history data of the surrounding rock of the water diversion tunnel, and acquire a complete characteristic data set through the deformation history data of the surrounding rock of the water diversion tunnel;

[0007] According to the complete characteristic data set, the extreme random forest model is used to predict the deformation range of the surrounding rock of the water diversion tunnel.

[0008] In combination with the first implementable manner of the first aspect, in a second implementable manner of the first aspect, obtaining a complete feature data set through the deformation history data of surrounding rock of the water diversion tunnel includes:

[0009] Extracting a feature data set from the deformation history data of the surrounding rock of the water diversion tunnel according to the influencing factors of the deformation of the surrounding rock of the water diversion tunnel during excavation and operation;

[0010] A missing data filling model is constructed through the feature data set, and the missing data in the feature data set is filled through the missing data filling model to obtain the complete feature data set.

[0011] In combination with the second implementable manner of the first aspect, in a third implementable manner of the first aspect, the feature data set includes: an excavation instantaneous effect feature set, a long-term creep feature set, a surrounding rock parameter feature set, and a tunneling parameter feature set.

[0012] In combination with the second implementable manner of the first aspect, in a fourth implementable manner of the first aspect, the missing data filling model is a KNN model based on feature correlation.

[0013] In combination with the fourth implementable manner of the first aspect, in the fifth implementable manner of the first aspect, the KNN model calculates the weighted distance between samples according to the correlation coefficients between different features, and selects the most similar sample data according to the weighted distance to estimate the missing data.

[0014] In combination with the first implementable manner of the first aspect, in a sixth implementable manner of the first aspect, an extreme random forest model is used to predict the deformation interval of surrounding rock of a diversion tunnel, including:

[0015] Constructing a training set and a validation set using the complete feature dataset;

[0016] The training set and the validation set are standardized respectively using the extreme value standardization method.

[0017] In combination with the first implementable manner of the first aspect, in a seventh implementable manner of the first aspect, it includes: verifying the trained extreme random forest model according to the root mean square error and the average error.

[0018] In combination with the first implementable manner of the first aspect, in an eighth implementable manner of the first aspect, an extreme random forest model is used to predict the deformation interval of surrounding rock of a water diversion tunnel, including:

[0019] According to the complete feature data set, multiple prediction results are obtained through the trained extreme random forest model;

[0020] All prediction results are arranged in order from small to large, and the deformation range of the surrounding rock of the water diversion tunnel is predicted based on the prediction results of different quantiles.

[0021] In a second aspect, a system for predicting deformation intervals of surrounding rock of a diversion tunnel based on an extreme random forest model is provided. In a first achievable manner of the second aspect, the system includes:

[0022] A data acquisition module is configured to acquire the deformation history data of the surrounding rock of the water diversion tunnel, and acquire a complete feature data set through the deformation history data of the surrounding rock of the water diversion tunnel;

[0023] The interval prediction module is configured to use an extreme random forest model to perform interval prediction of surrounding rock deformation of a water diversion tunnel based on the complete feature data set.

[0024] In combination with the first implementable manner of the second aspect, in a second implementable manner of the second aspect, the data acquisition module includes:

[0025] A data extraction unit configured to extract a characteristic data set from the deformation history data of the surrounding rock of the water diversion tunnel according to the influencing factors of the deformation of the surrounding rock of the water diversion tunnel during excavation and operation;

[0026] The data filling unit is configured to construct a missing data filling model through the feature data set, and fill the missing data in the feature data set through the missing data filling model to obtain the complete feature data set.

[0027] Beneficial effects:

[0028] 1. The extreme random forest prediction model based on the integrated algorithm is used to evaluate the nonlinear mapping relationship between the influencing factors of the diversion tunnel surrounding rock deformation and the diversion tunnel surrounding rock deformation, and to achieve the interval prediction of the diversion tunnel surrounding rock deformation. Compared with the traditional point prediction method, this method integrates the prediction results of multiple sub-random trees to construct a probability interval to obtain a relatively stable evaluation interval prediction result, which improves the reliability of the diversion tunnel surrounding rock deformation prediction results.

[0029] 2. By considering and constructing the instantaneous deformation effect and long-term creep effect during the excavation of the diversion tunnel, a characteristic set of deformation factors of the diversion tunnel surrounding rock with clear physical meaning is established, which makes up for the weak correlation of influencing factors in the traditional prediction of diversion tunnel surrounding rock deformation and the lack of timeliness of the diversion tunnel surrounding rock deformation.

[0030] 3. The present invention calculates the weighted distance between samples based on the correlation between different features, thereby taking into account the feature correlation to estimate the missing values ​​in the feature set, overcoming the problem that the traditional KNN filling algorithm cannot consider the impact of each feature on the missing feature. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the specific implementation of the present invention, the following will briefly introduce the drawings required for use in the specific implementation. In all the drawings, each element or part is not necessarily drawn according to the actual scale.

[0032] Figure 1 A flow chart of a method for predicting deformation intervals of surrounding rock of a diversion tunnel based on an extreme random forest model provided by an embodiment of the present invention;

[0033] Figure 2 A flowchart of a missing data filling method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.

[0035] like Figure 1 The flowchart of the method for predicting deformation interval of surrounding rock of diversion tunnel based on extreme random forest model is shown in FIG. 1 , and the prediction method includes:

[0036] Step 1: Obtaining the deformation history data of the surrounding rock of the water diversion tunnel, and obtaining a complete feature data set through the deformation history data of the surrounding rock of the water diversion tunnel;

[0037] Step 2: Based on the complete feature data set, an extreme random forest model is used to predict the deformation interval of the surrounding rock of the water diversion tunnel.

[0038] Specifically, first, the deformation history data of the surrounding rock of the water diversion tunnel can be obtained, and a corresponding complete feature data set can be constructed based on the deformation history data of the surrounding rock of the water diversion tunnel. Then, a training set and a validation set can be constructed based on the complete feature data set, and the extreme random forest model can be trained by the training set, and the extreme random forest model in the training process can be validated by the validation set, and finally the optimal extreme random forest model can be obtained.

[0039] The optimal extreme random forest model can be used to evaluate the nonlinear mapping relationship between the influencing factors of the diversion tunnel surrounding rock deformation and the diversion tunnel surrounding rock deformation, thereby realizing the interval prediction of the diversion tunnel surrounding rock deformation. Compared with the traditional point prediction method, this method integrates the prediction results of multiple sub-random trees to construct a probability interval to obtain a relatively stable evaluation interval prediction result, which improves the reliability of the diversion tunnel surrounding rock deformation prediction results.

[0040] In this embodiment, optionally, in step 1, obtaining a complete feature data set through the deformation history data of the surrounding rock of the water diversion tunnel includes:

[0041] Extracting a feature data set from the deformation history data of the surrounding rock of the water diversion tunnel according to the influencing factors of the deformation of the surrounding rock of the water diversion tunnel during excavation and operation;

[0042] A missing data filling model is constructed through the feature data set, and the missing data in the feature data set is filled through the missing data filling model to obtain the complete feature data set.

[0043] Specifically, firstly, a feature data set can be extracted from the historical data of the surrounding rock deformation of the water diversion tunnel according to the influencing factors of the surrounding rock deformation of the water diversion tunnel during excavation and operation. Then, a missing data filling model can be constructed through the feature data set, and the missing data in the feature data set can be estimated through the data filling model, so as to supplement the missing data in the feature data set and obtain complete feature data.

[0044] In this embodiment, the optional feature data sets include: an excavation instantaneous effect feature set, a long-term creep feature set, a surrounding rock parameter feature set, and a tunneling parameter feature set.

[0045] Specifically, the constructed feature data sets include excavation transient effect feature set, long-term creep feature set, surrounding rock parameter feature set and tunneling parameter feature set.

[0046] Among them, the excavation transient effect feature set is Where L is the distance between the section and the tunnel face, and R1 is the influence radius of the tunnel face. The influence radius of the tunnel face depends on the properties of the surrounding rock and the construction method, and can be set at 0.75 times the design radius of the tunnel.

[0047] Long-term creep characteristic set Y s ={T 1 , T 2}, where T 1 is the first creep characteristic, T 2 is the second creep feature. Specifically, the creep feature in the long-term creep feature set can be determined according to the following calculation formula:

[0048]

[0049]

[0050] Among them, T t is the creep factor at time t, τ j is a time between 0 and t, Δτ j is the time interval between two adjacent moments.

[0051] Surrounding rock parameter feature set Y r={σ, d, E, ρ, c, ω}, where σ is the surrounding rock strength, d is the rock layer thickness, E is the soil elastic modulus, ρ is the soil density, c is the soil cohesion, and ω is the soil moisture content. Excavation parameter feature set Y c ={v,u}, where v is the excavation speed and u is the surrounding rock reinforcement strength.

[0052] When constructing the characteristic data set, the instantaneous deformation effect and long-term creep effect during the excavation of the diversion tunnel are considered. The characteristic data set established is a characteristic set of deformation factors of the diversion tunnel surrounding rock with clear physical meaning, which makes up for the weak correlation of influencing factors in the traditional prediction of diversion tunnel surrounding rock deformation and the lack of timeliness of the diversion tunnel surrounding rock deformation.

[0053] In this embodiment, optionally, the missing data filling model is a KNN model based on feature correlation. Specifically, the missing data in the feature data set can be supplemented by using the KNN model as a missing data filling model. Figure 2 As shown, the specific steps are as follows:

[0054] Step S11, filtering out a complete sample set containing only complete records from the feature data set;

[0055] Step S12: Calculate the Pearson correlation coefficient between different features based on the complete sample set. The calculation method is as follows:

[0056]

[0057] where r p,q is the correlation coefficient between feature p and feature q, Cov(x p ,x q ) is the covariance between feature p and feature q, Var(x p ) and Var(x q ) are the variances of feature p and feature q respectively;

[0058] Step S13: For a sample i in the feature data set whose missing feature is p, the weighted distance between the sample i and the ath sample in the complete sample set is calculated according to the following formula:

[0059]

[0060] Among them, n1 is the number of features without missing values ​​in sample i, r p,w is the correlation coefficient between the missing value feature p and the non-missing value feature w, which can be determined by the Pearson correlation coefficient between different features calculated in step S12. i,w is the feature data corresponding to the feature w with no missing values ​​in sample i, x j,wIt is the feature data corresponding to the feature w without missing values ​​in the ath sample in the complete sample set.

[0061] In this way, the weighted distance between samples can be calculated based on the correlation between different features, so that the missing values ​​in the feature set can be estimated by taking into account the feature correlation, overcoming the problem that the traditional KNN filling algorithm cannot consider the impact of each feature on the missing feature.

[0062] Step S14: for a sample i in the feature set, feature p is missing, x i p , calculate the weighted distance between it and all complete samples and sort them, then take multiple complete samples with the smallest weighted distance to build the estimated data set X k =[X1,X2,…,X k ];

[0063] Step S15: Estimate missing data x by estimating the data set i p , the estimation method is as follows:

[0064]

[0065] In this embodiment, the number of samples k to be selected may be determined according to the amount of missing data in the feature data set.

[0066] In this embodiment, optionally, in step 2, an extreme random forest model is used to predict the deformation interval of the surrounding rock of the water diversion tunnel, including:

[0067] Constructing a training set and a validation set using the complete feature dataset;

[0068] The training set and the validation set are standardized respectively using the extreme value standardization method.

[0069] Specifically, after constructing a complete feature data set, first, feature data samples can be randomly extracted from the complete feature data set in a ratio of 8:2 to construct a training set and a validation set. Then, the extreme value normalization method can be used to perform linear transformation on the samples in the training set and the validation set, respectively, so as to standardize and map the sample data to the set extreme value interval, which is used to eliminate the influence of variables of different dimensions on parameter learning.

[0070] In this embodiment, optionally, during the training of the extreme random forest model, the trained extreme random forest model can be verified according to the root mean square error and the average error. The specific calculation formula is as follows:

[0071]

[0072]

[0073] Among them, m is the number of sub-random trees in the extreme random forest model, is the result of the b-th random tree for the s-th sample, y s is the true value of the sth sample, n is the number of samples in the validation set, Errors is the root mean square error of the prediction result of the sub-random tree in the extreme random forest model for sample s, and Error is the average error of the extreme random forest model for the entire validation set samples.

[0074] In this embodiment, optionally, after determining the optimal extreme random forest model according to the root mean square error and the average error, multiple prediction results can be obtained according to the complete feature data set through the trained extreme random forest model. Then, all the prediction results are arranged in order from small to large, and the deformation interval of the surrounding rock of the water diversion tunnel is predicted according to the prediction results of different quantiles, so as to integrate the prediction results of multiple sub-random trees to construct a probability interval to obtain a relatively stable evaluation interval prediction result, thereby improving the reliability of the prediction results of the deformation interval of the surrounding rock of the water diversion tunnel. In this embodiment, the prediction results of the 95% and 5% quantiles can be determined as the upper and lower limits of the deformation interval of the surrounding rock of the water diversion tunnel.

[0075] A water diversion tunnel surrounding rock deformation interval prediction system based on extreme random forest model, including:

[0076] A data acquisition module is configured to acquire the deformation history data of the surrounding rock of the water diversion tunnel, and acquire a complete feature data set through the deformation history data of the surrounding rock of the water diversion tunnel;

[0077] The interval prediction module is configured to use an extreme random forest model to perform interval prediction of surrounding rock deformation of a water diversion tunnel based on the complete feature data set.

[0078] Specifically, the prediction system includes a data acquisition module and an interval prediction module, wherein the data acquisition module can obtain the historical data of the surrounding rock deformation of the water diversion tunnel, and construct a corresponding complete feature data set based on the historical data of the surrounding rock deformation of the water diversion tunnel. The interval prediction module can construct a training set and a verification set based on the complete feature data set, train the extreme random forest model through the training set, and verify the extreme random forest model during the training process through the verification set, and finally obtain the optimal extreme random forest model. The nonlinear mapping relationship between the influencing factors of the surrounding rock deformation of the water diversion tunnel and the deformation of the surrounding rock of the water diversion tunnel can be evaluated through the optimal extreme random forest model, thereby realizing the interval prediction of the surrounding rock deformation of the water diversion tunnel. Compared with the traditional point prediction method, this method integrates the prediction results of multiple sub-random trees to construct a probability interval to obtain a relatively stable evaluation interval prediction result, thereby improving the reliability of the interval prediction result of the surrounding rock deformation of the water diversion tunnel.

[0079] In this embodiment, optionally, the data acquisition module includes:

[0080] A data extraction unit configured to extract a characteristic data set from the deformation history data of the surrounding rock of the water diversion tunnel according to the influencing factors of the deformation of the surrounding rock of the water diversion tunnel during excavation and operation;

[0081] The data filling unit is configured to construct a missing data filling model through the feature data set, and fill the missing data in the feature data set through the missing data filling model to obtain the complete feature data set.

[0082] Specifically, the data acquisition module is composed of a data extraction unit and a data filling unit. The data extraction unit can extract a feature data set from the historical data of the surrounding rock deformation of the water diversion tunnel according to the influencing factors of the surrounding rock deformation of the water diversion tunnel during excavation and operation. The data filling unit can construct a missing data filling model through the feature data set, and estimate the missing data in the feature data set through the data filling model, thereby supplementing the missing data in the feature data set to obtain complete feature data.

[0083] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A method for predicting deformation interval of surrounding rock of diversion tunnel based on extreme random forest model, characterized in that: include: Acquire the deformation history data of the surrounding rock of the water diversion tunnel, and acquire a complete characteristic data set through the deformation history data of the surrounding rock of the water diversion tunnel; Based on the complete characteristic data set, the extreme random forest model is used to predict the deformation interval of the surrounding rock of the water diversion tunnel, including: According to the complete feature data set, multiple prediction results are obtained through the trained extreme random forest model; Arrange all prediction results in order from small to large, and predict the deformation range of the surrounding rock of the water diversion tunnel according to the prediction results of different quantiles; A complete characteristic data set is obtained through the deformation history data of the surrounding rock of the water diversion tunnel, including: Extracting a feature data set from the deformation history data of the surrounding rock of the water diversion tunnel according to the influencing factors of the deformation of the surrounding rock of the water diversion tunnel during excavation and operation; Building a missing data filling model through the feature data set, and filling the missing data in the feature data set through the missing data filling model to obtain the complete feature data set; The feature data set includes: excavation instantaneous effect feature set, long-term creep feature set, surrounding rock parameter feature set and tunneling parameter feature set; Among them, the excavation transient effect feature set is Among them, L is the distance between the section and the tunnel face, and R1 is the influence radius of the tunnel face; Long-term creep characteristic set Y s ={T 1 , T 2 }, where T 1 is the first creep characteristic, T 2 is the second creep feature, and the creep feature in the long-term creep feature set is determined according to the following calculation formula: Among them, T t is the creep factor at time t, is a time between 0 and t, is the time interval between two adjacent moments; Surrounding rock parameter feature set Y r ={σ, d, E, ρ, c, ω}, where σ is the surrounding rock strength, d is the rock layer thickness, E is the soil elastic modulus, ρ is the soil density, c is the soil cohesion, and ω is the soil moisture content; the excavation parameter feature set Y c ={v,u}, where v is the excavation speed and u is the surrounding rock reinforcement strength.

2. The method for predicting deformation interval of surrounding rock of a diversion tunnel according to claim 1, characterized in that: The missing data filling model is a KNN model based on feature correlation.

3. The method for predicting deformation interval of surrounding rock of diversion tunnel according to claim 2, characterized in that: The KNN model calculates the weighted distance between samples according to the correlation coefficients between different features, and selects the most similar sample data according to the weighted distance to estimate the missing data.

4. The method for predicting deformation interval of surrounding rock of a diversion tunnel according to claim 1, characterized in that: The extreme random forest model is used to predict the deformation range of the surrounding rock of the diversion tunnel, including: Constructing a training set and a validation set using the complete feature dataset; The training set and the validation set are standardized respectively using the extreme value standardization method.

5. The method for predicting deformation interval of surrounding rock of diversion tunnel according to claim 1, characterized in that: include: The trained extreme random forest model was verified based on the root mean square error and the average error.

6. A water diversion tunnel surrounding rock deformation interval prediction system based on extreme random forest model, characterized in that: include: A data acquisition module is configured to acquire the deformation history data of the surrounding rock of the water diversion tunnel, and acquire a complete feature data set through the deformation history data of the surrounding rock of the water diversion tunnel; The interval prediction module is configured to use the extreme random forest model to perform interval prediction of surrounding rock deformation of the water diversion tunnel according to the complete feature data set, including: According to the complete feature data set, multiple prediction results are obtained through the trained extreme random forest model; Arrange all prediction results in order from small to large, and predict the deformation range of the surrounding rock of the water diversion tunnel according to the prediction results of different quantiles; The data acquisition module comprises: A data extraction unit configured to extract a characteristic data set from the deformation history data of the surrounding rock of the water diversion tunnel according to the influencing factors of the deformation of the surrounding rock of the water diversion tunnel during excavation and operation; a data filling unit, configured to construct a missing data filling model through the feature data set, and fill the missing data in the feature data set through the missing data filling model to obtain the complete feature data set; The feature data set includes: excavation instantaneous effect feature set, long-term creep feature set, surrounding rock parameter feature set and tunneling parameter feature set; Among them, the excavation transient effect feature set is Among them, L is the distance between the section and the tunnel face, and R1 is the influence radius of the tunnel face; Long-term creep characteristic set Y s ={T 1 , T 2 }, where T 1 is the first creep characteristic, T 2 is the second creep feature, and the creep feature in the long-term creep feature set is determined according to the following calculation formula: Among them, T t is the creep factor at time t, is a time between 0 and t, is the time interval between two adjacent moments; The surrounding rock parameter characteristic set Yr = {σ, d, E, ρ, c, ω}, where σ is the surrounding rock strength, d is the rock layer thickness, E is the soil elastic modulus, ρ is the soil density, c is the soil cohesion, and ω is the soil moisture content; the excavation parameter characteristic set Y c ={v,u}, where v is the excavation speed and u is the surrounding rock reinforcement strength.

Images