Advanced geological forecast multi-source data fusion unfavorable geological body identification method

Abstract

The invention discloses an unfavorable geologic body identification method based on advanced geological forecast multi-source data fusion, and the method comprises the steps: obtaining multi-source data corresponding to a forecast mileage segment of any tunnel face, and constructing a target identification framework of geological forecast; constructing a basic probability distribution model of each data source based on the target recognition framework and multi-source data, and performing conflict detection on the multi-source data; and generating a comprehensive probability distribution model through iterative fusion, and analyzing a result according to the comprehensive probability distribution model. Therefore, according to the method provided by the embodiment of the invention, the uncertainty, the local contradiction conflict of the multi-source data and the rule normalization conflict in the detection of the unfavorable geologic body can be effectively processed, the conclusion is corrected by combining the multi-source data, and the misjudgment of single data is avoided. The credibility of each data source is quantified through the probability function, the weight distribution of the multi-source data is dynamically adjusted by using the conflict coefficient, the dependence on the prior probability is not needed, and the unfavorable geologic body recognition is more efficient, simple and convenient.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geological prediction, and particularly relates to a method for identifying adverse geological bodies by fusing multi-source data of advanced geological prediction. BACKGROUND

[0002] In order to ensure tunnel excavation safety and reduce the frequency of adverse geological disasters, a comprehensive advanced geological prediction technology combining various prediction methods has become a mainstream trend in advanced geological prediction work. However, the prediction data of various advanced geological prediction methods often have noise, resolution differences and local contradictions, and the identification and detection results of the same adverse geological body are inconsistent, which reduces the reliability and accuracy of adverse geological body identification and is not conducive to construction safety and efficiency. SUMMARY

[0003] The present application aims to overcome the shortcomings of the prior art and provide a method for identifying adverse geological bodies by fusing multi-source data of advanced geological prediction, which solves the problem of low reliability and accuracy of adverse geological body identification results by using multi-source data.

[0004] The purpose of the present application is achieved by the following technical solutions: A method for identifying adverse geological bodies by fusing multi-source data of advanced geological prediction, comprising: obtaining multi-source data corresponding to any tunnel face prediction mileage section, and constructing a target identification framework for geological prediction; constructing a basic probability distribution model for each data source based on the target identification framework and multi-source data, and detecting conflicts in the multi-source data; if the identification results of the multi-source data for the adverse geological body in front are highly inconsistent, adjusting the probability distribution model until the fusion condition is met, and if the target identification results of the multi-source data are basically consistent, then performing the subsequent steps; generating a comprehensive probability distribution model by iterative fusion, and analyzing and deciding the identification results of the adverse geological body in front of the tunnel face according to the comprehensive probability distribution model M'.

[0005] Further, the method specifically comprises the following steps: S1, based on an arbitrary tunnel face forecast mileage section, using multiple detection methods for advanced detection to obtain the multi-source data of the tunnel face forecast mileage section; S2, constructing a target identification framework of geological prediction, the target identification framework representing all possible hypothesis sets of adverse geological body types in front of the tunnel face; S3, based on the multi-source data obtained in S1 and the target identification framework obtained in S2, respectively constructing the basic probability distribution model corresponding to each data source in the multi-source data source; S4, sequentially selecting two data sources in the multi-source data for fusion in order to calculate the conflict coefficient k; S5, comparing the conflict coefficient k calculated in step S4 with a preset threshold K; if the conflict coefficient k > K, performing S6, and if k ≤ K, directly performing S7; S6, multiplying the basic probability distribution model M of the conflicting data source by a decay coefficient to perform probability redistribution, and repeating S4-S5 until k ≤ K is met; S7, establishing a multi-source data fusion normalization factor; S8, based on the normalization factor, iteratively fusing the multi-source data to generate a comprehensive probability distribution model; S9, according to the comprehensive probability distribution model M finally fused in step S8, analyzing and deciding the identification result of the adverse geological body in front of the tunnel face.

[0006] Further, the multi-source data includes at least two of elastic wave forecast data, transient electromagnetic forecast data, geological radar forecast data, and borehole geological data.

[0007] Further, the target identification framework includes a soft interlayer fracture zone R, a water-bearing fault fracture zone P, and a water-rich karst Y, and the target identification framework is .

[0008] Further, the conflict coefficient k is calculated in the following manner: conflict coefficient kji=, where i and j are positive integers, representing the ith data source and the jth data source.

[0009] Further, the decay coefficient a satisfies a∈(0,1), and the basic probability distribution model is adjusted as follows: ni(A) = mi (A) ×α,A∈{R,P,Y}; ni (Ω) = 1 - α × (1 - mi(Ω)) to obtain a new probability distribution Ni{ni (R);ni (P);ni (Y);ni (Ω)}.

[0010] Further, the normalization factor Q is: Q = 1-k’ where k’ is the conflict coefficient finally satisfying k’ ≤ K.

[0011] Further, the type of the adverse geological body with the highest probability in the comprehensive probability distribution model M' is selected as the identification result; if the highest probability value is greater than or equal to the confidence threshold C, the determination result is reliable; if the highest probability value is less than C, the supplementary detection data is re-performed in S1, and then S2-S8 are re-performed.

[0012] The beneficial effects of the present application are that the method of the embodiments of the present application can effectively deal with the uncertainty and local conflict of multi-source data in adverse geological body detection, rule normalization conflict, and combine multi-source data to correct the conclusion, thereby avoiding single data misjudgment. The reliability of each data source is quantified by a probability function, and the multi-source data weight distribution is dynamically adjusted by using a conflict coefficient, without relying on prior probability, so that the adverse geological body identification is more efficient and simple. The uncertainty interval is retained to objectively reflect the confidence level of the fusion result, and the reliability and accuracy of the adverse geological body identification are improved. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 A flowchart of a method for identifying an adverse geological body by multi-source data fusion in advance geological prediction according to an embodiment of the present application. DETAILED DESCRIPTION

[0014] The technical solutions of the present application will be described in detail below with reference to the embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0015] Reference Figure 1 The present application provides a technical solution: Before the description, some professional terms are explained: Face: The face is the most front working face in the tunneling process, that is, the interface between the excavated and unexcavated strata.

[0016] Face prediction mileage: refers to the distance range of predicting the geological conditions and hydrological conditions of the unexcavated area in front of the face (tunneling working face).

[0017] The method comprises the following steps: acquiring multi-source data corresponding to any tunnel face prediction mileage section, and constructing a target identification framework for geological prediction; constructing a basic probability distribution model of each data source based on the target identification framework and the multi-source data, and performing conflict detection on the multi-source data; if the identification results of the multi-source data on the front adverse geological body are highly inconsistent, adjusting the probability distribution model until the fusion condition is met, and if the target identification results of the multi-source data are basically consistent, performing subsequent steps; generating a comprehensive probability distribution model through iterative fusion, and analyzing and deciding the identification result of the adverse geological body in front of the face according to the comprehensive probability distribution model M.

[0018] It can be understood that the method of the embodiment of the application can effectively handle the uncertainty and local conflict of multi-source data in adverse geological body detection, normalize the conflict, correct the conclusion combined with multi-source data, and avoid single data misjudgment. By quantifying the reliability of each data source through a probability function and dynamically adjusting the weight distribution of multi-source data through a conflict coefficient, the method does not need to rely on prior probability, and the adverse geological body identification is more efficient and simple. Moreover, the uncertainty interval is retained to objectively reflect the confidence level of the fusion result, and the reliability and accuracy of the adverse geological body identification are improved.

[0019] Next, combined with Figure 1 For the understanding of some exemplary embodiments, the method specifically comprises the following steps in subsequent embodiments: S1, based on any face prediction mileage, using multiple detection methods for advanced detection, acquiring multi-source advanced geological prediction data of the face prediction mileage section, and subsequently simplifying the multi-source advanced geological prediction data as multi-source parameters for ease of description.

[0020] Specifically, for the same tunnel face prediction mileage section, multiple geophysical prospecting methods are used for advanced detection to obtain multi-source data of the mileage section. The aforementioned multi-source refers to a plurality of types of parameters, for example, the advanced geological prediction data generally includes elastic wave prediction data (such as TSP / TGP, referred to as TGS), transient electromagnetic prediction data, geological radar prediction data, and borehole geological data (providing direct geological information) of the mileage section. These multi-source data jointly constitute the basis for subsequent fusion analysis.

[0021] S2, constructing a geological prediction target identification framework for representing all possible hypothesis sets of adverse geological body types that may exist in front of the face.

[0022] In the method of the embodiment of the application, common adverse geological body types are mainly divided into three categories: soft interlayer fracture zone R, water-bearing fault fracture zone P, and water-rich karst Y. The target identification framework provides a clear object set for subsequent probability distribution.

[0023] Subsequently, taking three types of data sources of geological radar, transient electromagnetic method, and TGS as examples, soft interlayer fracture zone R, water-bearing fault fracture zone P, and water-rich karst Y are identified through data fusion, and comprehensive decision analysis is performed. The target identification framework is .

[0024] S3, based on the multi-source data obtained in S1, a basic probability assignment model is constructed for each data source, which is used for identifying the adverse geological body.

[0025] Specifically, for each type of advanced geological prediction data source obtained in step S1, such as geological radar data, transient electromagnetic data, TGS data, and drilling data, a basic probability assignment model (BPA) is constructed for each data source. The model quantifies the degree of trust of each data source for each subset in the target identification framework, such as a single adverse geological body type such as R, P, and Y. The basic probability assignment must satisfy: the sum of the probabilities assigned to all subsets in the framework is 1.

[0026] For example, given the basic probability assignment of the identification results of the three groups of data sources for target body identification: M1{[m1(R)=a11;m1(P)=a12;m1(Y)=a13;m1(Ω)=a14]} and a11+a12+a13+a14=1; M2{[m2(R)=a21;m2(P)=a22;m2(Y)=a23;m2(Ω)=a24]} and a21+a22+a23+a24=1; M3{[m3(R)=a31;m3(P)=a32;m3(Y)=a33;m3(Ω)=a34]} and a31+a32+a33+a34=1; M4{[m4(R)=a41;m4(P)=a42;m4(Y)=a43;m4(Ω)=a44]} and a41+a42+a43+a44=1.

[0027] S4, data fusion is performed on the multi-source data, and the conflict coefficient k between the multi-source data is calculated.

[0028] When performing data fusion, two groups of data sources to be fused are selected in order; for example, data sources M1 and M2 are selected first. The conflict coefficient k between the two groups of data sources is calculated. The conflict coefficient k reflects the degree of contradiction between the two groups of data sources for the target identification result.

[0029] For example, when calculating the conflict coefficient k12 of data source M1 (geological radar data source) and data source M2 (transient electromagnetic data source), the conflict coefficient k12= The empty set includes m1(R) and m2(P), m1(R) and m2(Y), m1(P) and m2(R), m1(P) and m2(Y), m1(Y) and m2(R), m1(Y) and m2(P), k12=a11×a22+a11×a23+a12×a21+a12×a23+a13×a21+a13×a22.

[0030] The conflict coefficient k is between [0,1]. The larger the value of k, the more serious the conflict between the two data sources. The specific judgment process is as shown in S5.

[0031] S5. Perform target consistency determination, that is, compare the conflict coefficient k calculated in step S4 with the preset threshold K: If the conflict coefficient k > threshold K, it indicates that the two sets of multi-source data have highly inconsistent identification results for the adverse geological bodies ahead, and cannot be directly fused. Conversely, if the conflict coefficient k ≤ threshold K, it is considered that the target identification results of the two sets of multi-source data are basically consistent, and can support subsequent fusion and target identification.

[0032] In other words, if the identified target results are highly consistent, then proceed to the next steps; otherwise, improve the basic probability model until the conflict coefficient k is less than the threshold K.

[0033] Specifically, based on the determination result of step S5: If k ≤ K (i.e., the goals are consistent), skip this step and proceed directly to step S7.

[0034] If k > K (i.e., the target heights are inconsistent), then S6 needs to be executed to “discount the credibility” of data sources with large conflicts, that is, reduce their weight in the fusion, so as to redistribute the basic probability and improve their basic probability allocation model.

[0035] In S6, the probability assignment values ​​of the basic probability assignment model are multiplied by a decay coefficient α, where α ∈ (0, 1). A smaller α value indicates lower reliability of the forecast results from that data source. Then, this decay coefficient is used to adjust the basic probability assignment. For example, multiplying M1 by the attenuation coefficient α: n1(R) = m1(R) × α; n1(P) = m1(P) × α; n1(Y) = m1(Y) × α; n1(Ω) = 1 - α × (1 - m1(Ω)); thus, a new probability assignment N1{n1(R); n1(P); n1(Y); n1(Ω)} is obtained.

[0036] Further, the conflict coefficient k is recalculated using the new probability distribution value instead of the original probability distribution (e.g., N1 instead of the original M1), and the above process of recalculating k is repeated until k ≤ K (e.g., a new k12 is calculated, and k12 is less than the threshold K). At this time, the improved basic probability distribution model conflict is within an acceptable range, and fusion can be performed. In order to distinguish, the conflict coefficient after S5-S6 processing is defined as k'.

[0037] S7, a multi-source data fusion normalization factor is established.

[0038] When k ≤ K, or after the two groups of data sources are adjusted in step S5, k ≤ K is satisfied, the normalization factor Q required for fusion is calculated based on the determined conflict coefficient k'. The normalization factor Q is used to eliminate the influence of the conflict part and ensure that the sum of the probability distribution after fusion is 1. Q is defined as: Q = 1 - k'.

[0039] S8, based on the normalization factor, the multi-source data is iteratively fused.

[0040] For example, the normalization factor Q calculated in step S7 is used to fuse the selected two groups of data sources (e.g., M1 and M2). For each non-empty subset (e.g., {R}, {P}, {Y}, {Ω}) in the target recognition framework, all intersection combinations supporting the subset are found, the sum of the products of the basic probability distribution values is calculated, and then divided by the normalization factor Q.

[0041] Specifically, the fusion result of each adverse geological body identification is calculated: For target recognition {R}, it is composed of non-empty intersection pairs m1(R) and m2(R), m1(R) and m2(Ω), and m1(Ω) and m2(R), and m12(R) = (a11×a21+a11×a24+a14×a21) / Q. For target recognition {P}, it is composed of non-empty intersection pairs m1(P) and m2(P), m1(P) and m2(Ω), and m1(Ω) and m2(P), and m12(R) = (a12×a22+a12×a24+a14×a22) / Q. For target recognition {Y}, it is composed of non-empty intersection pairs m1(Y) and m2(Y), m1(Y) and m2(Ω), and m1(Ω) and m2(Y), and m12(R) = (a13×a23+a13×a24+a14×a23) / Q. For target recognition {Ω}, it is composed of non-empty intersection pairs m1(Ω) and m2(Ω), and m12(Ω) = (a14×a24) / Q.

[0042] The fusion result forms a set of new basic probability assignments of the synthetic data source Mji, where i, j are positive integers, representing the ith data source and the jth data source, for example, M12= {m12(R), m12(P), m12(Y), m12(Ω)}.

[0043] Then, the new synthetic data source is taken as a whole as one of the input data sources of the next round of fusion. The processes of steps S4, S5, S6, S7, and S8 are repeatedly performed. Such iteration is performed until all original data sources (such as M1, M2, M3, and M4) participate in the fusion process. Finally, a target body identification probability model M' = {M(R), M(P), M(Y), M(Ω)} that fuses information of all multi-source data is obtained, which is defined as a comprehensive probability assignment model.

[0044] S9, comprehensive analysis and decision of the identification result.

[0045] According to the comprehensive probability assignment model M' obtained by the final fusion of step S8, the identification result of the adverse geological body in front of the working face is analyzed and decided. Generally, the adverse geological body type with the highest probability value is selected as the final identification result.

[0046] In addition, in order to ensure the reliability of the result, a confidence threshold C (C is usually in the interval [0.7, 1], and the specific value needs to be determined in combination with the actual geological background and risk requirements of the engineering tunnel) is set. If the highest probability value is greater than or equal to the threshold C, it is considered that the possibility of the existence of the adverse geological body in front of the working face is very high, and the prediction result is reliable, and the construction decision should be guided according to the result. If the highest probability value is lower than the threshold C, it indicates that the uncertainty of the fusion result is large, and the comprehensive judgment needs to be combined with other geological information or supplementary detection.

[0047] Therefore, the method of the embodiment of the application solves the comprehensive decision-making problem of multi-source data in adverse geological body prediction. The application introduces an uncertainty interval, models multiple methods or multi-source data of advanced geological prediction, quantifies the reliability of different sources of advanced geological data in predicting adverse geological bodies by using trust functions and likelihood functions, fuses multi-source data in turn, combines conflicts, constantly updates the trust degree of multi-source data of advanced geological prediction in identifying adverse geological bodies, and obtains a more reliable comprehensive judgment result of advanced geological prediction. The application provides a fast comprehensive prediction method for multi-source data of advanced geological prediction, and improves the accuracy and reliability of the prediction result.

[0048] The foregoing is considered as illustrative only of the principles of the application. Further, since numerous modifications and changes will readily occur to those skilled in the art, it is not desired to limit the application to the exact construction and operation described. Accordingly, all such variations are intended to be included within the scope of the present application as defined in the claims below and their equivalents.

Claims

1. A method for identifying adverse geological bodies by fusing multi-source data of advanced geological prediction, characterized in that, The method comprises the following steps: S1, based on any tunnel face prediction mileage section, using multiple detection methods for advanced detection, obtaining the multi-source data of the tunnel face prediction mileage section; S2, constructing the target recognition framework of geological prediction, the target recognition framework representing all possible hypothesis sets of the type of adverse geological body in front of the tunnel face; S3, based on the multi-source data obtained in S1 and the target recognition framework obtained in S2, constructing the basic probability distribution model corresponding to each data source in the multi-source data source respectively; 2. The method according to claim 1, wherein, S4, selecting two data sources in the multi-source data in sequence for fusion, calculating the conflict coefficient k; S5, comparing the conflict coefficient k calculated in step S4 with the preset threshold K; if the conflict coefficient k > K, execute S6, if k ≤ K, execute S7 directly; S6, multiplying the basic probability distribution model M of the conflicting data source by the attenuation coefficient to redistribute the probability, repeating S4-S5 until k ≤ K is met; S7, establishing the multi-source data fusion normalization factor; S8, based on the normalization factor, iteratively fusing the multi-source data to generate the comprehensive probability distribution model; S9, according to the comprehensive probability distribution model M finally fused in step S8, analyzing and deciding the recognition result of the adverse geological body in front of the tunnel face. The multi-source data includes at least two of elastic wave prediction data, transient electromagnetic prediction data, geological radar prediction data, and borehole geological data. The calculation method of the conflict coefficient k is: The conflict coefficient kji=, where i and j are positive integers, representing the ith data source and the jth data source. The attenuation coefficient a satisfies a∈(0,1), and the basic probability distribution model is adjusted as follows:

3. The method according to claim 2, wherein, ni(A) = mi (A) ×α, A∈{R,P,Y}; 4. The method according to claim 2, wherein, The target identification framework includes a weak interlayer fracture zone R, a water-bearing fault fracture zone P, and a water-rich karst Y, and the target identification framework is .

5. The method according to claim 4, wherein, ni (Ω) = 1 - α × (1 - mi(Ω)) Get the new probability distribution Ni{ni (R);ni (P);ni (Y);ni (Ω)}.

6. The method according to claim 5, wherein, The normalization factor Q is: Q = 1-k', where k' is the conflict coefficient that finally satisfies k' ≤ K. Select the adverse geological body type with the highest probability in the comprehensive probability distribution model M' as the recognition result; If the highest probability value ≥ confidence threshold C, the result is reliable; If the highest probability value < C, it is necessary to re-detect data in S1, and then re-perform S2-S8.

7. The method according to claim 5, wherein, ​ 8. The method according to claim 5, wherein the method is characterized by: ​ ​ ​

Images