A tunnel front geological anomaly intelligent identification method and system based on multi-source data fusion

Abstract

This invention relates to an intelligent identification method and system for geological anomalies ahead of tunnels based on multi-source data fusion, belonging to the field of advanced geological prediction technology for tunnel engineering. The invention first constructs a two-level decision tree geological hypothesis set and assigns prior probabilities. Then, it establishes a likelihood matrix of observation characteristics from different geophysical methods. After obtaining multi-source geophysical evidence from the field, it uses Bayes' theorem for multiple rounds of dynamic iterative updates, outputting the quantitative confidence percentage of various geological anomalies. The system can also adaptively adjust model parameters based on the actual geological conditions revealed during excavation. This invention overcomes the shortcomings of traditional prediction methods, such as strong ambiguity and high dependence on human experience, achieving probabilistic intelligent identification of geological anomalies and providing clear and intuitive data support for safe tunnel construction and engineering decision-making.

Description

Technical Field

[0001] This invention relates to an intelligent identification method and system for geological anomalies ahead of tunnels based on multi-source data fusion, belonging to the field of advanced geological prediction technology for tunnel engineering. Background Technology

[0002] Accurate prediction of geological conditions ahead of the tunnel face is crucial for construction safety, efficiency, and cost control during tunnel excavation. In particular, the presence of adverse geological formations such as fault fracture zones, water-rich areas, and karst can easily trigger serious geological disasters such as sudden water and mud inrushes and landslides.

[0003] Currently, commonly used advanced geological prediction methods mainly rely on single geophysical methods, such as ground-penetrating radar (GPR), transient electromagnetic method (TEM), or seismic wave method (TSP). However, existing technologies have the following significant drawbacks: 1. High ambiguity and reliance on subjective experience: Different geological bodies may produce similar geophysical anomaly responses. Current "multi-source fusion" often involves technicians manually comparing several drawings, heavily relying on subjective experience and lacking objective mathematical evaluation standards. 2. Lack of quantitative confidence output: Existing methods typically only provide a qualitative conclusion that "anomalies may exist ahead," failing to provide precise probabilistic data support, leading to difficulties in engineering decisions (such as whether to incur high costs for advanced drilling verification). 3. Lack of dynamic learning and error correction mechanisms: As tunnel excavation progresses, the actual geological conditions revealed at the tunnel face cannot be systematically fed back to the prediction model, resulting in the prediction accuracy failing to adaptively improve with the construction progress. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention provides a method and system for intelligent identification of geological anomalies ahead of tunnels based on multi-source data fusion. By constructing a two-level decision tree and a Bayesian likelihood matrix, it realizes the quantitative confidence output of the geological conditions ahead and supports dynamic correction and updates.

[0005] The technical solution of this invention is: an intelligent identification method for geological anomalies ahead of tunnels based on multi-source data fusion, the specific steps of which are as follows: Step S1: Establish a set of geological hypotheses and a prior probability model. Construct a set of geological hypotheses for the possible geological conditions in front of the tunnel face into a two-level decision tree structure containing "background intact rock mass" and "existence of anomalies", and assign initial prior probabilities. Step S2: Construct the evidence-hypothesis likelihood matrix. For each geological hypothesis, determine the conditional probability of the likelihood matrix of physical evidence observed by different geophysical methods. Step S3: Dynamic Bayesian update and inference of multi-source evidence. When new geophysical evidence is obtained, the prior probability is updated iteratively using the Bayesian formula. The posterior probability of the previous round is used as the new prior probability for introducing the next geophysical evidence. Step S4: Output the quantitative identification results, output the posterior probability distribution after fusing all evidence, and quantitatively display the confidence level of various geological conditions in percentage form.

[0006] Furthermore, in the two-level decision tree structure in step S1, the second-level decision performs geological subdivision on the set of geological hypotheses with anomalies. The subdivision items include at least three of the following: "dry micro-fractures", "water-bearing micro-fractures", "lithological gradient zones", and "non-dense fracture zones".

[0007] Furthermore, the geophysical exploration method in step S2 includes at least two of the following: ground-penetrating radar, transient electromagnetic method, and seismic wave method; the physical evidence is obtained by feature discretization extraction of continuous geophysical signals by setting a threshold.

[0008] Furthermore, the physical evidence includes the electromagnetic wave reflection intensity characteristics of ground-penetrating radar and the apparent resistivity characteristics of transient electromagnetic methods.

[0009] Furthermore, step S4 also includes a model adaptive dynamic correction mechanism: after the tunnel excavation reveals the actual geological conditions ahead, the actual geological data is used as the truth input system to dynamically update the initial prior probability in step S1 and the likelihood matrix conditional probability in step S2.

[0010] The present invention also provides an intelligent identification system for geological anomalies ahead of tunnels based on multi-source data fusion, the system comprising: a module for executing the intelligent identification method for geological anomalies ahead of tunnels based on multi-source data fusion.

[0011] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the aforementioned intelligent identification method for geological anomalies ahead of tunnels based on multi-source data fusion.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Achieve objective quantification of predictions: Abandoning the traditional qualitative experience-based judgment that relies on manual map reading, a Bayesian inference model is introduced to output the confidence level of geological anomalies as a precise percentage, providing intuitive and reliable data support for engineering decisions.

[0013] 2. Two-level decision tree reduces computational dimensionality: It is the first to create a two-level hypothesis set structure that first distinguishes between complete and abnormal, and then further subdivides the abnormality type. This effectively filters out invalid calculations and improves the computational efficiency of real-time on-site forecasting.

[0014] 3. The model has "self-growth" capability: Through a dynamic update mechanism, the model can absorb the "truth" revealed after each blasting excavation, continuously correct its prior probability and likelihood matrix, so that the prediction accuracy continues to improve as the tunnel excavation depth increases. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall process steps of the intelligent recognition method described in this invention.

[0016] Figure 2 This is a schematic diagram of the two-level decision tree structure of the geological hypothesis set established for this invention.

[0017] Figure 3 A schematic diagram illustrating the principle of dynamic Bayesian iterative update for multi-source geophysical evidence. Detailed Implementation

[0018] Example 1: As Figures 1-3 As shown, a method for intelligent identification of geological anomalies ahead of tunnels based on multi-source data fusion is described, and the specific steps of the method are as follows: Step S1: Establish a set of geological hypotheses and a prior probability model. A two-level decision tree structure is constructed to represent the possible geological conditions within the predicted section ahead of the tunnel face as a set of geological hypotheses H. The first level of decision distinguishes between "background intact rock mass H0" and "existence of anomalies Hx". The second level of decision further subdivides "existence of anomalies Hx" into geological subdivisions, including but not limited to: "dry micro-fractures H1", "water-bearing micro-fractures H2", "lithological gradient zone H3", and "loose fracture zone H4". Based on the geological logging statistics of the excavated section of the tunnel and regional geological data, an initial prior probability P(Hi) is assigned to each geological hypothesis in the hypothesis set, satisfying that the sum of all probabilities is 1.

[0019] Step S2: Construct the evidence-hypothesis likelihood matrix. For each geological hypothesis Hi in the hypothesis set, determine the likelihood matrix conditional probability P(Ej|Hi) of physical evidence Ej observed by different geophysical methods, forming a likelihood matrix; the physical evidence Ej is obtained by threshold discretization of the electromagnetic wave reflection intensity of ground-penetrating radar, the apparent resistivity characteristics of transient electromagnetic methods, and the seismic wave reflection characteristics; the conditional probability is quantified based on geophysical exploration theory and a historical engineering case library.

[0020] Step S3: Dynamic Bayesian Update and Inference of Multi-Source Evidence. Multi-source geophysical data is collected at the tunnel face. When the first type of geophysical evidence E1 is obtained, the prior probability is updated using Bayes' theorem to obtain the posterior probability P(Hi|E1). When the second type of geophysical evidence E2 is obtained, the posterior probability P(Hi|E1) obtained from the first update is used as the new prior probability, and combined with the likelihood value P(E2|Hi) of evidence E2 for a second iteration update. This process continues, continuously updating the confidence level of each geological hypothesis through the sequential input of multi-source evidence.

[0021] Step S4: Output quantified identification results and adaptive dynamic correction. Output the posterior probability distribution after fusing all evidence. Quantify and display the confidence level of various geological conditions as a percentage; execute corresponding engineering verification or construction intervention measures based on the confidence level assessment results; when tunnel excavation reveals the actual geological conditions ahead, feed the actual geological data as the truth value back to steps S1 and S2, and adaptively correct the prior probability and likelihood matrix of the excavated section.

[0022] Furthermore, step S4 also includes a model adaptive dynamic correction mechanism: after the tunnel excavation reveals the actual geological conditions ahead, the actual geological data is used as the truth input system to dynamically update the initial prior probability in step S1 and the likelihood matrix conditional probability in step S2.

[0023] This invention also provides an intelligent identification system for geological anomalies ahead of tunnels based on multi-source data fusion, the system comprising: Establishment Module: Used to establish a set of geological hypotheses and a prior probability model. It constructs a set of geological hypotheses about the possible geological conditions in front of the tunnel face into a two-level decision tree structure containing "background intact rock mass" and "existence of anomalies", and assigns initial prior probabilities. The building module is used to construct the evidence-hypothesis likelihood matrix, and for each geological hypothesis, it determines the conditional probability of the likelihood matrix of physical evidence observed by different geophysical methods. Update Inference Module: Used for dynamic Bayesian update and inference of multi-source evidence. When new geophysical evidence is obtained, the prior probability is updated iteratively using Bayes' formula. The posterior probability of the previous round is used as the new prior probability for introducing the next geophysical evidence. Output module: Outputs the quantitative recognition results and the posterior probability distribution after fusing all evidence, and displays the confidence level of various geological conditions in percentage form.

[0024] Example 2: The present invention also provides an intelligent identification system for geological anomalies ahead of tunnels based on multi-source data fusion, the system comprising: The prior model building module is used to build a two-level decision tree for the geological hypothesis set and configure the prior probabilities; The likelihood matrix storage module is used to store the conditional probabilities of observation evidence from different geophysical methods; The Bayesian iterative inference module is used to receive multi-source geophysical input signals from the field and perform sequential iterative update calculations of probabilities according to the Bayesian formula. The confidence output and feedback error correction module is used to output the quantitative probability percentage of geological anomalies on the human-computer interaction interface and to receive real geological feedback after tunneling to update system parameters.

[0025] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the aforementioned intelligent identification method for geological anomalies ahead of tunnels based on multi-source data fusion.

[0026] Example 3: An intelligent identification method for geological anomalies ahead of a tunnel based on multi-source data fusion. A mountain tunnel is constructed using the drill-and-blast method. A 30-meter advance geological prediction is required ahead of the tunnel face. The specific steps of the method are as follows: S1: Establish the hypothesis set and prior probabilities. A two-level decision tree is established, containing H0 (intact), H1 (dry microfractures), H2 (water-bearing microfractures), H3 (lithological gradient zone), and H4 (loose zone). Based on statistics from the first 1000 meters of excavated sections, the prior probabilities for this round are set as follows: P(H0) = 0.60; P(H1) = 0.10; P(H2) = 0.15; P(H3) = 0.05; P(H4) = 0.10.

[0027] S2: Construct the likelihood matrix. Extract historical data and set conditional probabilities (taking ground-penetrating radar evidence E1 = weak reflection and transient electromagnetic evidence E2 = significantly low impedance as an example): For radar E1: P(E1|H0) = 0.1, P(E1|H2) = 0.8; For electromagnetic E2: P(E2|H0) = 0.05, P(E2|H2) = 0.9.

[0028] S3: Dynamic Bayesian Update Calculation. First Update (Incorporating Ground Penetrating Radar Evidence E1): The intermediate posterior probability including E1 is calculated using the Bayesian formula. At this point, due to the detection of "weak reflection," the probability of H2 increases from 0.15 to approximately 0.45. Second Update (Incorporating Transient Electromagnetic Evidence E2): The result of 0.45 from the first round is used as the new prior probability and substituted into the electromagnetic evidence E2 (low resistivity) for the second round of Bayesian calculation. After this iteration, the final confidence level for "water-bearing microfractures (H2)" soars to 93.5%.

[0029] S4: Decision Making and Dynamic Correction. The system interface directly outputs: "Warning: A water-bearing micro-fracture exists ahead of the tunnel face (H2), confidence level: 93.5%". After the water probe confirms the H2 hypothesis, the system captures this real feedback and automatically increases the initial prior probability of H2 in the next round of prediction, realizing the adaptive evolution of the algorithm.

[0030] The specific embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for intelligent identification of geological anomalies ahead of tunnels based on multi-source data fusion, characterized in that, The specific steps of the method are as follows: Step S1: Establish a set of geological hypotheses and a prior probability model. Construct a set of geological hypotheses for the possible geological conditions in front of the tunnel face into a two-level decision tree structure containing "background intact rock mass" and "existence of anomalies", and assign initial prior probabilities. Step S2: Construct the evidence-hypothesis likelihood matrix. For each geological hypothesis, determine the conditional probability of the likelihood matrix of physical evidence observed by different geophysical methods. Step S3: Dynamic Bayesian update and inference of multi-source evidence. When new geophysical evidence is obtained, the prior probability is updated iteratively using the Bayesian formula. The posterior probability of the previous round is used as the new prior probability for introducing the next geophysical evidence. Step S4: Output the quantitative identification results, output the posterior probability distribution after fusing all evidence, and quantitatively display the confidence level of various geological conditions in percentage form.

2. The intelligent identification method for geological anomalies ahead of tunnels based on multi-source data fusion according to claim 1, characterized in that: In the two-level decision tree structure in step S1, the second-level decision subdivides the set of geological hypotheses with anomalies. The subdivision items include at least three of the following: "dry micro-fractures", "water-bearing micro-fractures", "lithological gradient zones", and "non-dense fracture zones".

3. The intelligent identification method for geological anomalies ahead of tunnels based on multi-source data fusion according to claim 1, characterized in that: The geophysical exploration method in step S2 includes at least two of the following: ground-penetrating radar, transient electromagnetic method, and seismic wave method; the physical evidence is obtained by discretizing and extracting features from the continuous geophysical signal by setting a threshold.

4. The intelligent identification method for geological anomalies ahead of tunnels based on multi-source data fusion according to claim 3, characterized in that: The physical evidence includes the electromagnetic wave reflection intensity characteristics of ground-penetrating radar and the apparent resistivity characteristics of transient electromagnetic methods.

5. The intelligent identification method for geological anomalies ahead of tunnels based on multi-source data fusion according to claim 1, characterized in that: Step S4 also includes a model adaptive dynamic correction mechanism: after the tunnel excavation reveals the actual geological conditions ahead, the actual geological data is used as the truth input system to dynamically update the initial prior probability in step S1 and the likelihood matrix conditional probability in step S2.

6. A low-rank adaptive fine-tuning system based on trainable parameter projection constraints, characterized in that, The system includes a module for executing the intelligent identification method for geological anomalies ahead of tunnels based on multi-source data fusion as described in any one of claims 1-5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the intelligent identification method for geological anomalies ahead of tunnels based on multi-source data fusion as described in any one of claims 1-5.

Images