Tunnel engineering advanced geology forecasting method and system based on multi-source information fusion

Abstract

The invention provides a tunnel engineering advanced geological forecasting method and system based on multi-source information fusion, and the method comprises the steps: obtaining multi-source geological data, carrying out the preprocessing and data fusion of the multi-source geological data, generating comprehensive geological data, carrying out the deep learning model based on a convolutional neural network architecture, extracting the features of the comprehensive geological data, and generating geological data. And the geological data are dynamically updated through a feedback optimization mechanism according to new address data obtained in the construction process. According to the method, the influence of complex geological conditions can be overcome, missed judgment or misjudgment of unfavorable geologic bodies is avoided, the accuracy, real-time performance and reliability of advanced geological forecast are improved, the construction safety of tunnel engineering is fundamentally ensured, and a guarantee is provided for the construction progress and cost control.

Description

Technical Field

[0001] The invention belongs to the technical field of geological engineering and relates to an advanced geological prediction method and system. Background Art

[0002] In tunnel construction, advanced geological prediction is the key to ensure construction safety and progress; traditional advanced geological prediction methods mainly include geological radar detection, advanced drilling, etc., but under complex geological conditions, these methods often cannot meet the needs of accurate prediction. Especially when the tunnel project crosses multiple fault fracture zones, conventional advanced geological radar detection is used. Due to the influence of faults, radar image interpretation is difficult, and the prediction results are quite different from the actual geological conditions. Sometimes, fault fracture zones that were not predicted in advance will appear during construction. Failure to take effective support and prevention measures in time will lead to accidents, which will not only cause construction delays and economic losses, but may also endanger the lives of construction workers. Traditional advanced geological prediction methods, especially under complex geological conditions such as faults, caves, karst, and high ground stress, are very easy to miss or misjudge bad geological bodies, and the accuracy and reliability of the prediction are difficult to guarantee, which poses a major safety hazard to tunnel construction. Summary of the invention

[0003] In order to solve the problems described in the background technology that traditional advanced geological prediction methods under complex geological conditions are very likely to miss or misjudge bad geological bodies, and the accuracy and reliability of the prediction are difficult to guarantee, the present invention provides a tunnel engineering advanced geological prediction method and system based on multi-source information fusion.

[0004] The method of the present invention comprises the following steps:

[0005] Obtain multi-source geological data of tunnel engineering based on geological radar detection, drilling detection and seismic wave detection;

[0006] Preprocessing the multi-source geological data including data cleaning and standardization to obtain preprocessed multi-source geological data;

[0007] Determine the weight parameters of each data source according to its measurement accuracy, detection depth and sensitivity to the target geological features, and fuse the pre-processed multi-source geological data through a weighted fusion algorithm based on the weight parameters to generate comprehensive geological data;

[0008] The comprehensive geological data is input into the deep learning model of the convolutional neural network architecture, the comprehensive geological data features are extracted through the convolution layer, and the geological structure is classified and predicted using the fully connected layer to generate a geological forecast;

[0009] During the construction process, new geological data is continuously acquired based on the geological forecast. When the new geological data differs significantly from the original geological forecast, or the construction progress reaches the preset milestone, the deep learning model optimization is triggered. The deep learning model is trained based on the new geological data, and the deep learning model parameters are adjusted to generate an updated geological forecast.

[0010] Furthermore, the geological radar detection adopts high-density full-section radar scanning to obtain electrical difference information of the rock mass in front of the tunnel project face as geological data for geological radar detection; the drilling detection obtains core, lithology and groundwater information by laying a certain number of advanced horizontal boreholes in the tunnel project as geological data for drilling detection; the seismic wave detection obtains seismic wave velocity structure information within a certain longitudinal range of the tunnel project through TSP203 advanced geological forecast as geological data for seismic wave detection.

[0011] Furthermore, in the preprocessing of the multi-source geological data, the geological data detected by geological radar are preprocessed by methods including time zero point correction, background noise elimination, and gain recovery; the geological data detected by drilling are preprocessed by methods including core photo stitching, lithology interpretation, and RQD value statistics; the geological data detected by seismic waves are preprocessed by methods including first arrival correction, signal-to-noise ratio enhancement, and travel time tomography; the preprocessed geological data are stored in the form of a unified three-dimensional grid model.

[0012] Furthermore, in the deep learning model, a model for geological information interpretation is constructed based on a three-dimensional convolutional neural network, including an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer;

[0013] The input layer: 3D data block from the fusion model, size is 50x50x50;

[0014] There are three convolutional layers, with kernel sizes of 5x5x5, 3x3x3 and 1x1x1, and the number of feature maps is 32, 64 and 128 respectively. Suppose the input of the i-th convolutional layer is X i , the convolution kernel is W i , with a bias of b i , the activation function is f, then the output of this layer is X i+1 for:

[0015] X i+1 =f(W i *X i +b i ),

[0016] Among them, * represents the convolution operation;

[0017] The pooling layers are 3 maximum pooling layers, all with a size of 2x2x2, which are used for feature dimensionality reduction;

[0018] There are two fully connected layers, each containing 256 and 128 neurons;

[0019] The output layer uses Softmax activation to output the probability of various geological bodies;

[0020] The model is trained using the Adam optimizer, with the initial learning rate, batch size and training rounds set according to actual conditions. Multiple data blocks are randomly selected as training sets in the excavated sections of the tunnel project, and are labeled using the actual geological conditions revealed by the face. The training model can identify key geological features including rock interface, fault location, and cave distribution.

[0021] Furthermore, the method for adjusting the deep learning model parameters includes:

[0022] During the tunneling process, new geological data including real-time monitoring data and real-time drilling data are collected in real time. When the deviation between the new geological data and the original geological forecast exceeds the set threshold, or when the construction progress reaches the preset mileage, the deep learning model is optimized through incremental learning.

[0023] Incremental learning fine-tunes the original deep learning model based on new geological data, resets the learning rate and training rounds, and the updated deep learning model predicts the subsequent intervals of the tunnel project and adjusts the scope and frequency of geological forecasts as needed;

[0024] Suppose the new geological data obtained at time t is D t , the geological information corresponding to the original deep learning model is Y t , the geological information corresponding to the updated deep learning model is Y t+1 , then:

[0025] Y t+1 =U(Y t ,D t ),

[0026] Where U represents the forecast update function, which is calculated based on the deep learning model and new geological data;

[0027] The accuracy of geological prediction results is calculated as follows:

[0028] Accuracy=(T P +T N ) / (T P +T N +F P +F N ),

[0029] Among them, T P 、T N 、F P 、F N They represent the number of true positives, true negatives, false positives, and false negatives in the geological prediction results respectively.

[0030] The present invention also proposes an advanced geological prediction system for tunnel engineering based on multi-source information fusion, including a geological data acquisition module, a data preprocessing module, a comprehensive geological data generation module, a geological prediction generation module and a geological prediction update module.

[0031] The geological data acquisition module is used to acquire multi-source geological data of the tunnel project based on geological radar detection, drilling detection and seismic wave detection.

[0032] The data preprocessing module is used to perform preprocessing including data cleaning and standardization on the multi-source geological data to obtain the preprocessed multi-source geological data.

[0033] The comprehensive geological data generation module is used to determine the weight parameters of each data source according to the measurement accuracy, detection depth and sensitivity factors of each data source to the target geological characteristics, and fuse the pre-processed multi-source geological data through a weighted fusion algorithm based on the weight parameters to generate comprehensive geological data.

[0034] The geological forecast generation module is used to input the comprehensive geological data into the deep learning model of the convolutional neural network architecture, extract the characteristics of the comprehensive geological data through the convolution layer, and use the fully connected layer to classify and predict the geological structure to generate a geological forecast.

[0035] The geological forecast update module is used to continuously acquire new geological data during the construction process based on the geological forecast. When the new geological data differs significantly from the original geological forecast or the construction progress reaches a preset mileage, the deep learning model optimization is triggered, the deep learning model is trained based on the new geological data, the deep learning model parameters are adjusted, and an updated geological forecast is generated.

[0036] Compared with the prior art, the present invention obtains multi-source geological data and performs preprocessing and data fusion on it to generate comprehensive geological data. Based on the deep learning model of the convolutional neural network architecture, the comprehensive geological data features are extracted to generate geological data, and the geological data is dynamically updated according to the new address data obtained during the construction process through the feedback optimization mechanism. The present invention comprehensively and accurately depicts the geological conditions ahead of the tunnel line by fusing multi-source geological data, and overcomes the problem of insufficient prediction accuracy of conventional methods under complex geological environments; the proposed deep learning model can automatically extract complex geological features, improve the accuracy and efficiency of prediction, and reduce the uncertainty of manual interpretation; the dynamic update of geological prediction ensures the real-time and reliability of the prediction results, and meets the requirements of progress control and risk prevention of tunnel engineering construction. The present invention can overcome the influence of complex geological conditions, avoid the omission or misjudgment of bad geological bodies, improve the accuracy, real-time and reliability of advanced geological prediction, fundamentally ensure the construction safety of tunnel engineering, and provide guarantee for construction progress and cost control. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The present invention is a flow chart of the method.

[0038] Figure 2 It is a system architecture diagram of the present invention.

[0039] Figure 3 Detailed flowchart for preprocessing and fusion of multi-source geological data.

[0040] Figure 4 This is the architecture diagram of the deep learning model.

[0041] Figure 5 Schematic diagram of the deep learning model optimization process. DETAILED DESCRIPTION

[0042] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0043] The advanced geological prediction method for tunnel engineering based on multi-source information fusion is shown in the flowchart Figure 1 As shown, the details are as follows.

[0044] Firstly, multi-source geological data of the tunnel project are obtained based on geological radar detection, drilling detection and seismic wave detection.

[0045] Specifically, geological radar detection uses high-density full-section radar scanning to obtain electrical difference information of the rock mass in front of the tunnel face as geological data for geological radar detection. For example, a full-section radar antenna with a center frequency of 100MHz is used to scan once every 2m along the tunnel axis.

[0046] Drilling exploration is to obtain rock cores, lithology, and groundwater information by laying out a certain number of advanced horizontal boreholes in the tunnel project as geological data for drilling exploration. For example, an advanced horizontal borehole is laid out every 50m, with a hole depth of 30m and a core sampling interval of 1m.

[0047] Seismic wave detection uses TSP203 advanced geological prediction to obtain seismic wave velocity structure information within a certain range in the longitudinal direction of the tunnel project as geological data for seismic wave detection. For example, a 24-channel TSP203 full-section seismometer is used, with a line spacing of 5m and an excitation point spacing of 2.5m.

[0048] Then, the multi-source geological data are preprocessed including data cleaning and standardization to obtain preprocessed multi-source geological data.

[0049] Then, the weight parameters of each data source are determined according to its measurement accuracy, detection depth and sensitivity to the target geological features. The preprocessed multi-source geological data are fused through a weighted fusion algorithm based on the weight parameters to generate comprehensive geological data.

[0050] The detailed flow chart of preprocessing and fusion of multi-source geological data is as follows: Figure 3 shown.

[0051] Specifically, for geological data detected by geological radar, the preprocessing methods include: time zero point correction, background noise elimination, and gain recovery.

[0052] For the geological data of drilling exploration, the preprocessing methods include: core photo stitching, lithology interpretation, and RQD value statistics.

[0053] For geological data detected by seismic waves, the preprocessing methods include: first arrival correction, signal-to-noise ratio enhancement, and travel-time tomography.

[0054] The preprocessed geological data is stored in the form of a unified three-dimensional grid model.

[0055] Suppose there are n geological data D1, D2, ..., D n , whose corresponding weights are w1,w2,...,w n , satisfying w1+w2+...+w n =1, then the integrated geological data D after fusion can be expressed as:

[0056] D=w1D1+w2D2+...+w n D n .

[0057] Through weighted fusion, the comprehensive geological data D not only contains the information of each data source, but also reflects the difference in their importance, thereby improving the effectiveness of data fusion.

[0058] Then, the comprehensive geological data is input into the deep learning model of the convolutional neural network architecture, the comprehensive geological data features are extracted through the convolutional layer, and the geological structure is classified and predicted using the fully connected layer to generate a geological forecast.

[0059] The architecture diagram of the deep learning model is as follows Figure 2 As shown in the figure, in the deep learning model, a model for geological information interpretation is constructed based on a three-dimensional convolutional neural network, including an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer. The architecture diagram of the deep learning model is shown in Figure 4 shown.

[0060] Input layer: 3D data block from the fused model, size 50x50x50.

[0061] There are three convolutional layers, with kernel sizes of 5x5x5, 3x3x3, and 1x1x1, and the number of feature maps is 32, 64, and 128. Suppose the input of the i-th convolutional layer is X i , the convolution kernel is W i , with a bias of b i , the activation function is f, then the output of this layer is X i+1 for:

[0062] X i+1 =f(W i *X i +b i ),

[0063] Among them, * represents the convolution operation.

[0064] The pooling layers are 3 maximum pooling layers, all with a size of 2x2x2, which are used for feature dimensionality reduction.

[0065] There are two fully connected layers, containing 256 and 128 neurons respectively.

[0066] The output layer uses Softmax activation to output the probability of each type of geological body.

[0067] The model is trained using the Adam optimizer, and the initial learning rate, batch size, and training rounds are set according to the actual situation, such as an initial learning rate of 0.001, a batch size of 16, and 500 training rounds. Multiple data blocks are randomly selected as training sets in the excavated sections of the tunnel project, such as 500 data blocks, and are annotated using the actual geological conditions revealed by the face; the training model can identify key geological features including rock layer interfaces, fault locations, and cave distribution.

[0068] After training, the model can accurately identify complex geological bodies such as faults and caves from the integrated geological data with an accuracy of more than 85%.

[0069] Finally, new geological data are continuously acquired during the construction process based on the geological forecast. When the new geological data differs significantly from the original geological forecast, or the construction progress reaches the preset mileage, the deep learning model optimization is triggered, and the deep learning model is trained based on the new geological data. The deep learning model parameters are adjusted to generate an updated geological forecast.

[0070] During the tunneling process, new geological data including real-time monitoring data and real-time drilling data are collected in real time. When the deviation between the new geological data and the original geological forecast exceeds the set threshold, or when the construction progress reaches the preset mileage, the deep learning model is optimized through incremental learning. For example, during the collection of new geological data, SP203 automatically collects data every 10m of excavation and transmits it to the database; the set threshold can be 20%.

[0071] A schematic diagram of the deep learning model optimization process is shown below: Figure 5 Incremental learning fine-tunes the original deep learning model based on the new geological data, resetting the learning rate and training rounds, for example, the learning rate is set to 0.0001, and 100 rounds of training are performed; the updated deep learning model predicts the subsequent intervals of the tunnel project and adjusts the scope and frequency of geological forecasts according to demand.

[0072] Suppose the new geological data obtained at time t is D t , the geological information corresponding to the original deep learning model is Y t , the geological information corresponding to the updated deep learning model is Y t+1 , then:

[0073] Y t+1 =U(Y t ,D t ),

[0074] Where U represents the forecast update function, which is calculated based on the deep learning model and new geological data;

[0075] The accuracy of geological prediction results is calculated as follows:

[0076] Accuracy=(T P +T N ) / (T P +T N +F P +F N ),

[0077] Among them, T P 、T N 、F P 、F N They represent the number of true positives, true negatives, false positives, and false negatives in the geological prediction results respectively.

[0078] The present invention also proposes a tunnel engineering advanced geological prediction system based on multi-source information fusion, the architecture diagram of which is as follows: Figure 2 As shown, it consists of a geological data acquisition module, a data preprocessing module, a comprehensive geological data generation module, a geological forecast generation module and a geological forecast update module.

[0079] The geological data acquisition module is used to obtain multi-source geological data of tunnel engineering based on geological radar detection, drilling detection and seismic wave detection.

[0080] The data preprocessing module is used to preprocess the multi-source geological data, including data cleaning and standardization, to obtain the preprocessed multi-source geological data.

[0081] The comprehensive geological data generation module is used to determine the weight parameters of each data source according to the measurement accuracy, detection depth and sensitivity factors of the target geological characteristics of each data source, and to fuse the pre-processed multi-source geological data through a weighted fusion algorithm based on the weight parameters to generate comprehensive geological data.

[0082] The geological forecast generation module is used to input comprehensive geological data into the deep learning model of the convolutional neural network architecture, extract the characteristics of comprehensive geological data through the convolution layer, and use the fully connected layer to classify and predict the geological structure to generate a geological forecast.

[0083] The geological forecast update module is used to continuously obtain new geological data during the construction process based on the geological forecast. When the new geological data differs significantly from the original geological forecast or the construction progress reaches the preset mileage, the deep learning model optimization is triggered, the deep learning model is trained based on the new geological data, the deep learning model parameters are adjusted, and an updated geological forecast is generated.

[0084] The implementation method of each module in the system is consistent with that described in the above method and will not be repeated here.

[0085] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the advanced geological prediction method for tunnel engineering based on multi-source information fusion and the advanced geological prediction system for tunnel engineering based on multi-source information fusion as described above.

[0086] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the application can adopt the form of complete hardware embodiment, complete software embodiment, or the embodiment in combination with software and hardware. Moreover, the application can adopt the form of the computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiment of the present application can be implemented in various computer languages, for example, object-oriented programming language Java, C++, Python and literal scripting language JavaScript, etc.

[0087] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0088] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0089] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1The steps for the functions specified in one or more boxes.

[0090] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0091] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. The advanced geological prediction method for tunnel engineering based on multi-source information fusion is characterized by: The following steps are involved: Obtain multi-source geological data of tunnel engineering based on geological radar detection, drilling detection and seismic wave detection; Preprocessing the multi-source geological data including data cleaning and standardization to obtain preprocessed multi-source geological data; Determine the weight parameters of each data source according to its measurement accuracy, detection depth and sensitivity to the target geological features, and fuse the pre-processed multi-source geological data through a weighted fusion algorithm based on the weight parameters to generate comprehensive geological data; The comprehensive geological data is input into the deep learning model of the convolutional neural network architecture, the comprehensive geological data features are extracted through the convolution layer, and the geological structure is classified and predicted using the fully connected layer to generate a geological forecast; During the construction process, new geological data is continuously acquired based on the geological forecast. When the new geological data differs significantly from the original geological forecast, or the construction progress reaches the preset milestone, the deep learning model optimization is triggered. The deep learning model is trained based on the new geological data, and the deep learning model parameters are adjusted to generate an updated geological forecast.

2. The method for advanced geological prediction of tunnel engineering based on multi-source information fusion according to claim 1 is characterized in that: The geological radar detection adopts high-density full-section radar scanning to obtain the electrical difference information of the rock mass in front of the tunnel project face as the geological data of the geological radar detection; the drilling detection obtains the core, lithology and groundwater information by laying a certain number of advanced horizontal boreholes in the tunnel project as the geological data of the drilling detection; the seismic wave detection obtains the seismic wave velocity structure information within a certain range in the longitudinal direction of the tunnel project through the TSP203 advanced geological forecast as the geological data of the seismic wave detection.

3. The method for advanced geological prediction of tunnel engineering based on multi-source information fusion according to claim 2 is characterized in that: In the preprocessing of the multi-source geological data, the geological data detected by geological radar are preprocessed by methods including time zero point correction, background noise elimination, and gain recovery; the geological data detected by drilling are preprocessed by methods including core photo stitching, lithology interpretation, and RQD value statistics; the geological data detected by seismic waves are preprocessed by methods including first arrival correction, signal-to-noise ratio enhancement, and travel time tomography; the preprocessed geological data are stored in the form of a unified three-dimensional grid model.

4. The method for advanced geological prediction of tunnel engineering based on multi-source information fusion according to claim 1 is characterized in that: In the deep learning model, a model for geological information interpretation is constructed based on a three-dimensional convolutional neural network, including an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer; The input layer: 3D data block from the fusion model, size is 50x50x50; There are three convolutional layers, the kernel sizes are 5x5x5, 3x3x3 and 1x1x1 respectively, and the number of feature maps is 32, 64 and 128; let the input of the i-th convolutional layer be X i , the convolution kernel is W i , with a bias of b i , the activation function is f, then the output of this layer is X i+1 for: X i+1 =f(W i *X i +b i ), Among them, * represents the convolution operation; The pooling layers are 3 maximum pooling layers, all with a size of 2x2x2, which are used for feature dimensionality reduction; There are two fully connected layers, each containing 256 and 128 neurons; The output layer uses Softmax activation to output the probability of various geological bodies; The model is trained using the Adam optimizer, with the initial learning rate, batch size and training rounds set according to actual conditions. Multiple data blocks are randomly selected as training sets in the excavated sections of the tunnel project, and are labeled using the actual geological conditions revealed by the face. The training model can identify key geological features including rock interface, fault location, and cave distribution.

5. The method for advanced geological prediction of tunnel engineering based on multi-source information fusion according to claim 4 is characterized in that: The method for adjusting the deep learning model parameters includes: During the tunneling process, new geological data including real-time monitoring data and real-time drilling data are collected in real time. When the deviation between the new geological data and the original geological forecast exceeds the set threshold, or when the construction progress reaches the preset mileage, the deep learning model is optimized through incremental learning. Incremental learning fine-tunes the original deep learning model based on new geological data, resets the learning rate and training rounds, and the updated deep learning model predicts the subsequent intervals of the tunnel project and adjusts the scope and frequency of geological forecasts as needed; Suppose the new geological data obtained at time t is D t , the geological information corresponding to the original deep learning model is Y t , the geological information corresponding to the updated deep learning model is Y t+1 , then: Y t+1 =U(Y t ,D t ), Where U represents the forecast update function, which is calculated based on the deep learning model and new geological data; The accuracy of geological prediction results is calculated as follows: Accuracy=(T P +T N ) / (T P +T N +F P +F N ), Among them, T P 、T N 、F P 、F N They represent the number of true positives, true negatives, false positives, and false negatives in the geological prediction results respectively.

6. A tunnel engineering advanced geological prediction system based on multi-source information fusion implementing the method described in any one of claims 1 to 5, characterized in that: It includes a geological data acquisition module, a data preprocessing module, a comprehensive geological data generation module, a geological forecast generation module and a geological forecast update module; The geological data acquisition module is used to acquire multi-source geological data of the tunnel project based on geological radar detection, drilling detection and seismic wave detection; The data preprocessing module is used to preprocess the multi-source geological data including data cleaning and standardization to obtain the preprocessed multi-source geological data; The comprehensive geological data generation module is used to determine the weight parameters of each data source according to the measurement accuracy, detection depth and sensitivity factors of each data source to the target geological features, and fuse the pre-processed multi-source geological data through a weighted fusion algorithm according to the weight parameters to generate comprehensive geological data; The geological forecast generation module is used to input the comprehensive geological data into the deep learning model of the convolutional neural network architecture, extract the characteristics of the comprehensive geological data through the convolution layer, and use the fully connected layer to classify and predict the geological structure to generate a geological forecast; The geological forecast update module is used to continuously acquire new geological data during the construction process based on the geological forecast. When the new geological data differs significantly from the original geological forecast or the construction progress reaches a preset mileage, the deep learning model optimization is triggered, the deep learning model is trained based on the new geological data, the deep learning model parameters are adjusted, and an updated geological forecast is generated.

Images