TBM real-time slag monitoring system and method for advanced geological forecast

Abstract

The invention belongs to the field of geological engineering construction, discloses a TBM (Tunnel Boring Machine) real-time slag monitoring system and a TBM real-time slag monitoring method for advanced geological forecast, solves the problems of lack of real-time performance, continuity and accuracy and low processing efficiency of surrounding rock state analysis in the traditional technology, and provides powerful support for advanced geological forecast in a TBM tunneling process. The method comprises the following steps: continuously acquiring rock slag images to obtain a rock slag acquisition video; cutting the rock slag acquisition video into an image sequence with a fixed time interval, and aligning the timestamp of each rock slag image with the recording time of the TBM tunneling parameters; then the cut rock slag image is preprocessed; then identifying the preprocessed rock slag image to obtain a classification result of the rock slag; and finally, the classification result of the rock slag and TBM tunneling parameters are fused to serve as input, and the surrounding rock grade is calculated through a surrounding rock grade grading model so that the tunneling strategy can be adjusted in real time. The method is suitable for underground engineering construction scenes.

Description

Technical Field

[0001] The invention belongs to the field of geological engineering construction, and in particular relates to a TBM real-time slag monitoring system and method for advanced geological prediction. Background Art

[0002] In tunneling operations, when using TBM (tunnel boring machine) for underground engineering construction, real-time monitoring and evaluation of the geological conditions of the surrounding rock is crucial. Traditional technologies mainly rely on physical measurements and on-site inspections to understand the surrounding rock conditions. These methods include: (1) Tunneling parameter monitoring: By monitoring the TBM's excavation speed, thrust, cutting force and other parameters, engineers can indirectly understand the stability of the surrounding rock. Although these parameters can reflect changes in geological conditions, since excavation parameters are affected by many factors, these parameters alone are often unable to fully and accurately judge the surrounding rock conditions.

[0003] (2) Geological drilling and sampling: Before or during excavation, drilling sampling is used to obtain surrounding rock samples for laboratory analysis. This method can provide detailed geological information, but its sampling frequency is low and time-sensitive, and it cannot achieve real-time and continuous monitoring.

[0004] (3) Manual visual inspection: Manual visual analysis of rock debris after excavation is performed to check the size, shape and other characteristics of rock fragments. This method is highly subjective and inefficient, and it is difficult to obtain real-time and accurate data.

[0005] In summary, the traditional technology for obtaining surrounding rock information in TBM construction has the following defects: (1) Data timeliness is insufficient Traditional methods such as excavation parameter monitoring and geological drilling can only provide limited and intermittent data, and cannot achieve real-time monitoring and analysis. This leads to a slow response to surrounding rock changes and may miss the best time to adjust the excavation strategy.

[0006] (2) Lack of real-time and continuity Although excavation parameters can provide some indirect information, they lack direct and real-time analysis of the quality and characteristics of rock slag. Traditional manual visual inspections are also unable to achieve real-time continuous analysis, resulting in incomplete and delayed information collection.

[0007] (3) Highly subjective Manual visual inspection is affected by the experience and judgment of the inspector, and has a large degree of subjectivity and error. This method is difficult to ensure the accuracy and consistency of the data.

[0008] (4) Low data processing efficiency Traditional technologies for analyzing rock debris mainly rely on manual inspection or traditional image processing methods. These methods are inefficient when processing large-scale data and are difficult to adapt to the needs of rapid excavation.

[0009] (5) Lack of geological information integration Traditional technologies only focus on excavation parameters or local geological samples, and fail to fully integrate geological information and excavation data. This information isolation process may not accurately reflect the actual geological conditions. Summary of the invention

[0010] The technical problem to be solved by the present invention is to propose a TBM real-time slag monitoring system and method for advanced geological prediction, so as to solve the problems of lack of real-time, continuity, accuracy and low processing efficiency in surrounding rock status analysis in traditional technologies, and provide strong support for advanced geological prediction during TBM excavation.

[0011] The technical solution adopted by the present invention to solve the above technical problems is: In one aspect, the present invention provides a TBM real-time slag monitoring system for advanced geological prediction, comprising: An image acquisition unit, used for continuously acquiring rock slag images to obtain rock slag acquisition videos; A video processing unit, used to crop the rock slag acquisition video into a sequence of images at fixed time intervals, and align the timestamp of each rock slag image with the recording time of the TBM excavation parameters; An image preprocessing unit, used for preprocessing the cropped rock slag image; An image recognition unit is used to recognize the preprocessed rock slag image and obtain a classification result of the rock slag; The data fusion and analysis unit is used to fuse the classification results of rock debris with the TBM excavation parameters as input, and calculate the surrounding rock grade using the surrounding rock grade classification model for real-time adjustment of the excavation strategy.

[0012] Furthermore, the image acquisition unit is installed above the slag conveyor belt of the TBM and uses a high-resolution camera with a resolution of at least 1080p.

[0013] Furthermore, the video processing unit is specifically used to: use the FFMPEG tool to crop the rock slag acquisition video into an image sequence with a fixed time interval, and use the DTW (Dynamic Time Warping) algorithm to align the timestamp of each rock slag image with the recording time of the TBM excavation parameters.

[0014] Furthermore, the image preprocessing unit is specifically used to perform background trimming, grayscale conversion, noise reduction and image enhancement preprocessing on the cropped rock slag image.

[0015] Furthermore, the image recognition unit is specifically used to: use a CNN (convolutional neural network) model to identify the preprocessed rock slag image to obtain a classification result of the rock slag; the classification result of the rock slag includes four categories: large and uneven fragments, large and uniform fragments, small but uneven fragments, and small and uniform fragments.

[0016] Furthermore, the surrounding rock grade classification model uses a deep CNN algorithm to calculate the surrounding rock grade based on the fusion data of the rock slag classification results and the TBM excavation parameters.

[0017] On the other hand, based on the above system, the present invention also provides a TBM real-time slag monitoring method for advanced geological prediction, comprising the following steps: S1, continuously collect rock slag images to obtain rock slag collection videos; S2, cropping the rock slag acquisition video into a sequence of images at fixed time intervals, and aligning the timestamp of each rock slag image with the recording time of the TBM excavation parameters; S3, preprocessing the cropped rock slag image; S4, identifying the preprocessed rock slag image to obtain a classification result of the rock slag; S5. The classification results of rock debris are integrated with TBM excavation parameters as input, and the surrounding rock grade is calculated using the surrounding rock grade classification model to adjust the excavation strategy in real time.

[0018] Furthermore, in step S3, the cropped rock slag image is preprocessed, including: Background trimming: Use the YOLOv8 object detection model to identify the slag area in the slag image, obtain the bounding box of the slag area, extract the image of the slag area, and resize the extracted image of the slag area to the specified pixel size; Grayscale conversion: convert the image of the slag area into a grayscale image; Noise reduction: Gaussian filtering is used to suppress noise in grayscale images while retaining the detailed features of rock fragments; Image enhancement: CLAHE (Contrast Limited Adaptive Histogram Equalization) technology is used to enhance the edges and details in the image.

[0019] Furthermore, in step S4, a CNN model is used to identify the preprocessed rock slag image to obtain a rock slag classification result. The training method of the CNN model includes: obtaining a rock slag image classification training data set by classifying and marking the collected rock slag image samples, and training the CNN model based on the rock slag image classification training data set using a back propagation algorithm.

[0020] Furthermore, in step S5, the training method of the surrounding rock grade classification model includes: Firstly, an association mapping between the classification labels of rock slag image samples and the TBM excavation parameters synchronized with the acquisition time of the corresponding rock slag image samples is established. Then, the rock slag image samples and the corresponding TBM excavation parameters are marked with surrounding rock grade categories. Next, a surrounding rock grade classification training dataset is constructed based on the rock slag image samples and the corresponding TBM excavation parameters marked with surrounding rock grade categories. Finally, based on the surrounding rock grade classification training dataset, the back propagation algorithm is used to train the surrounding rock grade classification model.

[0021] The beneficial effects of the present invention are: (1) Realize real-time and continuous monitoring of rock slag Video image acquisition enables real-time and continuous acquisition of rock debris images, avoiding the problem of intermittent data acquisition in traditional methods and making the monitoring of surrounding rock conditions more comprehensive and timely.

[0022] (2) Improving the accuracy and efficiency of image analysis By performing a series of preprocessing on the collected rock slag images, such as background trimming, grayscale conversion, noise reduction and image enhancement, the amount of data processing can be reduced, noise interference can be reduced, and image details can be increased, which is conducive to improving the efficiency and accuracy of image analysis; and the use of convolutional neural networks for feature extraction and intelligent recognition of rock slag images can avoid the subjective errors caused by manual visual inspection in traditional methods, and achieve efficient automatic classification and evaluation.

[0023] (3) Improve comprehensive understanding and predictive ability of geological information By combining the results of rock slag image analysis with TBM excavation parameters to form a comprehensive data input, a surrounding rock grade intelligent classification model is generated through a deep learning algorithm, and the training and optimization of the model are driven by both geological information and excavation data. The model can not only reflect the geological state of the surrounding rock in real time, but also consider the actual parameters during the excavation process to provide more accurate surrounding rock grade predictions, improve the interpretability and robustness of the model, and enhance the model's adaptability and generalization capabilities under different geological conditions.

[0024] (4) Improving the safety and efficiency of TBM excavation By intelligently identifying the type of rock debris and the state of the surrounding rock, real-time feedback is provided to adjust the excavation strategy. When encountering unstable rock formations, cracks or changes in the hardness of the surrounding rock, the excavation parameters can be adjusted in time, thereby reducing the risk of tunnel collapse and excessive wear of equipment. The dynamic adjustment based on the excavation strategy also makes tunnel excavation operations safer and more efficient.

[0025] In summary, the present invention significantly improves the real-time, accuracy and processing efficiency of slag analysis through innovative real-time monitoring, intelligent identification and data fusion technologies, and provides strong support for geological prediction during TBM excavation. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a block diagram of a TBM real-time slag monitoring system for advanced geological prediction in Example 1 of the present invention; Figure 2 This is a flow chart of the TBM real-time slag monitoring method for advanced geological prediction in Example 2 of the present invention. DETAILED DESCRIPTION

[0027] The present invention aims to provide a TBM real-time slag monitoring system and method for advanced geological prediction, so as to solve the problems of lack of real-time, continuity, accuracy and low processing efficiency in surrounding rock state analysis in traditional technologies, and provide strong support for advanced geological prediction in the process of TBM excavation. Its core idea is: using video image acquisition technology to realize real-time and continuous monitoring of rock slag, thereby overcoming the lack of data timeliness in traditional methods; based on image preprocessing, the amount of data processing can be reduced, noise interference can be reduced, and image details can be increased, which is conducive to improving the efficiency and accuracy of image analysis; using deep learning algorithms to intelligently identify preprocessed rock slag images, reducing subjectivity and improving the efficiency and accuracy of data processing; combining rock slag image analysis and excavation parameters as inputs of the surrounding rock grade intelligent grading model to comprehensively predict the surrounding rock grade, thereby improving the comprehensive understanding and prediction capabilities of geological information, and being able to meet the needs of rapid excavation.

[0028] The scheme of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0029] Example 1

[0030] This embodiment provides a TBM real-time slag monitoring system for advanced geological prediction, see Figure 1 , which includes: image acquisition unit, video processing unit, image preprocessing unit, image recognition unit and data fusion and analysis unit; the implementation functions of each part are described as follows: (1) Image acquisition unit It is used to continuously collect rock slag images and obtain rock slag collection videos; in an exemplary embodiment, in order to capture the detailed features of the rock slag, the image acquisition unit can use a high-resolution camera with a resolution of at least 1080p, which is installed above the slag conveyor belt of the TBM to collect rock slag images in real time and continuously. It is understandable that other image acquisition devices, such as laser scanners or infrared cameras, can also be selected according to actual needs.

[0031] (2) Video processing unit It is used to crop the slag acquisition video into a sequence of images with fixed time intervals, and align the timestamp of each slag image with the recording time of the TBM excavation parameters; in an exemplary embodiment, the FFMPEG tool can be used to crop the slag acquisition video into a sequence of images with fixed time intervals, and then the DTW (Dynamic Time Warping) algorithm is used to align the timestamp of each slag image with the recording time of the TBM excavation parameters. The purpose of the alignment is to achieve the association mapping between the classification labels of the subsequent slag images and the excavation parameters at the same time, so as to realize the training and optimization of the surrounding rock grade classification model driven by geological information and excavation data, and obtain accurate surrounding rock grade prediction information. It is understandable that other video processing software or tools can also be selected according to actual needs to achieve the above-mentioned similar image cropping and time alignment functions.

[0032] (3) Image preprocessing unit It is used to pre-process the cropped rock debris image; in an exemplary embodiment, the pre-processing includes: background trimming, grayscale conversion, noise reduction and image enhancement; among them, background trimming is to remove the unnecessary background part in the image, and only retain the rock debris area to reduce the consumption of computing resources and the complexity of subsequent analysis; grayscale conversion is to convert the color image into a grayscale image to reduce the consumption of computing resources and increase the processing speed; noise reduction is to use Gaussian filtering technology to suppress noise while retaining image details as much as possible, improve image quality, and provide clearer input for subsequent processing; image enhancement is to use CLAHE (contrast limited adaptive histogram equalization) technology to enhance the edge features in the image, so as to better identify the size and shape of rock fragments. It can be understood that other pre-processing technologies that can reduce noise and enhance images can also be selected according to actual needs, such as mean filtering, edge detection, etc.

[0033] (4) Image recognition unit It is used to identify the preprocessed rock slag image and obtain the classification result of the rock slag; in an exemplary embodiment, a CNN (convolutional neural network) model is used to identify the preprocessed rock slag image to obtain the classification result of the rock slag; the convolutional neural network can effectively extract the essential features of the image, which includes 5 convolutional layers and 2 pooling layers, and finally connected to the fully connected layer. The convolutional layer and the pooling layer are used to capture the important features in the image, and the extracted features are input into the fully connected layer for classification to determine the classification of the rock fragments. It is understandable that other deep learning algorithms or machine learning models can also be selected according to actual needs to extract and classify features of the image, such as support vector machines, random forests, etc.

[0034] (5) Data fusion and analysis unit It is used to fuse the classification results of rock slag with TBM excavation parameters as input, and use the surrounding rock grade classification model to calculate the surrounding rock grade for real-time adjustment of the excavation strategy. In an exemplary embodiment, a weighted average algorithm is used to fuse the classification results of rock slag with TBM excavation parameters, so that the calculation of the surrounding rock grade takes into account both the geological state of the surrounding rock and the actual parameters during the excavation process, thereby obtaining a more accurate prediction result. It is understandable that other data fusion technologies, such as rule-based systems or hybrid models, can also be selected according to actual needs.

[0035] Example 2

[0036] Based on the system architecture of the above embodiment 1, this embodiment provides a TBM real-time slag monitoring method for advanced geological prediction, see Figure 2 , which includes the following implementation process: S1, continuously collect rock slag images to obtain rock slag collection videos; In this step, a high-resolution camera collects rock slag images in real time and continuously to obtain rock slag collection videos.

[0037] S2, cropping the rock slag acquisition video into a sequence of images at fixed time intervals, and aligning the timestamp of each rock slag image with the recording time of the TBM excavation parameters; In this step, the FFMPEG tool is used to crop the rock slag acquisition video into an image sequence with a fixed time interval (such as 1 image per second), and then the DTW algorithm is used to align the timestamps of the rock slag images with the recording time of the excavation parameters to ensure the synchronization of the image and excavation data.

[0038] DTW is an algorithm for nonlinear alignment of time series data. It is suitable for dealing with time stretching and distortion problems in time series data and can find the optimal time correspondence between two time series, even if they have different lengths and rhythms on the time axis. The specific steps of the DTW algorithm are: (1) Data preprocessing: remove noise and normalize the time series data; (2) Define distance metrics: use the Euclidean distance function to calculate the similarity between each point in the time series; (3) Construct a cost matrix: use the distance function to calculate the distance between all possible point pairs and construct a cost matrix; (4) Dynamic programming calculation: use the dynamic programming algorithm to fill in the cost matrix and calculate the cumulative cost of all possible paths from the excavation point to the monitoring point; (5) Find the optimal path: by backtracking the cost matrix, find the path with the minimum cumulative cost. This path represents the optimal time alignment between time series; (6) Timestamp mapping: according to the optimal path, map the timestamp of the video image of the monitoring point to the time of the excavation point.

[0039] S3, preprocessing the cropped rock slag image; In this step, in order to improve the efficiency and accuracy of subsequent image recognition processing, the rock slag image is preprocessed including background trimming, grayscale conversion, noise reduction and image enhancement, as follows: Background trimming: Use the YOLOv8 object detection model to identify the slag area in the slag image, obtain the bounding box of the slag area, and extract the image of the slag area. The extracted image is resized to a specified pixel size, such as 640x480 pixels, to accommodate subsequent processing.

[0040] Grayscale conversion: Convert the color image to grayscale. This is done to reduce the consumption of computing resources so that the edges and contours of rock fragments can be processed mainly.

[0041] Noise Reduction: A Gaussian filter (with a standard deviation of 1.0) was applied to suppress the noise in the image while retaining the detailed features of the rock fragments.

[0042] Image enhancement: Use CLAHE and set the contrast limit to 2.0 to enhance the edges and details in the image.

[0043] S4, identifying the preprocessed rock slag image to obtain a classification result of the rock slag; In this step, the trained CNN model is used to identify the preprocessed rock slag image to obtain the classification result of the rock slag. The training process of the CNN model is as follows: First, the collected rock slag image samples are classified and labeled. There are four types of classification labels in this embodiment: 1. The fragments are large and uneven (unstable rock formations appear in the surrounding rock mass, and excavation needs to be stopped for inspection); 2. The fragments are large and uniform (there are cracks, and the TBM advancement speed needs to be reduced); 3. The fragments are small but uneven (the surrounding rock has good hardness and integrity, and the TBM cutter head speed and thrust need to be reduced); 4. The fragments are small and uniform (the surrounding rock mass is in an ideal state, which can increase the advancement speed).

[0044] Then, a training data set is constructed based on the classified and labeled rock slag image samples, and the CNN model is trained using the back propagation algorithm using the training data set until the model converges.

[0045] Through the training process, the CNN model learns the characteristics of the rock slag image samples under the corresponding classification labels, so that after inputting the rock slag image to be identified, the classification result of the rock slag can be obtained.

[0046] S5. The classification results of rock debris are integrated with TBM excavation parameters as input, and the surrounding rock grade is calculated using the surrounding rock grade classification model to adjust the excavation strategy in real time.

[0047] In this step, the classification results of the rock slag image are combined with the TBM excavation parameters (excavation speed, thrust, cutting force, etc.) to form a comprehensive data input, and the surrounding rock grade information is predicted using the trained surrounding rock grade classification model to facilitate real-time adjustment of the excavation strategy and improve excavation efficiency and safety. The training process of the surrounding rock grade classification model is as follows: Firstly, an association mapping between the classification labels of the rock slag image samples and the TBM excavation parameters synchronized with the acquisition time of the corresponding rock slag image samples is established; Then, the rock debris image samples and the corresponding TBM excavation parameters are marked with surrounding rock grade categories; Next, based on the rock debris image samples labeled with surrounding rock grade categories and the corresponding TBM excavation parameters, a surrounding rock grade classification training dataset was constructed. Finally, based on the surrounding rock grade classification training data set, the back propagation algorithm is used to train the surrounding rock grade classification model.

[0048] Through the training process, the surrounding rock grade classification model learns the fusion features of rock slag image samples and TBM excavation parameters under the corresponding surrounding rock grade labels, so that after inputting the classification results of the identified rock slag images and the corresponding TBM excavation parameters, the surrounding rock grade can be predicted.

[0049] According to the records of the above embodiments, the present invention improves the real-time, accuracy and processing efficiency of rock slag analysis through innovative real-time monitoring, intelligent identification and data fusion technology, and provides strong support for geological prediction during TBM excavation. The scheme of the present invention has a wide range of application scenarios, including but not limited to: (1) Road construction: such as excavation monitoring of highways or railway tunnels. (2) Underground space development: such as rock slag monitoring in subway construction or underground commercial space development. (3) Mining: such as real-time rock fragment monitoring and geological early warning in mining.

[0050] Finally, it should be noted that the above embodiments are only preferred implementations and are not intended to limit the present invention. It should be pointed out that for those skilled in the art, several modifications, equivalent replacements, improvements, etc. can be made without departing from the scope of the present invention and the scope of protection of the claims, and all of these should be included in the protection scope of the present invention.

Claims

1. A TBM real-time slag monitoring system for advanced geological prediction, characterized in that: include: An image acquisition unit, used for continuously acquiring rock slag images to obtain rock slag acquisition videos; A video processing unit, used to crop the rock slag acquisition video into a sequence of images at fixed time intervals, and align the timestamp of each rock slag image with the recording time of the TBM excavation parameters; An image preprocessing unit, used for preprocessing the cropped rock slag image; An image recognition unit is used to recognize the preprocessed rock slag image and obtain a classification result of the rock slag; The data fusion and analysis unit is used to fuse the classification results of rock debris with the TBM excavation parameters as input, and calculate the surrounding rock grade using the surrounding rock grade classification model for real-time adjustment of the excavation strategy.

2. A TBM real-time slag monitoring system for advanced geological prediction according to claim 1, characterized in that: The image acquisition unit is installed above the slag conveyor belt of the TBM and uses a high-resolution camera with a resolution of at least 1080p.

3. A TBM real-time slag monitoring system for advanced geological prediction according to claim 1, characterized in that: The video processing unit is specifically used for: using the FFMPEG tool to cut the rock slag acquisition video into an image sequence with a fixed time interval, and using the DTW algorithm to align the timestamp of each rock slag image with the recording time of the TBM excavation parameters.

4. A TBM real-time slag monitoring system for advanced geological prediction according to claim 1, characterized in that: The image preprocessing unit is specifically used for preprocessing the cropped rock slag image by background trimming, grayscale conversion, noise reduction and image enhancement.

5. A TBM real-time slag monitoring system for advanced geological prediction according to claim 1, characterized in that: The image recognition unit is specifically used to: use the CNN model to recognize the preprocessed rock slag image to obtain the classification result of the rock slag; the classification result of the rock slag includes four categories: large and uneven fragments, large and uniform fragments, small but uneven fragments, and small and uniform fragments.

6. A TBM real-time slag monitoring system for advanced geological prediction according to claim 1, characterized in that: The surrounding rock grade classification model uses a deep CNN algorithm to calculate the surrounding rock grade based on the fusion data of the rock slag classification results and the TBM excavation parameters.

7. A TBM real-time slag monitoring method for advanced geological prediction, applied to the TBM real-time slag monitoring system for advanced geological prediction as claimed in any one of claims 1 to 6, characterized in that: The method comprises the following steps: S1, continuously collect rock slag images to obtain rock slag collection videos; S2, cropping the rock slag acquisition video into a sequence of images at fixed time intervals, and aligning the timestamp of each rock slag image with the recording time of the TBM excavation parameters; S3, preprocessing the cropped rock slag image; S4, identifying the preprocessed rock slag image to obtain a classification result of the rock slag; S5. The classification results of rock debris are integrated with TBM excavation parameters as input, and the surrounding rock grade is calculated using the surrounding rock grade classification model to adjust the excavation strategy in real time.

8. A TBM real-time slag monitoring method for advanced geological prediction according to claim 7, characterized in that: In step S3, the cropped rock slag image is preprocessed, including: Background trimming: Use the YOLOv8 object detection model to identify the slag area in the slag image, obtain the bounding box of the slag area, extract the image of the slag area, and resize the extracted image of the slag area to the specified pixel size; Grayscale conversion: convert the image of the slag area into a grayscale image; Noise reduction: Gaussian filtering is used to suppress noise in grayscale images while retaining the detailed features of rock fragments; Image enhancement: CLAHE technology is used to enhance the edges and detail features in the image.

9. A TBM real-time slag monitoring method for advanced geological prediction according to claim 7, characterized in that: In step S4, the preprocessed rock slag image is identified using a CNN model to obtain a rock slag classification result. The training method of the CNN model includes: obtaining a rock slag image classification training data set by classifying and marking the collected rock slag image samples, and training the CNN model using a back propagation algorithm based on the rock slag image classification training data set.

10. A TBM real-time slag monitoring method for advanced geological prediction according to claim 7, characterized in that: In step S5, the training method of the surrounding rock grade classification model includes: Firstly, an association mapping between the classification labels of rock slag image samples and the TBM excavation parameters synchronized with the acquisition time of the corresponding rock slag image samples is established. Then, the rock slag image samples and the corresponding TBM excavation parameters are marked with surrounding rock grade categories. Next, a surrounding rock grade classification training dataset is constructed based on the rock slag image samples and the corresponding TBM excavation parameters marked with surrounding rock grade categories. Finally, based on the surrounding rock grade classification training dataset, the back propagation algorithm is used to train the surrounding rock grade classification model.

Images