A method and system for intelligent treatment of muck combined with tunnel advance geological prediction

Abstract

The application discloses a kind of intelligent processing methods of muck combined with tunnel advance geological prediction, comprising: obtaining seismic wave training set;Geological analysis is carried out to the muck category in seismic wave training set, obtains the dynamic calculation parameter and physical data corresponding to each muck category, and the seismic wave feature is corresponded with dynamic calculation parameter and physical data, and the updated seismic wave training set is obtained;Seismic wave training set is used to train neural network, and the muck prediction processing model is obtained;Real-time geological prediction data are obtained, real-time geological prediction data are preprocessed, and the real-time geological prediction data after preprocessing are input into muck prediction processing model, and the muck type is output by muck prediction processing model;Combined with the working state of shield machine and geological prediction, muck processing suggestion is generated, to process the muck generated in shield construction process.The application can timely predict the muck type generated when tunnel is excavated, improve the timeliness of prediction.

Description

Technical Field

[0001] This invention relates to the field of shield tunnel excavation technology, and to a method and system for intelligent treatment of excavated soil and debris combined with advanced geological prediction of tunnels. Background Technology

[0002] With the deepening of urbanization in my country, the development and utilization of urban underground space has gradually become an important direction for urban development. Compared with traditional tunnel excavation methods, tunnel boring machines (TBMs) integrate excavation, support, propulsion, and lining operations, offering significant advantages in construction efficiency and cost-effectiveness. During construction, TBMs do not require large-scale demolition on the ground, do not disrupt traffic, generate no noise, do not cause ground subsidence, and do not affect the normal lives of residents, offering significant advantages in environmental protection and construction safety. Therefore, it is considered a commonly used technology in this field and is widely applied in major projects and infrastructure construction areas such as urban rail transit, underground utility tunnels, and railway and highway tunnels.

[0003] The use of shield tunneling technology inevitably generates a large amount of tunnel boring machine (TBM) excavation waste. Currently, in domestic tunnel engineering, the disposal of TBM excavation waste still mainly relies on traditional landfill methods, which occupy a large area and negatively impact the urban landscape. During TBM tunnel construction, surfactants, commonly known as foaming agents, are added to the soil chamber to protect the cutterhead, reduce cutterhead torque, and increase the fluidity of the strata. These foaming agents are adsorbed into the excavation waste, resulting in high moisture content, making it difficult to transport, and causing a series of physicochemical reactions, increasing the difficulty of environmentally friendly waste disposal. Direct discharge without proper treatment will cause soil, surface water, and groundwater pollution, resulting in significant damage to the ecological environment. Furthermore, differences in the strata properties at the excavation face will lead to the generation of various types of excavation waste by the TBM. These wastes have different particle size distributions, penetration indices, moisture content, and flow states, thus requiring different treatment methods.

[0004] Intelligent tunnel construction is currently a hot research topic in the field of railway tunnel construction. The shield tunneling method is a factory-style operation, with the construction process organized like an assembly line, boasting a high degree of mechanization and possessing the foundation and advantages for intelligentization. However, current intelligent technologies mainly focus on the centralized and unified management of visualization, digitalization, and intelligentization in tunnel exploration, design, and construction, with few technologies integrating exploration with construction waste management for the entire tunnel construction process into a fully intelligent system.

[0005] In modern tunnel engineering, excavated soil identification technology is crucial for optimizing the construction process, improving operational efficiency, and ensuring construction safety. Currently, excavated soil identification mainly relies on equipment such as computer vision, cutterhead sensors, and rheometers. Computer vision technology can provide a preliminary assessment of the geological conditions ahead through real-time analysis of the tunnel face images; while cutterhead sensors can monitor changes in various parameters during cutterhead operation, indirectly reflecting the geological characteristics of the tunnel face. As for the application of rheometers, it is mainly limited to measuring the physical properties of excavated soil, such as viscosity and density. This information helps to understand the state of the soil within the excavation chamber, but it cannot provide direct data on unexcavated areas.

[0006] However, the aforementioned methods share a common limitation: they primarily focus on analyzing current or completed work areas, lacking the ability to proactively predict future geological conditions. This lack of timeliness limits the effectiveness of waste soil identification technology in guiding waste soil disposal plans. Especially under complex geological conditions, timely and accurate prediction of future geological conditions is crucial for preventing potential risks. In recent years, with the development of artificial intelligence technology, machine learning-based advanced geological prediction has become possible. This technology, by integrating multiple geological data sources and training deep learning algorithm models, can, to some extent, predict the geological characteristics of undeveloped areas. However, existing research and practice focus more on applying AI technology to the excavation process itself to improve excavation efficiency and safety, while the integration of advanced geological prediction with waste soil disposal remains in the exploratory stage. Summary of the Invention

[0007] To address the aforementioned shortcomings in existing technologies, this invention provides an intelligent method and system for treating excavated soil and debris based on advanced geological prediction of tunnels, which solves the problem of poor timeliness in predicting excavated soil and debris generated during tunnel excavation in existing technologies.

[0008] To achieve the aforementioned objectives, the technical solution adopted by this invention is as follows: Firstly, a method for intelligent treatment of excavated soil combined with advanced tunnel geological prediction, comprising:

[0009] S1. Obtain the seismic wave training set;

[0010] The seismic wave training set includes seismic wave features and the corresponding soil and debris category for each seismic wave feature;

[0011] S2. Perform geological analysis on the soil and debris categories in the seismic wave training set, obtain the dynamic calculation parameters and physical data corresponding to each soil and debris category, and match the seismic wave characteristics with the dynamic calculation parameters and physical data to obtain the updated seismic wave training set.

[0012] S3. Train the neural network using the seismic wave training set to obtain the waste soil prediction and treatment model;

[0013] S4. Obtain real-time geological forecast data, preprocess the real-time geological forecast data, and input the preprocessed real-time geological forecast data into the slag and soil prediction and processing model. Output the slag and soil type through the slag and soil prediction and processing model.

[0014] S5. Based on the type of excavated soil and in conjunction with the working status of the tunnel boring machine and geological forecasts, generate excavated soil treatment suggestions to handle the excavated soil generated during the tunnel boring process.

[0015] The beneficial effects of the above scheme are:

[0016] (1) This invention combines horizontal acoustic / seismic profiling (HSP) with deep learning technology to achieve high-precision geological prediction, which can accurately predict the geological conditions ahead and improve the safety and efficiency of construction.

[0017] (2) By intelligently predicting the type of waste soil, appropriate treatment methods can be selected to reduce environmental damage, which is in line with the concept of sustainable development.

[0018] Furthermore, following S1, the method also includes:

[0019] The seismic wave training set is preprocessed to obtain the preprocessed seismic wave training set.

[0020] S2 specifically includes:

[0021] Geological analysis was performed on the soil and debris categories in the preprocessed seismic wave training set to obtain the dynamic calculation parameters and physical data corresponding to each soil and debris category. The seismic wave characteristics were then mapped to the dynamic calculation parameters and physical data to obtain the updated seismic wave training set.

[0022] The benefits of the above further solutions are: after removing noise, the data more accurately reflects the true situation. Imputing missing values ​​ensures the integrity of the dataset, allowing the data to be fully utilized in subsequent analysis and model training. Data standardization maps the value ranges of different features to similar intervals, enabling the gradient descent algorithm to converge to the optimal solution more quickly, thus improving the model's convergence speed.

[0023] Furthermore, in S2, the dynamic calculation parameters include P-wave velocity, S-wave velocity, wave velocity ratio, density, Poisson's ratio, Young's modulus, and bulk modulus;

[0024] Physical data include porosity, water content, and particle size distribution.

[0025] Furthermore, in S3, the waste soil prediction and processing model includes an input layer, a convolutional layer, a flattened layer, a fully connected layer, and an output layer;

[0026] The input layer receives data in the form of N*8*3;

[0027] The convolutional layer includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, a fourth convolutional layer, and a fourth pooling layer;

[0028] The first convolutional layer has 64 kernels, the second convolutional layer has 128 kernels, the third convolutional layer has 256 kernels, and the fourth convolutional layer has 512 kernels. The kernel size of the first, second, third, and fourth convolutional layers is 3*3, and the activation function of the first, second, third, and fourth convolutional layers is the ReLU function.

[0029] The pooling window size for the first, second, third, and fourth pooling layers is 2*2.

[0030] The fully connected layer includes a first fully connected layer and a second fully connected layer;

[0031] The first fully connected layer has 1024 nodes, and the second fully connected layer has 512 nodes;

[0032] The activation functions for both the first and second fully connected layers are ReLU functions.

[0033] The Dropout rates of both the first and second fully connected layers are 0.5.

[0034] The number of nodes in the output layer is the same as the number of waste soil categories.

[0035] The benefits of the above further solutions are: smaller convolutional kernels can capture local features while reducing the number of parameters and improving the computational efficiency of the model. The convolutional kernel moves pixel-by-pixel across the input data, ensuring that as much detail as possible is captured. Padding ensures that the output feature map size is the same as the input, avoiding size reduction caused by convolution operations. The ReLU activation function helps the model learn more complex patterns. Max pooling layers are used for dimensionality reduction, reducing the size of the feature map while retaining the most important features. Pooling layers can extract higher-level features and continue dimensionality reduction. Fully connected layers with a large number of nodes can capture more feature combinations, improving the model's expressive power. Dropout randomly discards a portion of neurons to prevent overfitting and improve the model's generalization ability.

[0036] Furthermore, in S3, the loss function of the waste soil prediction and treatment model is Categorical Crossentropy;

[0037] The optimizer for the waste soil prediction and treatment model is the Adam optimizer, and the initial learning rate of the Adam optimizer is 0.001.

[0038] L2 regularization is applied in the fully connected layer, and the weight decay coefficient is 0.001.

[0039] The beneficial effects of the above further schemes are: Categorical Crosssentropy, used for multi-class classification tasks, measures the difference between the probability distribution predicted by the model and the true labels. The Adam optimizer combines the advantages of momentum and adaptive learning rate, enabling faster convergence and making it suitable for training deep learning models.

[0040] Secondly, a combined tunnel advanced geological prediction intelligent waste soil treatment system, implemented based on the combined tunnel advanced geological prediction intelligent waste soil treatment method according to any one of claims 1-5, includes:

[0041] The model training module is used to acquire a seismic wave training set; to perform geological analysis on the waste soil categories in the seismic wave training set, to obtain the dynamic calculation parameters and physical data corresponding to each waste soil category, and to map the seismic wave characteristics with the dynamic calculation parameters and physical data to obtain an updated seismic wave training set; and to use the seismic wave training set to train the neural network to obtain a waste soil prediction and processing model.

[0042] The seismic wave training set includes seismic wave features and the corresponding soil and debris category for each seismic wave feature.

[0043] The real-time data acquisition module is used to acquire real-time geological forecast data, preprocess the real-time geological forecast data, and input the preprocessed real-time geological forecast data into the slag prediction and treatment model to obtain the type of slag generated during shield tunneling.

[0044] The slag analysis module is used to acquire real-time geological forecast data, input the real-time geological forecast data into the slag prediction and processing model, and obtain the type of slag generated during shield tunneling.

[0045] The waste disposal module is used to generate waste disposal suggestions based on the type of waste, combined with the working status of the tunnel boring machine and geological forecasts, in order to handle the waste generated during the tunnel boring process.

[0046] Furthermore, the real-time data acquisition module includes a seismic wave exciter and a detector;

[0047] The seismic wave generator is used to generate active source seismic waves, and the detector is used to receive vibration signals throughout space.

[0048] Furthermore, the intelligent waste disposal system also includes:

[0049] The data preprocessing module is used to preprocess the seismic wave training set to obtain the preprocessed seismic wave training set.

[0050] The beneficial effects of the above-mentioned further scheme are: by combining the horizontal acoustic / seismic wave profiling method with deep learning technology, high-precision geological prediction can be achieved, and the type of waste soil can be intelligently predicted based on the prediction results, and appropriate waste soil treatment methods can be selected. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating a method for intelligent treatment of construction waste based on advanced geological prediction in tunnels.

[0052] Figure 2 This is a schematic diagram of a combined tunnel advanced geological prediction intelligent waste disposal system. Detailed Implementation

[0053] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0054] Example 1

[0055] like Figure 1 As shown, a smart method for treating excavated soil based on advanced tunnel geological prediction includes:

[0056] S1. Obtain the seismic wave training set; the seismic wave training set includes seismic wave features and the corresponding soil and debris category for each seismic wave feature.

[0057] For example, the seismic wave training set can be obtained by collecting historical geological exploration data (seismic wave characteristics) and the analysis results of soil samples (soil categories) during actual shield tunneling construction, and then constructing the seismic wave training set.

[0058] S2. Perform geological analysis on the soil and debris categories in the seismic wave training set, obtain the dynamic calculation parameters and physical data corresponding to each soil and debris category, and map the seismic wave characteristics with the dynamic calculation parameters and physical data to obtain the updated seismic wave training set.

[0059] In this embodiment, after S1, the method further includes: preprocessing the seismic wave training set to obtain a preprocessed seismic wave training set; S2 specifically includes:

[0060] Geological analysis was performed on the soil and debris categories in the preprocessed seismic wave training set to obtain the dynamic calculation parameters and physical data corresponding to each soil and debris category. The seismic wave characteristics were then mapped to the dynamic calculation parameters and physical data to obtain the updated seismic wave training set.

[0061] In this embodiment, in S2, the dynamic calculation parameters include longitudinal wave velocity, transverse wave velocity, wave velocity ratio, density, Poisson's ratio, Young's modulus, and bulk modulus; the physical data include porosity, water content, and particle size distribution.

[0062] Optionally, the seismic wave training set can be preprocessed, including noise removal, missing value imputation, and data standardization, to ensure data quality and consistency.

[0063] For example, to remove noise from a seismic wave training set, a low-pass filter can be constructed using the Python programming language by calling `scipy.signa1`. A function is defined to create and apply a low-pass Butterworth filter. This function generates filter coefficients based on the given cutoff frequency, sampling rate, and filter order, and then filters the signal.

[0064] Missing value imputation on a seismic wave training set can be accomplished using the `fillna()` function from the Pandas library in Python. First, define a sample variable, then iterate through each column, calculate the average value for that column, and use this average value to fill in all empty values.

[0065] To standardize the seismic wave training set, the `sklearn.preprocessing.StandardScaler` class can be used. The `fit_transform` method automatically calculates the mean μ and standard deviation σ for each feature and transforms the data. , where X represents the original data point.

[0066] Detailed geological analysis was conducted on the waste soil samples, recording their dynamic calculation parameters (P-wave velocity, S-wave velocity, wave velocity ratio, density, Poisson's ratio, Young's modulus, bulk modulus) and physical data (porosity, water content, particle size distribution). Simultaneously, the data was accurately labeled, and the dynamic calculation parameters and physical data of different types of waste soil samples were matched with corresponding seismic wave characteristics. A dictionary can be created in Python to match the dynamic calculation parameters, physical data, and corresponding seismic wave characteristics of different types of waste soil samples.

[0067] S3. Use the seismic wave training set to train the neural network to obtain the waste soil prediction and treatment model.

[0068] In this embodiment, in S3, the slag prediction and processing model includes an input layer, a convolutional layer, a flattened layer, a fully connected layer, and an output layer;

[0069] The input layer receives data in the form of N*8*3;

[0070] The convolutional layer includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, a fourth convolutional layer, and a fourth pooling layer;

[0071] The first convolutional layer has 64 kernels, the second convolutional layer has 128 kernels, the third convolutional layer has 256 kernels, and the fourth convolutional layer has 512 kernels. The kernel size of the first, second, third, and fourth convolutional layers is 3*3, and the activation function of the first, second, third, and fourth convolutional layers is the ReLU function.

[0072] The pooling window size for the first, second, third, and fourth pooling layers is 2*2.

[0073] The fully connected layer includes a first fully connected layer and a second fully connected layer;

[0074] The first fully connected layer has 1024 nodes, and the second fully connected layer has 512 nodes;

[0075] The activation functions for both the first and second fully connected layers are ReLU functions.

[0076] The Dropout rate of both the first and second fully connected layers is 0.5.

[0077] The number of nodes in the output layer is the same as the number of waste soil categories.

[0078] In this embodiment, the loss function of the slag and soil prediction and treatment model is Categorical Crossentropy.

[0079] The optimizer for the waste soil prediction and treatment model is the Adam optimizer, and the initial learning rate of the Adam optimizer is 0.001.

[0080] L2 regularization is applied in the fully connected layer, and the weight decay coefficient is 0.001.

[0081] Optionally, after obtaining the updated seismic wave training set, the updated seismic wave training set can be divided into a training set, a test set, and a validation set. In the subsequent model training process, the training set is used to train the model, the test set is used to test the model, and the validation set is used to evaluate the model.

[0082] For example, the framework and model design of the neural network can be done using Python, specifically a convolutional neural network (CNN) model designed based on the TensorFlow framework. The input layer receives data in the form of N*8*3, where N is the time series length, 8 represents 8 different seismic wave data points, and 3 represents the three components of each seismic wave data point. The flattening layer can convert the multidimensional feature maps generated by the convolutional layer into one-dimensional vectors, ready for input to the fully connected layer.

[0083] Optionally, the stride of the first, second, third, and fourth convolutional layers is all 1, and the padding method is 'same' to keep the output size constant. When training the neural network using the seismic wave training set, the batch size is set to 32, the number of iterations is set to 50, and an early stopping strategy is adopted, that is, training is terminated early when the performance on the validation set no longer improves, in order to avoid overfitting.

[0084] For example, the input dimension of the input layer can be a CNN architecture containing spatiotemporal convolutional layers, which can accept inputs of shape [number of samples, time length, spatial height, spatial width, number of channels]. The spatial height and width can correspond to the spatial layout of the 8 locations of the seismic wave detection device, and the number of channels can be 3.

[0085] The parameter settings for the first convolutional layer, the first pooling layer, the second convolutional layer, the second pooling layer, the third convolutional layer, the third pooling layer, the fourth convolutional layer, and the fourth pooling layer can be as follows:

[0086] In the first convolutional layer, the kernel size is 3x3. Small kernels capture local features while reducing the number of parameters and improving computational efficiency. The number of kernels is 64, and the stride is 1. A stride of 1 means the kernel moves pixel-by-pixel across the input data, ensuring as much detail as possible is captured. Padding is 'same'. The purpose of 'same' padding is to ensure the output feature map is the same size as the input, avoiding size reduction caused by convolution operations. The activation function is ReLU, which helps the model learn more complex patterns. The activation function converts all negative values ​​to 0 and retains positive values.

[0087] In the first pooling layer, the kernel size is 2x2, and the stride is 2. This first pooling layer is used for dimensionality reduction, decreasing the size of the feature map while retaining the most important features. A stride of 2 means moving two pixels at a time, further reducing the amount of data.

[0088] In the second convolutional layer, the kernel size is 3x3. The number of kernels is 128. The stride is 1. The padding is 'same'. The activation function is ReLU.

[0089] In the second pooling layer, the kernel size is 2x2, and the stride is 2. This further extracts higher-level features while continuing dimensionality reduction.

[0090] In the third convolutional layer, the kernel size is 3x3. The number of kernels is 256. The stride is 1. The padding is 'same'. The activation function is ReLU.

[0091] In the third pooling layer, the kernel size is 2x2, and the stride is 2. More complex features are then extracted, and dimensionality is further reduced.

[0092] In the fourth convolutional layer, the kernel size is 3x3. The number of kernels is 512. The stride is 1. The padding is 'same'. The activation function is ReLU.

[0093] In the fourth pooling layer, the kernel size is 2x2, and the stride is 2. The objective is to further extract the deepest features and continue dimensionality reduction.

[0094] The parameter settings for the first fully connected layer and the second fully connected layer can be as follows:

[0095] In the first fully connected layer, there are 1024 nodes. A large number of nodes allows for the capture of more feature combinations, improving the model's expressive power. The activation function is ReLU. Dropout is 0.5; Dropout randomly discards a portion of neurons to prevent overfitting and improve the model's generalization ability.

[0096] In the second fully connected layer, the number of nodes is 512. The activation function is ReLU. Dropout is 0.5. Further feature processing reduces the model's complexity while maintaining high expressive power.

[0097] The output layer parameters can be set as follows: Number of nodes: C (where C is the number of different types of construction waste). Activation function: Softmax.

[0098] Optionally, training can be stopped early when the accuracy on the validation set no longer improves. After each model training iteration, the trends of training loss, training accuracy, and validation loss and accuracy can be analyzed to evaluate the model's fit. If overfitting is found, the model training process should be restarted, and parameters such as the number of iterations, learning rate and its decay strategy, and batch size should be reset based on the difference between training and validation performance. Adjustments to the activation function should also be considered to reduce overfitting. The model should then be trained again until a satisfactory performance level is achieved.

[0099] Next, check whether the accuracy on the training and validation sets meets the expected standard. If not, return to the model training step to continue optimization; otherwise, once the required accuracy is met, save the current optimal weight configuration.

[0100] The model then enters the performance testing phase, using data independent of the training process (the test set) to evaluate its performance. By inputting samples from the test set into the trained model, predictions are obtained and compared with actual values ​​to measure the consistency between the model's output and the real-world situation. This step aims to test the model's ability to generalize to unknown data. If the accuracy obtained in the testing phase is lower than the preset target value, the model needs to be returned to the training phase for further adjustments and retraining.

[0101] S4. Obtain real-time geological forecast data, preprocess the real-time geological forecast data, and input the preprocessed real-time geological forecast data into the slag and soil prediction and processing model. Output the slag and soil type through the slag and soil prediction and processing model.

[0102] For example, real-time geological forecast data can be obtained by deploying seismic wave exciters and detectors inside the tunnel to ensure the effective excitation of active source seismic waves and the reception of vibration signals throughout the space.

[0103] Optionally, after acquiring real-time geological forecast data, data preprocessing can be performed on the real-time geological forecast data. Data preprocessing may include noise removal, missing value imputation, and data standardization.

[0104] Optionally, during the tunnel boring machine (TBM) construction process, the prediction results and actual construction conditions can be recorded to continuously optimize the spoil prediction and treatment model, thereby improving the accuracy and reliability of the prediction. The optimization method can be to fine-tune the spoil prediction and treatment model.

[0105] For example, one excitation point and eight three-component geophones are deployed inside the tunnel. The excitation point is located on the tunnel lining segments to ensure the effective excitation of active source seismic waves. The geophones are spaced 2 meters apart, with a minimum offset of 2.5 meters and a maximum offset of 16.5 meters.

[0106] S5. Based on the type of excavated soil and in conjunction with the working status of the tunnel boring machine and geological forecasts, generate excavated soil treatment suggestions to handle the excavated soil generated during the tunnel boring process.

[0107] For example, appropriate treatment methods are selected based on the characteristics of different types of construction waste and environmental protection requirements.

[0108] Optionally, appropriate processing equipment can be provided according to the actual needs of different processing methods, including curing agent spraying devices, screening machines, cleaning machines, etc.

[0109] Alternatively, waste disposal recommendations may include: For soil-type waste, which contains a large amount of fine particles (clay, sand, etc.) and has a high moisture content, usually exceeding the liquid limit and approaching saturation. This type of waste typically originates from excavation in completely weathered rock or soil strata. Treatment methods: Due to its high moisture content and good fluidity, moisture can be reduced using natural air drying, additive solidification methods (such as cement and lime), filtration separation (by setting up waste disposal pits), and mechanical dewatering methods (such as using filter presses). The equipment used includes: screening systems (for separating coarse particles), slurry mixing devices (to adjust fluidity), and filter presses (plate and frame filter presses, belt thickeners, centrifugal dewatering machines, or screw presses). For intermediate-type waste: This type of waste has a particle size distribution between soil-type and stony-type waste, with a stony content between 40% and 70%, exhibiting mixed characteristics and forming a dense-skeleton structure. This type of waste typically originates from moderately weathered rock strata. Treatment methods: Natural air drying, additive solidification (such as cement or lime), filtration, and mechanical dewatering can all be used. The machinery used is similar to that for soil-type waste soil, but more attention may be needed to process the coarse aggregate to ensure the quality of the final product. For stone-type waste soil: This type of waste soil contains a large amount of coarse aggregate (crushed stone), has a low moisture content, and high permeability. This type of waste soil typically comes from excavation in weakly weathered rock strata. Treatment methods: The main method is to separate the coarse aggregate from the fine aggregate using a screening system. The remaining fine aggregate is then processed using natural air drying, additive solidification (such as cement or lime), or mechanical dewatering. The machinery used includes: a screening system (for separating coarse particles), a slurry mixing device (to adjust fluidity), and a filter press (plate and frame filter press, belt thickener / dewatering machine, centrifugal dewatering machine, or screw press dewatering machine).

[0110] In this embodiment, the waste soil prediction and processing model of the present invention can be loaded into the computer of the tunnel boring machine control system. A stable data interface is developed to enable the model to acquire real-time sensor data from the tunnel boring machine's central control system. A feedback mechanism is designed to ensure that the processed analysis results are promptly fed back to the control system to guide waste soil processing. The trained waste soil prediction and processing model is adaptively adjusted to better suit the data characteristics under actual working conditions, and the entire process is thoroughly tested in a simulation environment to ensure that the model's accuracy meets engineering requirements. Finally, a regular inspection mechanism is established to track the system's operating status, collect feedback information, continuously improve the algorithm's accuracy, and continuously update and improve the model based on project progress and actual application effects, thereby achieving more efficient and accurate waste soil prediction.

[0111] Example 2

[0112] like Figure 2As shown, a smart waste disposal system based on advanced tunnel geological prediction is implemented using a smart waste disposal method based on advanced tunnel geological prediction. The system includes:

[0113] The model training module is used to acquire a seismic wave training set; to perform geological analysis on the waste soil categories in the seismic wave training set, to obtain the dynamic calculation parameters and physical data corresponding to each waste soil category, and to map the seismic wave characteristics with the dynamic calculation parameters and physical data to obtain an updated seismic wave training set; and to use the seismic wave training set to train the neural network to obtain a waste soil prediction and processing model.

[0114] The seismic wave training set includes seismic wave features and the corresponding soil and debris category for each seismic wave feature.

[0115] The real-time data acquisition module is used to acquire real-time geological forecast data, preprocess the real-time geological forecast data, and input the preprocessed real-time geological forecast data into the slag prediction and treatment model to obtain the type of slag generated during shield tunneling.

[0116] The waste soil analysis module is used to acquire real-time geological forecast data, input the real-time geological forecast data into the waste soil prediction and processing model, and obtain the type of waste soil generated during shield tunneling.

[0117] The waste disposal module is used to generate waste disposal suggestions based on the type of waste, combined with the working status of the tunnel boring machine and geological forecasts, in order to handle the waste generated during the tunnel boring process.

[0118] Furthermore, the real-time data acquisition module includes a seismic wave exciter and a detector;

[0119] The seismic wave generator is used to generate active source seismic waves, and the detector is used to receive vibration signals throughout space.

[0120] Furthermore, the intelligent waste disposal system also includes:

[0121] The data preprocessing module is used to preprocess the seismic wave training set to obtain the preprocessed seismic wave training set.

[0122] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of the invention.

Claims

1. A method for intelligent treatment of construction waste soil combined with advanced geological prediction in tunnels, characterized in that, The method includes: S1. Obtain the seismic wave training set; The seismic wave training set includes seismic wave features and the corresponding soil and debris category for each seismic wave feature. S2. Perform geological analysis on the soil and debris categories in the seismic wave training set, obtain the dynamic calculation parameters and physical data corresponding to each soil and debris category, and map the seismic wave characteristics with the dynamic calculation parameters and physical data to obtain an updated seismic wave training set. S3. Use the earthquake wave training set to train the neural network to obtain the slag and soil prediction and treatment model; S4. Obtain real-time geological forecast data, preprocess the real-time geological forecast data, input the preprocessed real-time geological forecast data into the slag soil prediction and processing model, and output the slag soil type through the slag soil prediction and processing model. S5. Based on the type of excavated soil, and in conjunction with the working status of the tunnel boring machine and geological forecasts, generate excavated soil treatment suggestions to treat the excavated soil generated during the tunnel boring process.

2. The method according to claim 1, characterized in that, Following S1, the method further includes: The seismic wave training set is preprocessed to obtain a preprocessed seismic wave training set; S2 specifically includes: Geological analysis is performed on the soil and debris categories in the preprocessed seismic wave training set to obtain the dynamic calculation parameters and physical data corresponding to each soil and debris category. The seismic wave characteristics are then mapped to the dynamic calculation parameters and physical data to obtain an updated seismic wave training set.

3. The method according to claim 2, characterized in that, In S2, the dynamic calculation parameters include longitudinal wave velocity, transverse wave velocity, wave velocity ratio, density, Poisson's ratio, Young's modulus, and bulk modulus. The physical data include porosity, water content, and particle size distribution.

4. The method according to claim 3, characterized in that, In S3, the slag and soil prediction and processing model includes an input layer, a convolutional layer, a flattened layer, a fully connected layer, and an output layer. The data obtained by the input layer is in the form of N*8*3; The convolutional layer includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, a fourth convolutional layer, and a fourth pooling layer; The first convolutional layer has 64 kernels, the second convolutional layer has 128 kernels, the third convolutional layer has 256 kernels, and the fourth convolutional layer has 512 kernels. The kernel size of the first, second, third, and fourth convolutional layers is 3*3, and the activation function of the first, second, third, and fourth convolutional layers is the ReLU function. The pooling window size of the first pooling layer, the second pooling layer, the third pooling layer, and the fourth pooling layer is 2*2. The fully connected layer includes a first fully connected layer and a second fully connected layer; The first fully connected layer has 1024 nodes, and the second fully connected layer has 512 nodes; The activation functions of both the first fully connected layer and the second fully connected layer are ReLU functions; The Dropout rates of both the first fully connected layer and the second fully connected layer are 0.

5. The number of nodes in the output layer is the same as the number of waste soil categories.

5. The method according to claim 4, characterized in that, In S3, the loss function of the waste soil prediction and treatment model is Categorical Crossentropy. The optimizer for the waste soil prediction and treatment model is the Adam optimizer, and the initial learning rate of the Adam optimizer is 0.

001. L2 regularization is applied in the fully connected layer, and the weight decay coefficient is 0.

001.

6. A combined tunnel advanced geological prediction intelligent waste soil treatment system, implemented based on the combined tunnel advanced geological prediction intelligent waste soil treatment method according to any one of claims 1-5, characterized in that, include: The model training module is used to acquire a training set of seismic waves; Geological analysis is performed on the waste soil categories in the seismic wave training set to obtain the dynamic calculation parameters and physical data corresponding to each waste soil category. The seismic wave characteristics are then mapped to the dynamic calculation parameters and physical data to obtain an updated seismic wave training set. The seismic wave training set is used to train the neural network to obtain a waste soil prediction and processing model. The seismic wave training set includes seismic wave features and the corresponding soil and debris category for each seismic wave feature; The real-time data acquisition module is used to acquire real-time geological forecast data, preprocess the real-time geological forecast data, and input the preprocessed real-time geological forecast data into the slag prediction and processing model to obtain the type of slag generated during shield tunneling. The slag analysis module is used to acquire real-time geological forecast data, input the real-time geological forecast data into the slag prediction and processing model, and obtain the type of slag generated during shield tunneling. The waste disposal module is used to generate waste disposal suggestions based on the type of waste, combined with the working status of the tunnel boring machine and geological forecasts, in order to handle the waste generated during the tunnel boring process.

7. The intelligent waste disposal system according to claim 6, characterized in that, The real-time data acquisition module includes a seismic wave exciter and a detector; The seismic wave exciter is used to excite active source seismic waves, and the detector is used to receive vibration signals throughout space.

8. The intelligent waste disposal system according to claim 7, characterized in that, The intelligent waste disposal system also includes: The data preprocessing module is used to preprocess the seismic wave training set to obtain a preprocessed seismic wave training set.

Images