Targeted positioning grouting efficient plugging reinforcement method and system based on deep learning

Abstract

The invention discloses a targeted positioning grouting efficient plugging reinforcement method and system based on deep learning, and the method comprises the steps: inputting geological data into a three-dimensional geological weakness prediction model for prediction, obtaining three-dimensional geological weakness distribution, inputting the three-dimensional geological weakness distribution into a targeted grouting prediction model for prediction, and obtaining an optimal grouting point location and grouting parameters of a current engineering region; a set threshold value of the error is reconstructed based on the targeted grouting prediction model, and when the error is within the set threshold value, targeted grouting is conducted through the optimal grouting point location and the grouting parameters; when the error exceeds a set threshold value, the geological environment is updated through the three-dimensional geological weakness prediction model, and weakness distribution of the targeted grouting prediction model is updated again; and setting operation parameters of the grouting equipment based on the grouting parameters so as to realize a target grouting effect. According to the method, whether the grouting abnormal condition exists on the basis of the established grouting strategy or not is periodically judged through error threshold analysis, and the originally predicted grouting optimal point location and grouting parameters can be adjusted in time according to the abnormal condition.

Description

Technical Field

[0001] The present invention relates to the technical field of grouting plugging and reinforcement, and particularly to a method and system for efficient plugging and reinforcement of targeted grouting based on deep learning. Background Art

[0002] The statements in this section merely provide background technical information related to the present disclosure and do not necessarily constitute prior art.

[0003] With the rapid and vigorous development of the economy, the construction of underground projects, including tunnel projects, has entered a new stage. As the scale of engineering construction becomes larger, the depth of advancement becomes deeper, and the geological conditions passed through become more complex, the geological disasters encountered during the construction process are also more strongly sudden and destructive, and it is becoming increasingly important to effectively prevent and control complex geological disasters. The grouting project can effectively treat the sudden disasters of underground projects. Through the solidification and cementation of the slurry with the surrounding rock mass structure, it can effectively block the recharge and runoff of the aquifer, transform the water-bearing property of the aquifer, and reinforce the soft rock mass. However, the ineffective diffusion of the slurry is relatively serious during the construction of conventional grouting technology for different stratum medium conditions, especially in the strata with dynamic water plugging and complex fracture connectivity. The reinforcement of rock and soil masses and the effective plugging of sudden water inrush are not ideal.

[0004] The targeted construction research of existing grouting projects mainly focuses on the parameter control of grouting pressure, flow rate, and material ratio, which is a global adjustment of grouting. The problems that occur during the grouting process are difficult to solve in real time, easy to appear grouting blind spots, and it is difficult to control the penetration path and diffusion depth of the slurry, resulting in poor regional grouting effects. In the adjustment of the grouting plan, the problems that occur during the grouting process in real time, the operation of grouting equipment, and the timely feedback of grouting effects have not been comprehensively considered. Summary of the Invention

[0005] To overcome the deficiencies of the above-mentioned prior art, the present invention provides a method and system for efficient plugging and reinforcement of targeted grouting based on deep learning. By analyzing the error threshold, it periodically determines whether there is an abnormal grouting situation based on the established grouting strategy, and can timely adjust the original predicted optimal grouting point and grouting parameters according to the abnormal situation. At the same time, combined with the operation status of the grouting equipment parameters, it compares the real-time status of the equipment with the target grouting effect, and adjusts the output power and pressure of the grouting equipment to ensure that the slurry can accurately reach the targeted area.

[0006] To achieve the above object, one or more embodiments of the present invention provide the following technical solutions:

[0007] In the first aspect, the present invention provides a method for efficient plugging and reinforcement of targeted grouting based on deep learning, including:

[0008] Obtain the geological data of the current engineering area and perform preprocessing to obtain the preprocessed geological data;

[0009] Input the preprocessed geological data into a pre-trained three-dimensional geological weakness prediction model for prediction to obtain the three-dimensional geological weakness distribution;

[0010] Input the three-dimensional geological weakness distribution into a pre-trained targeted grouting prediction model for prediction to obtain the optimal grouting points and grouting parameters in the current engineering area; Based on the set threshold of the reconstruction error of the targeted grouting prediction model, when the error is within the set threshold, use the optimal grouting points and grouting parameters for targeted grouting; when the error exceeds the set threshold, update the geological environment through the three-dimensional geological weakness prediction model and perform re-update of the weakness distribution of the targeted grouting prediction model;

[0011] Set the operating parameters of the grouting equipment based on the grouting parameters. Based on the basic principle of multi-agent deep reinforcement learning, regard each grouting equipment as an individual agent, and optimize the operating parameters of each agent through the reinforcement learning strategy to achieve the target grouting effect.

[0012] In a further technical solution, the three-dimensional geological weakness prediction model is constructed using a generative adversarial network, including a generator and a discriminator. The variational autoencoder is used as the backbone structure of the network generator, and the Transformer is used as the backbone structure of the discriminator.

[0013] In a further technical solution, the data processing process in the three-dimensional geological weakness prediction model is specifically as follows:

[0014] Take the preprocessed geological data as the input data and input it into the generator with the variational autoencoder as the backbone structure to output the three-dimensional geological weakness distribution. At the same time, associate the three-dimensional geological weakness distribution with the geological data to obtain the geological data carrying parameter information;

[0015] Then pass through the discriminator with the Transformer as the backbone structure to compare the distribution of bad geology in the actual stratum with the generated three-dimensional geological weakness distribution, and at the same time use mutual information as the loss term;

[0016] Based on backpropagation for update and optimization to solve, when the model generated by the generator is close to the real model distribution and the loss term meets the requirements, complete the model training.

[0017] In a further technical solution, the targeted grouting prediction model is constructed based on a convolutional neural network. Different three-dimensional geological weakness distributions are output through the three-dimensional geological weakness prediction model, including single weakness and multi-weakness combinations.

[0018] Further technical solution: the single weakness is targeted grouting according to the characteristics of different weaknesses, and the combination of multiple weaknesses is predicted according to a pre-trained targeted grouting prediction model. The neural network learns from the historical grouting project data to learn the relationship between the distribution of different three-dimensional geological weaknesses and the optimal grouting points and grouting parameters in the presence of multiple weaknesses;

[0019] The formula is expressed as:

[0020] S(z i ) t =v(x i ,S(z i ) t-1 ) t +b(t)

[0021] where S represents the different distribution situations of weaknesses at the current time step, z i represents the different distribution situations of weaknesses, t and t - 1 represent time steps, i represents different weakness classifications, x i represents the actual grouting parameters corresponding to the weaknesses, v represents the situation of the weakness parameters at the current time step under the combined action of the selection of grouting parameters at the current time step and the weakness parameters at the previous time step, and b(t) represents the influence of the surrounding environment on grouting at each time step.

[0022] Further technical solution: optimize the parameters of each agent through a reinforcement learning strategy, and the set reward function is expressed as:

[0023]

[0024] where R is the state quantity corresponding to the actual grouting progress after quantization of different data for each action, m represents the total number of devices, n represents the total number of actions of device j, j represents different devices, k represents a certain action of device j, t is the time step, α(t) and β(t) respectively represent the adjustment coefficients at different times, w j (t) represents the loss of device j at a certain time, λ j (t) is the reward coefficient corresponding to the actual grouting progress of the state quantity of different data for each action of each device at different times, Y j represents the state quantity of a certain action of device j at a certain time, y k represents the state quantity of device j at a certain time step under action k, Y j (y k ,t) is the state quantity of each action of each device at different times.

[0025] Further technical solution: determine the action combination value required for the grouting parameters corresponding to each agent, and the combination value calculation is expressed as:

[0026]

[0027] Among them, x i represents the actual grouting parameters corresponding to the weaknesses, j represents different devices, k represents a certain action of device j, n represents the total number of actions of device j, and γ j,k represents the weight of the contribution value of a certain action k of device j to the parameter, and y k represents the state quantity of device j at a certain time step under action k, and Z j (y k ) represents the contribution value corresponding to the state quantity.

[0028] In a second aspect, the present invention provides a high-efficiency plugging and reinforcement system for targeted positioning grouting based on deep learning, including:

[0029] A data acquisition module, which is configured to: acquire geological data of the current engineering area and perform preprocessing to obtain preprocessed geological data;

[0030] A weakness distribution prediction module, which is configured to: input the preprocessed geological data into a pre-trained three-dimensional geological weakness prediction model for prediction to obtain a three-dimensional geological weakness distribution;

[0031] A grouting prediction module, which is configured to: input the three-dimensional geological weakness distribution into a pre-trained targeted grouting prediction model for prediction to obtain the optimal grouting points and grouting parameters in the current engineering area; based on a set threshold of the reconstruction error of the targeted grouting prediction model, when the error is within the set threshold, perform targeted grouting using the optimal grouting points and grouting parameters; when the error exceeds the set threshold, update the geological environment through the three-dimensional geological weakness prediction model and perform re-update of the weakness distribution of the targeted grouting prediction model;

[0032] A grouting equipment operation module, which is configured to: set the operation parameters of the grouting equipment based on the grouting parameters, and based on the basic principle of multi-agent deep reinforcement learning, regard each grouting equipment as an independent agent, and optimize the operation parameters of each agent through the reinforcement learning strategy to achieve the target grouting effect.

[0033] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps in the method for high-efficiency plugging and reinforcement of targeted positioning grouting based on deep learning as described in the first aspect.

[0034] In a fourth aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the program, it implements the steps in the method for high-efficiency plugging and reinforcement of targeted positioning grouting based on deep learning as described in the first aspect.

[0035] The above one or more technical solutions have the following beneficial effects:

[0036] The present invention obtains geological data of the engineering area, analyzes the poor geological attributes (i.e., weaknesses) of the construction area, preprocesses the data, establishes a geological database obtained by different monitoring means, and simultaneously performs three-dimensional tagging of the data. A three-dimensional geological weakness prediction model is established, and through a generative adversarial network, local estimation is carried out on actual void channels, loose strata, water-rich areas, etc., providing accurate positioning information for targeted reinforcement of grouting treatment. A targeted grouting prediction model is established. Based on the determined distribution of three-dimensional geological weaknesses, the optimal grouting points and grouting parameters in this area are predicted. At the same time, a threshold value for the reconstruction error of the targeted grouting prediction model is set to meet the adjustment of the grouting strategy in abnormal situations, fully considering abnormal situations during the grouting process, and being able to timely adjust the originally predicted optimal grouting points and grouting parameters according to abnormal situations.

[0037] The present invention also, based on the basic principle of multi-agent deep reinforcement learning, sets the consistency of the operating parameters and grouting parameters between multiple devices, compares the real-time state of the devices with the target grouting effect, and adjusts the output power and pressure of the grouting equipment to ensure that the grout can accurately reach the targeted area. After the grouting of each cycle section is completed, an evaluation and feedback of the grouting effect is carried out. By comparing the parameter characteristics of adjacent grouting cycles, the success rate of grouting and the long-term stability of rock and soil reinforcement are verified.

[0038] The present invention obtains the poor geological area of the formation by establishing a three-dimensional geological weakness prediction model, constructs a targeted grouting prediction model to obtain the optimal grouting points and grouting parameters, and at the same time, based on the targeted grouting prediction model, a new error threshold is set (different from the loss for judging convergence during the training of the original model). Through the error threshold analysis, it is periodically determined whether there are abnormal grouting situations based on the established grouting strategy and adjusted. At the same time, in combination with the operating conditions of the grouting equipment parameters, the real-time state of the equipment is compared with the target grouting effect, and the output power and pressure of the grouting equipment are adjusted to ensure that the grout can accurately reach the targeted area. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings forming a part of this specification are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0040] Figure 1 It is a flowchart of the plugging and reinforcement method of the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0042] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0043] In the case of no conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0044] Embodiment 1

[0045] As Figure 1 shown, this embodiment discloses an efficient plugging and reinforcement method for targeted positioning grouting based on deep learning. The method includes the following steps:

[0046] S1: Obtain the geological data of the current engineering area and perform preprocessing to obtain the preprocessed geological data;

[0047] In this embodiment, by using the propagation characteristics of seismic waves and acoustic waves in different geological formations, reflection and refraction data are collected to reveal information such as density and porosity in the formation. Through ground penetrating radar, internal imaging of the formation is carried out to obtain a cross-sectional view of the geological structure, providing formation structure information in the form of images, and identifying fissures, cavities, etc. in the formation. Through borehole monitoring, information such as pressure and temperature in the formation is obtained to judge the compactness of the formation and the uniformity of the structure, and physical tests and element information identification of borehole core samples are carried out to analyze the poor geological properties of the construction area, providing basic data for local identification.

[0048] The missing values in the obtained geological data are filled by mean filling, and the outlier detection method is used to identify and remove outliers to ensure data quality. The obtained different types of data are standardized and normalized to adapt to the model input format, ensuring that the model converges quickly and improving the prediction accuracy. A geological database containing geological data obtained by different monitoring means is established.

[0049] The actual formation grouting area is divided three-dimensionally, and the three-dimensional data in the database, i.e., the obtained geological data, is labeled to mark the formation properties of each area, such as density, porosity, element properties, etc. Specifically, the geological data is partitioned according to different scales, and three-dimensional solid meshes with different thicknesses and degrees of structuring are divided in different scale areas. The centroid of each three-dimensional grid point is used as the data storage point. By labeling the three-dimensional data in the database, the formation properties of each area, such as density, porosity, element properties, etc., are marked and assigned to the data storage points of each grid. By aligning all the data to a unified three-dimensional coordinate system, data noise reduction is performed through convolutional filtering, data is completed using GAN, and data labeling is carried out.

[0050] S2: Input the preprocessed geological data into a pre-trained three-dimensional geological weakness prediction model for prediction to obtain the three-dimensional geological weakness distribution;

[0051] In this embodiment, the three-dimensional geological weakness prediction model is constructed using a generative adversarial network, including a generator and a discriminator. The variational autoencoder VAE is used as the backbone structure of the network generator, and the Transformer is used as the backbone structure of the discriminator.

[0052] The data processing process in the three-dimensional geological weakness prediction model is specifically as follows:

[0053] (1) The preprocessed geological data (three-dimensional data) is used as input data and input into the three-dimensional geological weakness prediction model. After passing through the generator with the variational autoencoder as the backbone structure, the three-dimensional geological weakness distribution (such as loose formations, water-rich areas, karst pipelines, etc.) is output. At the same time, the three-dimensional geological weakness distribution is associated with the data in the database to obtain geological data carrying parameter information, that is, each generated three-dimensional geological weakness distribution carries corresponding geological parameter information. For example, a higher porosity is associated with a loose formation, while lower density and resistivity are associated with voids, fractures, and water-rich areas. The carried parameter information is subjected to normalized data processing and data noise reduction.

[0054] (2) Then, through the discriminator with the Transformer as the backbone structure, the distribution of bad geology in the actual formation is compared with the generated three-dimensional geological weakness distribution, and mutual information is used as the loss term, that is:

[0055] L a = I(M; N) (1)

[0056] Among them, M and N are the parameter characteristics carried by different three-dimensional geological weaknesses, and I(M; N) represents the mutual information between them. By minimizing the information between different parameters, accurate determination of different three-dimensional geological weakness distributions is ensured.

[0057] (3) Based on the update and optimization of backpropagation, when the model generated by the generator is close to the real model distribution and the loss term meets the requirements, the model training is completed.

[0058] Geological weaknesses refer to the poor geological areas of the strata. Based on the three-dimensional geological weakness distribution, combined with the area, depth, and connectivity of the regional distribution, the actual void channels, loose strata, water-rich areas, etc. are locally estimated to provide accurate positioning information for the targeted reinforcement of grouting treatment.

[0059] S3: Input the three-dimensional geological weakness distribution into the pre-trained targeted grouting prediction model for prediction to obtain the best grouting points and grouting parameters in the current engineering area; based on the set threshold of the reconstruction error of the targeted grouting prediction model, when the error is within the set threshold, use the best grouting points and grouting parameters for targeted grouting; when the error exceeds the set threshold, update the geological environment through the three-dimensional geological weakness prediction model and re-update the weakness distribution of the targeted grouting prediction model;

[0060] In this embodiment, the targeted grouting prediction model is constructed based on a convolutional neural network. Different three-dimensional geological weakness distributions are output through the three-dimensional geological weakness prediction model, including single weakness and multi-weakness combination situations. Starting from engineering experience and historical data of actual engineering grouting effects, the input is three-dimensional geological weakness distribution information. The model performs convolutional operations on this input data to represent it in the three-dimensional data space, and the output data is data such as the best grouting points and grouting parameters.

[0061] Based on the obtained three-dimensional geological weakness distribution information, predict the best grouting points and grouting parameters in the current engineering area. Consider the specific classifications of weaknesses including loose strata, fractures, karst pipelines, water-rich areas, high temperature, etc.

[0062] For a single weakness, such as using different slurry types for loose strata to achieve better penetration effect and form stable slurry veins and stone bodies. For water-rich fractures or karst pipelines, consider their fracture apertures, extension degrees, pipe diameters, and branch situations of karst pipelines for targeted grouting.

[0063] For the situation of multiple weaknesses, it is predicted according to the pre-trained targeted grouting prediction model. The neural network learns from the historical grouting engineering data, learns the relationship between different three-dimensional geological weakness distributions carrying different geological parameter information and the best grouting points and grouting parameters in the case of multiple weaknesses, and trains the model to be able to quickly output the best grouting points and grouting parameters when different three-dimensional geological weakness distributions are input.

[0064] The formula is expressed as:

[0065] S(z i )t = v(x i , S(z i ) t-1 ) t + b(t) (2)

[0066] where S represents different distribution situations of weaknesses at the current time step, z i represents different distribution situations of weaknesses, t and t - 1 represent time steps, i represents different weakness classifications, x i represents the actual grouting parameters corresponding to the weaknesses, such as slurry selection, grouting points, etc. v represents the situation of the grouting parameter selection at the current time step and the parameters of the weaknesses (poor geological areas) at the previous time step acting together to determine the weakness parameters at the current time step. b(t) represents the influence of grouting on the surrounding environment at each time step. It can be considered that, in the set time step loop, based on the obtained weakness distribution information, the three-dimensional geological weakness distribution is complemented through the optimal grouting points and grouting parameters at each time step. Numerically, it can be expressed as within the set time step loop, assuming that the initial three-dimensional geological weakness is assigned a value of 1, the grouting points and grouting parameters are dynamically adjusted at each time step, and finally the three-dimensional geological weakness distribution is assigned a value of 0, that is, the three-dimensional geological weakness can be completed through grouting.

[0067] Use three-dimensional pooling to retain important features and improve the computational efficiency and spatial perception ability of the model. Perform multi-layer convolutional pooling, and at the same time add residual blocks and skip connections in the deep network to maintain the depth of the model, avoid gradient disappearance, and further improve the model's representation ability for complex geological structures.

[0068] Adopt an adaptive optimization strategy to adjust the parameters of the training data and obtain a targeted grouting prediction model under normal conditions.

[0069] In this embodiment, the targeted grouting prediction model will have corresponding time step data at each time step. Through the set threshold of the reconstruction error, it is periodically determined whether the geological weakness changes under the conditions of the predicted optimal grouting points and grouting parameters. Periodicity means setting the error threshold judgment under different time periods. That is to say, the original loss is used to train the model, and now the reconstruction error threshold is used to periodically determine whether the original optimal grouting points and grouting parameters can be completed under the existing grouting parameter conditions. The error threshold is set as the percentage error from the predicted value of the original model according to the actual grouting process.

[0070] Based on the established targeted grouting prediction model, set the threshold of the reconstruction error to meet the adjustment of the grouting strategy under abnormal conditions. And set the period of each threshold update. Specifically, after the targeted grouting prediction model outputs the optimal grouting points and grouting parameters in the current engineering area, perform the error threshold determination again (this threshold determination is different from the loss for judging convergence during the training of the original targeted grouting prediction model). Analyze the periodicity through the error threshold to determine whether there are abnormal grouting situations based on the established grouting strategy and make adjustments.

[0071] (1) When the error is within the set threshold, it meets the targeted grouting standard, and perform targeted grouting according to the output of the targeted grouting prediction model.

[0072] (2) When the error exceeds the set threshold, that is, an abnormal situation is detected during the grouting process. The abnormal grouting situation is divided into two types: First, as the grouting progresses, the three-dimensional geological weaknesses in the grouted area will be continuously improved, that is, it is necessary to periodically update and adjust the above-generated optimal parameter ratio; Second, during the grouting process, there are new three-dimensional geological weaknesses generated in the grouted area due to grouting disturbance, or the grouting effect does not meet the ideal requirements due to human and environmental factors during the actual grouting process. Adjust the best grouting points and grouting parameters predicted originally.

[0073] When the error is within the set threshold, there is no need to continue updating. When the error is outside the set threshold, determine the situation of exceeding the error threshold for different geological weaknesses or comprehensive multiple weaknesses through the situations such as the water output in the grouted area not decreasing in time, abnormal hole collapse, abnormal fluctuation of the slurry concentration, slurry loss and blockage, and sudden change of temperature during the grouting process, and determine whether it belongs to the first abnormal situation or the second abnormal situation. Update and analyze the geological environment ahead through the three-dimensional geological weakness prediction model, consider the adaptation relationship between the grouting effect and the environmental impact, and perform re-updating of the weakness distribution of the targeted grouting prediction model.

[0074] At the same time, consider the periodic update of the targeted grouting area required ahead under the condition of the grouting plugging that has been carried out, that is, if the abnormal situation is that the targeted grouting plugging effect is good, resulting in an increase in the grouting plugging and reinforcement efficiency, then adjust the time step of the targeted grouting prediction model to adapt to the actual grouting effect on site.

[0075] S4: Set the operating parameters of the grouting equipment based on the grouting parameters. Based on the basic principle of multi-agent deep reinforcement learning, regard each grouting equipment as an individual agent, and optimize the operating parameters of each agent through the reinforcement learning strategy to achieve the target grouting effect.

[0076] In this embodiment, grouting equipment is also combined for multi-modal regulation. Specifically, based on the basic principle of multi-agent deep reinforcement learning, each grouting equipment and monitoring component is regarded as an individual agent. The operating parameters of each agent are optimized through reinforcement learning strategies. At the same time, the coordinated operation of multiple grouting equipments is monitored and controlled, and the consistency of the operating parameters and grouting parameters between multiple equipments is set. The operating parameters are parameters such as the pressure, flow rate, and monitoring angle and azimuth of the grouting equipment.

[0077] Through the reinforcement learning strategy, continuously optimize the changes in parameters such as the pressure, flow rate, and monitoring angle and azimuth of each agent, analyze the independent actions and mutual influences of multiple agents, process non-linear coordination problems, achieve multi-source information fusion, precisely adjust each link of the grouting process, take actions according to real-time state data (i.e., the data of the grouting equipment and monitoring components themselves and the acquired data), and achieve the overall grouting goal through the coordinated operation of the agents.

[0078] The specific steps of the reinforcement learning strategy are as follows:

[0079] (1) Establish a continuous action space, including the startup of each equipment, and establish different continuous action models for different equipments (such as the startup of grouting equipment, grouting flow rate, grouting pressure, slurry selection, the position and pipe type of the grouting pipe, the embedding and monitoring angle of the sensing equipment, etc.). At the same time, set the continuous movement of different equipments in the space.

[0080] (2) Quantify the data of each step of the actions of the equipments in the continuous space (such as setting the rotation angle to 360°, the grouting rate to 5m 3 / h, etc.), and display the state quantities of each equipment in a dataized manner.

[0081] (3) Synthesize the three-dimensional geological weakness distribution, the best grouting points, and grouting parameters obtained above, and combine historical data to analyze the actual progress of grouting treatment for three-dimensional geological weaknesses under the state quantities of different data quantifications for each action, and analyze the optimal actions and their state quantity selections under different combinations. The reward function corresponding to the state quantity of different data quantifications for different actions can be set as follows:

[0082]

[0083] Among them, R is the actual grouting progress corresponding to the state quantity of different data quantifications for each action, m represents the total number of equipments, n represents the total number of actions of equipment j, j represents different equipments, k represents a certain action of equipment j, t is the time step, α(t) and β(t) respectively represent the adjustment coefficients at different times, w j (t) represents the loss amount of equipment j at a certain time, λ j(t) is the reward coefficient corresponding to the actual grouting progress of each device's different data quantization status quantities for each action at different times, Y j represents the status quantity of a certain action of device j at a certain time, y k represents the status quantity of device j at a certain time step under action k, Y j (y k , t) are the status quantities of each device's each action at different times.

[0084] Compare the real-time status of the device with the target grouting effect, and adjust the output power and pressure of the grouting device to ensure that the slurry can accurately reach the target area.

[0085] Based on the status quantities of the existing devices' data quantization reaching the grouting parameters predicted by the target grouting prediction model, optimize the total grouting progress R, that is, the operation of each grouting device corresponds to achieving the actual grouting parameters. After determining the grouting parameters through the target grouting prediction model, determine the combined values of each action required for the corresponding grouting parameters of each agent. For example, after determining the slurry type, adjust the slurry-water ratio and conveying speed of the slurry preparation platform to reach the target value. The combined value calculation can be expressed as:

[0086]

[0087] Among them, x i represents the actual grouting parameters corresponding to the weak points such as slurry type selection, grouting point location, etc. j represents different devices, k represents a certain action of device j, n represents the total number of actions of device j, γ j,k represents the weight of the contribution value of a certain action k of device j to the parameter, y k represents the status quantity of device j at a certain time step under action k, Z j (y k ) represents the contribution value corresponding to the status quantity.

[0088] Based on the known actions meeting the constraints of the above formula, obtain the maximum value of the total grouting progress R based on the Lagrange multiplier method.

[0089] The grouting footage cycle refers to the distance that needs to be advanced to complete a grouting cycle during the grouting construction process. After each grouting footage cycle, re-conduct geophysical exploration and data acquisition of the front formation, predict the front formation weaknesses through the three-dimensional geological weakness prediction model, and at the same time combine the water flow path and pressure distribution after each cycle of grouting, etc., and compare the regions and characteristics in the three-dimensional space with the previous grouting cycle, analyze the grouting plugging and reinforcement effect, take geological area samples for geotechnical physical property tests, verify the success rate of grouting and the long-term stability of geotechnical reinforcement, evaluate the grouting effect and give feedback.

[0090] After each grouting cycle footage, by means of advanced detection methods such as induced polarization, borehole television, cross-hole CT, advanced horizontal drilling, and data obtained from the engineering site, analyze the water flow path and pressure distribution after each cycle of grouting. At the same time, based on the geological data obtained from the current cycle section, obtain the three-dimensional geological weakness distribution of the current grouting cycle section through the three-dimensional geological weakness prediction model. Compare the three-dimensional geological weakness distribution, water flow path, and pressure distribution of the current grouting cycle section with the previous grouting cycle, analyze the grouting plugging and reinforcement effect, take geological area samples for geotechnical physical property tests, verify the success rate of grouting and the long-term stability of geotechnical reinforcement, evaluate the grouting effect and give feedback.

[0091] Example Two

[0092] This embodiment provides a high-efficiency plugging and reinforcement system for targeted positioning grouting based on deep learning, including:

[0093] A data acquisition module, which is configured to: acquire geological data of the current engineering area and perform preprocessing to obtain preprocessed geological data;

[0094] A weakness distribution prediction module, which is configured to: input the preprocessed geological data into a pre-trained three-dimensional geological weakness prediction model for prediction to obtain a three-dimensional geological weakness distribution;

[0095] A grouting prediction module, which is configured to: input the three-dimensional geological weakness distribution into a pre-trained targeted grouting prediction model for prediction to obtain the optimal grouting points and grouting parameters of the current engineering area; based on the set threshold of the reconstruction error of the targeted grouting prediction model, when the error is within the set threshold, perform targeted grouting using the optimal grouting points and grouting parameters; when the error exceeds the set threshold, update the geological environment through the three-dimensional geological weakness prediction model and perform re-updating of the weakness distribution of the targeted grouting prediction model;

[0096] A grouting equipment operation module, which is configured to: set the operation parameters of the grouting equipment based on the grouting parameters, and based on the basic principle of multi-agent deep reinforcement learning, regard each grouting equipment as an individual agent, and optimize the operation parameters of each agent through the reinforcement learning strategy to achieve the target grouting effect.

[0097] Example Three

[0098] The purpose of this embodiment is to provide a computing device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the steps of the method in Example One.

[0099] Example Four

[0100] The purpose of this embodiment is to provide a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it performs the steps of the method in Embodiment 1.

[0101] The steps involved in the devices in the above Embodiments 3 and 4 correspond to those in Method Embodiment 1, and for the specific implementation manners, reference may be made to the relevant description part of Embodiment 1. The term "computer-readable storage medium" should be understood to include a single medium or multiple media containing one or more instruction sets; it should also be understood to include any medium that can store, encode, or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0102] Those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computer device. Optionally, they can be implemented by program codes executable by a computing device, so that they can be stored in a storage device for execution by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0103] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

[0104] Although the specific implementation manners of the present invention have been described above in conjunction with the accompanying drawings, it is not a limitation on the protection scope of the present invention. Those skilled in the art should understand that based on the technical solutions of the present invention, various modifications or deformations that can be made by those skilled in the art without creative efforts are still within the protection scope of the present invention.

Claims

1. A targeted positioning grouting and efficient plugging and reinforcement method based on deep learning, characterized in that: include: Acquire the geological data of the current project area and preprocess it to obtain preprocessed geological data; The preprocessed geological data is input into a pre-trained three-dimensional geological weakness prediction model for prediction to obtain a three-dimensional geological weakness distribution; The three-dimensional geological weakness distribution is input into a pre-trained targeted grouting prediction model for prediction, and the optimal grouting point and grouting parameters of the current engineering area are obtained; based on the set threshold of the reconstruction error of the targeted grouting prediction model, when the error is within the set threshold, the optimal grouting point and grouting parameters are used for targeted grouting; when the error exceeds the set threshold, the geological environment is updated through the three-dimensional geological weakness prediction model, and the weakness distribution of the targeted grouting prediction model is updated again; The operating parameters of the grouting equipment are set based on the grouting parameters. Based on the basic principles of multi-agent deep reinforcement learning, each grouting equipment is treated as a separate agent. The operating parameters of each agent are optimized through reinforcement learning strategy to achieve the target grouting effect.

2. The targeted positioning grouting efficient plugging and reinforcement method based on deep learning as claimed in claim 1 is characterized in that: The three-dimensional geological weakness prediction model is constructed by using a generative adversarial network, including a generator and a discriminator. A variational autoencoder is used as the backbone structure of the network generator, and a Transformer is used as the backbone structure of the discriminator.

3. The deep learning-based targeted grouting efficient plugging and reinforcement method according to claim 2, characterized in that: The data processing process in the three-dimensional geological weakness prediction model is specifically as follows: The preprocessed geological data is input as input data, and a generator with a variational autoencoder as the main structure is used to output a three-dimensional geological weakness distribution. At the same time, the three-dimensional geological weakness distribution is associated with the geological data to obtain geological data carrying parameter information; Afterwards, the discriminator with Transformer as the backbone structure compares the bad geological distribution in the actual stratum with the generated 3D geological weakness distribution, and uses mutual information as the loss term; Based on the back-propagation update optimization solution, when the model generated by the generator is close to the distribution of the true model and the loss term meets the requirements, the model training is completed.

4. The deep learning-based targeted grouting efficient plugging and reinforcement method according to claim 1, characterized in that: The targeted grouting prediction model is constructed based on a convolutional neural network, and outputs different three-dimensional geological weakness distributions, including single weaknesses and multiple weakness combinations, through a three-dimensional geological weakness prediction model.

5. The targeted positioning grouting efficient plugging and reinforcement method based on deep learning as claimed in claim 4 is characterized in that: The single weakness is targeted grouting according to the characteristics of different weaknesses, and the multiple weakness combination is predicted according to the pre-trained targeted grouting prediction model, and the historical grouting engineering data is learned through the neural network to learn the relationship between the distribution of different three-dimensional geological weaknesses and the optimal grouting points and grouting parameters when multiple weaknesses exist; The formula is: S(z i ) t =v(x i ,S(z i ) t-1 ) t +b(t) Among them, S represents the different distribution of weaknesses at the current time step, z i represents different distributions of weaknesses, t and t-1 represent time steps, i represents different weakness classifications, and x i It represents the actual grouting parameters corresponding to the weakness, v represents the grouting parameter selection at the current time step and the weakness parameter situation at the previous time step, and b(t) represents the influence of the surrounding environment on the grouting at each time step.

6. The deep learning-based targeted grouting efficient plugging and reinforcement method according to claim 1, characterized in that: The operating parameters of each agent are optimized through reinforcement learning strategy, where the reward function is expressed as: Where R is the actual grouting progress corresponding to the state quantity quantified by different data of each action, m is the total number of equipment, n is the total number of actions of equipment j, j is different equipment, k is an action of equipment j, t is the time step, α(t) and β(t) are the adjustment coefficients at different times, respectively, and w is j (t) represents the loss of equipment j at a certain time, λ j (t) is the reward coefficient of the actual grouting progress corresponding to the different data quantified state quantities of each action of each device at different times, Y j Represents the state quantity of a certain action of device j at a certain time, y k represents the state quantity of device j at a certain time step under action k, Y j (y k ,t) is the state quantity of each action of each device at different times.

7. The deep learning-based targeted grouting efficient plugging and reinforcement method according to claim 6, characterized in that: Determine the action combination value required for the grouting parameters corresponding to each agent. The combination value calculation is expressed as: Among them, x i represents the actual grouting parameters corresponding to the weakness, j represents different equipment, k represents an action of equipment j, n represents the total number of actions of equipment j, γ j,k represents the weight of the contribution value of a certain action k of device j to the parameter, y k It is represented as the state quantity of device j at a certain time step under action k, Z j (y k ) represents the contribution value corresponding to the state quantity.

8. A targeted positioning grouting and efficient plugging and reinforcement system based on deep learning, characterized in that: include: A data acquisition module is configured to: acquire geological data of the current engineering area and perform preprocessing to obtain preprocessed geological data; A weakness distribution prediction module is configured to: input the preprocessed geological data into a pre-trained three-dimensional geological weakness prediction model for prediction to obtain a three-dimensional geological weakness distribution; The grouting prediction module is configured to: input the three-dimensional geological weakness distribution into a pre-trained targeted grouting prediction model for prediction, and obtain the optimal grouting point and grouting parameters for the current engineering area; based on the set threshold of the reconstruction error of the targeted grouting prediction model, when the error is within the set threshold, use the optimal grouting point and grouting parameters for targeted grouting; when the error exceeds the set threshold, update the geological environment through the three-dimensional geological weakness prediction model, and re-update the weakness distribution of the targeted grouting prediction model; The grouting equipment operation module is configured to: set the operation parameters of the grouting equipment based on the grouting parameters, treat each grouting equipment as a separate agent based on the basic principles of multi-agent deep reinforcement learning, and optimize the operation parameters of each agent through reinforcement learning strategies to achieve the target grouting effect.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the targeted grouting efficient plugging and reinforcement method based on deep learning are implemented as described in any one of claims 1 to 7.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the targeted grouting efficient plugging and reinforcement method based on deep learning are implemented as described in any one of claims 1-7.

Images