Grouting pressure intelligent adjusting method and system based on deep learning

Abstract

The invention discloses an intelligent grouting pressure adjusting method and system based on deep learning, and relates to the technical field of grouting pressure adjustment. Geological and geophysical prospecting data of a to-be-grouted area before grouting construction are obtained; all parameters of the surrounding environment influence, the active regulation and control pressure pulse and the pressure pulse of the grouting equipment serve as the state S for describing grouting diffusion, all the parameters in the state S are input into a pre-trained grouting pressure influence prediction model, and the corresponding grouting effect is obtained; according to geological and geophysical prospecting data, the needed grouting diffusion threshold value state is calculated, all parameters in the state S, the corresponding grouting effect and the grouting diffusion threshold value state are input into a pre-trained grouting pressure regulation and control model, and the model takes the grouting diffusion threshold value state as the target of the grouting effect and outputs grouting pressure regulation operation; and according to the actual grouting condition and the real-time feedback of the actual grouting parameters, the grouting pressure regulation and control model is updated, so that the grouting pressure is regulated, and the grouting pressure is automatically and accurately regulated.

Description

Technical Field

[0001] The present invention relates to the technical field of grouting pressure regulation, and in particular to an intelligent grouting pressure regulation method and system based on deep learning. Background Technique

[0002] The statements in this part only provide background technical information related to the present invention, and do not necessarily constitute prior art.

[0003] During the construction of underground projects, many complex geological sections such as extremely water-rich or high-pressure conditions are often faced. The regional rock and soil masses often show relatively weak properties and are prone to local collapse under disturbance, which will threaten the safety of construction personnel and equipment and the normal production of the project. At present, disasters in poor geological sections of underground projects are often treated by grouting. The slurry is injected into the formation through a grouting pipe to cement the original loose structure or cracks into a whole, achieving the effects of sealing water inrush and reinforcing the surrounding rock.

[0004] In the actual grouting process, affected by the grouting equipment and environment, the regulation of grouting pressure is often dominated by manual operation. The manual dominance is strong, and it is difficult to accurately judge the required grouting pressure according to the actual formation conditions, resulting in difficult control of the diffusion range of the slurry and the actual grouting effect. That is, at present, for the pressure regulation during the grouting process, it is often achieved by manually controlling part of the grouting equipment and manually judging the grouting situation and then manually adjusting the change of grouting pressure intermittently. However, in the actual grouting process, in addition to the certain fluctuation impact of manual regulation on the pressure itself (i.e., the active regulation pressure pulse), due to the limitations of the grouting equipment itself, the transportation of the slurry is not stable and continuous, and this pressure pulse of the grouting equipment itself will also affect the pressure regulation during the grouting process. The existing manual pressure regulation method is difficult to accurately consider the influence of the pressure pulse of the grouting equipment itself, and does not comprehensively consider the comprehensive influence of the equipment and the automatic regulation pressure pulse during the grouting process, and it is difficult to achieve the automatic intelligent regulation of grouting pressure considering the formation environment. Summary of the Invention

[0005] To solve the deficiencies of the above-mentioned prior art, the present invention provides an intelligent grouting pressure regulation method and system based on deep learning. By constructing a grouting pressure influence prediction model and a grouting pressure regulation model, and then according to the actual geological conditions on site and the real-time feedback of grouting data, the automatic, real-time and accurate regulation of grouting pressure is realized.

[0006] In the first aspect, the present invention provides an intelligent grouting pressure regulation method based on deep learning.

[0007] An intelligent grouting pressure regulation method based on deep learning includes:

[0008] Obtain the geological geophysical exploration data of the area to be grouted before grouting construction;

[0009] Taking the parameters of the surrounding environment impact, actively regulating the pressure pulse, and the pressure pulse of the grouting equipment itself as the state S describing the grouting diffusion, input the parameters in the state S into the pre-trained grouting pressure impact prediction model to obtain the corresponding grouting effect;

[0010] According to the geological geophysical exploration data, calculate the required grouting diffusion threshold state, and input the parameters in the state S, their corresponding grouting effects, and the grouting diffusion threshold state into the pre-trained grouting pressure regulation model. This model takes the grouting diffusion threshold state as the target of the grouting effect and outputs the grouting pressure adjustment operation, which includes the change values of the parameters in the state S;

[0011] According to the actual grouting situation and the real-time feedback of the actual grouting parameters, update the grouting pressure regulation model, and use the updated model to adjust the grouting pressure.

[0012] In a second aspect, the present invention provides an intelligent grouting pressure regulation system based on deep learning.

[0013] An intelligent grouting pressure regulation system based on deep learning includes:

[0014] A data acquisition module for acquiring geological geophysical exploration data of the area to be grouted before grouting construction;

[0015] A grouting effect feedback module for taking the parameters of the surrounding environment impact, actively regulating the pressure pulse, and the pressure pulse of the grouting equipment itself as the state S describing the grouting diffusion, inputting the parameters in the state S into the pre-trained grouting pressure impact prediction model to obtain the corresponding grouting effect;

[0016] A grouting pressure regulation module for calculating the required grouting diffusion threshold state according to the geological geophysical exploration data, and inputting the parameters in the state S, their corresponding grouting effects, and the grouting diffusion threshold state into the pre-trained grouting pressure regulation model. This model takes the grouting diffusion threshold state as the target of the grouting effect and outputs the grouting pressure adjustment operation, which includes the change values of the parameters in the state S;

[0017] A grouting pressure regulation update module for updating the grouting pressure regulation model according to the actual grouting situation and the real-time feedback of the actual grouting parameters, and using the updated model to adjust the grouting pressure.

[0018] In a third aspect, the present invention further provides an electronic device, including: a memory for storing executable instructions; a processor for implementing the above-mentioned intelligent grouting pressure regulation method based on deep learning when executing the executable instructions stored in the memory.

[0019] Fourthly, the present invention also provides a computer-readable storage medium storing executable instructions, which, when causing a processor to execute the executable instructions, implement the above-described intelligent adjustment method for grouting pressure based on deep learning.

[0020] Fifthly, the present invention also provides a computer program product, which includes executable instructions stored in a computer-readable storage medium; wherein, when a processor of an electronic device reads and executes the executable instructions from the computer-readable storage medium, the above-described intelligent adjustment method for grouting pressure based on deep learning is implemented.

[0021] The above one or more technical solutions have the following beneficial effects:

[0022] 1. The present invention provides an intelligent adjustment method and system for grouting pressure based on deep learning, which considers the active adjustment and the pressure pulse of the grouting equipment itself during the actual grouting process, analyzes the slurry diffusion situation affected by the two pressure pulse conditions and the surrounding energy exchange, and establishes a prediction model for the influence of grouting pressure; at the same time, considering the limiting relationship between the active adjustment and the pressure pulse of the grouting equipment itself, ensuring the efficiency of the grouting diffusion and reflux processes, realizing the plugging of the orifice pipe during the grouting process, and establishing a control model for grouting pressure; also considering the connectivity relationship between different grouting holes, adjusting each parameter of the pressure pulse actively controlled by the grouting pressure control model to reduce the slurry reflux and the loss of grouting materials, and at the same time, according to the real-time feedback of the on-site data, updating the pressure threshold of the grouting pressure control model in real time; the present invention realizes the automatic, real-time and accurate adjustment of the grouting pressure by constructing a prediction model for the influence of grouting pressure and a control model for grouting pressure, and according to the actual geological conditions on site and the real-time feedback of grouting data.

[0023] 2. The present invention establishes a grouting database based on geological data, grouting parameters and pressure pulse information, and uses this grouting database to train the prediction model for the influence of grouting pressure. During the training process, the slurry diffusion range, slurry flow rate and slurry reflux situation corresponding to each pressure pulse are considered, and the multivariate non-linear change situation of the slurry diffusion range and slurry reflux under the conditions of the influence of the surrounding environment, the actively controlled pressure pulse and the pressure pulse of the grouting equipment itself superimposed multiple times is set. At the same time, it is considered that the pressure pulse of the grouting equipment itself is affected by the actively controlled pressure pulse in terms of cycle time, so as to ensure the accuracy of the constructed and trained prediction model for the influence of grouting pressure.

[0024] 3. Before the actual grouting construction, the present invention pre - processes data such as geological geophysical exploration data and grouting parameter data, etc. Considering the relationship between active regulation and the pressure pulse limitation of the grouting equipment itself, it ensures the efficiency of the grouting diffusion and reflux process, realizes the plugging of the orifice pipe during the grouting process, and establishes a grouting pressure regulation model. At the same time, considering the connectivity relationship between different grouting holes, it adjusts each parameter of the pressure pulse of the grouting pressure regulation model actively to reduce the slurry reflux and the loss of grouting materials. According to the real - time feedback of on - site data, it updates the pressure threshold of the grouting pressure regulation model in real - time to obtain a real - time and accurate grouting pressure adjustment value closer to the actual situation.

[0025] Advantages of additional aspects of the present invention will be given in part in the following description, become apparent in part from the following description, or be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The schematic diagrams forming a part of the present invention are used to provide a further understanding of the present invention. The illustrative embodiments and descriptions thereof are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0027] Figure 1 It is a flowchart of the intelligent grouting pressure adjustment method based on deep learning in the embodiment of the present invention;

[0028] Figure 2 It is a cyclic framework diagram of multi - task learning with weight update added in a single time step in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] It should be noted that the following detailed description is exemplary only, and is only for describing the specific embodiments, aiming to provide further explanation of the present invention, and is not intended to limit the exemplary embodiments according to 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. In addition, it should also be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0030] Embodiment 1

[0031] This embodiment provides an intelligent grouting pressure adjustment method based on deep learning, which specifically includes the following steps:

[0032] Step S1: Obtain the geological geophysical exploration data of the area to be grouted before the grouting construction;

[0033] Step S2: Take the parameters of the surrounding environment impact, the actively regulated pressure pulse, and the pressure pulse of the grouting equipment itself as the state S describing the grouting diffusion, input the parameters in the state S into the pre-trained grouting pressure impact prediction model, and obtain the corresponding grouting effect;

[0034] Step S3: Calculate the required grouting diffusion threshold state based on the geological geophysical exploration data, and input the parameters in the state S, their corresponding grouting effects, and the grouting diffusion threshold state into the pre-trained grouting pressure regulation model. This model takes the grouting diffusion threshold state as the target of the grouting effect and outputs the grouting pressure adjustment operation, which includes the change values of the parameters in the state S;

[0035] Step S4: Update the grouting pressure regulation model according to the actual grouting situation and the real-time feedback of the actual grouting parameters, and use the updated model to adjust the grouting pressure.

[0036] The intelligent grouting pressure adjustment method based on deep learning proposed in this embodiment is introduced in more detail through the following content.

[0037] First, before grouting construction in the area to be grouted, build and pre-train the grouting pressure impact prediction model and the grouting pressure regulation model. The specific process is as Figure 1 shown, including:

[0038] (1) Grouting pressure impact prediction model

[0039] (1.1) Obtain the grouting information during the historical grouting construction process, including: geological geophysical exploration data, grouting parameter data, and pressure pulse information.

[0040] Specifically, through geological exploration and other methods, obtain the geological geophysical exploration data of the grouting area, that is, collect the geological characteristics of the area, including geological minerals, geological elements, voids, and fracture-related information. Through data analysis, obtain the distribution and connectivity relationship of the void and fracture spaces; at the same time, obtain the corresponding grouting parameter data, that is, the flow rate, diffusion range, filling effect, etc. of the slurry under different medium conditions and grouting pressures, collect the pressure change curve during the grouting process, and record the fluctuation trend under different geological conditions and different slurry material concentrations; in addition, simultaneously collect the pressure pulse information generated by each manual adjustment during the grouting process (abbreviated as actively regulated pressure pulse) and the pressure pulse information generated by the grouting equipment itself (abbreviated as pressure pulse of the grouting equipment itself). This pressure pulse information includes time, frequency, amplitude, and duration, etc.

[0041] Further, according to the different degrees of influence of each parameter data on the final grouting, the above-mentioned multiple parameters are divided into key parameters and non-key parameters. The key parameters include the actively adjusted pressure pulse, the pressure pulse of the grouting equipment itself, the flow rate, diffusion range, filling effect, etc. of the grout under different medium conditions and grouting pressures; the non-key parameters include the grouting volume, geological minerals, geological elements, etc.

[0042] (1.2) Preprocess the obtained data to construct a grouting database.

[0043] Specifically, first, the autoencoder is used to learn the change patterns of the data to identify outliers and fill in missing values. Among them, the outliers of non-key parameters can be directly removed, and the abnormal data of key parameters are marked for judging construction problems existing in the actual grouting process.

[0044] Further, the numerical data with different dimensions and dimensions are standardized, and the data is normalized to ensure the quality and consistency of the data.

[0045] After the above preprocessing, using the surrounding environment influence (i.e., the influence of the formation environment after actual grouting, reflected as real-time geological data, such as the spatial distribution of fractures, connectivity relationships, etc.), the actively regulated pressure pulse, and the pressure pulse of the grouting equipment itself as the parameters to describe the state S of grouting diffusion, and using the state S data and its corresponding grouting diffusion (including the grout diffusion range, grout flow rate, grout reflux range) data as training data, thus constructing a grouting database and storing the preprocessed data in the database.

[0046] (1.3) Construct and train a grouting pressure influence prediction model

[0047] First, construct a grouting pressure influence prediction model. Specifically, as Figure 2 shown, adopt the basic framework of a multi-task learning model considering weights, set different branch networks corresponding to the grout diffusion range, grout velocity, and grout reflux range of grouting respectively, use the input layer of each branch network as the shared layer, and use the surrounding environment influence, actively regulated pressure pulse, and pressure pulse of the grouting equipment itself under different medium conditions and grout material concentrations as the shared input values E1 - E3 of the neural network shared layer and input them into the model. Then map the shared layer data to the fully connected layers based on non-linear activation functions in multiple branch networks to learn the relationship between various input data and the grout diffusion range, grout velocity, and grout reflux range (i.e., output 1 - 3) finally output by each branch network; in addition, in each branch network, add the initial gain weights W 1n ~W 3n, update each weight through the loss error during the training process.

[0048] Secondly, select data from the database and divide it into a training set and a validation set. Use the data in the training set to train the grouting pressure influence prediction model built above. The training process is as follows: At each time step calculation, consider the slurry diffusion range, slurry flow rate, and slurry backflow situation corresponding to each pressure pulse. Set the multi-frequency superposition conditions of the surrounding environment influence, actively regulated pressure pulse, and the pressure pulse of the grouting equipment itself, and consider the multi-dimensional non-linear variation of the slurry diffusion range, slurry velocity, and slurry backflow range. At the same time, consider that the pressure pulse of the grouting equipment itself is affected by the actively regulated pressure pulse in the cycle time. The training process of the above grouting pressure influence prediction model can be expressed as:

[0049] F(y 1 ,y 2 ,y 3 ) = W(ε(i,t),X i,T ,x(t,X i,T ))^b(T,t);

[0050] Among them, y 1 , y 2 , y 3 represent the diffusion range of the grouting slurry, slurry velocity, and slurry backflow situation respectively. ε(i,t) represents the energy exchange coefficient affected by the surrounding environment. X i,T represents the parameter values of the pressure pulse actively regulated at the i-th frequency within the T-time change. x(t,X i,T ) represents the parameter values of the pressure pulse of the grouting equipment itself affected by the pressure pulse X actively regulated at the i-th frequency within t time. b(T,t) represents the gain value of the parameter values of each pressure pulse within t time. W = [W 1n , W 2n , W 3n represents the gain weights of the parameter values of the surrounding environment influence, actively regulated pressure pulse, and the pressure pulse of the grouting equipment itself. n represents the number of lower-level parameters of each parameter value. Taking the pulse parameters as an example, it includes time, frequency, amplitude, and duration, etc.

[0051] During the training process, based on different medium conditions and slurry material concentrations, on the basis of the initial values of the input initial surrounding environment impact, actively regulated pressure pulse, and the pressure pulse of the grouting equipment itself, and the initial gain weights and gain values of the three parameters corresponding to different branch networks, the deviation between the predicted value and the true value output by the model is used as the loss (i.e., Loss 1-3), and continuous iterative training is carried out to update the gain weights and gain values on different branch networks, and continuously adjust the coefficient values of bias ε(i,t) and b(T,t) until the error is within the allowable range, and the accuracy of the model is verified through the validation set to obtain the finally trained grouting pressure impact prediction model.

[0052] (2) Grouting pressure regulation model

[0053] (2.1) Construct the grouting pressure regulation model and its training dataset. Specifically, taking the parameters of the surrounding environment impact, actively regulated pressure pulse, and the pressure pulse of the grouting equipment itself as the state S describing the grouting diffusion, and taking the changes of the parameters in the state S as the continuous grouting operation A, the grouting effect feedback under different grouting operations A can be obtained through the above-mentioned pre-trained grouting pressure impact prediction model, and the grouting effect includes the slurry diffusion range, slurry flow rate, and slurry reflux range. Furthermore, through the above method, the grouting effects corresponding to different states S under different grouting operations A can be obtained, and the training dataset can be constructed accordingly.

[0054] (2.2) Obtain the geological geophysical exploration data of the area to be grouted before the actual grouting construction, etc., and preprocess the obtained data, and analyze the information of the front voids, fractures and their connectivity according to the preprocessed geological geophysical exploration data.

[0055] (2.3) Calculate the required grouting diffusion threshold state S according to the recorded information such as void size, porosity, fracture aperture, and extension length i , this state S i includes the grouting diffusion radius threshold R m and the grouting diffusion time threshold T m , and the calculation formulas for each threshold are as follows:

[0056]

[0057] Among them, R m is the grouting diffusion radius threshold, α 1 and α 2 are the corresponding weights of voids and fractures respectively, V 1 and V 2 are the volumes of slurry that can be injected into voids and fractures respectively, φ is the porosity, h is the rock layer thickness, ω is the fracture aperture, L f is the fracture extension length.

[0058]

[0059] Among them, T m is the grouting diffusion time threshold, μ is the viscosity of the injected grout, and k is the permeability.

[0060] (2.4) Take the preprocessed geological geophysical exploration data and the required grouting diffusion threshold state S i as the initial data, set pulses of alternating high and low pressure on the active regulation, set the high and low pressure thresholds according to the grouting pressure range required by the construction, and set the initial periodic change of the high and low pressures. Then, the pressure pulse of the grouting equipment itself has different periodic pressure changes under each active regulation pressure pulse; use the initial data and the training data set to train the grouting pressure regulation model based on the basic architecture of the deep Q network, continuously adjust the parameter values of the two pressure pulses, and at the same time appropriately adjust the pressure pulse of the active regulation in the second half of the grouting to achieve the promotion and hindrance effects on the pressure pulse of the grouting equipment itself. By taking the active regulation as the limiter of the pressure pulse of the equipment itself, ensure that the slurry backflow during the grouting process does not exceed the boundary of the injected hole, and achieve the plugging of the orifice pipe during the grouting process.

[0061] Specifically, take the initial values of the parameters in the state S as the input of the model, and at the same time input the geological geophysical exploration data and the grouting diffusion threshold state S i , take the grouting effect as the output of the model, and train the model until the finally output grouting effect reaches the required grouting diffusion threshold state S i . The training process of this grouting pressure regulation model can be expressed as:

[0062] X = w X (m j , T 1 ), x = w x (n j , T 2 , X), z = f(X, x);

[0063] Among them, X is the action value of the active regulation pressure pulse, x is the action value of the pressure pulse of the grouting equipment itself, w X and w x are the reward coefficients under the conditions of the active regulation and the pressure pulse of the grouting equipment itself respectively, m j is the parameter value of the pressure pulse of the active regulation in the jth cycle, n j is the parameter value of the pressure pulse of the grouting equipment itself in the jth cycle of the active regulation, T 1 and T 2They are adjustable cycle parameters for active regulation and the pressure pulse of the equipment itself. f is the limiting relationship between active regulation and the pressure pulse of the equipment itself, and z is the output result of the model. This output result is the grouting effect, including the diffusion range of the grouting slurry, the slurry velocity, and the slurry reflux situation.

[0064] Finally, based on the above initial data for training, continuously adjust the parameter values of the two pressure pulses so that the grouting effect obtained from the grouting pressure influence prediction model matches that of S i to complete the training of the grouting pressure regulation model. Among them, this matching means: normalizing the numerical values of the slurry diffusion range, slurry velocity, and slurry reflux range of the two parts of grouting and performing numerical subtraction, and performing weighted averaging on the differences in the three states. When the final weighted average value is within the set threshold range, it indicates that the matching is completed.

[0065] After completing the above model construction and training, based on the pre-trained grouting pressure influence prediction model and grouting pressure regulation model, perform real-time pressure intelligent regulation on the area to be grouted, including:

[0066] First, obtain the geological geophysical exploration data of the area to be grouted before grouting construction; second, take the surrounding environment influence, active regulation pressure pulse, and each parameter of the pressure pulse of the grouting equipment itself as the state S describing grouting diffusion, and input each parameter in the state S into the pre-trained grouting pressure influence prediction model to obtain the corresponding grouting effect; then, according to the geological geophysical exploration data, calculate the required grouting diffusion threshold state, and input each parameter in the state S, its corresponding grouting effect, and the grouting diffusion threshold state into the pre-trained grouting pressure regulation model. This model takes the grouting diffusion threshold state as the target of the grouting effect and automatically outputs the grouting pressure adjustment operation. This operation includes the change values of each parameter in the state S, mainly the change values of the active regulation pressure pulse and each parameter of the pressure pulse of the grouting equipment itself, that is, the grouting pressure adjustment value. On this basis, automatically adjust the grouting operation according to the change values of each parameter to achieve intelligent regulation of the grouting pressure.

[0067] Preferably, when using the grouting pressure regulation model to output the grouting pressure adjustment value, it is necessary to continuously update the grouting pressure regulation model according to the actual situation of this area and the real-time feedback of the actual grouting parameters on site, and use the updated model to adjust the grouting pressure so that the output grouting pressure adjustment value is more accurate. Specifically, it includes the following situations:

[0068] ①Based on the geological geophysical exploration data of the area, considering the connectivity between multiple pores, that is, considering the connectivity relationship between multiple different grouting holes, in the area with higher connectivity, adjust to increase the frequency of active regulation in the grouting pressure regulation model, and at the same time adjust the amplitude of this active regulation, so that the grouting slurry can penetrate, diffuse and fill in a certain area at an appropriate speed, and can still achieve a higher cumulative reward under the condition of a higher active regulation pressure pulse and meet the matching with the grouting diffusion threshold state S i ;

[0069] ②According to the grouting situation of each grouting hole, obtain the diffusion situation of the slurry in the area to be grouted in the front, and combine the setting time and cementation strength of the slurry to update and adjust the area to be grouted in the front, that is, update the grouting diffusion threshold state S according to the calculation formula in the above (2.3) i , and import the updated value into the grouting pressure regulation model to perform real-time adjustment of the grouting pressure.

[0070] ③For combined grouting of multiple holes, consider the slurry diffusion range and flow situation of multiple grouting holes at the same time, that is, perform multiple calculations according to the calculation formula in the above (2.3), and obtain the corresponding grouting diffusion threshold states S 1 , S 2 ,..., S n in sequence, determine different slurry diffusion threshold states (mainly the slurry diffusion radius) in the spatial dimension, obtain the flow velocity of the slurry in the time dimension, adjust the frequency of active regulation of the grouting pressure regulation model, and at the same time adjust its amplitude, and accurately obtain the time when the slurry of each hole flows to the overlapping area of the slurry diffusion radius of each hole, so as to reduce the slurry backflow and the loss of grouting materials.

[0071] That is, under complex geological structures, predict the overall diffusion situation under different grouting pressures under the initial threshold conditions, aiming at achieving uniform filling, and then predict the pressure range based on the real-time feedback of on-site data, and update the pressure threshold for the grouting pressure regulation model to realize the real-time regulation of the grouting pressure based on the actual situation and ensure the accuracy of the final grouting.

[0072] Through the above methods, real-time and accurate grouting pressure regulation closer to the actual situation can be realized, and a better grouting effect can be achieved.

[0073] Embodiment 2

[0074] This embodiment provides a grouting pressure intelligent regulation system based on deep learning, including:

[0075] A data acquisition module, used to acquire the geological geophysical exploration data of the area to be grouted before grouting construction;

[0076] The grouting effect feedback module is used to use the parameters of the surrounding environment impact, active regulation pressure pulse, and the pressure pulse of the grouting equipment itself to describe the state S of grouting diffusion, input the parameters in the state S into the pre-trained grouting pressure impact prediction model, and obtain the corresponding grouting effect;

[0077] The grouting pressure regulation module is used to calculate the required grouting diffusion threshold state according to the geological geophysical exploration data, and input the parameters in the state S, their corresponding grouting effects, and the grouting diffusion threshold state into the pre-trained grouting pressure regulation model. This model takes the grouting diffusion threshold state as the target of the grouting effect and outputs the grouting pressure adjustment operation, and this operation includes the change values of the parameters in the state S;

[0078] The grouting pressure regulation update module is used to update the grouting pressure regulation model according to the actual situation of grouting and the real-time feedback of the actual grouting parameters, and use the updated model to adjust the grouting pressure.

[0079] Embodiment III

[0080] This embodiment provides an electronic device, including: a memory for storing executable instructions; a processor for implementing the above method provided in this embodiment when executing the executable instructions stored in the memory.

[0081] Embodiment IV

[0082] This embodiment also provides a computer-readable storage medium storing executable instructions, which when executed by a processor will cause the processor to execute the above method provided in this embodiment.

[0083] Embodiment V

[0084] This embodiment provides a computer program product, which includes executable instructions, and the executable instructions are a kind of computer instructions; the executable instructions are stored in a computer-readable storage medium. When the processor of an electronic device reads the executable instructions from the computer-readable storage medium and the processor executes the executable instructions, the electronic device is caused to execute the above method provided in this embodiment.

[0085] The steps involved in Embodiments II to V above correspond to those in Method Embodiment I, and the specific implementation manners can refer to the relevant description part of Embodiment I. 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 cause the processor to execute any method in the present invention.

[0086] 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 and executed 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.

[0087] The above are only the preferred embodiments of the present invention. Although the specific implementation manners of the present invention are described in conjunction with the accompanying drawings, it is not a limitation to 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 without creative efforts by those skilled in the art are still within the protection scope of the present invention.

Claims

1. A grouting pressure intelligent adjustment method based on deep learning, characterized in that: include: Obtain geological and geophysical data of the area to be grouting before grouting construction; The parameters of the surrounding environment, active pressure pulse regulation, and pressure pulse of the grouting equipment itself are used as the state S describing the grouting diffusion. The parameters in the state S are input into the pre-trained grouting pressure influence prediction model to obtain the corresponding grouting effect. According to the geological and geophysical data, the required grouting diffusion threshold state is calculated, and each parameter in the state S and its corresponding grouting effect and the grouting diffusion threshold state are input into the pre-trained grouting pressure control model. The model takes the grouting diffusion threshold state as the target of the grouting effect and outputs the grouting pressure adjustment operation, which includes the change value of each parameter in the state S; According to the actual grouting situation and real-time feedback of actual grouting parameters, the grouting pressure control model is updated, and the grouting pressure is adjusted using the updated model.

2. A method for intelligently adjusting grouting pressure based on deep learning as claimed in claim 1, characterized in that: The training process of the grouting pressure impact prediction model includes: Obtain geological and geophysical data, grouting parameter data and pressure pulse information; Preprocess the acquired data and build a grouting database; A grouting pressure influence prediction model based on a multi-task learning basic architecture considering weights is constructed, including: setting different branch networks corresponding to the diffusion range, slurry velocity and slurry reflux range of the grouting slurry, respectively; using the input layer of each branch network as a shared layer; inputting the surrounding environment influence under different medium conditions, slurry material concentration, active regulation pressure pulse, and the pressure pulse of the grouting equipment itself as the shared input value of the neural network shared layer into the model; and then mapping the shared layer data to the fully connected layer based on the nonlinear activation function in multiple branch networks to learn the relationship between multiple input data and the slurry diffusion range, slurry velocity and slurry reflux range finally output by each branch network; in each branch network, adding the initial gain weights corresponding to the parameters of the surrounding environment influence, active regulation pressure pulse, and the pressure pulse of the grouting equipment itself in different branch networks; The training data set in the grouting database is used for model training. After continuous iterative training and adjustment, a trained grouting pressure influence prediction model is obtained.

3. A method for intelligently adjusting grouting pressure based on deep learning as claimed in claim 2, characterized in that: The training process of the grouting pressure impact prediction model is as follows: In the calculation of each time step, the corresponding slurry diffusion range, slurry flow rate and slurry reflux range under each pressure pulse are considered, and the multivariate nonlinear changes of the slurry diffusion range, slurry velocity and slurry reflux range under the influence of the surrounding environment, the active control pressure pulse and the multi-frequency superposition of the pressure pulse of the grouting equipment itself are set. At the same time, the pressure pulse of the grouting equipment itself is considered to be affected by the actively controlled pressure pulse in the cycle time; The training process of the grouting pressure impact prediction model is expressed as: F(y1,y2,y3)=W(ε(i,t),X i,T ,x(t,X i,T ))^b(T,t); Among them, y1, y2, and y3 represent the diffusion range, slurry velocity, and slurry reflux of the grouting slurry, respectively; ε(i, t) represents the energy exchange coefficient affected by the surrounding environment; X i,T represents the parameter values ​​of the pressure pulse of the ith frequency actively regulated within the time variation of T, x(t,X i,T ) represents the pressure pulse parameter values ​​of the grouting equipment itself affected by the pressure pulse X with the i-th frequency active control within time t, b(T,t) represents the gain value of each pressure pulse parameter value within time t, W = [W 1n ,W 2n ,W 3n ] represents the gain weight of the surrounding environment influence, active regulation of pressure pulse, and pressure pulse parameter value of the grouting equipment itself, and n represents the number of next-level parameters for each parameter value.

4. The method for intelligently adjusting grouting pressure based on deep learning according to claim 1, characterized in that: The training process of the grouting pressure control model is as follows: The parameters of the surrounding environment, active pressure pulse regulation, and pressure pulse of the grouting equipment itself are used as the state S to describe the grouting diffusion, and the changes of the parameters in the state S are used as the continuous grouting operation A. The grouting effects under different grouting operations are obtained through the pre-trained grouting pressure influence prediction model; a training data set is constructed according to the grouting effects under different grouting operations A under different states S; the grouting effects include slurry diffusion range, slurry flow rate, and slurry reflux range; Obtain the geological and geophysical data of the area to be grouting before the actual grouting construction, and calculate the required grouting diffusion threshold state S based on the geological and geophysical data i ; Based on geological and geophysical data and grouting diffusion threshold state S i As the initial data, high and low pressure alternating pulses are set in the active regulation, the high and low pressure thresholds are set according to the grouting pressure range required by the construction, and the initial periodic changes of high and low pressure are set. The grouting pressure control model based on the basic architecture of the deep Q network is trained using the initial data and the training data set, and the parameter values ​​of the two pressure pulses are continuously adjusted to make the grouting effect obtained based on the grouting pressure influence prediction model consistent with the grouting diffusion threshold state S i Match and complete the training of the grouting pressure control model.

5. A method for intelligently adjusting grouting pressure based on deep learning as claimed in claim 4, characterized in that: The grouting diffusion threshold state S i Including the grouting diffusion radius threshold R m and grouting diffusion time threshold T m , where the calculation formula for the grouting diffusion radius threshold is: In the above formula, R m is the grouting diffusion radius threshold, α1 and α2 are the weights of voids and fractures, V1 and V2 are the slurry volumes available for injection in voids and fractures, φ is the porosity, h is the rock thickness, ω is the fracture aperture, and L f is the crack extension length; The calculation formula for the grouting diffusion time threshold is: In the above formula, T m is the grouting diffusion time threshold, μ is the injected slurry viscosity, and k is the permeability.

6. A method for intelligently adjusting grouting pressure based on deep learning as claimed in claim 4, characterized in that: The training process of the grouting pressure control model is expressed as: X=w X (m j ,T1),x=w x (n j ,T2,X),z=f(X,x); Among them, X is the action value of the active adjustment pressure pulse, x is the action value of the pressure pulse of the grouting equipment itself, and w X and w x are the bonus coefficients under active regulation and pressure pulse conditions of the grouting equipment itself, m j are the values ​​of the pressure pulse parameters actively adjusted in the jth cycle, n j In order to actively adjust the parameter values ​​of the pressure pulse of the grouting equipment itself in the jth cycle, T1 and T2 are the adjustable period parameters of active adjustment and the pressure pulse of the equipment itself, respectively, f is the limiting relationship between active adjustment and the pressure pulse of the equipment itself, and z is the output result of the model, that is, the grouting effect.

7. A grouting pressure intelligent adjustment system based on deep learning, characterized in that: include: A data acquisition module is used to obtain geological and geophysical data of the area to be grouting before grouting construction; The grouting effect feedback module is used to describe the state S of grouting diffusion by using the surrounding environment influence, active pressure pulse regulation, and the pressure pulse parameters of the grouting equipment itself, and input each parameter in the state S into the pre-trained grouting pressure influence prediction model to obtain the corresponding grouting effect; The grouting pressure control module is used to calculate the required grouting diffusion threshold state according to the geological and geophysical data, and input each parameter in the state S and its corresponding grouting effect and the grouting diffusion threshold state into the pre-trained grouting pressure control model. The model takes the grouting diffusion threshold state as the target of the grouting effect and outputs the grouting pressure adjustment operation, which includes the change value of each parameter in the state S; The grouting pressure control update module is used to update the grouting pressure control model according to the actual grouting situation and real-time feedback of actual grouting parameters, and use the updated model to adjust the grouting pressure.

8. An electronic device, characterized in that: include: A memory for storing executable instructions; The processor is used to implement the deep learning-based intelligent adjustment method of grouting pressure as described in any one of claims 1 to 6 when executing the executable instructions stored in the memory.

9. A computer-readable storage medium, characterized in that: Executable instructions are stored, which are used to cause the processor to execute the executable instructions to implement the deep learning-based intelligent adjustment method for grouting pressure as described in any one of claims 1-6.

10. A computer program product, characterized in that The computer program product includes executable instructions stored in a computer-readable storage medium; When the processor of the electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the deep learning-based intelligent adjustment method for grouting pressure described in any one of claims 1 to 6 is implemented.

Images