Method and system for intelligently predicting plugging effect of end of working face based on network model

Abstract

The invention provides a method and system for intelligently predicting the end plugging effect of a working face based on a network model, and the method comprises the steps: simulating the end plugging effect of Anus type foam slurry under different working condition parameters through computational fluid dynamics, obtaining goaf oxygen concentration distribution image data, carrying out the preprocessing of the data obtained through simulation, and obtaining a goaf oxygen concentration distribution image; then, a deep learning model is trained based on the preprocessed data set, the trained deep learning model fits the mapping relation between different working condition parameters and goaf oxygen concentration distribution characteristics, and after the grouting pressure, the grouting flow, the grouting position and other parameter values of the current actual working face end plugging material are input, the goaf oxygen concentration distribution characteristics can be obtained; and the plugging process effect can be quickly predicted. According to the method, a working face end plugging material with plugging, oxygen isolation, cooling and fire extinguishing effects is provided, and on the basis, the end plugging process effect is predicted in time in combination with a network model.

Description

Technical Field

[0001] The present invention relates to the technical field of mine fire prevention and extinguishing, and particularly to a method and system for intelligently predicting the plugging effect at the end of a working face based on a network model. Background Art

[0002] Fires caused by coal spontaneous combustion not only result in the loss of coal resources, disrupt the normal production of mines, but also pose a threat to the lives and safety of miners. At the fully mechanized caving face, end leakage is one of the key factors leading to coal spontaneous combustion in the goaf. For this reason, researchers have developed a variety of end plugging technologies aimed at preventing and controlling coal spontaneous combustion in the goaf. For example, the prior art introduces a balloon-type rapid air leakage plugging device applied to the end of a fully mechanized caving face. This device is made of thermoplastic rubber material, which has good plasticity, high elasticity, heat resistance, compression set resistance, and excellent mechanical properties. Although this balloon can play a role in plugging, it does not have the function of cooling and extinguishing fires. Moreover, since the balloon itself is combustible, its effect in preventing and controlling coal spontaneous combustion in the goaf of a high-temperature fully mechanized caving face is limited. In addition, the prior art also uses solidified foam to block the air leakage channels at the end of a fully mechanized caving face. In order to further improve the effect of the solidified foam plugging application process, researchers usually use Computational Fluid Dynamics (CFD) to pre-simulate the accumulation and diffusion patterns of solidified foam at the end of the working face to guide the design of the plugging process. However, CFD simulation still has problems such as long calculation cycles and insufficient timeliness. Especially in the event of an emergency fire, it is difficult to provide an effective emergency response plan in a timely manner.

[0003] Therefore, it is urgent to research and develop a new material that simultaneously has the dual functions of plugging and oxygen isolation and cooling and extinguishing fires. However, how to intelligently, quickly, and efficiently predict the end plugging process effect of this new material has become a key technical problem that needs to be urgently solved by those skilled in the art. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for intelligently predicting the end plugging process effect based on a network model, so as to provide an end plugging material for the working face with both plugging and oxygen isolation and cooling and extinguishing effects. On this basis, combined with the network model, it can intelligently, quickly, and efficiently predict the end plugging process effect.

[0005] In the first aspect, an embodiment of the present invention provides a method for predicting the end plugging process effect based on a network model. The method includes:

[0006] Use CFD to simulate the effect of Janus-type foam slurry on plugging the end of the fully-mechanized caving face under different working condition parameters, and obtain the image data of the oxygen concentration distribution in the goaf; among them, the Janus-type foam slurry is composed of cement-based slurry, Janus-type water-based foam and coagulant;

[0007] Perform data normalization and data augmentation preprocessing on the data obtained from the CFD simulation to construct a data set;

[0008] Train a deep learning model based on the data set to obtain a trained deep learning model; the deep learning model includes an encoder, a discriminator, a decoder and a deep regression neural network model, and the trained deep learning model fits the mapping relationship between different working condition parameters and the oxygen concentration distribution characteristics in the goaf;

[0009] Input the current actual working condition parameters into the trained deep learning model, and the trained deep learning model outputs the image data of the oxygen concentration distribution in the goaf, and the image data of the oxygen concentration distribution in the goaf is used to evaluate the effect of the end plugging process.

[0010] In a possible implementation, training the deep learning model based on the data set specifically includes:

[0011] Input the working condition data set I n into the deep regression neural network model, and the output end of the deep regression neural network model is connected to the encoder;

[0012] Input the image data set P of the oxygen concentration distribution in the goaf n into the encoder, and the encoder applies a deep convolutional neural network to downsample the data set P n to compress the data set P n into reduced-dimensional data features as negative samples; and generate random data with the same dimension as the reduced-dimensional data features from a determined random normal distribution as positive samples, and input the positive samples and the negative samples together into the discriminator, and the discriminator adjusts the encoder to compress the oxygen concentration distribution characteristics in the goaf so that the decoder is infinitely close to the true random normal distribution;

[0013] The encoder inputs the reduced-dimensional data features into the decoder, and the decoder applies a transposed convolutional neural network to upsample the reduced-dimensional data features to obtain a reconstructed data set, and finally the trained deep learning model outputs the predicted image of the oxygen concentration distribution in the goaf.

[0014] In other possible implementations, the objective function corresponding to the adversarial auto-encoding algorithm adopted by the encoder is as follows:

[0015]

[0016] In the formula, x is the data in the real dataset, an array; z is the random data in the random distribution, an array; P is the data distribution probability function; d represents the discriminator, and g represents the decoder, min g max d V(d,g) means that the generator needs to minimize this cross-entropy loss under the condition that the discriminator maximizes the cross-entropy loss V(d,g) of real and fake pictures; represents the expectation of the real dataset; represents the expectation of the random distribution dataset.

[0017] In another possible implementation, the cement-based slurry is formed by uniformly stirring two incombustible inorganic cementitious materials, namely coal-based solid waste and portland cement, and water. The coal-based solid waste accounts for 29.3-35% of the mass of these two inorganic cementitious materials, and the mass ratio of water to these two inorganic cementitious materials is 0.5-0.6;

[0018] The Janus-type water-based foam is composed of a hydrophilic polymer, hydrophobic nanoparticles and water; the hydrophilic polymer accounts for 1-3.2% of the mass of the Janus-type water-based foam, the hydrophobic nanoparticles account for 1.2-2% of the mass of the Janus-type water-based foam, and the addition amount of the Janus-type water-based foam is 2-2.21 times the volume of the cement-based slurry;

[0019] The coagulant includes one or more of liquid silicate, powdered silicate and aluminum sulfate, and the addition amount of the coagulant is 2.9-6.1% of the mass of the cement-based slurry.

[0020] In other possible implementations, the hydrophilic polymer includes one or more of polyvinyl alcohol, xanthan gum and sodium alginate, and the hydrophobic nanoparticles include one or more of silicon dioxide, titanium dioxide and carbon nanotubes; the working conditions parameters include grouting pressure, grouting flow rate and the position of the grouting pipe.

[0021] In another possible implementation, data normalization and data augmentation preprocessing are performed on the data obtained by CFD simulation, including: normalizing the working conditions parameters in the CFD simulation by using the z-score normalization method; performing data augmentation on the goaf oxygen concentration distribution image data through the Transforms tool of the TorchVision module in PyTorch.

[0022] In another possible implementation, the encoder is composed of 7 convolutional neural network layers, the decoder is composed of 8 transposed convolutional neural network layers, and the discriminator uses 2 convolutional neural network layers for downsampling and 1 fully connected layer for feature compression.

[0023] In a second aspect, the present invention also provides a system for predicting the effect of the end plugging process based on a network model, wherein:

[0024] A simulation data module is used to obtain image data of the oxygen concentration distribution in the gob area blocked by Janus-type foam slurry under different working condition parameters through CFD simulation; wherein, the Janus-type foam slurry is composed of a cement-based slurry, a Janus-type water-based foam, and a coagulant.

[0025] A dataset construction module is used to perform data normalization and data augmentation preprocessing on the data obtained from CFD simulation to construct a dataset.

[0026] A training module is used to train a deep learning model based on the dataset to obtain a trained deep learning model; wherein the deep learning model includes an encoder, a discriminator, a decoder, and a deep regression neural network model, and the trained deep learning model fits the mapping relationship between different working condition parameters and the oxygen concentration distribution characteristics in the gob area.

[0027] A model application module is used to input the current actual working condition parameters into the trained deep learning model, and the trained deep learning model outputs image data of the oxygen concentration distribution in the gob area, and the image data of the oxygen concentration distribution in the gob area is used to evaluate the effect of the end plugging process.

[0028] In a third aspect, an embodiment of the present invention further provides a computer-readable storage medium, which includes a computer program. When the computer program runs on a processor, the processor is caused to execute the method performed in the first aspect above.

[0029] In a fourth aspect, an embodiment of the present invention provides an electronic device, including a processor and a memory. Among them, the memory is used to store one or more computer programs; when one or more computer programs stored in the memory are executed by the processor, the processor is enabled to implement the method of any possible embodiment in the first aspect above.

[0030] In a fifth aspect, an embodiment of the present invention further provides a computer program product. When the computer program product runs on a processor, the processor is caused to execute the method of any possible embodiment in any of the above aspects.

[0031] The method and system for intelligently predicting the effect of the end plugging process based on a network model provided by the embodiments of the present invention have the following beneficial effects:

[0032] (1) Developed Janus-type foam slurry, endowing the material with the performance advantages of plugging and oxygen isolation and cooling and fire extinguishing.

[0033] (2) This method can predict the leakage stoppage effect at the end of the fully-mechanized caving face intelligently, quickly and efficiently, solve the problems of long CFD simulation calculation cycle and poor timeliness, and efficiently respond to possible emergency fires in the goaf of the fully-mechanized caving face.

[0034] (3) A discriminator is constructed in the encoder-decoder process, and the encoder is used as the generator. The mapped result of the oxygen concentration distribution characteristics in the goaf and a random normal distribution are adversarially trained by the discriminator, prompting the oxygen concentration distribution characteristics in the goaf to generate high-quality reconstructed images through the decoder.

[0035] (4) A convolutional autoencoder model is established to realize the dimensionality reduction and reconstruction of oxygen concentration distribution data. The encoder uses a deep convolutional neural network instead of a linear layer, and the decoder up-samples by applying a transposed convolutional neural network, solving the problems of weak representation ability of the linear layer for high-dimensional data and overfitting or underfitting of feature extraction. Brief Description of the Drawings

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0037] Figure 1 It is a schematic diagram of the goaf of the working face of a "U"-shaped ventilation system provided by the embodiment of the present invention;

[0038] Figure 2 It is a schematic diagram of the method flow for intelligently predicting the effect of the leakage stoppage process at the end based on a network model provided by the embodiment of the present invention;

[0039] Figure 3 It is a schematic diagram of the structure of a deep learning model provided by the embodiment of the present invention;

[0040] Figure 4 It is a schematic diagram of the structure of an electronic device provided by the embodiment of the present invention. Detailed Embodiments

[0041] With the research and progress of artificial intelligence technology, artificial intelligence has been studied and applied in multiple fields. For example, common ones include smart home, intelligent customer service, virtual assistant, smart speaker, intelligent marketing, driverless, autonomous driving, robot, intelligent medical treatment, etc. It is believed that with the development of technology, artificial intelligence will be applied in more fields and play an increasingly important role.

[0042] Machine learning is an interdisciplinary field that involves multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve their own performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence, and its applications cover all fields of artificial intelligence. Machine learning and deep learning usually include technologies such as artificial neural networks, belief networks, reinforcement learning, transfer learning, and inductive learning.

[0043] In the description of the embodiments of the present invention, the terms used in the following embodiments are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification and claims of the present invention, the singular forms "a", "the", "above-mentioned", "this", and "such" are also intended to include forms such as "one or more", unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of the present invention, "at least one" and "one or more" mean one or more than two (including two). The term "and / or" is used to describe the association relationship of associated objects and means that three relationships can exist; for example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone, where A and B can be singular or plural. The character " / " generally means that the associated objects before and after are in an "or" relationship.

[0044] References to "one embodiment" or "some embodiments" etc. described in this specification mean that specific features, structures, or characteristics described in connection with that embodiment are included in one or more embodiments of the present invention. Thus, the statements "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments" etc. that appear in different places in this specification are not necessarily all referring to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "comprise", "include", "have" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways. The term "connection" includes direct connection and indirect connection, unless otherwise stated. "First" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features.

[0045] In the embodiments of the present invention, words such as "exemplarily" or "for example" are used to give examples, illustrations or explanations. Any embodiment or implementation described as "exemplarily" or "for example" in the embodiments of the present invention should not be construed as more preferred or more advantageous than other embodiments or implementations. Rather, the use of words such as "exemplarily" or "for example" is intended to present relevant concepts in a specific manner.

[0046] In view of the problems existing in the prior art, the present invention provides a new type of leak - plugging material applied to the ends of fully - mechanized caving faces. This material has both the effects of leak - plugging and oxygen isolation and temperature reduction and fire extinguishing to solve the problem of spontaneous combustion of residual coal in the gob caused by air leakage at the upper and lower ends of the fully - mechanized caving face.

[0047] As Figure 1 shown, for the gob of the "U" - shaped ventilation system, according to the magnitude of air leakage and the possibility of residual coal occurrence, the gob can be divided into three zones: the heat - dissipation zone, the spontaneous - combustion zone (oxidation and temperature - rising zone) and the asphyxiation zone. According to past mining experience, certain free spaces 1 are likely to form at both ends of the fully - mechanized caving face, which further exacerbates air leakage. If not plugged in time, it is extremely likely to cause spontaneous combustion in the gob, posing a great hidden danger to safe production. The new material provided by the present invention is precisely used to plug the free spaces 1 at both ends of the fully - mechanized caving face.

[0048] The new leak - plugging material provided by the embodiments of the present invention is a Janus - type foam slurry, which is formed by mixing a cement - based slurry, a Janus - type water - based foam and a coagulant. The interface formed by the hydrophobic nanoparticles and the hydrophilic polymer in the slurry has both hydrophilicity and hydrophobicity, enhancing the stability of the bubbles. In the fields of modern science and medicine, Janus - type may be used to describe certain molecules, proteins or drugs with two different functional or structural characteristics.

[0049] The characteristics of the Janus-type foam slurry are as follows: The cement-based slurry is formed by uniformly mixing two incombustible inorganic cementitious materials, namely coal-based solid waste and portland cement, with water. The coal-based solid waste accounts for 29.3 - 35% of the mass of these two inorganic cementitious materials, and the mass ratio of water to these two inorganic cementitious materials is 0.5 - 0.6. The Janus-type water-based foam is composed of a hydrophilic polymer, hydrophobic nanoparticles, and water. Among them, the hydrophilic polymer includes one or more of polyvinyl alcohol, xanthan gum, and sodium alginate, and the hydrophobic nanoparticles include one or more of silica, titanium dioxide, and carbon nanotubes. The hydrophilic polymer accounts for 1 - 3.2% of the mass of the Janus-type water-based foam, the hydrophobic nanoparticles account for 1.2 - 2% of the mass of the Janus-type water-based foam, and the addition amount of the Janus-type water-based foam is 2 - 2.21 times the volume of the cement-based slurry. The coagulant includes one or more of liquid silicate, powdered silicate, and aluminum sulfate, and the addition amount of the coagulant is 2.9 - 6.1% of the mass of the cement-based slurry. In addition, the Janus-type foam slurry will turn into a solidified foam within 10 - 30 seconds, and has the properties of high-efficiency stacking and long-term stability.

[0050] Based on the above Janus-type foam slurry with both plugging and oxygen isolation and cooling and fire extinguishing effects, in this embodiment, a deep learning model for predicting the plugging effect of the Janus-type foam slurry at the end of the fully-mechanized caving face under different working conditions is further trained. The main process is as follows: First, CFD simulation is used to obtain the image data of the oxygen concentration distribution in the goaf under different working condition parameters, and the simulated image data is preprocessed. Then, a deep learning model is trained based on the preprocessed data set. The trained deep learning model fits the mapping relationship between different working condition data and the oxygen concentration distribution characteristics in the goaf, and can quickly predict the plugging process effect after inputting the current actual parameter values such as grouting pressure, grouting flow rate, and grouting position.

[0051] As Figure 2 shown, an embodiment of the present invention provides a method for intelligently predicting the end plugging process effect based on a network model. The process of this method can be executed by an electronic device, and this method includes the following steps:

[0052] S201, use CFD simulation to obtain the effect of the Janus-type foam slurry plugging the end of the fully-mechanized caving face under different working condition parameters, and obtain the image data of the oxygen concentration distribution in the goaf.

[0053] In this embodiment, the different working condition parameters simulated by CFD involve grouting pressure, grouting flow rate, and the position of the grouting pipe, with at least 27 working conditions. CFD simulation: First, based on the actual situation of the goaf in the working face and the application technology of plugging the end with Janus-type foam slurry, an equal-proportion physical model is established. Second, the physical model is meshed. Finally, the distribution of oxygen concentration in the goaf of the fully-mechanized caving face under different working conditions is obtained. Through CFD simulation research, the distribution of oxygen concentration in the goaf of the fully-mechanized caving face under different working conditions can be obtained, providing data support for the training of the subsequent deep learning model.

[0054] S202. Perform data normalization and data augmentation preprocessing on the data obtained from the CFD simulation to construct a data set.

[0055] Specifically, in the first step, data normalization: The working condition parameters in the CFD simulation are normalized using, for example, the z-score normalization method to meet the input requirements of the deep regression neural network. In the second step, data augmentation: For example, the data is augmented using the Transforms tool in the TorchVision module of PyTorch to improve the generalization ability of the deep learning model. In the third step, data set construction: Combine the normalized and augmented data to obtain the input data set for the entire deep learning algorithm model. After determining the input data set for the entire deep learning algorithm model, construct a training set and a validation set, and load the training data and validation data by rewriting the Dataset and Dataloader in PyTorch, providing a data basis for the training and validation of the deep learning model.

[0056] S203. Train a deep learning model based on the preprocessed data set to obtain a trained deep learning model. The deep learning model includes an encoder, a discriminator, a decoder, and a deep regression neural network model. The trained deep learning model fits the mapping relationship between different working condition parameters and the distribution characteristics of oxygen concentration in the goaf.

[0057] S204. Input the current actual working condition parameters into the trained deep learning model, and the trained deep learning model outputs the distribution image data of oxygen concentration in the goaf. The distribution image data of oxygen concentration in the goaf is used to evaluate the effect of the end plugging process.

[0058] Combined with the figure, Figure 3 shows the structure of a deep learning model. The model includes two parts. The first part is: the deep convolutional adversarial auto-encoding algorithm; the second part is: the deep regression neural network. Among them, the preprocessed working condition parameter data set is Figure 3 the working condition data set I input into the deep regression neural network in n10. The image data of the oxygen concentration distribution in the goaf under different preprocessed working condition parameters is Figure 3 the dataset P of the oxygen concentration distribution in the goaf input into the input encoder 2 n 1.

[0059] From Figure 3 It can be seen that the working condition dataset I n 10 is input into the deep regression neural network model 9 and then connected to the encoder 2. The dataset P n 1 is input into the encoder 2. The encoder 2 applies a deep convolutional neural network to downsample the dataset P n 1 and compresses the dataset P n 1 into the dimensionality-reduced data feature 3. On the one hand, the dimensionality-reduced data feature 3 output by the encoder 2 is used as a negative sample (fake-), and random data with the same dimension as the dimensionality-reduced data feature 3 is generated from a certain random normal distribution 7 as a positive sample (real+). The positive sample (real+) and the negative sample (fake-) are input into the discriminator 8 together. The discriminator 8 adjusts the encoder 2 to compress the oxygen concentration distribution feature in the goaf, causing the decoder 4 to approach the real random normal distribution 7. On the other hand, the encoder 2 inputs the dimensionality-reduced data feature 3 into the decoder 4. The decoder 4 applies a transposed convolutional neural network to upsample the dimensionality-reduced data feature 3 to obtain the reconstructed dataset 5. Finally, the model finally outputs the predicted image 6 of the oxygen concentration distribution in the goaf. By applying the above model, it is possible to fit different working condition data and the oxygen concentration distribution feature in the goaf. When the actual parameters of the grouting pressure, grouting flow rate, and grouting position are input into the trained network model, it is possible to intelligently and quickly predict the effect of the high-efficiency air leakage prevention process.

[0060] Figure 3 Among them, a discriminator 8 is constructed between the encoder 2 and the decoder 4, and the decoder 4 is used as a generator. Through the discriminator 8, the mapping result of the oxygen concentration distribution feature in the goaf is adversarially trained with a random normal distribution, so that the oxygen concentration distribution feature generates a high-quality reconstructed image through the decoder. The objective function of the generative adversarial autoencoder algorithm model is as follows. Among them, a convolutional layer is used for downsampling, a transposed convolutional layer is used for upsampling, the encoder 2 consists of 7 convolutional neural network layers, the decoder 4 consists of 8 transposed convolutional neural network layers, the discriminator 8 uses 2 convolutional neural network layers for downsampling, and 1 fully connected layer outputs the compressed feature.

[0061]

[0062] In the formula, x is the data in the real dataset, an array; z is the random data in the random distribution, an array; P is the data distribution probability function; d represents the discriminator, g represents the decoder, min g max dV(d,g) represents that the decoder needs to minimize this cross - entropy loss while the discriminator maximizes the cross - entropy loss of real and fake pictures, V(d,g). represents the expectation of the real - data set; represents the expectation of the randomly - distributed data set.

[0063] The deep - learning model provided in this embodiment mainly executes three processes: data dimensionality reduction, data regression, and data reconstruction. Among them, based on the CFD simulation results, the model is trained to fit the data characteristics of the oxygen - concentration distribution results in the gob area of the fully - mechanized caving face, realizing the reconstruction of the oxygen - concentration distribution data in the gob area. By training the data - regression model to fit the distribution characteristics of different working - condition data and the oxygen concentration in the gob area, the deep - learning proxy model loads the weights of each part of the data - dimensionality reduction, data - regression, and data - reconstruction models. When inputting the grouting pressure, grouting flow rate, and grouting - position parameters, the oxygen - concentration feature vector after air leakage in the gob area is mapped through the deep - regression neural - network model, and then the encoder in the deep - convolutional adversarial auto - encoding algorithm is used to reconstruct the oxygen - concentration distribution image in the gob area, realizing the intelligent and rapid prediction of the efficient air - leakage - blocking process.

[0064] The beneficial effects of this embodiment compared with the prior art are as follows: (1) A Janus - type foam grout is developed, endowing the material with the performance advantages of air - leakage blocking, oxygen isolation, temperature reduction, and fire extinguishing. (2) This method can intelligently, rapidly, and efficiently predict the air - leakage - blocking effect at the end of the fully - mechanized caving face, solve the problems of long CFD simulation calculation cycle and poor timeliness, and efficiently respond to possible emergency fires in the gob area of the fully - mechanized caving face. (3) A discriminator is constructed in the encoder - decoder process, and the encoder is used as the generator. The mapped result of the oxygen - concentration distribution characteristics in the gob area and a random normal distribution are adversarially trained through the discriminator, prompting the oxygen - concentration distribution characteristics in the gob area to generate high - quality reconstructed images through the decoder. (4) A convolutional auto - encoder model is established to realize the dimensionality reduction and reconstruction of the oxygen - concentration distribution data. The encoder uses a deep convolutional neural network instead of a linear layer, and the decoder up - samples using a transposed convolutional neural network, solving the problems of weak representation ability of the linear layer for high - dimensional data and over - fitting or under - fitting of feature extraction.

[0065] All relevant contents of each step involved in the above - mentioned method embodiment can be cited in the function description of the corresponding functional module, and will not be elaborated here.

[0066] The above - mentioned system can be executed by a chip or a chip module. Regarding each module / unit included in each device and product described in the above - mentioned embodiments, it can be a software module / unit, a hardware module / unit, or it can also be partially a software module / unit and partially a hardware module / unit.

[0067] In some other embodiments of the present invention, the present - invention embodiment discloses an electronic device, such asFigure 4 As shown, the electronic device may integrate the above system. In terms of hardware composition, it includes: one or more processors 401; a memory 402; a display 403; one or more applications (not shown); and one or more computer programs 404. The above components may be connected through one or more communication buses 405. Among them, the one or more computer programs 404 are stored in the memory 402 and configured to be executed by the one or more processors 401. The one or more computer programs 404 include instructions.

[0068] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a computer, it implements the method described in the above method embodiment. The specific beneficial effects can be referred to the above method embodiment.

[0069] The present invention also provides a computer program product. When the computer program product is executed by a computer, it implements the method described in the above method embodiment. The specific beneficial effects can be referred to the above method embodiment.

[0070] Through the description of the above embodiments, those skilled in the art can clearly understand that for the convenience and conciseness of description, only the above division of each functional module is used as an example. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be repeated here.

[0071] In each embodiment of the present invention, the functional units may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0072] If the above integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as flash memory, mobile hard disk, read-only memory, random access memory, magnetic disk, or optical disk that can store program codes.

[0073] As described above, it is only the specific implementation manner of the embodiments of the present invention, but the protection scope of the embodiments of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed in the embodiments of the present invention should be covered within the protection scope of the embodiments of the present invention. Therefore, the protection scope of the embodiments of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for intelligently predicting the plugging effect of the working face end based on a network model, characterized in that: include: Computational fluid dynamics (CFD) was used to simulate the effect of Janus-type foam slurry on plugging the end of the fully mechanized caving working face under different working parameters, and the image data of oxygen concentration distribution in the goaf was obtained; the Janus-type foam slurry is a mixture of cement-based slurry, Janus-type water-based foam and accelerator; Perform data normalization and data enhancement preprocessing on the data obtained from CFD simulation to construct a data set; A deep learning model is trained based on the data set to obtain a trained deep learning model; wherein the deep learning model includes an encoder, a discriminator, a decoder and a deep regression neural network model, and the trained deep learning model fits the mapping relationship between different working condition parameters and the oxygen concentration distribution characteristics of the goaf; The current actual working condition parameters are input into the trained deep learning model, and the trained deep learning model outputs the goaf oxygen concentration distribution image data, and the goaf oxygen concentration distribution image data is used to evaluate the effect of the plugging process at the end of the fully mechanized caving working face.

2. The method according to claim 1, characterized in that Training a deep learning model based on the data set specifically includes: The working condition data set I n Input to the deep regression neural network model, the output end of the deep regression neural network model is connected to the encoder; The oxygen concentration distribution image dataset P of the goaf n Input the encoder, which applies a deep convolutional neural network to the dataset P n Downsample the dataset P n Compress into reduced dimension data features as negative samples; and generate random data of the same dimension as the reduced dimension data features from a determined random normal distribution as positive samples, and input the positive samples and the negative samples together into the discriminator, and the discriminator adjusts the encoder to compress the oxygen concentration distribution features of the goaf so that the decoder is infinitely close to the real random distribution normal distribution; The encoder inputs the reduced-dimensionality data features to the decoder, and the decoder applies a transposed convolutional neural network to upsample the reduced-dimensionality data features to obtain a reconstructed data set. Finally, the trained deep learning model outputs a predicted goaf oxygen concentration distribution image.

3. The method according to claim 1, characterized in that The encoder adopts the following formula for the objective function of the adversarial self-encoding algorithm: In the formula, x is the data in the real data set, array; z is the random data in the random distribution, array; P is the data distribution probability function; d represents the discriminator, g represents the decoder, min g max d V(d,g) means that the generator should minimize the cross entropy loss V(d,g) while maximizing the cross entropy loss between real and fake images in the discriminator; Represents the expectation of the real data set; Represents the expectation of a randomly distributed dataset.

4. The method according to claim 2, characterized in that: The cement-based slurry is formed by mixing coal-based solid waste and silicate cement, two flame-retardant inorganic cementitious materials, and water, wherein the coal-based solid waste accounts for 29.3-35% of the mass of the two inorganic cementitious materials, and the mass ratio of water to the two inorganic cementitious materials is 0.5-0.6; The Janus-type water-based foam is composed of a hydrophilic polymer, hydrophobic nanoparticles and water; the hydrophilic polymer accounts for 1 to 3.2% of the mass of the Janus-type water-based foam, the hydrophobic nanoparticles account for 1.2 to 2% of the mass of the Janus-type water-based foam, and the addition amount of the Janus-type water-based foam is 2 to 2.21 times the volume of the cement-based slurry; The coagulant includes one or more of liquid silicate, powdered silicate and aluminum sulfate, and the added amount of the coagulant is 2.9-6.1% of the mass of the cement-based slurry.

5. The method according to any one of claims 1 to 4, characterized in that The hydrophilic polymer includes one or more of polyvinyl alcohol, xanthan gum and sodium alginate, and the hydrophobic nanoparticles include one or more of silicon dioxide, titanium dioxide and carbon nanotubes; the operating parameters include grouting pressure, grouting flow rate and grouting pipe position.

6. The method according to any one of claims 1 to 4, characterized in that The data obtained from CFD simulation is preprocessed by data normalization and data enhancement, including: The z-score normalization method is used to normalize the operating parameters in CFD simulation; The data of oxygen concentration distribution image data in the goaf is enhanced through the Transforms tool of the TorchVision module in PyTorch.

7. The method according to any one of claims 1 to 4, characterized in that: The encoder is composed of 7 convolutional neural network layers, the decoder is composed of 8 transposed convolutional neural network layers, and the discriminator uses 2 convolutional neural network layers for downsampling and 1 fully connected layer for feature compression.

8. A system for intelligently predicting the plugging effect of the working face end based on a network model, characterized in that: include: A simulation data module is used to simulate different working parameters using CFD, as well as the image data of oxygen concentration distribution in the goaf of the fully mechanized caving working face end blocked by Janus-type foam slurry under different working parameters; wherein the Janus-type foam slurry is a mixture of cement-based slurry, Janus-type water-based foam and accelerator; Construct a data set module to perform data normalization and data enhancement preprocessing on the data obtained from CFD simulation and construct a data set; A training module, used for training a deep learning model based on the data set to obtain a trained deep learning model; wherein the deep learning model includes an encoder, a discriminator, a decoder and a deep regression neural network model, and the trained deep learning model fits the mapping relationship between different working condition parameters and the oxygen concentration distribution characteristics of the goaf; The model application module is used to input the current actual operating parameters into the trained deep learning model, and the trained deep learning model outputs the oxygen concentration distribution image data of the goaf, and the oxygen concentration distribution image data of the goaf is used to evaluate the effect of the end plugging process.

9. A computer-readable storage medium, wherein a program is stored in the computer-readable storage medium, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory stores a program executable on the processor, and when the program is executed by the processor, the electronic device implements the method as claimed in any one of claims 1 to 7.

Images