Deep learning-based rock failure time prediction method and system

Abstract

The invention discloses a rock failure time prediction method and system based on deep learning, and the method comprises the steps: obtaining monitoring data collected by a plurality of sensors when a target type of rock is damaged, carrying out the preprocessing of the monitoring data, obtaining the preprocessed data, and generating a data set according to the preprocessed data; constructing a deep learning combination model, and training and testing the deep learning combination model by using the data set to obtain a target prediction model; and obtaining real-time state data of a to-be-predicted target type rock, and inputting the real-time state data into the target prediction model for prediction to obtain the residual damage time of the to-be-predicted rock. The rock damage time can be predicted in real time, the spatio-temporal characteristics in the rock damage process can be accurately captured, the early warning signal can be sent out, the method has important application prospects, and the method has important value in rock damage monitoring and safety management especially in mining, tunneling and other projects.

Description

Technical Field

[0001] The present invention relates to the technical field of resource mining disaster prevention and control, and in particular to a rock failure time prediction method, system, terminal and computer-readable storage medium based on deep learning. Background Art

[0002] The study of rock failure process is of great significance in engineering fields such as geological engineering, mining, tunneling and reservoir dams. Rock failure is usually accompanied by complex physical and mechanical changes, such as stress, strain and acoustic emission. In particular, acoustic emission signals, as a characterization of micro-fractures inside rock materials, can provide rich dynamic information about the evolution of rock failure. However, due to the heterogeneity of rock materials and the complexity of the failure process, accurately predicting the failure time of rocks has always been a challenging research topic.

[0003] Existing studies mostly use monitoring data such as acoustic emission for early warning and prediction of rock failure, but such methods usually assume that the rock failure process is a stable and uniform process, which is difficult to cope with the complex heterogeneity and randomness of rock materials in practice. In order to improve the prediction accuracy, artificial intelligence and deep learning methods have been gradually introduced into the field of rock mechanics in recent years. Among them, long short-term memory networks have been used to analyze and predict rock failure time due to their superiority in processing time series data.

[0004] However, the long short-term memory network can only learn the dynamic change characteristics of time series data, and cannot effectively capture the complex topological relationship between multidimensional data in rock acoustic emission signals. It does not fully utilize the spatiotemporal characteristics of the acoustic emission multidimensional information sequence, resulting in insufficient prediction accuracy for rock failure time.

[0005] Therefore, the prior art still needs to be improved and developed. Summary of the invention

[0006] The main purpose of the present invention is to provide a rock failure time prediction method, system, terminal and computer-readable storage medium based on deep learning, aiming to solve the problem that when predicting rock failure time based on deep learning network in the prior art, the complex topological relationship between multidimensional data in rock acoustic emission signals cannot be effectively captured, and the spatiotemporal characteristics of the acoustic emission multidimensional information sequence are not fully utilized, resulting in insufficient prediction accuracy of rock failure time.

[0007] To achieve the above object, the present invention provides a rock failure time prediction method based on deep learning, and the rock failure time prediction method based on deep learning comprises the following steps: Acquire monitoring data of target type rock destruction collected by multiple sensors, preprocess the monitoring data to obtain preprocessed data, and generate a data set based on the preprocessed data; Constructing a deep learning combination model, and using the data set to train and test the deep learning combination model to obtain a target prediction model; The real-time status data of the target type of rock to be predicted is obtained, and the real-time status data is input into the target prediction model for prediction to obtain the remaining failure time of the rock to be predicted.

[0008] Optionally, in the rock failure time prediction method based on deep learning, the multiple sensors include a stress sensor, a strain sensor and an acoustic emission sensor, the stress sensor is used to obtain stress data, the strain sensor is used to obtain strain data, and the acoustic emission sensor is used to obtain acoustic emission data; The monitoring data includes stress data, strain data and acoustic emission data, wherein the acoustic emission data includes basic waveform characteristic parameters, continuous signal parameters, spectrum analysis data and waveform analysis data.

[0009] Optionally, the rock failure time prediction method based on deep learning, wherein the monitoring data is preprocessed to obtain preprocessed data, and a data set is generated according to the preprocessed data, specifically comprising: Performing noise reduction and standardization processing on the monitoring data to obtain first intermediate data, and performing missing value deletion on the first intermediate data to obtain second intermediate data; Performing sliding window movement processing on the time series of the second intermediate data to obtain the preprocessed data; The preprocessed data is divided into a training set, a test set and a validation set according to a preset ratio, the training set is used to train the deep learning combination model, the test set is used to evaluate the deep learning combination model in each round of training, and the validation set is used to evaluate the trained deep learning combination model.

[0010] Optionally, in the rock failure time prediction method based on deep learning, the deep learning combined model includes a long short-term memory network, a graph neural network, a fusion layer and a fully connected layer; Among them, the long short-term memory network is used to capture the dynamic timing characteristics of time series data in the rock failure process, and the graph neural network is used to learn the topological relationship between acoustic emission multidimensional data sequences.

[0011] Optionally, the rock failure time prediction method based on deep learning, wherein the use of the data set to train and test the deep learning combined model to obtain a target prediction model specifically includes: The long short-term memory network and the graph neural network are combined in a serial or parallel manner, and the training set is input into the deep learning combination model to extract time series data features and spatial features, so as to obtain the time series features of rock data and the topological relationship between rock data sequences of different dimensions; Inputting the time series features and the topological relationship into the fusion layer for weighted feature fusion or hierarchical feature fusion to obtain fusion features, inputting the fusion features into the fully connected layer for mapping, and outputting the predicted remaining damage time; Calculate a mean square error according to the predicted remaining damage time and the corresponding label, use the mean square error as a loss function, set a back propagation gradient according to the loss function, adjust network hyperparameters of the long short-term memory network and the graph neural network according to the back propagation gradient and based on a back propagation algorithm, and evaluate the model based on the test set; Continue to train the deep learning combined model based on the adjusted network parameters until the loss function reaches a preset convergence condition, thereby obtaining a trained deep learning combined model; The prediction result of the trained deep learning combination model is judged using the validation set. If the error between the prediction result and the label in the validation set is lower than a preset threshold, the target prediction model is obtained.

[0012] Optionally, the rock failure time prediction method based on deep learning, wherein the obtaining of the remaining failure time of the rock to be predicted further includes: The remaining destruction time is compared with safety thresholds of different levels. If the remaining destruction time is lower than a certain level of safety threshold, a safety warning signal of a corresponding level is issued according to the level of the safety threshold, and corresponding emergency measures are triggered according to the level of the safety warning signal.

[0013] Optionally, the rock failure time prediction method based on deep learning, wherein the rock failure time prediction method based on deep learning further includes: Adding an adaptive learning mechanism to the target prediction model and acquiring monitoring data and prediction data of the target prediction model over a long period of time; According to the monitoring data and the prediction data, based on the adaptive learning mechanism, the target prediction model is continuously updated and optimized to maintain high-precision prediction.

[0014] In addition, to achieve the above-mentioned purpose, the present invention also provides a rock failure time prediction system based on deep learning, wherein the rock failure time prediction system based on deep learning includes: A data acquisition and processing module is used to obtain monitoring data of target type rock destruction collected by various sensors, preprocess the monitoring data to obtain preprocessed data, and generate a data set based on the preprocessed data; A model building and training module is used to build a deep learning combination model, train and test the deep learning combination model using the data set, and obtain a target prediction model; The rock failure time prediction module is used to obtain the real-time status data of the target type of rock to be predicted, input the real-time status data into the target prediction model for prediction, and obtain the remaining failure time of the rock to be predicted.

[0015] In addition, to achieve the above-mentioned purpose, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a deep learning-based rock failure time prediction program stored in the memory and runnable on the processor, and when the deep learning-based rock failure time prediction program is executed by the processor, the steps of the deep learning-based rock failure time prediction method as described above are implemented.

[0016] In addition, to achieve the above-mentioned objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a rock failure time prediction program based on deep learning, and when the rock failure time prediction program based on deep learning is executed by a processor, the steps of the rock failure time prediction method based on deep learning as described above are implemented.

[0017] In the present invention, monitoring data of target type rock destruction collected by multiple sensors are obtained, the monitoring data are preprocessed to obtain preprocessed data, and a data set is generated according to the preprocessed data; a deep learning combination model is constructed, and the deep learning combination model is trained and tested using the data set to obtain a target prediction model; the real-time status data of the target type rock to be predicted is obtained, and the real-time status data is input into the target prediction model for prediction to obtain the remaining destruction time of the rock to be predicted. The present invention can predict the rock destruction time in real time, accurately capture the spatiotemporal characteristics of the rock destruction process, and issue an early warning signal, which has important application prospects, especially in mining, tunneling, reservoir dams and other projects, and has important theoretical and practical value for monitoring and safety management of rock destruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1It is a flow chart of a preferred embodiment of the rock failure time prediction method based on deep learning of the present invention; Figure 2 It is a principle diagram of a preferred embodiment of the rock failure time prediction method based on deep learning of the present invention; Figure 3 It is a flowchart of training and testing the deep learning combined model in the rock failure time prediction method based on deep learning of the present invention; Figure 4 It is a structural schematic diagram of a preferred embodiment of a rock failure time prediction system based on deep learning of the present invention; Figure 5 Schematic diagram of the operating environment of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0019] The present application provides a rock failure time prediction method, system, terminal and computer-readable storage medium based on deep learning. In order to make the purpose, technical solution and effect of the present application clearer and more specific, the present application is further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0020] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with those in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as here.

[0021] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0022] The rock failure time prediction method based on deep learning described in the preferred embodiment of the present invention is as follows: Figure 1 and Figure 2 As shown, the rock failure time prediction method based on deep learning includes the following steps: Step S10: Acquire monitoring data of target type rock destruction collected by multiple sensors, preprocess the monitoring data to obtain preprocessed data, and generate a data set based on the preprocessed data.

[0023] In this embodiment, a variety of sensors, including stress sensors, strain sensors, and acoustic emission sensors, are arranged in the rock failure area to monitor the overall state parameters of the rock and obtain monitoring data for subsequent model training. The stress sensor is used to obtain stress data, the strain sensor is used to obtain strain data, and the acoustic emission sensor is used to obtain acoustic emission data.

[0024] Specifically, firstly, a variety of sensors are used to collect multi-dimensional acoustic emission signal data (monitoring data) during the rock failure process, including stress data, strain data and acoustic emission data; the acoustic emission data includes basic waveform characteristic parameters (such as wave impact count, ring count, amplitude, energy count, rise time, duration and time difference, etc.), continuous signal parameters (such as ring count, average signal level and effective value voltage, etc.), spectrum analysis data and waveform analysis data. It can be understood that a variety of sensors are arranged at key positions of rock samples to ensure the integrity and accuracy of the data.

[0025] Furthermore, the monitoring data is subjected to noise reduction and standardization processing to obtain first intermediate data, and missing values ​​are deleted from the first intermediate data to obtain second intermediate data; and a sliding window moving processing is performed on the time series of the second intermediate data to obtain the preprocessed data.

[0026] It can be understood that the monitoring data is first subjected to noise reduction processing. Data noise reduction processing refers to preprocessing the data to eliminate or reduce the noise in the data, thereby improving the quality, accuracy and credibility of the data. Common noise reduction methods include mean filtering, singular value decomposition and wavelet transform. Next, the noise-reduced data is standardized. The standardized processing may include steps such as data cleaning, data conversion and data normalization to ensure the consistency, accuracy and comparability of the data. Then, the first intermediate data is subjected to missing value deletion, that is, directly deleting samples containing missing values, which is sometimes the simplest and most effective, but the premise is that the proportion of missing data is small and the missing data appears randomly, so that the deletion of missing data has little effect on the analysis results, and the second intermediate data is obtained. Finally, the time series of the second intermediate data is subjected to sliding window movement processing to obtain the preprocessed data: In time series analysis, sliding window is a common and effective representation technology. The sliding window technology divides the time series data into multiple windows of fixed size, which can be overlapping or non-overlapping. The time series data in each window is used to generate a feature vector to represent the state or characteristics of the window.

[0027] These data preprocessing steps provide high-quality input for subsequent model learning. The data preprocessing steps ensure the accuracy and consistency of the data, laying the foundation for the effective training of the subsequent model. By denoising and standardizing the data, environmental noise interference can be eliminated and the prediction accuracy of the model can be improved. Acoustic emission data provides dynamic information on the formation and expansion of microcracks inside the rock.

[0028] Furthermore, the preprocessed data is divided into a training set, a test set and a validation set according to a preset ratio, the training set is used to train the deep learning combination model, the test set is used to evaluate the deep learning combination model in each round of training, and the validation set is used to evaluate the trained deep learning combination model.

[0029] The preprocessed data is divided into a training set, a test set, and a validation set according to a preset ratio (for example, the data ratio of the training set, the test set, and the validation set is 6:2:2). The training set is used to train the deep learning combination model and optimize the parameters of the model, while the test set is used to detect the accuracy of the prediction results of the trained deep learning combination model, thereby evaluating the performance of the model.

[0030] Step S20: construct a deep learning combination model, use the data set to train and test the deep learning combination model, and obtain a target prediction model.

[0031] In this embodiment, the deep learning combined model includes a long short-term memory network, a graph neural network, a fusion layer and a fully connected layer (multi-layer perceptron); wherein the long short-term memory network is used to capture the dynamic timing characteristics of time series data during rock failure, and the graph neural network is used to learn the topological relationship between acoustic emission multi-dimensional data sequences.

[0032] like Figure 3 As shown, the deep learning combined model is trained and tested using the data set to obtain a target prediction model, which specifically includes: Step S201: Combine the long short-term memory network and the graph neural network in a serial or parallel manner, input the training set into the deep learning combination model to perform time series data feature extraction and spatial feature extraction, and obtain the time series characteristics of the rock data and the topological relationship between the rock data sequences of different dimensions.

[0033] It can be understood that the long short-term memory network (LSTM) can effectively process multi-dimensional time series and learn the time evolution characteristics of rock damage; the graph neural network (GNN) regards the multi-dimensional acoustic emission signal sequence as a graph structure, and GNN can perform convolution operations on the graph structure to learn the complex associations between data. The present invention uniquely applies GNN to the acoustic emission multi-dimensional information sequence to construct a graph model of the rock destruction process to learn the topological relationship between multi-dimensional data. Traditional methods are usually unable to effectively model the complex spatial correlation in acoustic emission signals, and the present invention achieves this through GNN. GNN can perform graph modeling on the acoustic emission multi-dimensional information sequence and capture the spatial relationship between data, thereby providing a deep understanding of the internal destruction mechanism of the rock. By utilizing the topological structure in the acoustic emission data, the model can accurately identify the key changes in the rock destruction process and improve the prediction accuracy of the remaining destruction time.

[0034] The present invention combines LSTM and GNN for the first time to predict the remaining destruction time of rocks. LSTM is good at capturing long-term dependencies in time series data, while GNN is good at learning the topological relationships of multidimensional data. By combining these two models, the present invention can simultaneously capture the temporal dynamic timing characteristics and spatial topological characteristics of the acoustic emission data sequence in the rock destruction process, thereby improving the prediction accuracy of the model and enhancing the robustness of the model.

[0035] Furthermore, there are many combinations of LSTM and GNN, including serial models and parallel models: LSTM and GNN can be used in series, first extracting temporal features through LSTM, then extracting spatial features (topological relationships) through GNN, or first extracting spatial features through GNN, then extracting temporal features through LSTM. The model can better mine the features of data layer by layer. Furthermore, LSTM and GNN can run in parallel, extracting temporal and spatial features separately, and then performing comprehensive analysis at the fusion layer. Parallel models can learn simultaneously in temporal and spatial dimensions, improving the efficiency and accuracy of the model.

[0036] Step S202: input the time series features and the topological relationship into the fusion layer for weighted feature fusion or hierarchical feature fusion to obtain fused features, input the fused features into the fully connected layer for mapping, and output the predicted remaining damage time.

[0037] In this embodiment, a feature fusion strategy is proposed to fuse the temporal features and spatial features extracted by LSTM and GNN to form a comprehensive feature representation for predicting rock failure time. The fused features contain time series and topological structure information, making the model's representation of the rock failure process more comprehensive and detailed. By fusing different types of features, the model has better generalization ability under different geological conditions and can cope with the diversity and complexity of the rock failure process.

[0038] Furthermore, there are multiple combinations of the time series features and the topological relationships, including weighted fusion and hierarchical fusion: weighted fusion is performed according to the importance of the time series features and the topological relationships to optimize the decision-making process of the model (for example, the weights of the fusion layer can be automatically adjusted through learning so that the model pays more attention to the features that contribute more to the prediction results). Alternatively, feature fusion is performed at multiple levels, such as fusion after each layer output of LSTM and GNN, to build a more hierarchical feature representation layer by layer.

[0039] Step S203, calculate the mean square error according to the predicted remaining damage time and the corresponding label, use the mean square error as the loss function, set the back propagation gradient according to the loss function, adjust the network hyperparameters of the long short-term memory network and the graph neural network according to the back propagation gradient and based on the back propagation algorithm, and evaluate the model based on the test set.

[0040] It is understandable that during the model training process, a large-scale rock damage data set is used to train the model. The network parameters of LSTM and GNN are adjusted through the back propagation algorithm, and the mean square error (MSE) is used as the loss function to ensure that the error between the predicted value and the true value is minimized. The back propagation gradient is set according to the loss function, and the network parameters are adjusted based on the back propagation gradient and the back propagation algorithm.

[0041] Step S204: Continue to train the deep learning combination model based on the adjusted network parameters until the loss function reaches a preset convergence condition, thereby obtaining a trained deep learning combination model.

[0042] Step S205: Use the verification set to judge the prediction result of the trained deep learning combination model. If the error between the prediction result and the label in the verification set is lower than a preset threshold, the target prediction model is obtained.

[0043] In this embodiment, the main purpose of setting the validation set is to evaluate the final performance and generalization ability of the model. The validation set is used to evaluate the performance of the model on unseen data and is the final test of the model's generalization ability. If the error between the prediction result and the label is lower than the preset threshold (such as 5% or 10%), it means that the effect is good.

[0044] Step S30: acquiring real-time status data of the target type of rock to be predicted, inputting the real-time status data into the target prediction model for prediction, and obtaining the remaining destruction time of the rock to be predicted.

[0045] In this embodiment, the trained target prediction model is deployed to the system, and the sensor data is received in real time for calculation. According to the input data, the model outputs the remaining failure time of the rock.

[0046] Furthermore, the remaining destruction time is compared with safety thresholds of different levels. If the remaining destruction time is lower than a certain level of safety threshold, a safety warning signal of a corresponding level is issued according to the level of the safety threshold, and corresponding emergency measures are triggered according to the level of the safety warning signal.

[0047] Specifically, multiple safety thresholds can be set according to actual conditions and needs. For example, in a certain environment, three safety thresholds are set, namely the first safety threshold, the second safety threshold and the third safety threshold. The first safety threshold, the second safety threshold and the third safety threshold are 1 hour, 2 hours and 3 hours from small to large, respectively. Among them, the safety warning signal corresponding to the first safety threshold is level one, the safety warning signal corresponding to the second safety threshold is level two, and the safety warning signal corresponding to the third safety threshold is level three. When the predicted remaining destruction time is obtained, the remaining destruction time is compared with safety thresholds of different levels. For example, if the remaining destruction time is 50 minutes, and the remaining destruction time is lower than the first safety threshold (1 hour), the safety warning signal (level one) corresponding to the first safety threshold is obtained, and a level one safety warning signal is issued, and corresponding emergency measures are triggered (for example, evacuation is completed within 30 minutes).

[0048] Furthermore, the rock failure time prediction method based on deep learning also includes: Adding an adaptive learning mechanism to the target prediction model and acquiring monitoring data and prediction data of the target prediction model over a long period of time; According to the monitoring data and the prediction data, based on the adaptive learning mechanism, the target prediction model is continuously updated and optimized to maintain high-precision prediction.

[0049] It can be understood that after adding an adaptive learning mechanism to the target prediction model, the system has an adaptive learning function. Through long-term monitoring and data accumulation, the LSTM+GNN model is continuously updated and optimized, so that it can always maintain high-precision predictions under different geological conditions. It can automatically adjust the model according to different geological conditions to ensure accuracy in long-term use and adapt to complex and changeable rock environments.

[0050] Furthermore, if Figure 4 As shown, based on the above-mentioned rock failure time prediction method based on deep learning, the present invention also provides a rock failure time prediction system based on deep learning, wherein the rock failure time prediction system based on deep learning includes: The data acquisition and processing module 51 is used to obtain monitoring data of target type rock failure collected by various sensors, preprocess the monitoring data to obtain preprocessed data, and generate a data set according to the preprocessed data; A model building and training module 52 is used to build a deep learning combination model, train and test the deep learning combination model using the data set, and obtain a target prediction model; The rock failure time prediction module 53 is used to obtain the real-time status data of the target type of rock to be predicted, input the real-time status data into the target prediction model for prediction, and obtain the remaining failure time of the rock to be predicted.

[0051] Furthermore, if Figure 5 As shown, based on the above-mentioned rock failure time prediction method and system based on deep learning, the present invention also provides a terminal accordingly, and the terminal includes a processor 10, a memory 20 and a display 30. Figure 5 Only some components of the terminal are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0052] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal. Furthermore, the memory 20 may also include both an internal storage unit of the terminal and an external storage device. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code of the installation terminal. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a rock failure time prediction program 40 based on deep learning is stored on the memory 20, and the rock failure time prediction program 40 based on deep learning can be executed by the processor 10, thereby realizing the rock failure time prediction method based on deep learning in the present application.

[0053] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code or process data stored in the memory 20, such as executing the rock failure time prediction method based on deep learning.

[0054] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 30 is used to display information on the terminal and to display a visual user interface. The components of the terminal communicate with each other via a system bus.

[0055] In one embodiment, when the processor 10 executes the rock failure time prediction program 40 based on deep learning in the memory 20, the following steps are implemented: Acquire monitoring data of target type rock destruction collected by multiple sensors, preprocess the monitoring data to obtain preprocessed data, and generate a data set based on the preprocessed data; Constructing a deep learning combination model, and using the data set to train and test the deep learning combination model to obtain a target prediction model; The real-time status data of the target type of rock to be predicted is obtained, and the real-time status data is input into the target prediction model for prediction to obtain the remaining failure time of the rock to be predicted.

[0056] Wherein, the multiple sensors include stress sensors, strain sensors and acoustic emission sensors, the stress sensors are used to obtain stress data, the strain sensors are used to obtain strain data, and the acoustic emission sensors are used to obtain acoustic emission data; The monitoring data includes stress data, strain data and acoustic emission data, wherein the acoustic emission data includes basic waveform characteristic parameters, continuous signal parameters, spectrum analysis data and waveform analysis data.

[0057] The preprocessing of the monitoring data to obtain preprocessed data and generating a data set according to the preprocessed data specifically includes: Performing noise reduction and standardization processing on the monitoring data to obtain first intermediate data, and performing missing value deletion on the first intermediate data to obtain second intermediate data; Performing sliding window movement processing on the time series of the second intermediate data to obtain the preprocessed data; The preprocessed data is divided into a training set, a test set and a validation set according to a preset ratio, the training set is used to train the deep learning combination model, the test set is used to evaluate the deep learning combination model in each round of training, and the validation set is used to evaluate the trained deep learning combination model.

[0058] The deep learning combined model includes a long short-term memory network, a graph neural network, a fusion layer and a fully connected layer; Among them, the long short-term memory network is used to capture the dynamic timing characteristics of time series data in the rock failure process, and the graph neural network is used to learn the topological relationship between acoustic emission multidimensional data sequences.

[0059] The step of using the data set to train and test the deep learning combined model to obtain a target prediction model specifically includes: The long short-term memory network and the graph neural network are combined in a serial or parallel manner, and the training set is input into the deep learning combination model to extract time series data features and spatial features, so as to obtain the time series features of rock data and the topological relationship between rock data sequences of different dimensions; Inputting the time series features and the topological relationship into the fusion layer for weighted feature fusion or hierarchical feature fusion to obtain fusion features, inputting the fusion features into the fully connected layer for mapping, and outputting the predicted remaining damage time; Calculate a mean square error according to the predicted remaining damage time and the corresponding label, use the mean square error as a loss function, set a back propagation gradient according to the loss function, adjust network hyperparameters of the long short-term memory network and the graph neural network according to the back propagation gradient and based on a back propagation algorithm, and evaluate the model based on the test set; Continue to train the deep learning combined model based on the adjusted network hyperparameters until the loss function reaches a preset convergence condition, thereby obtaining a trained deep learning combined model; The prediction result of the trained deep learning combination model is judged using the validation set. If the error between the prediction result and the label in the validation set is lower than a preset threshold, the target prediction model is obtained.

[0060] Wherein, the remaining failure time of the rock to be predicted is obtained, and then the following steps are further included: The remaining destruction time is compared with safety thresholds of different levels. If the remaining destruction time is lower than a certain level of safety threshold, a safety warning signal of a corresponding level is issued according to the level of the safety threshold, and corresponding emergency measures are triggered according to the level of the safety warning signal.

[0061] Wherein, the rock failure time prediction method based on deep learning also includes: Adding an adaptive learning mechanism to the target prediction model and acquiring monitoring data and prediction data of the target prediction model over a long period of time; According to the monitoring data and the prediction data, based on the adaptive learning mechanism, the target prediction model is continuously updated and optimized to maintain high-precision prediction.

[0062] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a rock failure time prediction program based on deep learning, and when the rock failure time prediction program based on deep learning is executed by a processor, the steps of the rock failure time prediction method based on deep learning as described above are implemented.

[0063] In summary, the present invention provides a method and system for predicting rock failure time based on deep learning, the method comprising: obtaining monitoring data of target type rock failure collected by multiple sensors, preprocessing the monitoring data to obtain preprocessed data, and generating a data set based on the preprocessed data; constructing a deep learning combination model, using the data set to train and test the deep learning combination model to obtain a target prediction model; obtaining real-time status data of the target type rock to be predicted, inputting the real-time status data into the target prediction model for prediction, and obtaining the remaining failure time of the rock to be predicted. The present invention can predict the rock failure time in real time, accurately capture the spatiotemporal characteristics of the rock failure process, and issue an early warning signal, which has important application prospects, especially in mining, tunneling, reservoir dams and other projects, and has important value for monitoring and safety management of rock failure.

[0064] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or terminal including the element.

[0065] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0066] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A rock failure time prediction method based on deep learning, It is characterized in that The rock failure time prediction method based on deep learning includes: Acquire monitoring data of target type rock destruction collected by multiple sensors, preprocess the monitoring data to obtain preprocessed data, and generate a data set based on the preprocessed data; Constructing a deep learning combination model, and using the data set to train and test the deep learning combination model to obtain a target prediction model; The real-time status data of the target type of rock to be predicted is obtained, and the real-time status data is input into the target prediction model for prediction to obtain the remaining failure time of the rock to be predicted.

2. According to the rock failure time prediction method based on deep learning in claim 1, It is characterized in that The multiple sensors include a stress sensor, a strain sensor and an acoustic emission sensor, wherein the stress sensor is used to obtain stress data, the strain sensor is used to obtain strain data, and the acoustic emission sensor is used to obtain acoustic emission data; The monitoring data includes stress data, strain data and acoustic emission data, wherein the acoustic emission data includes basic waveform characteristic parameters, continuous signal parameters, spectrum analysis data and waveform analysis data.

3. According to the rock failure time prediction method based on deep learning in claim 1, It is characterized in that The preprocessing of the monitoring data to obtain preprocessed data and generating a data set according to the preprocessed data specifically includes: Performing noise reduction and standardization processing on the monitoring data to obtain first intermediate data, and performing missing value deletion on the first intermediate data to obtain second intermediate data; Performing sliding window movement processing on the time series of the second intermediate data to obtain the preprocessed data; The preprocessed data is divided into a training set, a test set and a validation set according to a preset ratio, the training set is used to train the deep learning combination model, the test set is used to evaluate the deep learning combination model in each round of training, and the validation set is used to evaluate the trained deep learning combination model.

4. According to the rock failure time prediction method based on deep learning in claim 3, It is characterized in that The deep learning combined model includes a long short-term memory network, a graph neural network, a fusion layer and a fully connected layer; Among them, the long short-term memory network is used to capture the dynamic timing characteristics of time series data in the rock failure process, and the graph neural network is used to learn the topological relationship between acoustic emission multidimensional data sequences.

5. The rock failure time prediction method based on deep learning according to claim 4, It is characterized in that The using the data set to train and test the deep learning combined model to obtain a target prediction model specifically includes: The long short-term memory network and the graph neural network are combined in a serial or parallel manner, and the training set is input into the deep learning combination model to extract time series data features and spatial features, so as to obtain the time series features of rock data and the topological relationship between rock data sequences of different dimensions; Inputting the time series features and the topological relationship into the fusion layer for weighted feature fusion or hierarchical feature fusion to obtain fusion features, inputting the fusion features into the fully connected layer for mapping, and outputting the predicted remaining damage time; Calculate a mean square error according to the predicted remaining damage time and the corresponding label, use the mean square error as a loss function, set a back propagation gradient according to the loss function, adjust network hyperparameters of the long short-term memory network and the graph neural network according to the back propagation gradient and based on a back propagation algorithm, and evaluate the model based on the test set; Continue to train the deep learning combination model based on the adjusted network parameters until the loss function reaches a preset convergence condition, thereby obtaining a trained deep learning combination model; The prediction result of the trained deep learning combination model is judged using the validation set. If the error between the prediction result and the label in the validation set is lower than a preset threshold, the target prediction model is obtained.

6. The rock failure time prediction method based on deep learning according to claim 1, It is characterized in that The remaining failure time of the rock to be predicted is obtained, and then the following steps are further included: The remaining destruction time is compared with safety thresholds of different levels. If the remaining destruction time is lower than a certain level of safety threshold, a safety warning signal of a corresponding level is issued according to the level of the safety threshold, and corresponding emergency measures are triggered according to the level of the safety warning signal.

7. The rock failure time prediction method based on deep learning according to claim 1, It is characterized in that The rock failure time prediction method based on deep learning also includes: Adding an adaptive learning mechanism to the target prediction model and acquiring monitoring data and prediction data of the target prediction model over a long period of time; According to the monitoring data and the prediction data, based on the adaptive learning mechanism, the target prediction model is continuously updated and optimized to maintain high-precision prediction.

8. A rock failure time prediction system based on deep learning, It is characterized in that The rock failure time prediction system based on deep learning includes: A data acquisition and processing module is used to obtain monitoring data of target type rock destruction collected by various sensors, preprocess the monitoring data to obtain preprocessed data, and generate a data set based on the preprocessed data; A model building and training module is used to build a deep learning combination model, train and test the deep learning combination model using the data set, and obtain a target prediction model; The rock failure time prediction module is used to obtain the real-time status data of the target type of rock to be predicted, input the real-time status data into the target prediction model for prediction, and obtain the remaining failure time of the rock to be predicted.

9. A terminal, It is characterized in that The terminal includes: a memory, a processor, and a deep learning-based rock failure time prediction program stored in the memory and executable on the processor. When the deep learning-based rock failure time prediction program is executed by the processor, the steps of the deep learning-based rock failure time prediction method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a deep learning-based rock failure time prediction program, and when the deep learning-based rock failure time prediction program is executed by a processor, the steps of the deep learning-based rock failure time prediction method as described in any one of claims 1 to 7 are implemented.

Images