Microseismic signal identification method, training method, device, equipment and medium
Through the one-dimensional convolutional neural network training model based on transfer learning, the problems of low efficiency and poor consistency in the traditional microseismic signal recognition method are solved, and efficient and accurate microseismic signal recognition is achieved, which improves the accuracy and consistency of the recognition.
Patent Information
- Application Number
- CN202510416305.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-03
- Publication Date
- 2025-07-18
AI Technical Summary
Traditional microseismic signal recognition methods rely on manual empirical analysis, have low efficiency and poor consistency, making it difficult to meet the needs of real-time monitoring and accuracy.
A one-dimensional convolutional neural network based on transfer learning is used to train the model through seismic signal data and microseismic signal data to classify and identify microseismic signals, pre-training with seismic signal data, and fine-tuning of microseismic signal data to improve the recognition accuracy and consistency of the model.
It realizes efficient, accurate and consistent recognition of micro-seismic signals, meets the needs of real-time monitoring and accuracy, and improves the accuracy and recall of recognition.
Smart Images
Figure CN120336853A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of geological engineering, and in particular to a method for identifying microseismic signals, a training method, a device, equipment, and a medium. Background Art
[0002] As a core technology for the safety monitoring of underground engineering, a microseismic monitoring system captures minute vibration signals of rock masses or structures in real time through a high-precision sensor network, enabling early warning of disasters, structural health assessment, and construction optimization. In the underground engineering environment, signals are often mixed with noises such as mechanical vibrations and electromagnetic interferences. Therefore, identifying microseismic signals from complex signals is a fundamental task in microseismic monitoring.
[0003] Traditional signal classification methods mainly rely on manual empirical analysis, where experts subjectively judge waveform characteristics (such as amplitude, frequency, duration, etc.). However, manual analysis is inefficient, lacks consistency, and requires extremely high experience from operators, making it difficult to meet the requirements of real-time monitoring and monitoring accuracy. Summary of the Invention
[0004] The present application provides a method for identifying microseismic signals, a training method, a device, equipment, and a medium, which can solve one of the problems existing in the background art.
[0005] To achieve the above object, the present application adopts the following technical solutions:
[0006] In a first aspect, a training method for a microseismic signal recognition model is provided. The training method includes:
[0007] Obtaining training data, where the training data includes: seismic signal data and microseismic set signal data. The seismic signal data includes: strong earthquake signal data and weak earthquake signal data; the microseismic set signal data includes: microseismic signal data and interference signal data; and
[0008] Using the training data to train a microseismic signal recognition model, where the microseismic signal recognition model is used to identify the microseismic signal data through transfer learning.
[0009] Based on the above technical solution, by mainly establishing an initial model for microseismic signal recognition and using seismic signal data, microseismic signal data, and interference signal data to perform transfer learning training on the model, the classification and recognition of microseismic signals are carried out. In this way, the identification of microseismic signals does not rely on manual empirical analysis, and using an intelligent model can ensure the accuracy, timeliness, and consistency of the identification, meeting the requirements of real-time monitoring and accuracy.
[0010] In a possible design of the first aspect, the microseismic signal recognition model is a one-dimensional convolutional neural network, and the one-dimensional convolutional neural network includes: an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer arranged in sequence from the input direction to the output direction. The training process includes:
[0011] Using the seismic signal data to perform the first-stage training on the initialized one-dimensional convolutional neural network; and
[0012] Freezing the weights of some convolutional layers and pooling layers of the one-dimensional convolutional neural network obtained through the first-stage training, and using the microseismic set signal data to train the frozen one-dimensional convolutional neural network to adjust the weights of the remaining convolutional layers, pooling layers, and fully connected layers in the frozen one-dimensional convolutional neural network.
[0013] In a possible design of the first aspect, the loss functions used in the first-stage training and the second-stage training are binary cross-entropy.
[0014] In a possible design of the first aspect, the convolutional layer uses 4 layers of convolution with filter numbers of 16, 32, 64, and 128 respectively, and the convolution kernel size is 3; the pooling layer uses max pooling with a window size of 2; the fully connected layer uses 64 neurons, and the activation function is ReLU; the activation function of the output layer is Sigmoid.
[0015] In a possible design of the first aspect, the training data is obtained through the following processing:
[0016] Performing normalization processing on the seismic original signal, microseismic original signal, and interference original signal; and
[0017] Cropping or supplementing the signals obtained through the normalization processing to align the seismic signal data, microseismic signal data, and interference signal data.
[0018] In the second aspect, a method for recognizing microseismic signals is provided. The recognition method includes:
[0019] Obtaining the signal data to be recognized; and
[0020] Using the trained microseismic signal recognition model as described above to process the signal data to be recognized to obtain a recognition result.
[0021] In the third aspect, a training device for a microseismic signal recognition model is provided. The training device includes:
[0022] A first acquisition unit, configured to acquire training data, where the training data includes: seismic set signal data and microseismic set signal data, the seismic set signal data includes: strong earthquake signal data and weak earthquake signal data; the microseismic set signal data includes: microseismic signal data and interference signal data; and
[0023] A training unit, configured to use the training data to train a microseismic signal recognition model, where the microseismic signal recognition model is used to recognize the microseismic signal data through transfer learning.
[0024] In a fourth aspect, a microseismic signal recognition device is provided, and the recognition device includes:
[0025] A second acquisition unit, configured to acquire signal data to be recognized; and
[0026] A recognition unit, configured to use the trained microseismic signal recognition model as described above to process the signal data to be recognized, and obtain a recognition result.
[0027] In a fifth aspect, an electronic device is provided, and the electronic device includes: a processor, and a memory coupled to the processor, where the memory is configured to store a computer program; the processor is configured to execute the computer program stored in the memory, so that the electronic device executes the training method according to any possible implementation manner in the first aspect, or executes the recognition method according to the second aspect.
[0028] In a sixth aspect, a computer-readable storage medium is provided, including a computer program or instruction, when the computer program or instruction runs on a computer, the computer is caused to execute the training method according to any possible implementation manner in the first aspect, or execute the recognition method according to the second aspect.
[0029] In a seventh aspect, a computer program product is provided, including: a computer program or instruction, when the computer program or instruction runs on a computer, the computer is caused to execute the training method according to any possible implementation manner in the first aspect, or execute the recognition method according to the second aspect. Description of the Drawings
[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings in the following description are only some embodiments of the embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.
[0031] Figure 1 is a schematic flowchart of the method provided by the embodiment of the present application;
[0032] Figure 2 It is a schematic diagram of the model principle of the method provided by the embodiments of the present application;
[0033] Figure 3 It is a schematic diagram of the performance indicators of the method provided by the embodiments of the present application and the comparative example;
[0034] Figure 4 It is a schematic diagram of the system for implementing the microseismic signal recognition method of the one-dimensional convolutional neural network based on transfer learning provided by the embodiments of the present application. Detailed implementation manners
[0035] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. 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.
[0036] It should be noted that although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order from the module division in the device or the flowchart. The terms "first", "second", etc. in the specification, claims and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence.
[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.
[0038] The embodiments of the present application propose a one-dimensional 1D-convolutional neural network CNN model based on transfer learning, which solves the problem of efficient classification of microseismic signals in small-sample scenarios by combining seismic signal pre-training and microseismic data fine-tuning, and at the same time improves the robustness of the model in a low signal-to-noise ratio environment.
[0039] One of the objectives of the embodiments of the present application is to provide a microseismic signal recognition method based on a one-dimensional convolutional neural network based on transfer learning that is efficient, robust and has good performance in small samples.
[0040] The second objective of the embodiments of the present application is to provide a system for a microseismic signal recognition method based on a one-dimensional convolutional neural network based on transfer learning.
[0041] The microseismic signal recognition method based on a one-dimensional convolutional neural network based on transfer learning provided by the embodiments of the present application includes the following steps:
[0042] S1: Obtain seismic signal data from relevant public databases as pre-training data and construct a source domain dataset.
[0043] S2: Collect microseismic signals and blasting signals at the mine site, and after preprocessing, form a small-sample target dataset to construct a target domain dataset.
[0044] S3: Normalize the signals in the source domain and target domain datasets obtained in steps S1 and S2 to enhance the generalization ability of the model.
[0045] S4: For the source domain and target domain datasets obtained in step S3 after preliminary processing, supplement signals through window sliding to make them of the same length.
[0046] S5: Use the source domain dataset obtained in step S4 to train a one-dimensional convolutional neural network to classify strong earthquake signals and weak earthquake signals, and save the weight file.
[0047] S6: Fine-tune the weights of the last three layers of the convolutional neural network in the weight file with the target domain dataset obtained in step S4 to obtain a microseismic signal recognition model.
[0048] S7: Use the microseismic signal recognition model obtained in step S6 to complete the microseismic signal recognition of the one-dimensional convolutional neural network based on transfer learning.
[0049] The collected seismic data described in step S1 is divided into strong earthquake signals with a magnitude of 7 and weak earthquake signals with a magnitude of 1, and is downloaded from the public website of the United States Geological Survey.
[0050] The normalization process of the signal data collected in S1 and S2 described in step S3 includes the following steps:
[0051] Determine that the signal forms of the source domain and target domain are X i ={x0,x1,…,x j ,…,x n}, and perform normalization processing through the following calculation formula:
[0052]
[0053] Among them, x min is the minimum value of the signal, x max is the maximum value of the signal, and the processed signal form is X i ' = {x0',x1',…,x j ',…,x n '}.
[0054] The process of supplementing or deleting the signal data processed in S3 described in step S4 to make it of the same length includes the following steps:
[0055] Judge the signal length. If the length is greater than 13,200 sampling points, use the window sliding method. Pass through Calculate the window variance, where μ is the mean of the intercepted window. Sequentially select the signal segment with the smallest variance for truncation until the signal length is fixed at 13,200 sampling points; if the length is less than 13,200 sampling points, select the signal segment with the smallest variance through the same method and supplement the selected signal segment with the smallest variance at the end of the signal.
[0056] The steps described in step S5 use the source domain dataset obtained in step S4 to train a one-dimensional convolutional neural network to classify strong earthquake signals and weak earthquake signals, and save the weight file, which specifically includes the following steps:
[0057] Construct a one-dimensional convolutional neural network (1D-CNN) model. This model includes four perception layers, namely: input layer: receive one-dimensional time series signals; convolutional layer: 4 convolutional layers, the number of filters are 16, 32, 64, 128 respectively, and the convolutional kernel size is 3; pooling layer: use max pooling, and the window size is 2; fully connected layer: 64 neurons, and the activation function is ReLU. Its mathematical expression is as follows:
[0058]
[0059] where x is the signal input; f(x) is the corresponding output, and the same below.
[0060] Finally, it reaches the output layer, and its activation function is Sigmoid. This activation function is often used in binary classification tasks (microseismic events / blasting events), and its mathematical expression is as follows:
[0061]
[0062] During the training process, the source domain database is randomly divided into a training set and a validation set according to 8:2. The loss function uses binary cross-entropy, and its expression is:
[0063]
[0064] where, predicted value; y is the actual value.
[0065] Preferably, the number of times the source domain cyclic training feature 1D-CNN model is 30. At this time, the model loss has dropped to the minimum, and the weights of the model at this time are retained. The weight file is source_model_weights.weights.
[0066] The steps described in step S6 use the target domain dataset obtained in step S4 to train a one-dimensional convolutional neural network to classify microseismic signals and blasting signals, and fine-tune the weights of the last three layers of the convolutional neural network in the weight file to obtain a microseismic signal recognition model, which specifically includes the following steps:
[0067] The one-dimensional convolutional neural network after migration has the same structure as that described in step S5. Import the weight file obtained in step S5 and freeze the first four layers of the weight file, including three convolutional layers and one pooling layer. Fine-tune the weights of the last three layers, including one convolutional layer, one pooling layer, and one fully connected layer, using the target domain signal data.
[0068] Finally, an input signal data is obtained, and the output is the classification result of the input signal, which is a microseismic signal recognition model based on transfer learning of a one-dimensional convolutional neural network.
[0069] This method is tested on the test set together with existing schemes (Support Vector Machine (SVM), Naive Bayes, Logistics Regression). Several metrics such as accuracy, recall rate, and F1 score are used to comprehensively compare the performance. The comparison data is shown in Table 1 and Figure 3 as follows:
[0070] Table 1 Schematic Table of Performance Comparison
[0071]
[0072] Through Figure 3 Table 1, it can be seen that the accuracy of this method reaches 99.22%. Compared with the Support Vector Machine (SVM) among the other models, the accuracy is increased by 5.41%, the recall rate is increased by 16.11%, and the F1 value is increased by 11.05%. Therefore, this method has good accuracy and evaluation effects, leading comprehensively in various evaluation metrics.
[0073] As Figure 4The figure shows a schematic diagram of the functional modules of a system for an application example of the present application. This system for implementing the microseismic signal recognition method based on transfer learning using a one-dimensional convolutional neural network includes a data collection module, a data processing module, a training set construction module, a source domain training module, a weight transfer module, a target domain training module, and a signal classification module; the data collection module, the training set construction module, the source domain training module, the weight transfer module, the target domain training module, and the signal classification module are connected in series in sequence; the data collection module is used to obtain strong earthquake and weak earthquake data signal information, as well as on-site microseismic data and blasting data of the application project, and upload the data information to the data processing module; the data processing module is used to process the received signal data, normalize, supplement or delete the signal data according to the obtained signal data, and upload the processed signal data to the training set construction module; the training set construction module is used to divide the received signal data into a source domain data set and a target domain data set according to the signal data category, and upload the signal data to the source domain training module and the target domain training module respectively; the source domain training module uses the signal data of the received source domain data set to train a one-dimensional convolutional neural network and saves the weight file; the weight transfer module extracts the weight file of the source domain training module and uploads it to the target domain training module; the target domain training module uses the received weight file to construct a one-dimensional convolutional neural network, and fine-tunes some weights of the one-dimensional convolutional neural network using the signal data of the received target domain data set to obtain a microseismic signal recognition model based on transfer learning; the signal classification module uses the constructed microseismic signal recognition model based on transfer learning according to the input signal data of unknown type to obtain a classification result, and completes the microseismic signal recognition system based on transfer learning using a one-dimensional convolutional neural network.
[0074] An embodiment of the present application also provides a training device for a microseismic signal recognition model, and the training device includes:
[0075] A first acquisition unit, configured to acquire training data, where the training data includes: strong earthquake signal data and a microseismic set signal data, and the microseismic set signal data includes: microseismic signal data and interference signal data; and
[0076] A training unit, configured to use the training data to train a microseismic signal recognition model, and the microseismic signal recognition model is used for: performing a first classification on the strong earthquake signal data and the microseismic set signal data, and performing a second classification on the microseismic signal data and the interference signal data obtained from the first classification.
[0077] An embodiment of the present application also provides a recognition device for a microseismic signal, and the recognition device includes:
[0078] A second acquisition unit, configured to acquire signal data to be recognized; and
[0079] An identification unit for processing the signal data to be identified by using the microseismic signal identification model trained as above to obtain an identification result.
[0080] An embodiment of the present application further provides an electronic device, including: a processor, and a memory coupled to the processor, where the memory is used to store a computer program; the processor is used to execute the computer program stored in the memory so that the electronic device executes the method described in any one of the above embodiments.
[0081] The electronic device may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The electronic device may include, but is not limited to, a processor and a memory.
[0082] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the electronic device, and connects various parts of the entire device through various interfaces and lines.
[0083] The memory may be used to store the computer program. The processor realizes various functions of the electronic device by running or executing the computer program stored in the memory and calling the data stored in the memory.
[0084] The memory may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system, application programs required for at least one function, etc.; the data storage area may store data created according to the use of the mobile phone, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0085] An embodiment of the present application also provides a storage medium, which is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc.
[0086] An embodiment of the present application also provides a computer program product, including: a computer program or instruction. When the computer program or instruction runs on a computer, the computer is enabled to execute the method of any of the above possible implementation manners.
[0087] The above is the preferred implementation manner of the present application. It should be noted that for those of ordinary skill in the art of the present technology, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the present application.
Claims
1. A training method for a microseismic signal recognition model, characterized in that, The training method includes: Obtaining training data, where the training data includes seismic signal data and microseismic set signal data, the seismic signal data includes strong earthquake signal data and weak earthquake signal data; the microseismic set signal data includes microseismic signal data and interference signal data; and Using the training data to train a microseismic signal recognition model, which is used to recognize the microseismic signal data through transfer learning.
2. The training method according to claim 1, characterized in that, The microseismic signal recognition model is a one-dimensional convolutional neural network, and the one-dimensional convolutional neural network includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer arranged in sequence from the input to the output direction. The training process includes: Using the seismic signal data to perform the first-stage training on the initialized one-dimensional convolutional neural network; and Freezing the weights of some convolutional layers and pooling layers of the one-dimensional convolutional neural network obtained after the first-stage training, and using the microseismic set signal data to train the frozen one-dimensional convolutional neural network to adjust the weights of the remaining convolutional layers, pooling layers, and fully connected layers in the frozen one-dimensional convolutional neural network.
3. The training method according to claim 2, characterized in that, The loss functions used in the first-stage training and the second-stage training are binary cross-entropy.
4. The training method according to claim 2, wherein The convolutional layer uses 4 layers of convolution with filter numbers of 16, 32, 64, and 128 respectively, and the convolution kernel size is 3; the pooling layer uses max pooling with a window size of 2; the fully connected layer uses 64 neurons, and the activation function is ReLU; the activation function of the output layer is Sigmoid.
5. The training method according to claim 1, wherein The training data is obtained through the following processing: Performing normalization processing on the original seismic signal, original microseismic signal, and original interference signal; and Performing cropping and / or supplementation on the signals obtained after the normalization processing to align the seismic signal data, the microseismic signal data, and the interference signal data.
6. A method for identifying microseismic signals, characterized in that, The recognition method includes: Obtaining signal data to be recognized; and Using the trained microseismic signal recognition model as described in any one of claims 1-5 to process the signal data to be recognized to obtain a recognition result.
7. A training device for a microseismic signal recognition model, characterized in that, The training device includes: A first acquisition unit for obtaining training data, where the training data includes seismic signal data and microseismic set signal data, the seismic signal data includes strong earthquake signal data and weak earthquake signal data; the microseismic set signal data includes microseismic signal data and interference signal data; and A training unit for using the training data to train a microseismic signal recognition model, which is used to recognize the microseismic signal data through transfer learning.
8. An identification device for microseismic signals, characterized in that, The recognition device includes: A second acquisition unit for obtaining signal data to be recognized; and A recognition unit for using the trained microseismic signal recognition model as described in claims 1-5 to process the signal data to be recognized to obtain a recognition result.
9. An electronic device, characterized in that, The electronic device includes: a processor and a memory coupled to the processor, The memory is used to store a computer program; and The processor is configured to execute the computer program stored in the memory, so that the electronic device executes the training method according to any one of claims 1-5, or executes the recognition method according to claim 6.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a computer program or instructions, which, when running on a computer, cause the computer to execute the training method according to any one of claims 1-5, or execute the recognition method according to claim 6.