Microseismic signal identification method, training method, device, equipment and medium

Abstract

The invention provides a micro-seismic signal identification method, a training method, a device, equipment and a medium, a micro-seismic signal identification model is obtained through training by mainly utilizing training data such as signal waveform duration time, waveform interval time and a main frequency value, and micro-seismic signals and interference signals can be classified and identified by utilizing the model, so that the identification accuracy of the micro-seismic signals is improved. The accuracy and efficiency of classification identification can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of rock mass engineering, and particularly to a microseismic signal recognition method, a training method, a device, equipment and a medium. Background Technique

[0002] The monitoring and early warning of rock mass dynamic disasters in deep wells with high in-situ stress mines are difficult technical problems. For a specific monitoring and early warning problem of a rock mass structure project in a mine, after the microseismic / acoustic emission monitoring system is built, in the complex working environment of the mine, waveform recognition is of primary importance: the effective microseismic signals released by the seismic source reflecting rock fracture and damage can be identified and extracted without omission; the vibration signals generated by interference sources in the environment where the rock mass structure is located cannot be misidentified as effective signals, so as to obtain the most accurate data interpretation of the micro-fracture gestation, development state and trend of the rock mass structure.

[0003] Using the method of manual experience recognition to distinguish noise signals from effective microseismic signals requires technicians to have high technical qualities and spend a lot of time and energy, and the accuracy rate (the accuracy rate is not necessarily low) and recognition efficiency are very low. Summary of the Invention

[0004] The present application provides a microseismic signal recognition method, a training method, a device, equipment and a medium, which can solve one of the problems existing in the background technique.

[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, and the training method includes:

[0007] Obtaining training data, where the training data includes: signal waveform duration, waveform interval time and main frequency value; and

[0008] Using the training data to train a microseismic signal recognition model, where the microseismic signal recognition model is used to classify and recognize microseismic signals and interference signals.

[0009] Based on the above technical solutions, mainly using training data such as signal waveform duration, waveform interval time and main frequency value, a microseismic signal recognition model is trained, and this model can be used to classify and recognize microseismic signals and interference signals. In this way, the accuracy rate and efficiency of classification and recognition can be improved.

[0010] In a possible design of the first aspect, the microseismic signal recognition model includes: a first classifier for classifying and recognizing the microseismic signal and the interference signal, a second classifier for classifying and recognizing the large-energy signal and the small-energy signal from the microseismic signal classified by the first classifier, and a third classifier for classifying and recognizing the blasting signal, the electrical interference signal, and the rock drilling signal from the interference signal classified by the first classifier.

[0011] In a possible design of the first aspect, the first classifier, the second classifier, or the third classifier is a multi-layer feedforward fully connected neural network.

[0012] In a possible design of the first aspect, the training method further includes:

[0013] Using an eigenvalue calculation algorithm to determine the start point and the end point of the signal waveform; and

[0014] According to the start point and the end point of the signal waveform, calculating the duration of the signal waveform and the main frequency value, and determining the waveform interval time.

[0015] In a possible design of the first aspect, using an eigenvalue calculation algorithm to determine the start point and the end point of the signal waveform specifically includes:

[0016] Using a short window average value detection algorithm to determine whether there is an average value of the amplitudes within the window that satisfies a first preset condition.

[0017] If so, using the AIC algorithm to determine the start point of the signal waveform, and using the short window average value detection algorithm to determine the end point of the signal waveform.

[0018] Otherwise, using a single-point amplitude algorithm to determine the start point and the end point of the signal waveform by detecting whether the amplitude satisfies a second preset condition.

[0019] In the second aspect, a microseismic signal recognition method is provided. The recognition method includes:

[0020] Obtaining a signal to be recognized; and

[0021] Using the trained microseismic signal recognition model as described above to process the signal to be recognized to obtain a recognition result.

[0022] In the third aspect, a training device for a microseismic signal recognition model is provided. The training device includes:

[0023] A first acquisition unit for obtaining training data, where the training data includes: the duration of the signal waveform, the waveform interval time, and the main frequency value; and

[0024] A training unit for training a microseismic signal recognition model using the training data, where the microseismic signal recognition model is used to classify and recognize microseismic signals and interference signals.

[0025] In a fourth aspect, a microseismic signal recognition device is provided, and the recognition device includes:

[0026] A second acquisition unit for acquiring a signal to be recognized; and

[0027] A recognition unit for processing the signal to be recognized using the trained microseismic signal recognition model as described above to obtain a recognition result.

[0028] In a fifth aspect, an electronic device is provided, and the electronic device includes: a processor and a memory coupled to the processor. 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 training method according to any possible implementation manner in the first aspect, or executes the recognition method according to the second aspect.

[0029] 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 made 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.

[0030] 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 made 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. Description of the Drawings

[0031] 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 described below 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.

[0032] Figure 1 It is a schematic diagram of various noises and valid signals provided by the embodiments of the present application;

[0033] Figure 2 It is an automatic calculation flowchart of parameter values of typical features provided by the embodiments of the present application;

[0034] Figure 3 is the characteristic parameter value of a waveform to be recognized calculated by the method provided in the embodiments of the present application;

[0035] Figure 4 is the overall classification and recognition idea diagram provided in the embodiments of the present application;

[0036] Figure 5 is the specific structure diagram of the convolutional neural network provided in the embodiments of the present application;

[0037] Figure 6 is the schematic diagram for determining the input parameters of the convolutional neural network provided in the embodiments of the present application;

[0038] Figure 7 is the schematic diagram of the waveform sample for learning provided in the embodiments of the present application. Detailed implementation manners

[0039] 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, rather than to limit the present application.

[0040] 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. Terms such as "first" and "second" in the specification, claims and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence.

[0041] 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.

[0042] Before introducing the embodiments of the present application, a brief description of the current technical research of the present application will be given first:

[0043] The method of manual experience recognition is used to distinguish the noise signal and the effective signal waveform, which requires technicians to have high technical qualities and spend a lot of time and energy. Although the accuracy can be well guaranteed, the recognition efficiency is very low and the practicability is very poor.

[0044] In the prior art, there is no solution to automatically, effectively, and accurately calculate the characteristic parameters of the waveforms of the pattern classes to be recognized. Through theoretical analysis and experience accumulation, the waveform duration, the interval time between independent waveforms, and the main frequency of the waveform are important features for distinguishing each pattern class to be recognized. However, there is no relevant technology on which algorithm and algorithm process to use to automatically and accurately calculate these characteristic parameter values, which also results in poor accuracy of subsequent various automatic recognition algorithms.

[0045] Since, in actual application working conditions, the waveforms generated under the same pattern class also have great differences, a large number of learning samples should be included in the recognition and classification algorithm, and a scientific and efficient deep learning model should be constructed to improve the recognition accuracy.

[0046] Since the above points cannot be achieved in the current applications in mines, it often fails to timely and accurately recognize and analyze a large number of important precursor signals starting from weakness, resulting in irreparable losses in safety and severely restricting the application effect of microseismic / acoustic emission monitoring technology.

[0047] The embodiments of this application mainly include the following three parts:

[0048] First, based on a large amount of practice, in the mine application environment, the pattern classes to be recognized that mainly need to be automatically recognized and classified are sorted out, and the typical features of each pattern class to be recognized are extracted.

[0049] Second, to achieve automatic and accurate recognition and classification, it is necessary to automatically and accurately calculate the parameter values of the typical features of the waveforms of the pattern classes to be recognized. In this embodiment, a systematic calculation method and algorithm process are established.

[0050] Third, a scientific and efficient convolutional neural network is constructed, which belongs to a feedforward multi-layer fully connected neural network. The main content of this embodiment is how to scientifically and accurately determine key parameters such as input parameters, convolution kernel types, number of hidden layers, number of neurons in each layer, activation function types, etc., conduct a large amount of learning and testing, obtain a high recognition accuracy, and maintain stable technical performance.

[0051] 1. As Figure 1 shown, in the strong interference environment of mines, it is difficult to identify and classify various noises and effective signals through simple methods. The signal waveforms of each pattern class shown in the figure overlap and cross each other in terms of frequency, amplitude, etc., and it is impossible to filter and classify the noises through simple methods such as band-pass filtering.

[0052] 2. Extraction of main pattern classes to be recognized and their main features

[0053] In the application of more than a decade in multiple mines, technicians collected and sorted a large number of waveforms of various types of patterns to be recognized, and obtained the results shown in Table 1 below, including the typical recognition features of each pattern to be recognized and the distribution ranges of their characteristic values.

[0054] Table 1 Automatic Recognition Decision Function for Microseismic Waveform Features in Mines

[0055]

[0056] 3. Automatic Calculation Method for Main Features

[0057] When automatically calculating the waveforms of each pattern to be recognized, there are mainly four algorithms for calculating their characteristic values, namely the AIC algorithm, short window average, short-long window ratio, and single-point amplitude trigger. The specific descriptions are as follows:

[0058] 1) AIC algorithm

[0059] For the signal waveform to be recognized, its original data is a one-dimensional sequence with a length of N, and the amplitude of each sampling point is x i (i = 1, 2, ···, n)

[0060] For the AIC algorithm, its calculation formula is as follows:

[0061] AIC = i·log(var(x(1,i))) + (N - i - 1)log(var(x(i + 1,N)))

[0062] In the formula: N is the length of the signal x, that is, the total number of data points; i is a variable representing a certain moment point in the time series, which is used to divide the signal x into two parts. x(1:i) represents the subsequence from the starting point of the signal to the moment i, and x(i + 1:N) represents the subsequence from the moment i + 1 to the end point of the signal; log represents the logarithm with base 10; var represents variance. For example, var(x(1:i)) represents the variance of the subsequence of the signal x from the starting point to the moment i.

[0063] The definition of the variance function var is:

[0064]

[0065] In the formula: σ 2 represents the population variance; N is the length of the signal x, that is, the total number of data points; i is a variable representing a certain moment point in the time series, x i represents the data point at the i-th moment; represents the average value of all data points.

[0066] 2) Short-long window ratio

[0067] When a burst signal is detected, there are two moving time windows on the time axis, namely the short-term window (SAT) and the long-term window (LAT). The end times of these two windows are the same. According to experience, the time length of (SAT) is 10 ms, and the time length of (LAT) is 35 ms. The amplitude of each sampling point is x i (i = 1, 2, ···, n).

[0068] Definition:

[0069]

[0070] In the formula: λ represents the short-to-long window ratio; SAT represents the short window length; LAT represents the long window length; i is a variable representing a certain moment point in the time series, and x i represents the data point at the i-th moment.

[0071] The physical meaning of Equation (2) is the ratio of the average value of the absolute values of the amplitudes of all sampling points in the short-term window to the average value of the absolute values of the amplitudes of all sampling points in the long-term window.

[0072] 3) Short window average value

[0073] Definition:

[0074]

[0075] In the formula: ε represents the short window average value; SAT represents the short window length; i is a variable representing a certain moment point in the time series, and x i represents the data point at the i-th moment.

[0076] That is, the average value of the absolute values of the amplitudes of all sampling points in the short-term window.

[0077] 4) Single-point amplitude trigger

[0078] It is only applicable to electromagnetic interference signals with a very short duration and only a small number of or only one oscillation process in amplitude.

[0079] The automatic calculation process of the parameter values of the typical characteristics is specifically as Figure 2 shown.

[0080] It should be noted specifically that: Figure 2 The process of is gradually worked out by the inventors of this application through many deliberations and attempts. The core reason and the technical problem to be solved are: for the actual vibration signal waveforms generated by various vibration sources in the real mine environment, their seismic phases are different, the amplitude ranges vary greatly, and they overlap and interfere with each other, and cannot be generally processed. The designed process Figure 2It specifically grasps the mechanism and key features of each recognition mode, obtains the waveform duration of a single burst signal, and then obtains the interval time between single burst signals, masters the time characteristics of the signal, and the frequency characteristics of the main frequency. Extracting these features that can depict the overall outline of each signal and accurately calculating the parameter values are the key basis for the subsequent convolutional neural network recognition model.

[0081] Furthermore, Figure 2 The designed automatic calculation process can bring the following effects: The method selected for calculating the parameter values of typical features is appropriate, and its sequence is scientific. It can clearly and straightforwardly target the typical features of each pattern to be recognized, and capture its key features and parameter values.

[0082] 1) Waveform duration:

[0083] First, use a window with a length of 10 ms to move across the entire waveform and calculate whether there is a window position where the average value is greater than 3 mv. If so, the starting point of the waveform is calculated using the AIC method (select the signal starting point to the point with the largest energy change as the detection interval, and take the point with the minimum value of the AIC function within the detection interval as the starting point of the waveform). The end point of the waveform is calculated using the window mean method (when the window mean is less than 7 mv, take the end point of the current window as the end point of the waveform). Then, calculate the waveform duration by subtracting the waveform starting time from the waveform ending time.

[0084] First, use a window with a length of 10 ms to move across the entire waveform and calculate whether there is a window position where the average value is greater than 3 mv. If not, it belongs to the electromagnetic interference waveform. However, this type of electromagnetic interference waveform is bursty and the waveform duration is very short, and the window mean method cannot detect the waveform. Therefore, each waveform is first monitored using the window method. If the average value greater than 3 mv cannot be detected, it means it is an electromagnetic interference wave. Use the single-point amplitude method to obtain its starting point and end point. When the amplitude at a certain moment is first detected to be greater than 15 mv for the first time, record this position as the starting point, and then the position where the amplitude is first detected to be less than 15 mv further back is recorded as the end point.

[0085] It should be noted that the values listed in the text, such as the window with a length of 10 ms, 3 mv, 7 mv, and 15 mv, etc., are preferred solutions. These values are comprehensively determined based on the sensors used, the amplitude of the signal white noise floor noise, the application environment noise, and the amplitude level of the effective signal, etc. They are parameter values that have been adjusted and tried many times and finally have better and more stable effects. 2) Waveform main frequency:

[0086] Use the fast Fourier transform method to calculate the waveform main frequency from the starting point to the end point of the waveform.

[0087] Figure 3It is the characteristic parameter value of a waveform to be recognized calculated by the above method.

[0088] 4. Design of Convolutional Neural Network with Deep Learning Intelligent Classification Function

[0089] Classification process: First, a waveform is classified into two categories of valid signals and interference signals through an effective signal classifier, and then the subsequent classifier is used to further distinguish the category to which the waveform belongs, as Figure 4 shown.

[0090] Among them, the output layer of classifier 3 has 3 nodes, corresponding to 3 types of waveforms: blasting, electromagnetic interference, and rock drilling. The output layers of classifier 1 and classifier 2 still have 2 nodes.

[0091] The convolutional neural network in each classifier consists of 2 convolutional layers + 3 fully connected layers, as Figure 5 shown.

[0092] The main design parameters of the convolutional neural network for automatic classification of microseismic waveforms are as Figure 6 shown.

[0093] Training data includes signal waveforms of various types to be recognized. The optimization objective function adopted by the neural network is the cross-entropy loss function, and the definition of this function is as follows:

[0094] The cross-entropy loss function is a loss function used to calculate the prediction accuracy of the model in classification tasks. It measures the gap between the prediction result of the model and the true result, and is usually used together with the softmax function to calculate the loss of multi-classification problems. The smaller the value of the cross-entropy, the better the prediction effect of the model.

[0095] Cross-entropy: Cross-entropy is a measure of the difference between two probability distributions. Let P(x) be the true distribution and Q(x) be the predicted distribution, then the cross-entropy H(P, Q) between P and Q is defined as H(P, Q) = -∑ x P(x) log Q(x). In classification problems, P(x) represents the true class distribution of the samples, and Q(x) represents the class distribution predicted by the model. The cross-entropy loss function is based on the concept of cross-entropy and is used to measure the difference between the model prediction result and the true result.

[0096] For a classification problem with n samples, assume the true label of each sample is y i (usually represented by one-hot encoding, that is, if the sample belongs to the kth class, then y i,k = 1, and the rest of y i,j = 0, j ≠ k), and the probability distribution predicted by the model is Then the formula for the cross-entropy loss function L is:

[0097]

[0098] where m is the number of categories.

[0099] The cross-entropy loss function has the following characteristics:

[0100] Monotonicity: The value of the cross-entropy loss function increases as the difference between the model's prediction result and the true result increases, and decreases as the difference decreases. It has good monotonicity, which enables it to well reflect the performance of the model.

[0101] Non-negativity: Due to the properties of the logarithmic function, the value of the cross-entropy loss function is always non-negative. When the model's prediction result is exactly the same as the true result, the value of the cross-entropy loss function is 0, which is the ideal minimum loss situation.

[0102] Differentiability: The cross-entropy loss function is differentiable, which allows it to use gradient-based optimization algorithms (such as stochastic gradient descent) to adjust the model's parameters to minimize the loss function.

[0103] The convolutional neural network for automatic microseismic waveform classification includes:

[0104] 1) Multiple convolutions, using a one-dimensional convolutional kernel Conv1D;

[0105] 2) Using a multi-layer feedforward fully connected neural network;

[0106] 3) The activation function of the neurons uses the ReLU (Rectified Linear Unit) function;

[0107] 4) There are a total of 3 classifiers for two-stage classification;

[0108] 5) Classifier I: Input layer: 33750 neurons, Hidden layer 1: 125 neurons, Hidden layer 2: 25 neurons, Hidden layer 3: 16 neurons, Output layer: 2 neurons, namely effective signal and noise signal;

[0109] 6) Classifier II: Input layer same as above, Output layer: 2 neurons, namely effective large-energy signal and effective small-energy signal;

[0110] 7) Classifier III: Input layer same as above, Output layer: 3 neurons, namely rock drilling signal, blasting signal and electrical interference signal.

[0111] The core design of the above convolutional neural network is: Selecting accurate and appropriate recognition features, namely waveform duration, waveform interval time and main frequency value, and inputting them as input parameters into the convolutional neural network, which can significantly improve the recognition accuracy of the algorithm.

[0112] 5. Learning of Convolutional Neural Network

[0113] Select 137 pieces of data from five categories of blasting, electromagnetic interference, rock drilling, effective small and effective large, with good waveforms, and randomly divide them into training data and test data according to a ratio of 7:3. Use the original waveform data, waveform duration, and waveform main frequency as feature inputs into the model.

[0114] Each time, train the model for 5000 rounds on the training data, and then use the test data to test the classification accuracy of the model.

[0115] Table 2 Waveform Data Statistics

[0116] Waveform type Number of training sets Number of test sets Total number Blasting 21 9 30 Electromagnetic interference 21 9 30 Rock drilling 11 6 17 Effective small 21 9 30 Effective large 21 9 30 Total 95 42 137

[0117] Typical waveform samples for learning are as Figure 7 shown.

[0118] Calculation methods for learning and test metrics:

[0119] Precision: For a certain category, among the samples predicted as this category, the proportion that is actually this category.

[0120]

[0121] Recall: Among the samples that are actually of a certain category, the proportion that is correctly predicted.

[0122]

[0123] F1 value: The harmonic mean of precision and recall.

[0124]

[0125] Accuracy: The proportion of the number of samples with correct classification to the total number of samples.

[0126]

[0127] (6) Recognition Accuracy

[0128] Learning and testing: Select 1370 pieces of data from five categories of blasting, electromagnetic interference, rock drilling, effective small energy signals, and effective large energy signals, and randomly divide them into training data and test data according to a ratio of 7:3. Use the original waveform data, waveform duration, and waveform main frequency as features to input into the model for 5000 rounds of learning and 10 tests.

[0129] Waveform classification accuracy: The recognition accuracy of the automatic classification software reached 87.14%

[0130] The output results in Table 3 below. The first row is the number of waveforms of interference signals and valid signals in the test data. The second row is the number of correct classifications of the corresponding classes by the model. The third row is the actual number of classifications of each class by the model.

[0131] Table 3: 10 test results of Classifier 1

[0132]

[0133]

[0134] The output results in Table 4 below. The first row is the number of waveforms of large valid and small valid in the test data. The second row is the number of correct classifications of the corresponding classes by the model. The third row is the actual number of classifications of each class by the model.

[0135] Table 4: 10 test results of Classifier 2

[0136]

[0137] The output results in Table 5 below. The first row is the number of waveforms of blasting, electromagnetic interference, and rock drilling in the test data. The second row is the number of correct classifications of the corresponding classes by the model. The third row is the actual number of classifications of each class by the model.

[0138] Table 5: 10 test results of Classifier 3

[0139]

[0140] The beneficial effects of this embodiment are as follows:

[0141] (1) The recognition accuracy of the automatic classification software has reached 87%.

[0142] (2) It achieves the technical purpose of accurately extracting valid signals from a large number of interference noise signals automatically or with as little human intervention as possible. It can capture the precursor information of the gestation and evolution of disasters in real time and accurately, realizing the full automation of monitoring and early warning. It liberates the energy of technical personnel from a large number of complex waveform manual identifications to a great extent, and instead devotes more energy to data verification and data analysis, improving the effectiveness and value of microseismic monitoring technology, and playing an important role in overcoming the key technical problems of monitoring and prevention of rock mass dynamic disasters faced in the development of deep mineral resources in the future.

[0143] This embodiment also provides a training device for a microseismic signal recognition model. The training device includes:

[0144] A first acquisition unit for obtaining training data, where the training data includes: signal waveform duration, waveform interval time, and main frequency value; and

[0145] A training unit for training a microseismic signal recognition model using the training data, where the microseismic signal recognition model is used to classify and identify microseismic signals and interference signals.

[0146] This embodiment also provides a microseismic signal recognition device, which includes:

[0147] A second acquisition unit for acquiring a signal to be recognized; and

[0148] A recognition unit for processing the signal to be recognized using the trained microseismic signal recognition model as described above to obtain a recognition result.

[0149] This application embodiment also 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.

[0150] 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.

[0151] 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, connecting various parts of the entire device using various interfaces and lines.

[0152] 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.

[0153] 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.

[0154] The embodiment of the present application also provides a storage medium, which is a computer-readable storage medium. 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 method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code may 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, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disc, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0155] The 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 one of the above possible implementation manners.

[0156] The above is the preferred embodiment of the present application. It should be noted that for those of ordinary skill in the art, 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: signal waveform duration, waveform interval time, and main frequency value; and Using the training data to train a microseismic signal recognition model, where the microseismic signal recognition model is used to classify and recognize microseismic signals and interference signals.

2. The training method according to claim 1, wherein The microseismic signal recognition model includes: a first classifier for classifying and recognizing microseismic signals and interference signals, a second classifier for classifying large-energy signals and small-energy signals from the microseismic signals classified by the first classifier, and a third classifier for classifying blast signals, electrical interference signals, and rock drilling signals from the interference signals classified by the first classifier.

3. The training method according to claim 2, wherein The first classifier, the second classifier, or the third classifier is a multi-layer feedforward fully connected neural network.

4. The training method according to claim 1, wherein The training method further includes: Using an eigenvalue calculation algorithm to determine the start and end points of the signal waveform; and Calculating the signal waveform duration and the main frequency value according to the start and end points of the signal waveform, and determining the waveform interval time.

5. The training method according to claim 1, wherein Using an eigenvalue calculation algorithm to determine the start and end points of the signal waveform specifically includes: Using a short window average value detection algorithm to determine whether the average value of the amplitudes within the window satisfies a first preset condition. If so, using the AIC algorithm to determine the start point of the signal waveform, and using the short window average value detection algorithm to determine the end point of the signal waveform. Otherwise, using a single-point amplitude algorithm to determine the start and end points of the signal waveform by detecting whether the amplitude satisfies a second preset condition.

6. A method for identifying microseismic signals, characterized in that, The recognition method includes: Obtaining a signal to be recognized; and Using the trained microseismic signal recognition model according to any one of claims 1-5 to process the signal to be recognized and 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: signal waveform duration, waveform interval time, and main frequency value; and A training unit for using the training data to train a microseismic signal recognition model, where the microseismic signal recognition model is used to classify and recognize microseismic signals and interference signals.

8. A microseismic signal recognition device, characterized in that, The recognition device includes: A second acquisition unit for obtaining a signal to be recognized; and A recognition unit for using the trained microseismic signal recognition model according to any one of claims 1-5 to process the signal to be recognized and 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 used 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 instruction, and when the computer program or instruction runs on a computer, it causes the computer to execute the training method according to any one of claims 1-5, or execute the recognition method according to claim 6.

Images