Microseismic event diagnosis and early warning method and device based on zero sample learning

Abstract

The invention provides a microseismic event diagnosis and early warning method and device based on zero sample learning, and the method comprises the steps: obtaining a first signal collected at a target monitoring point; performing feature extraction on the first signal to obtain a first feature vector corresponding to the first signal; determining a target event type corresponding to the first signal based on the first feature vector, a reference feature vector and a reference event type corresponding to the reference feature vector; according to the first signal, the first feature vector and the target event type, predicting an occurrence probability and a danger level of a micro-seismic event corresponding to the first signal; and when the occurrence probability is greater than a probability threshold value and / or the insurance grade is greater than a grade threshold value, sending early warning information. Therefore, the classification of the microseismic signals can be realized without a large amount of labeled data, the robustness of microseismic event diagnosis in a cross-domain scene is improved, the improvement of the efficiency and reliability of disaster early warning is facilitated, and the life and property safety is further guaranteed.

Description

Technical Field

[0001] This application relates to the fields of geological disaster monitoring and early warning technology and machine learning, and particularly relates to a microseismic event diagnosis and early warning method and device based on zero-shot learning. Background Art

[0002] A microseismic event refers to a phenomenon of relatively weak seismic waves generated by rock fractures caused by changes in the stress field within a rock mass. Such seismic waves are also known as weak seismic signals. By collecting, monitoring, and analyzing microseismic events, it is possible to predict and evaluate the stability and safety of the rock mass during production activities, warn of potential disaster risks, and play an important role in fields such as safe mining of mines, tunnel excavation, earthquake early warning, and geological disaster monitoring. Summary of the Invention

[0003] This application aims to at least partly solve one of the technical problems in the related art.

[0004] To this end, the first objective of this application is to propose a microseismic event diagnosis and early warning method based on zero-shot learning, so as to determine the type of microseismic event corresponding to the real-time monitored microseismic signal through feature mapping, and then predict the occurrence probability and level of the disaster and give an early warning.

[0005] The second objective of this application is to propose a microseismic event diagnosis and early warning device based on zero-shot learning.

[0006] The third objective of this application is to propose an electronic device.

[0007] The fourth objective of this application is to propose a computer-readable storage medium.

[0008] The fifth objective of this application is to propose a computer program product.

[0009] To achieve the above objectives, the first aspect embodiment of this application proposes a microseismic event diagnosis and early warning method based on zero-shot learning, including:

[0010] Obtain a first signal collected at a target monitoring point;

[0011] Extract features from the first signal to obtain a first feature vector corresponding to the first signal;

[0012] Based on the first feature vector, a reference feature vector, and the reference event type corresponding to the reference feature vector, determine the target event type corresponding to the first signal;

[0013] Predict the occurrence probability and risk level of the microseismic event corresponding to the first signal according to the first signal, the first feature vector, and the target event type;

[0014] Send a warning message when the occurrence probability is greater than the probability threshold and / or the risk level is greater than the level threshold.

[0015] To achieve the above object, an embodiment of the second aspect of the present application provides a microseismic event diagnosis and warning device based on zero-shot learning, including:

[0016] An acquisition module, configured to acquire a first signal collected at a target monitoring point;

[0017] A feature extraction module, configured to extract features from the first signal to obtain a first feature vector corresponding to the first signal;

[0018] A classification module, configured to determine a target event type corresponding to the first signal based on the first feature vector, a reference feature vector, and a reference event type corresponding to the reference feature vector;

[0019] A prediction module, configured to predict the occurrence probability and risk level of the microseismic event corresponding to the first signal according to the first signal, the first feature vector, and the target event type;

[0020] A warning module, configured to send a warning message when the occurrence probability is greater than the probability threshold and / or the risk level is greater than the level threshold.

[0021] To achieve the above object, an embodiment of the third aspect of the present application provides an electronic device, including: a processor and a memory communicatively connected to the processor;

[0022] The memory stores computer execution instructions;

[0023] The processor executes the computer execution instructions stored in the memory to implement the microseismic event diagnosis and warning method based on zero-shot learning provided in the embodiment of the first aspect.

[0024] To achieve the above object, an embodiment of the fourth aspect of the present application provides a computer-readable storage medium, in which computer execution instructions are stored, and when the computer execution instructions are executed by a processor, they are used to implement the microseismic event diagnosis and warning method based on zero-shot learning provided in the embodiment of the first aspect.

[0025] To achieve the above object, an embodiment of the fifth aspect of the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the microseismic event diagnosis and warning method based on zero-shot learning provided in the embodiment of the first aspect.

[0026] The microseismic event diagnosis and early warning method and device based on zero-shot learning provided by this application collect microseismic signals by setting target monitoring points, classify the microseismic signals based on the mapping and similarity between the feature vectors extracted from the microseismic signals and the reference feature vectors, and then predict the disaster occurrence probability and danger level corresponding to the microseismic event that generates the microseismic signal according to the classification result and the microseismic signal, and further determine whether early warning is needed. Thus, without a large amount of labeled data, the classification of microseismic signals can be achieved, the robustness of microseismic event diagnosis in cross-domain scenarios is improved, which is beneficial to improving the efficiency and reliability of disaster early warning, and further ensuring life and property safety.

[0027] Additional aspects and advantages of this application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of this application. Description of the Drawings

[0028] The above-mentioned and / or additional aspects and advantages of this application will become apparent and easy to understand from the following description of the embodiments in conjunction with the drawings, where:

[0029] Figure 1 is a schematic flow chart of a microseismic event diagnosis and early warning method based on zero-shot learning provided by an embodiment of this application;

[0030] Figure 2 is a schematic flow chart of a microseismic event diagnosis and early warning method based on zero-shot learning provided by another embodiment of this application;

[0031] Figure 3 is a schematic flow chart of a microseismic event diagnosis and early warning method based on zero-shot learning provided by another embodiment of this application;

[0032] Figure 4 is a schematic structural diagram of a microseismic event diagnosis and early warning device based on zero-shot learning provided by an embodiment of this application. Detailed Embodiments

[0033] The embodiments of this application will be described in detail below. Examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are intended to explain this application and should not be construed as a limitation of this application.

[0034] The microseismic event diagnosis and early warning method and device based on zero-shot learning according to the embodiments of this application will be described below with reference to the drawings.

[0035] In related technologies, microseismic disaster early warning usually relies on a large number of labeled samples for supervised learning. However, in practical applications, it is often difficult to obtain a sufficient amount of high-quality labeled data. Moreover, due to the complex and variable geological environment, the characteristics of microseismic signals are diverse. The early warning methods trained based on labeled data have poor applicability in different environments and are difficult to accurately and effectively identify newly emerging abnormal signal types, resulting in the inability to predict disaster events in a timely manner and low safety.

[0036] To address the above problems, the present application proposes a microseismic event diagnosis and early warning method based on zero-shot learning. A feature mapping function is constructed to map the characteristics of the collected microseismic signals into the feature vector space of known categories, so that the classification of microseismic signals can be achieved without a large amount of labeled data, improving the cross-domain adaptability and reliability of microseismic event diagnosis in different environments and ensuring the timeliness of disaster early warning.

[0037] Figure 1 It is a schematic flowchart of a microseismic event diagnosis and early warning method based on zero-shot learning provided by an embodiment of the present application.

[0038] As Figure 1 shown, the microseismic event diagnosis and early warning method based on zero-shot learning includes the following steps:

[0039] Step 101, obtain a first signal collected at a target monitoring point.

[0040] Among them, the target monitoring point is at least one microseismic monitoring point reasonably selected and deployed in wells and on the ground in the area to be detected, such as mining areas, oil and gas fields, according to geological structures, historical seismic activity conditions, and monitoring requirements, etc.

[0041] Among them, the first signal refers to the seismic wave signal generated by the tiny vibrations caused by rock fractures or fluid disturbances during the production construction or the change of the natural stress field, and can also be called a microseismic signal. The intensity of the first signal is weak and can contain information such as arrival time, amplitude, and phase, and can reflect the processes of internal deformation and failure of rock masses. The characteristics of the first signal can change over time.

[0042] In an embodiment of the present application, a data acquisition device can be configured at each target monitoring point, and the data acquisition device can collect the seismic waves at the target monitoring point in real time to obtain the first signal.

[0043] It should be noted that the data acquisition device may include a seismometer and a data collector. Among them, the seismometer is used to capture seismic wave signals in real time and convert seismic wave information into electrical signals. The selection of the seismometer should meet the requirements of low noise, high sensitivity, high precision, and a frequency band width covering the main frequency range of microseismic signals, such as from 200 Hertz (Hz) to 1500 Hz, so that the data acquisition device can capture weak seismic wave signals. The data collector is used to amplify, perform analog-to-digital conversion, and filter the weak seismic signals output by the seismometer, so as to store and analyze the microseismic signals. The data collector can set appropriate sampling rates and dynamic ranges according to the signal frequency range to be captured and the signal variation amplitude to be processed, to ensure the accuracy and reliability of the signals.

[0044] In the embodiments of the present application, after the data acquisition device has collected the seismic wave signals of each monitoring point in real time, it can transmit the collected first signal to the data center for storage and processing.

[0045] Optionally, at least one target monitoring point in the area to be monitored can be determined according to the first parameter of the area to be monitored.

[0046] Among them, the first parameter may include at least one of the geological structure of the area to be monitored, the intensity, frequency, and distribution of historical seismic activities, and the requirements such as monitoring objectives, scope, and accuracy.

[0047] In the embodiments of the present application, the area to be monitored refers to an area where microseismic event diagnosis and early warning are required, such as mining areas, oil and gas fields, etc. Since in different areas to be monitored, the geological conditions, production activities, etc. that cause microseismic signals are different, in order to more effectively capture seismic signals, improve the accuracy and timeliness of early warning, and meet the requirements of different monitoring objectives, it is necessary to specifically analyze the positions where microseismic signals are likely to occur according to the first parameter corresponding to each area to be monitored, and reasonably select monitoring points.

[0048] It should be noted that the geological structure can reflect the law of seismic activities. Therefore, according to the geological structure, the positions in the area to be monitored that may generate vibrations or affect the occurrence and propagation of earthquakes can be determined and set as target monitoring points. For example, monitoring points can be set near active faults to capture the tiny vibrations caused by fault activities.

[0049] It should be noted that since the positions with high seismic activity intensity and frequency have relatively high seismic risks, more monitoring points need to be set in areas with high intensity and frequency of historical seismic activities to strengthen monitoring. And, target monitoring points can be set according to the distribution of historical seismic activities.

[0050] It should be noted that in the actual production environment, the monitoring requirements can be set according to experience and work needs, etc. The monitoring requirements can include monitoring targets, scopes, or precisions, etc. The monitoring target can include one or more specified monitoring points. The monitoring scope determines the number of monitoring points that need to be set. The larger the monitoring scope, the more monitoring points need to be set. The monitoring precision determines the number of monitoring points that need to be set and the density of the distribution of the monitoring points. The higher the precision, the more monitoring points are selected and the distribution is more dense. Therefore, it can be known that the first parameter for selecting the target monitoring points can include at least one of the geological structure of the area to be monitored, the intensity, frequency, and distribution of historical seismic activities, and the requirements such as monitoring targets, scopes, and precisions.

[0051] Step 102, extract features from the first signal to obtain a first feature vector corresponding to the first signal.

[0052] In the embodiments of the present disclosure, the time domain features of the first signal can be extracted, and the features of each dimension extracted can be combined into a vector in a certain order and format, so as to obtain a first feature vector corresponding to the first signal. The first feature vector can be an array or matrix of a fixed length, and each element in the feature vector corresponds to a feature value.

[0053] It should be noted that the features extracted from the first signal can include at least one of statistical features, time features, and waveform features. Specifically, the mean, standard deviation, skewness, kurtosis, time difference between P waves and S waves, etc. of the signal can be calculated to obtain statistical features. The rise time, fall time, duration, etc. of the signal can also be extracted to obtain time features describing microseismic events. And, the maximum peak value, minimum valley value, number of zero crossings, etc. can also be extracted to obtain waveform features describing microseisms.

[0054] It should be noted that in some possible embodiments, the first signal may be continuous time series data. At this time, the first signal may include multiple independent microseismic events, and each microseismic event has different durations and features. Then, directly extracting features from the first signal may result in inaccurate feature extraction due to the change of features such as amplitude and frequency over time.

[0055] Therefore, in the present application, the start and end of microseismic events can be identified based on the change of features such as the amplitude, frequency, or other features of the signal, the duration of each microseismic event can be determined, and then the first signal can be segmented according to the start and end of each microseismic event, and then feature extraction is performed on each segment of the signal respectively to obtain the first feature vector.

[0056] It should be noted that in the present application, wavelet transform and Hilbert transform can also be used to enhance signal features and improve the accuracy of subsequent processing.

[0057] It should be noted that in this application, before extracting features from the first signal, the first signal can also be preprocessed to improve the accuracy of feature extraction. Specifically, a high-pass filter and a low-pass filter can be used to remove background noise to ensure the purity of the first signal, and wavelet denoising technology can be used to further eliminate high-frequency noise. The detrending technique can also be applied to eliminate the long-term drift phenomenon in the signal to make the first signal more stable.

[0058] Step 103: Determine the target event type corresponding to the first signal based on the first feature vector, the reference feature vector, and the reference event type corresponding to the reference feature vector.

[0059] The reference feature vector refers to the main features extracted from the microseismic signals collected historically after clustering, which reflects the attributes of the event types corresponding to this type of microseismic signals. The reference feature vector can include information such as focal depth, focal mechanism, and waveform features.

[0060] In the embodiments of this application, the first feature vector can be input into a pre-constructed zero-shot learning model to obtain the result of classifying the first feature vector by the zero-shot learning model, that is, the target event type corresponding to the first signal.

[0061] It should be noted that the zero-shot learning model is based on attribute embedding and can use the data of known classes and their attributes to understand and infer unknown classes. The zero-shot learning model can include a Convolutional Neural Networks (CNN) and a Long Short-Term Memory (LSTM) to learn the time series features of microseismic signals, and combine an attribute knowledge base and unsupervised learning for zero-shot classification to obtain and store at least one reference feature vector and the reference event type corresponding to each reference feature vector. The zero-shot learning model can also include a feature mapping function, which is trained by the time series features of microseismic signals and the corresponding reference feature vectors after classification.

[0062] Therefore, in the embodiments of this application, the zero-shot learning model can use the trained feature mapping function to first map the first feature vector, calculate the cosine similarity between the mapped feature vector and the stored multiple reference feature vectors respectively, and determine the event type corresponding to the reference feature vector with the highest similarity as the class label of the first signal of the unknown class. Thus, the classification of microseismic signals across different domains and environments can be realized without obtaining historical annotation data of specific types, improving the applicability and efficiency of microseismic event diagnosis.

[0063] It should be noted that in some embodiments, there may be a situation where the similarity between the mapped feature vector and all reference feature vectors is extremely low. In this case, determining the event type corresponding to the reference feature vector with the highest similarity as the target event type cannot guarantee the credibility of microseismic event diagnosis, thereby affecting the reliability of disaster warning. Therefore, in the embodiments of the present application, a confidence threshold can be set according to experience, etc. After calculating the similarity corresponding to each reference feature vector, the similarity is compared with the confidence threshold. If the maximum similarity is greater than or equal to the confidence threshold, it can be determined that the result of the zero-shot learning model for the first signal is valid. Conversely, if the maximum similarity is less than the confidence threshold, the first signal can be marked as an uncertain category to facilitate subsequent use of the first signal to optimize and update the parameters of the zero-shot learning model.

[0064] Step 104, predict the occurrence probability and risk level of the microseismic event corresponding to the first signal according to the first signal, the first feature vector, and the target event type.

[0065] In the embodiments of the present application, the first signal, the first feature vector, and the target event type can be input into a pre-trained microseismic disaster prediction model. The prediction model analyzes and dynamically models the first signal according to the first feature vector, and combines the target event type to predict the future trend and intensity of microseismic activities at the target monitoring point where the first signal is collected, so as to obtain the probability and risk level of the microseismic event corresponding to the first signal.

[0066] It should be noted that the microseismic disaster prediction model can be constructed based on historical microseismic data and geological structure information, and time series analysis and machine learning algorithms, such as long short-term memory network (LSTM), etc., are introduced to be able to perform real-time prediction on the collected microseismic signals and improve the reliability of the prediction results.

[0067] In the embodiments of the present application, after determining the occurrence probability and risk level, the reliability of the prediction result can also be evaluated by combining a geomechanics model and numerical simulation.

[0068] Step 105, send a warning message when the occurrence probability is greater than the probability threshold and / or the risk level is greater than the level threshold.

[0069] It should be noted that the probability threshold and the level threshold can be defined and set according to the actual warning accuracy requirements. The probability threshold is the critical value for judging whether a warning is needed from the perspective of the occurrence probability of a microseismic event. When the occurrence probability is greater than the probability threshold, it can be determined that a disaster may occur, and a warning is needed. The level threshold is the critical value for judging whether a warning is needed from the perspective of the severity of the losses that a microseismic event may cause. When the risk level is greater than the level threshold, it can be determined that the microseismic event is a disaster event that affects safety, and a warning is needed.

[0070] It should be noted that in this application, in order to ensure the rational allocation of resources for disaster early warning and reduce ineffective early warnings for events with a high occurrence probability but a low danger level, early warning information can be sent only when the two conditions of the occurrence probability being greater than the probability threshold and the danger level being greater than the level threshold are met.

[0071] In the embodiments of this application, the early warning information may include the location information of the target monitoring point where the first signal is collected, the predicted microseismic event type, the occurrence probability, the danger level, the cause of the first signal, and parameters such as the source location, depth, and intensity of the earthquake.

[0072] It should be noted that the sending method of the early warning information can be determined according to the actual application scenario of the microseismic diagnosis and early warning system, and can be at least one channel such as radio broadcast, text message, mobile application, etc., to ensure the wide spread and rapid response of the early warning information, thereby ensuring life and property safety.

[0073] In this embodiment, microseismic signals are collected by setting target monitoring points, and the microseismic signals are classified based on the mapping and similarity between the feature vectors extracted from the microseismic signals and the reference feature vectors. Then, according to the classification results and the microseismic signals, the occurrence probability and danger level of the microseismic event corresponding to the microseismic signal are predicted, and then it is judged whether early warning is needed. Thus, the classification of microseismic signals can be achieved without a large amount of labeled data, improving the robustness of microseismic event diagnosis in cross-domain scenarios, which is beneficial to improving the efficiency and reliability of disaster early warning, and thus ensuring life and property safety.

[0074] This embodiment provides another microseismic event diagnosis and early warning method based on zero-shot learning. Figure 2 It is a schematic flowchart of a microseismic event diagnosis and early warning method based on zero-shot learning provided by the embodiments of this application.

[0075] As Figure 2 shown, the microseismic event diagnosis and early warning method based on zero-shot learning may include the following steps:

[0076] Step 201, obtain the first signal collected at the target monitoring point.

[0077] Step 202, perform feature extraction on the first signal to obtain the first feature vector corresponding to the first signal.

[0078] For the detailed description of the above steps 201 and 202, reference can be made to the above embodiments of this application, and details will not be repeated here.

[0079] Step 203, use the pre-constructed feature mapping function to map the first feature vector to the second feature vector.

[0080] In the embodiments of the present application, a mapping function can be constructed based on a Support Vector Machine (SVM), which can transfer the feature vectors of unknown classes to the attribute space of known classes. That is to say, using the pre-constructed feature mapping function to map the first feature vector to obtain a second feature vector, and the second feature vector should belong to the attribute space of known classes or be similar to the attributes of known classes.

[0081] It should be noted that the feature mapping function is constructed based on unsupervised learning (such as clustering analysis, etc.) and zero-shot classification of the historically collected microseismic signals.

[0082] Optionally, the second signals collected historically can be clustered first to determine at least one event type and at least one second signal corresponding to each event type.

[0083] Among them, the second signal can be the historically collected microseismic signal at any environment and any location. Since the second signal is historical data, the causes, types, and source locations of the microseismic events corresponding to the second signal can be determined.

[0084] In the embodiments of the present application, a clustering analysis method can be used to preliminarily classify the second signals collected historically, obtain at least one event type after clustering and at least one second signal corresponding to each event type, and potential abnormal patterns in the second signals can be identified, that is, determine which type of microseismic signal can reflect that the rock mass is at the edge of failure and has a high degree of danger.

[0085] Then, feature extraction can be performed on at least one second signal corresponding to each event type to obtain a third feature vector.

[0086] It should be noted that when extracting the first feature vector from the first signal and the third feature vector from the second signal, the consistency of the extracted feature dimensions should be ensured.

[0087] In the embodiments of the present application, a zero-shot learning model based on attribute embedding can be constructed, and the zero-shot learning model extracts the third feature vector corresponding to each second signal through a multi-layer perceptron, that is, maps the attribute description of the second signal to a high-dimensional feature space.

[0088] After that, a reference feature vector corresponding to the event type can be constructed based on at least one third feature vector.

[0089] In the embodiments of the present application, domain knowledge or geological expert knowledge can be used to select the main features that can describe this type from at least one third feature vector corresponding to the same event type to construct the reference feature vector corresponding to this event type.

[0090] It should be noted that multiple second signals and their corresponding feature vectors can be divided into a training set, a validation set, and a test set. Then, the zero-shot learning model based on attribute embedding can be trained using the second signals of known categories and the corresponding reference feature vectors. And through the validation set and the test set, the model parameters can be optimized so that the model can map the category attribute description to the high-dimensional feature space.

[0091] Finally, a feature mapping function can be constructed based on the third feature vector and the reference feature vector.

[0092] In the embodiments of the present application, a mapping function can be constructed based on a support vector machine, and the mapping function can be trained using the third feature vectors of known categories and the corresponding reference feature vectors. And through the third feature vectors and reference feature vectors of the validation set and the test set, the mapping function can be optimized so that the mapping function can accurately map the feature vectors to the attribute spaces corresponding to each category. Thus, in the present application, the feature mapping function can map the microseismic signals of unknown categories to the known feature space, and further realize the classification of the microseismic signals of unknown categories.

[0093] It should be noted that in the embodiments of the present application, domain adaptation techniques, such as methods of adversarial training, maximum mean discrepancy minimization, covariance matrix matching, etc., are used to adjust the differences between the source domain and the target domain. By using the adversarial training strategy, through the game between the generator and the discriminator, the features learned by the model are made more robust and generalizable. An auxiliary classifier is added during the training process. This auxiliary classifier attempts to distinguish between source domain data and target domain data, while the main classifier tries to confuse this auxiliary classifier, thereby achieving feature-level alignment, being able to adjust the model to reduce the gap between different mine environments, and enabling the model to adapt to the new data distribution.

[0094] Step 204, determine the similarity between the second feature vector and the reference feature vector.

[0095] In the embodiments of the present application, the cosine similarity between the second feature vector and the reference feature vector can be calculated, and by comparing the magnitudes of the similarities, it can be determined which type of signal feature the microseismic signal collected in real time better conforms to.

[0096] Optionally, in the case where the maximum value in the similarities is less than the confidence threshold, the event type corresponding to the first signal is marked with a preset identifier. Then, based on the first signal and the first feature vector corresponding to the preset identifier, the reference event type, the reference feature vector, and the feature mapping function can be updated.

[0097] Among them, the confidence threshold is a critical value used to determine the effectiveness of classification. Only when the similarity between the second feature vector and any reference feature vector is greater than this threshold can it be determined that the classification result based on this similarity is effective. The confidence threshold can be set according to experience, and this application does not limit it.

[0098] Among them, the preset identifier can be any string used to mark that the first signal cannot be classified, such as "uncertain category" or "unknown category", etc.

[0099] In the embodiments of this application, there may be a situation where the similarities between the mapped feature vectors and all reference feature vectors are particularly low. At this time, determining the event type corresponding to the reference feature vector with the highest similarity as the target event type cannot guarantee the credibility of microseismic event diagnosis, and thus will affect the reliability of disaster warning. Therefore, when the maximum value in the similarities is less than the confidence threshold, the event type corresponding to the first signal can be marked with the preset identifier. Then, when the number of first signals marked with the preset identifier reaches a certain value, or after a certain time interval since the last update, etc., based on the first signals corresponding to the preset identifier and the first feature vectors, the reference event types, reference feature vectors, and feature mapping functions can be updated to enrich the reference event types and reference feature vectors stored in the model, optimize the feature mapping function, and ensure the robustness of the model for classifying microseismic signals.

[0100] Step 205: Determine the reference event type corresponding to the reference feature vector with the highest similarity as the target event type corresponding to the first signal.

[0101] In this embodiment, after using the feature mapping function to map the feature vector extracted from the first signal to a new feature vector, the similarity between the mapped feature vector and the reference feature vectors of known categories is calculated, and the reference event type corresponding to the reference feature vector with the highest similarity is determined as the target event type corresponding to the first signal. Thus, the model can classify microseismic signals across different fields and environments without obtaining specific types of historical annotation data, improving the applicability and efficiency of microseismic event diagnosis.

[0102] Step 206: Predict the occurrence probability and danger level of the microseismic event corresponding to the first signal according to the first signal, the first feature vector, and the target event type.

[0103] Step 207: Send a warning message when the occurrence probability is greater than the probability threshold and / or the danger level is greater than the level threshold.

[0104] For the detailed descriptions of the above steps 206 and 207, reference can be made to the above embodiments of this application, and details are not described here again.

[0105] This embodiment provides another microseismic event diagnosis and early warning method based on zero-shot learning. Figure 3 It is a schematic flowchart of a microseismic event diagnosis and early warning method provided by an embodiment of the present application based on zero-shot learning.

[0106] As Figure 3 shown, the microseismic event diagnosis and early warning method based on zero-shot learning may include the following steps:

[0107] Step 301, obtain a first signal collected at a target monitoring point.

[0108] Step 302, perform feature extraction on the first signal to obtain a first feature vector corresponding to the first signal.

[0109] Step 303, determine a target event type corresponding to the first signal based on the first feature vector, a reference feature vector, and a reference event type corresponding to the reference feature vector.

[0110] Step 304, predict the occurrence probability and risk level of the microseismic event corresponding to the first signal according to the first signal, the first feature vector, and the target event type.

[0111] For the detailed descriptions of the above steps 301 to 304, reference may be made to the above embodiments of the present application, which will not be elaborated here.

[0112] Step 305, in the case where the occurrence probability is greater than a probability threshold and / or the risk level is greater than a level threshold, determine at least one early warning method according to the current business scenario.

[0113] In the embodiments of the present application, since the methods that can achieve or meet the early warning requirements are different in different business scenarios, appropriate early warning methods can be selected in different business scenarios. The early warning methods may include radio broadcasts, text messages, mobile applications, etc.

[0114] Step 306, generate and send an early warning message based on at least one early warning method.

[0115] It should be noted that for different early warning methods, in order to ensure the timeliness and accuracy of early warnings, the content of the early warning messages included may not be exactly the same. For example, when publishing an early warning message through the radio broadcast channel, the early warning message should be concise and to the point, so the early warning message may only include the location and cause of the disaster event. Or, when displaying an early warning message on a display screen for monitoring the working state of a mine, the early warning message can be as detailed and intuitive as possible. Then, the early warning message may include the waveform diagram of the first signal, the diagnosed event type, the cause, the disaster occurrence probability, the risk level, as well as specific disaster early warning suggestions, 3D display diagrams, etc.

[0116] Optionally, based on the first signal, a second parameter corresponding to the first signal may be determined, where the second parameter includes the source location, depth, and intensity.

[0117] In the embodiments of the present application, the source location of a microseismic event that has occurred may be located to determine parameters such as the source location, depth, and intensity, that is, the second parameter is obtained. Specifically, an inversion algorithm (such as a grid search method, a conjugate gradient method, etc.) may be used to process the first signal to obtain accurate second parameters.

[0118] Then, based on the location information of the target monitoring point and the second parameter, the geometric characteristics of the crack are determined, and a three-dimensional display diagram is generated. After that, an early warning message may be generated based on the three-dimensional display diagram.

[0119] In the embodiments of the present application, since the second parameter describes the source location of the microseismic event, and the location information of the target monitoring point and the time when the first signal is collected can reflect the spatio-temporal distribution law of the microseismic event, so based on the location information of the target monitoring point and the second parameter, the geometric characteristics such as the geometric shape, size, and distribution of the underground crack can be explained, and then the geological model and numerical simulation method are used to perform three-dimensional visualization of the crack, so that a three-dimensional display diagram can be generated, providing an intuitive scientific basis for disaster prevention and production guidance.

[0120] In the embodiments of the present application, by selecting a suitable early warning method based on the business scenario of the diagnosis and early warning method provided by the present application, the early warning information is generated and sent, which can ensure the wide spread and rapid response of the early warning information, and then ensure the safety of life and property.

[0121] Optionally, after sending the early warning information, the score of the current early warning system under a preset index may also be determined based on the feedback data for the early warning information, where the early warning index may include at least one of accuracy rate, response time, and false alarm rate. Then, based on the score under the preset index, the early warning system may be updated to ensure the reliability of the early warning system and optimize the use experience.

[0122] It should be noted that the early warning system refers to a system that uses the microseismic event diagnosis and early warning method based on zero-shot learning provided by the present application, and may include the zero-shot learning model, prediction model, threshold, etc. mentioned in the above embodiments. Therefore, after receiving the feedback data for the early warning information, the performance and effect of the early warning system can be evaluated, and when the score of any preset index is low, the parameters or threshold size of the model can be adjusted.

[0123] To implement the above embodiments, the present application also proposes a microseismic event diagnosis and early warning device based on zero-shot learning.

[0124] Figure 4Schematic diagram of a microseismic event diagnosis and early warning device based on zero-shot learning provided by an embodiment of the present application.

[0125] As Figure 4 shown, the microseismic event diagnosis and early warning device based on zero-shot learning includes:

[0126] An acquisition module 401, configured to acquire a first signal collected at a target monitoring point;

[0127] A feature extraction module 402, configured to extract features from the first signal to obtain a first feature vector corresponding to the first signal;

[0128] A classification module 403, configured to determine a target event type corresponding to the first signal based on the first feature vector, a reference feature vector, and a reference event type corresponding to the reference feature vector;

[0129] A prediction module 404, configured to predict the occurrence probability and risk level of a microseismic event corresponding to the first signal according to the first signal, the first feature vector, and the target event type;

[0130] An early warning module 405, configured to send an early warning message when the occurrence probability is greater than a probability threshold, and / or the risk level is greater than a level threshold.

[0131] Further, in a possible implementation manner of the embodiment of the present application, the acquisition module 401 may further be configured to:

[0132] Determine at least one target monitoring point in the area to be monitored according to a first parameter of the area to be monitored.

[0133] Further, in a possible implementation manner of the embodiment of the present application, the first feature vector includes at least one of the following: statistical features, time features, and waveform features, where the statistical features include at least one of the following: mean, standard deviation, skewness, kurtosis, time difference between P wave and S wave;

[0134] The time features include at least one of the following: rise time, fall time, duration;

[0135] The waveform features include at least one of the following: maximum peak value, minimum valley value, number of zero-crossing points.

[0136] Further, in a possible implementation manner of the embodiment of the present application, the classification module 403 may specifically be configured to:

[0137] Use a pre-constructed feature mapping function to map the first feature vector to a second feature vector;

[0138] Determine the similarity between the second feature vector and the reference feature vector;

[0139] Determine the target event type corresponding to the first signal as the reference event type corresponding to the reference feature vector with the highest similarity.

[0140] Further, in a possible implementation manner of the embodiment of the present application, the classification module 403 may further be used to:

[0141] Cluster the second signals collected historically to determine at least one event type and at least one second signal corresponding to each event type;

[0142] Extract features from at least one second signal corresponding to each event type respectively to obtain a third feature vector;

[0143] Based on at least one third feature vector, construct a reference feature vector corresponding to the event type;

[0144] Based on the third feature vector and the reference feature vector, construct a feature mapping function.

[0145] Further, in a possible implementation manner of the embodiment of the present application, the classification module 403 may further be used to:

[0146] In the case where the maximum value in the similarity is less than the confidence threshold, mark the event type corresponding to the first signal as a preset identifier;

[0147] Based on the first signal and the first feature vector corresponding to the preset identifier, update the reference event type, the reference feature vector, and the feature mapping function.

[0148] Further, in a possible implementation manner of the embodiment of the present application, the warning module 405 may specifically be used to:

[0149] Determine at least one warning method according to the current business scenario;

[0150] Generate and send a warning message based on at least one warning method.

[0151] Further, in a possible implementation manner of the embodiment of the present application, the warning module 405 may specifically be used to:

[0152] Based on the first signal, determine a second parameter corresponding to the first signal, where the second parameter includes the epicenter position, depth, and intensity;

[0153] According to the position information of the target monitoring point and the second parameter, determine the geometric characteristics of the crack and generate a three-dimensional display diagram;

[0154] Generate a warning message based on the three-dimensional display diagram.

[0155] Further, in a possible implementation manner of the embodiment of the present application, the warning module 405 may further be configured to:

[0156] Determine the score of the current warning system under a preset index based on the feedback data for the warning information, where the warning index includes at least one of accuracy rate, response time, and false alarm rate;

[0157] Update the warning system based on the score under the preset index.

[0158] It should be noted that the foregoing explanation of the embodiment of the microseismic event diagnosis and warning method based on zero-shot learning is also applicable to the microseismic event diagnosis and warning device based on zero-shot learning of this embodiment, and will not be repeated here.

[0159] In the embodiment of the present application, microseismic signals are collected by setting target monitoring points, and the microseismic signals are classified based on the mapping and similarity between the feature vectors extracted from the microseismic signals and the reference feature vectors. Then, according to the classification results and the microseismic signals, the disaster occurrence probability and risk level corresponding to the microseismic event that generates the microseismic signals are predicted, and then it is determined whether a warning is needed. Thus, the classification of microseismic signals can be realized without a large amount of labeled data, the robustness of microseismic event diagnosis in cross-domain scenarios is improved, which is beneficial to improving the efficiency and reliability of disaster warning, and further ensuring life and property safety.

[0160] To implement the above embodiment, the present application also proposes an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the method provided in the foregoing embodiment.

[0161] To implement the above embodiment, the present application also proposes a computer-readable storage medium, in which computer execution instructions are stored, and when the computer execution instructions are executed by a processor, they are used to implement the method provided in the foregoing embodiment.

[0162] To implement the above embodiment, the present application also proposes a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the method provided in the foregoing embodiment.

[0163] The collection, storage, use, processing, transmission, provision, and disclosure of the user's personal information involved in the present application and other processing all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.

[0164] It should be noted that personal information from users should be collected for legal and reasonable purposes and not shared or sold outside of such legal uses. In addition, such collection / sharing should be carried out after obtaining the informed consent of the users, including but not limited to notifying the users to read the user agreement / user notice before using the function and signing an agreement / authorization including authorizing the relevant user information. In addition, any necessary steps should be taken to defend and safeguard access to such personal information data and ensure that others with access to the personal information data comply with their privacy policies and procedures.

[0165] This application is expected to provide an implementation for users to selectively block the use or access of personal information data. That is, this disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by restricting data collection and deleting the data. In addition, when applicable, personal identifiers are removed from such personal information to protect the privacy of the users.

[0166] In the description of the foregoing embodiments, the descriptions referring to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0167] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of this application, "a plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0168] Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment, or part of code including one or more executable instructions for implementing a customized logical function or process, and the scope of the preferred implementation of this application includes additional implementations, where the functions may be executed in a substantially simultaneous manner or in a reverse order according to the involved functions, rather than in the order shown or discussed, which should be understood by those skilled in the art to which the embodiments of this application belong.

[0169] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definitional sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection portion having one or more wirings (electronic device), a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then storing it in a computer memory.

[0170] It should be understood that various parts of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0171] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the method of implementing the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0172] In addition, each functional unit in various embodiments of the present application may be integrated into a processing module, may exist physically alone for each unit, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0173] The above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disc, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present application.

Claims

1. A microseismic event diagnosis and early warning method based on zero-sample learning, characterized in that: The method comprises: Acquire a first signal collected at a target monitoring point; Performing feature extraction on the first signal to obtain a first feature vector corresponding to the first signal; Determine a target event type corresponding to the first signal based on the first feature vector, a reference feature vector, and a reference event type corresponding to the reference feature vector; Predicting the occurrence probability and hazard level of a microseismic event corresponding to the first signal according to the first signal, the first eigenvector and the target event type; When the occurrence probability is greater than a probability threshold, and / or the danger level is greater than a level threshold, a warning message is sent.

2. The method according to claim 1, characterized in that Before acquiring the first signal collected at the target monitoring point, the method further includes: At least one target monitoring point in the area to be monitored is determined according to a first parameter of the area to be monitored.

3. The method according to claim 1, characterized in that The first feature vector includes at least one of the following: statistical features, time features, and waveform features, wherein the statistical features include at least one of the following: mean, standard deviation, skewness, kurtosis, and arrival time difference between P wave and S wave; The time characteristics include at least one of the following: rise time, fall time, and duration; The waveform feature includes at least one of the following: maximum peak value, minimum valley value, and number of zero crossing points.

4. The method according to claim 1, characterized in that The determining, according to the first feature vector, a reference feature vector, and a reference event type corresponding to the reference feature vector, a target event type corresponding to the first signal includes: Mapping the first feature vector to a second feature vector using a pre-constructed feature mapping function; determining a similarity between the second feature vector and the reference feature vector; The reference event type corresponding to the reference feature vector with the highest similarity is determined as the target event type corresponding to the first signal.

5. The method according to claim 4, characterized in that Before determining the target event type corresponding to the first signal according to the first feature vector, the reference feature vector, and the reference event type corresponding to the reference feature vector, the method further includes: Clustering the second signals collected historically to determine at least one event type and at least one second signal corresponding to each of the event types; Performing feature extraction on at least one second signal corresponding to each event type to obtain a third feature vector; Based on the at least one third feature vector, construct a reference feature vector corresponding to the event type; A feature mapping function is constructed based on the third feature vector and the reference feature vector.

6. The method according to claim 4, characterized in that After determining the similarity between the second feature vector and the reference feature vector, the method further includes: When the maximum value among the similarities is less than the confidence threshold, marking the event type corresponding to the first signal as a preset identifier; Based on the first signal corresponding to the preset identifier and the first feature vector, a reference event type, a reference feature vector and a feature mapping function are updated.

7. The method according to claim 1, characterized in that The sending of warning information includes: Determine at least one early warning method based on the current business scenario; Based on the at least one warning method, generate and send warning information.

8. The method according to claim 7, characterized in that The generating and sending warning information based on the at least one warning method includes: Based on the first signal, determining a second parameter corresponding to the first signal, wherein the second parameter includes a source location, depth, and intensity; Determine the geometric features of the crack according to the location information of the target monitoring point and the second parameter, and generate a three-dimensional display diagram; Based on the three-dimensional display diagram, early warning information is generated.

9. The method according to any one of claims 1 to 8, characterized in that: After sending the warning information, the method further includes: Based on the feedback data for the warning information, determine the score of the current warning system under the preset indicators, wherein the warning indicators include at least one of accuracy, response time and false alarm rate; Based on the scores under the preset indicators, the early warning system is updated.

10. A microseismic event diagnosis and early warning device based on zero-sample learning, characterized in that: The device comprises: An acquisition module, used to acquire a first signal collected at a target monitoring point; A feature extraction module, used to extract features from the first signal to obtain a first feature vector corresponding to the first signal; a classification module, configured to determine a target event type corresponding to the first signal based on the first feature vector, a reference feature vector, and a reference event type corresponding to the reference feature vector; A prediction module, configured to predict the occurrence probability and hazard level of a microseismic event corresponding to the first signal according to the first signal, the first characteristic vector and the target event type; The early warning module is used to send early warning information when the probability of occurrence is greater than a probability threshold and / or the danger level is greater than a level threshold.

Images