CN107784276A - Microseismic event recognition methods and device - Google Patents

Abstract

The present invention provides a microseismic event identification method and device, the method comprising: S1, based on the waveform images of each channel of the event to be identified, using a preset convolutional neural network model to extract the waveform features of each channel of the event to be identified; S2, combining the waveform features of all channels of the event to be identified to obtain the combined waveform feature of the event to be identified; S3, based on the combined waveform feature of the event to be identified, using a preset support vector machine model to analyze the event to be identified sort. The waveform features of each channel of the event to be identified are extracted through the preset convolutional neural network, and then the waveform features of each channel are combined into a whole, that is, the combined waveform feature, and the preset support vector machine model is input to realize the classification of the event to be identified, and finally realize the Identification of microseismic events in events. The automatic identification of microseismic events is realized without relying on the knowledge level and experience of operators, with high accuracy, not affected by application scenarios, and strong generalization ability.

Description

Method and device for identifying microseismic events

technical field

Embodiments of the present invention relate to the technical field of geophysical detection, and more specifically, to a method and device for identifying microseismic events.

Background technique

Microearthquakes refer to earthquakes with a magnitude less than three. People generally cannot feel such earthquakes and can only be monitored with specific instruments. Microseisms will generate corresponding excitations to the underground medium, and such intensity may change the mechanical state of the underground medium. In the field of mine safety monitoring, microseisms are the precursor of mine dynamic disasters, and real-time monitoring of microseisms can effectively predict and prevent the occurrence of dynamic disasters. In addition, during the exploration and development of unconventional oil and gas such as shale gas and coalbed methane, microseismic-based fracture monitoring technology has become one of the most accurate technologies for fracturing crack monitoring at home and abroad. In the process of microseismic monitoring, in addition to the microseismic signals generated by microseismic events, the signals received by monitoring instruments often include signals generated by blasting events, noise time, etc. Therefore, in the process of microseismic monitoring, microseismic events can be identified from many events is the basis for monitoring.

A microseismic event has the characteristics of single event and multiple channels. Specifically, when a microseismic event occurs, multiple sensors are usually triggered, and the channels corresponding to these sensors will collect and store corresponding waveform data. Such a waveform caused by a microseismic event and triggered by multiple channels is called a single-event multi-channel waveform. At present, microseismic events are mainly identified by information technology personnel. Microseismic monitoring instruments receive a large number of signals transmitted by multiple channels, and technicians extract the waveform characteristics of these signals through certain signal processing methods. Based on practical experience, it is judged whether the event corresponding to the received signal is a microseismic event.

However, the method of manually extracting signal waveform features to identify microseismic events requires a large workload and depends on the knowledge and experience of the operator, and according to the manually extracted waveform features, the recognition accuracy is low, and it is generally only suitable for specific application scenarios , poor generalization ability.

Contents of the invention

Aiming at the problems of heavy workload, low recognition accuracy and poor generalization ability in the microseismic event recognition technology in the prior art. Embodiments of the present invention provide a microseismic event identification method and device for overcoming the above problems or at least partially solving the above problems.

On the one hand, the embodiment of the present invention provides a method for identifying microseismic events, the method comprising:

S1, based on the waveform image of each channel of the event to be identified, using a preset convolutional neural network model to extract the waveform features of each channel of the event to be identified;

S2. Combining the waveform features of all channels of the event to be identified to obtain the combined waveform feature of the event to be identified;

S3. Based on the combined waveform features of the event to be identified, use a preset support vector machine model to classify the event to be identified.

Wherein, before step S1 also includes:

The waveform signal of each channel of the event to be identified is converted into an image form, and the waveform image is obtained through preprocessing.

Wherein, step S1 specifically includes:

The waveform images of each channel of the event to be identified are respectively randomly cut into several small blocks, and the waveform features of each small block are respectively extracted by using the preset convolutional neural network model;

Calculate the mean value of the waveform features of the several small tiles corresponding to each channel of the event to be identified to obtain the waveform feature of each channel of the event to be identified.

Wherein, step S2 specifically includes:

Converting the waveform feature of each channel of the event to be identified into a 1×M dimension, where M is an integer greater than zero;

Combine the waveform features of all channels of the event to be identified to obtain a combined waveform feature of N×M dimensions, where N is the number of channels of the event to be identified.

Wherein, the preset convolutional neural network model is obtained through the following steps:

Construct a convolutional neural network, wherein the input of the convolutional neural network is the waveform image of the event to be identified, and the softmax layer is connected after the fully connected layer;

Using the first training data set to train the convolutional neural network to obtain a trained convolutional neural network; wherein the first training data set includes the waveform image of the event to be identified and the corresponding event category.

Wherein, the preset support vector machine model is obtained through the following steps:

Constructing a support vector machine, the input dimension of the support vector machine is N×M;

The support vector machine is trained by using the second training data set to obtain a preset support vector machine model; wherein the second data set includes the combined waveform features of the event to be identified and the corresponding event category.

Wherein, the preset support vector machine model includes a plurality of support vector machines with different input dimensions N.

On the other hand, an embodiment of the present invention provides a device for identifying microseismic events, the device comprising:

The feature extraction module is used to extract the waveform features of each channel of the event to be identified based on the waveform image of each channel of the event to be identified using a preset convolutional neural network model;

A feature combination module, configured to combine the waveform features of all channels of the event to be identified to obtain the combined waveform feature of the event to be identified;

The identification module is configured to classify the event to be identified based on the combined waveform features of the event to be identified by using a preset support vector machine model.

In the third aspect, an embodiment of the present invention provides a computer program product, the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed When executed by a computer, the computer is made to execute the microseismic event identification method.

In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, where the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the microseismic event identification method.

The embodiment of the present invention provides a method and device for identifying microseismic events. The waveform features of each channel of the event to be identified are extracted through a preset convolutional neural network, and then the waveform features of each channel are combined into a whole, that is, the combined waveform feature. The support vector machine model is set to classify the events to be identified, and finally realize the identification of microseismic events in the event. The automatic identification of microseismic events is realized without relying on the knowledge level and experience of operators, with high accuracy, not affected by application scenarios, and strong generalization ability.

Description of drawings

Fig. 1 is a flow chart of a method for identifying microseismic events provided by an embodiment of the present invention;

2 is a schematic diagram of obtaining the preset convolutional neural network model and preset support vector machine model in an embodiment of the present invention;

Fig. 3 is a structural block diagram of a device for identifying microseismic events provided by an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are the Some, but not all, embodiments are invented. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

Fig. 1 is a flowchart of a microseismic event recognition method provided by an embodiment of the present invention. As shown in Fig. 1, the method includes: S1, based on the waveform images of each channel of the event to be recognized, using a preset convolutional neural network model Extracting the waveform features of each channel of the event to be identified; S2, combining the waveform features of all channels of the event to be identified to obtain the combined waveform feature of the event to be identified; S3, based on the combined waveform feature of the event to be identified, using The preset support vector machine model classifies the event to be recognized.

Convolutional Neural Networks (CNN) is an efficient recognition method that has been developed in recent years and has attracted widespread attention. Generally, the basic structure of CNN includes two layers, one is the feature extraction layer, the input of each neuron is connected to the local receptive field of the previous layer, and the local features are extracted. Once the local feature is extracted, the positional relationship between it and other features is also determined; the second is the feature map layer, each calculation layer of the network is composed of multiple feature maps, each feature map is a plane, All neurons on the plane have equal weights. The feature map structure uses a nonlinear activation function as the activation function of the convolutional neural network, so that the feature map has displacement invariance. Preferably, the Relu activation function can be used. In addition, since neurons on a mapping plane share weights, the number of free parameters of the network is reduced. Each convolutional layer in the convolutional neural network is followed by a calculation layer for local averaging and secondary extraction. This unique two-time feature extraction structure reduces the feature resolution.

CNNs are primarily used to identify two-dimensional graphics that are invariant to displacement, scaling, and other forms of distortion. Since the feature detection layer of CNN learns through training data, when using CNN, it avoids the explicit feature extraction, and learns implicitly from the training data; moreover, because the weights of neurons on the same feature map are the same , so the network can learn in parallel, which is also a major advantage of the convolutional network over the network of neurons connected to each other. Convolutional neural network has unique advantages in speech recognition and image processing with its special structure of local weight sharing. Its layout is closer to the actual biological neural network. Weight sharing reduces the complexity of the network, especially multi-dimensional The feature that the image of the input vector can be directly input into the network avoids the complexity of data reconstruction in the process of feature extraction and classification.

However, since the signal in the microseismic monitoring technology has the characteristics of single event and multiple channels, if the convolutional neural network is used to identify microseismic events, since the number of channels for each event is different, it is necessary to construct and train multiple different convolutional neural networks. The training of convolutional neural network takes a long time, which cannot meet the requirements of quickly identifying microseismic events.

Support Vector Machine Support Vector Machine (Support Vector Machine, SVM) was first proposed by Cortes and Vapnik in 1995. It shows many unique advantages in solving small sample, nonlinear and high-dimensional pattern recognition, and can be extended to Function fitting and other machine learning problems. The support vector machine method is based on the VC dimension theory of statistical learning theory and the principle of structural risk minimization, according to the complexity of the model (that is, the learning accuracy for specific training samples, Accuracy) and learning ability (that is, The ability to identify any sample without error) seeks the best compromise in order to obtain the best generalization ability (or generalization ability).

The construction and training of support vector machines is very convenient, so in the process of identifying single-event and multi-channel events, we only need to build and train support vector machines of corresponding dimensions to classify events with the number of channels. The support vector machine realizes the classification of events with different channel numbers well, and at the same time avoids the time-consuming problem of constructing and training the convolutional neural network when using the convolutional neural network for event classification.

Specifically, in microseismic monitoring, waveform signals of multiple channels are received in one event, corresponding to multiple waveform images. After using the preset convolutional neural network model to extract the waveform features of each channel, the waveform features of each channel are combined to obtain the combined waveform features of the event to be identified, so that the waveform features of each channel of the event to be identified Combined into a whole as the input of the preset vector machine, it realizes the classification of the event to be identified, and judges whether the event to be identified is a microseismic event.

A microseismic event recognition method provided by an embodiment of the present invention extracts the waveform features of each channel of the event to be identified through a preset convolutional neural network, and then combines the waveform features of each channel into a whole, that is, the combined waveform feature, and inputs the preset support The vector machine model realizes the classification of the events to be identified, and finally realizes the identification of microseismic events in the event. The automatic identification of microseismic events is realized without relying on the knowledge level and experience of operators, with high accuracy, not affected by application scenarios, and strong generalization ability.

In the above embodiment, before step S1, it also includes:

The waveform signal of each channel of the event to be identified is converted into an image form, and the waveform image is obtained through preprocessing.

In the above embodiment, step S1 specifically includes:

The waveform images of each channel of the event to be identified are respectively randomly cut into several small blocks, and the waveform features of each small block are respectively extracted by using the preset convolutional neural network model;

Calculate the mean value of the waveform features of the several small tiles corresponding to each channel of the event to be identified to obtain the waveform feature of each channel of the event to be identified.

Preferably, the waveform image of each channel can be randomly cropped into 10 small tiles.

In the embodiment of the present invention, the waveform features of each channel can be extracted more accurately by randomly cutting out the waveform images of each channel and extracting the waveform features respectively, and then calculating the mean value features of these small blocks.

In the above embodiment, step S2 specifically includes:

Converting the waveform feature of each channel of the event to be identified into a 1×M dimension, where M is an integer greater than zero;

Combine the waveform features of all channels of the event to be identified to obtain a combined waveform feature of N×M dimensions, where N is the number of channels of the event to be identified.

In the above embodiment, the preset convolutional neural network model is obtained through the following steps:

Construct a convolutional neural network, wherein the input of the convolutional neural network is the waveform image of the event to be identified, the fully connected layer is connected to the softmax layer, and the output is the waveform feature of the event to be identified;

Using the first training data set to train the convolutional neural network to obtain a trained convolutional neural network; wherein the first training data set includes the waveform image of the event to be identified and the corresponding event category.

Wherein, the softmax layer is used to classify the events corresponding to the waveform pictures input to the convolutional neural network. During the training process, it is necessary to judge the accuracy of the classification through the classification results to determine whether to obtain the trained convolutional neural network. . After getting the trained convolutional neural network, because it is only necessary to use the trained convolutional neural network to extract the features of the waveform image, when identifying microseismic events, directly in the corresponding convolutional layer of the trained convolutional neural network It is enough to extract the waveform features of each channel at the time to be detected, instead of using the classification results of the softmax layer of the trained neural network at the time to be detected. Then, the preset convolutional neural network model not only utilizes the function of the convolutional neural network to extract waveform features of waveform graphics, but also avoids the problem that it is only applicable to events with a specific number of channels when using the softmax layer for classification.

In addition, as shown in Figure 2, when training the convolutional neural network, select typical waveform images of known events to train the convolutional neural network, for example, select 30,000 waveform images of known events to convolute The neural network is trained, and the classification accuracy of the trained convolutional neural network can reach 95.13%.

In the above embodiment, as shown in Figure 2, the preset support vector machine model is obtained through the following steps:

Constructing a support vector machine, the input dimension of the support vector machine is N×M;

The support vector machine is trained by using the second training data set to obtain a preset support vector machine model; wherein the second data set includes the combined waveform features of the event to be identified and the corresponding event category.

Wherein, the dimension of the input quantity of the constructed support vector machine is determined by the dimension of the combination feature of the event to be recognized. The combined waveform features in the second data set are obtained from the first data set after step S1 and step S2.

In the above embodiment, the preset support vector machine model includes a plurality of support vector machines with different input dimensions N.

Due to the relatively convenient construction and training of support vector machines, in order to facilitate the identification of events with different channel numbers, the preset support vector machine model can include a plurality of support vector machines, and the N value in the dimension of each support vector machine input quantity Different, events with different channel numbers can be classified.

An embodiment of the present invention provides a device for identifying microseismic events. As shown in FIG. 3 , the device includes: a feature extraction module 1 , a feature combination module 2 and a recognition module 3 . in:

The feature extraction module 1 is used to extract the waveform features of each channel of the event to be identified based on the waveform image of each channel of the event to be identified using a preset convolutional neural network model; the feature combination module 2 is used to combine all channels of the event to be identified The combined waveform features of the event to be identified are obtained from the waveform features of the event to be identified; the identification module 3 is used to classify the event to be identified by using a preset support vector machine model based on the combined waveform feature of the event to be identified.

Since the signal in the microseismic monitoring technology has the characteristics of single event and multiple channels, if the convolutional neural network is used to identify the microseismic event, since the number of channels of each event is different, it is necessary to construct and train multiple different convolutional neural networks. The training of convolutional neural network takes a long time, which cannot meet the requirements of quickly identifying microseismic events. The construction and training of support vector machines is very convenient, so in the process of identifying single-event and multi-channel events, we only need to build and train support vector machines of corresponding dimensions to classify events with the number of channels. The support vector machine realizes the classification of events with different channel numbers well, and at the same time avoids the time-consuming problem of constructing and training the convolutional neural network when using the convolutional neural network for event classification.

Specifically, in microseismic monitoring, waveform signals of multiple channels are received in one event, corresponding to multiple waveform images. After using the preset convolutional neural network model to extract the waveform features of each channel, the waveform features of each channel are combined to obtain the combined waveform features of the event to be identified, so that the waveform features of each channel of the event to be identified Combined into a whole as the input of the preset vector machine, it realizes the classification of the event to be identified, and judges whether the event to be identified is a microseismic event.

The microseismic event recognition device provided by the embodiment of the present invention uses the feature extraction module to extract the waveform features of each channel of the event to be identified, and then uses the feature combination module to combine the waveform features of each channel into a whole, that is, the combined waveform feature, and input it to the classification module Realize the classification of the events to be identified, and finally realize the identification of microseismic events in the event. The automatic identification of microseismic events is realized without relying on the knowledge level and experience of operators, with high accuracy, not affected by application scenarios, and strong generalization ability.

An embodiment of the present invention discloses a computer program product, the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by a computer, The computer can execute the methods provided by the above method embodiments, for example, including: based on the waveform images of each channel of the event to be identified, using a preset convolutional neural network model to extract the waveform features of each channel of the event to be identified; Identifying the waveform features of all channels of the event to obtain the combined waveform feature of the event to be identified; based on the combined waveform feature of the event to be identified, using a preset support vector machine model to classify the event to be identified.

An embodiment of the present invention provides a non-transitory computer-readable storage medium, the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the methods provided in the above method embodiments, for example Including: based on the waveform image of each channel of the event to be identified, using a preset convolutional neural network model to extract the waveform features of each channel of the event to be identified; combining the waveform features of all channels of the event to be identified to obtain the event to be identified Combining waveform features: based on the combined waveform features of the event to be identified, the event to be identified is classified using a preset support vector machine model.

Those of ordinary skill in the art can understand that all or part of the steps for realizing the above-mentioned method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the It includes the steps of the above method embodiments; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other various media that can store program codes.

Through the above description of the implementations, those skilled in the art can clearly understand that each implementation can be implemented by means of software plus a necessary general hardware platform, and of course also by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the prior art can be embodied in the form of software products, and the computer software products can be stored in computer-readable storage media, such as ROM/RAM, magnetic discs, optical discs, etc., including several instructions to make a computer device (which may be a personal computer, server, or network device, etc.) execute the methods described in various embodiments or some parts of the embodiments.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent replacements are made to some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims (10)

1. a kind of microseismic event recognition methods, it is characterised in that methods described includes：

S1, based on the waveform image of each passage of event to be identified, using to be identified described in default convolutional neural networks model extraction The wave character of each passage of event；

S2, the wave character for combining all passages of event to be identified obtain the combined waveform feature of the event to be identified；

S3, based on the combined waveform feature of the event to be identified, using default supporting vector machine model to the thing to be identified Part is classified.

2. method according to claim 1, it is characterised in that also include before step S1：

The waveform signal of each passage of event to be identified is converted into image format, and the oscillogram is obtained by pretreatment Picture.

3. method according to claim 1, it is characterised in that step S1 is specifically included：

It is several thumbnails by the waveform image difference random cropping of each passage of event to be identified, utilizes the default volume Product neural network model extracts the wave character of each thumbnail respectively；

Being worth to for the wave character of several thumbnails is asked corresponding to each passage of event to be identified described to wait to know The wave character of other each passage of event.

4. method according to claim 1, it is characterised in that step S2 is specifically included：

The wave character of each passage of event to be identified is converted into 1 × M dimensions, wherein M is the integer more than zero；

The wave character of all passages of event to be identified is combined to obtain the combined waveform feature of N × M dimensions, wherein N is the port number of the event to be identified.

5. method according to claim 1, it is characterised in that the default convolutional neural networks model is obtained by following steps ：

Convolutional neural networks are built, wherein, the input of the convolutional neural networks is the waveform image of the event to be identified, entirely Softmax layers are connected after articulamentum；

The convolutional neural networks trained are trained using convolutional neural networks described in the first training data set pair；Its In, first training dataset includes the waveform image of the event to be identified and corresponding event category.

6. method according to claim 4, it is characterised in that the default supporting vector machine model is obtained by following steps ：

SVMs is built, the input quantity dimension of the SVMs is N × M；

It is trained using SVMs described in the second training data set pair, obtains default supporting vector machine model；Wherein, institute Stating the second data set includes the combined waveform feature of the event to be identified and the classification of corresponding event.

7. method according to claim 6, it is characterised in that the default supporting vector machine model is tieed up including multiple input quantities Spend the different SVMs of N.

8. a kind of microseismic event identification device, it is characterised in that described device includes：

Characteristic extracting module, for the waveform image based on each passage of event to be identified, utilize default convolutional neural networks model Extract the wave character of each passage of event to be identified；

Feature combination module, the wave character for combining all passages of event to be identified obtain the event to be identified Combined waveform feature；

Identification module, for the combined waveform feature based on the event to be identified, using default supporting vector machine model to institute Event to be identified is stated to be classified.

9. a kind of computer program product, it is characterised in that the computer program product includes being stored in non-transient computer Computer program on readable storage medium storing program for executing, the computer program include programmed instruction, when described program is instructed by computer During execution, the computer is set to perform the method as described in claim 1 to 7 is any.

10. a kind of non-transient computer readable storage medium storing program for executing, it is characterised in that the non-transient computer readable storage medium storing program for executing is deposited Computer instruction is stored up, the computer instruction makes the computer perform the method as described in claim 1 to 7 is any.