CN113095364B - High-speed rail seismic event extraction method, medium and equipment using convolutional neural network - Google Patents
High-speed rail seismic event extraction method, medium and equipment using convolutional neural network Download PDFInfo
- Publication number
- CN113095364B CN113095364B CN202110269510.9A CN202110269510A CN113095364B CN 113095364 B CN113095364 B CN 113095364B CN 202110269510 A CN202110269510 A CN 202110269510A CN 113095364 B CN113095364 B CN 113095364B
- Authority
- CN
- China
- Prior art keywords
- speed rail
- seismic
- training
- samples
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000605 extraction Methods 0.000 title claims abstract description 26
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 66
- 238000012360 testing method Methods 0.000 claims abstract description 33
- 238000000034 method Methods 0.000 claims description 30
- 238000012545 processing Methods 0.000 claims description 14
- XEEYBQQBJWHFJM-UHFFFAOYSA-N iron Substances [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 claims description 13
- 229910052742 iron Inorganic materials 0.000 claims description 13
- 239000013598 vector Substances 0.000 claims description 10
- 238000011176 pooling Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 10
- 230000006870 function Effects 0.000 description 7
- 238000001514 detection method Methods 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 4
- 238000002955 isolation Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a high-speed rail seismic event extraction method, medium and equipment by utilizing a convolutional neural network, which normalize corresponding original seismic records of detectors according to statistical parameters of the detectors; setting a single sample time sequence length according to the time sequence length parameter of the high-speed rail seismic event; intercepting a complete seismic record section according to the time sequence length of a single sample, and constructing a training sample set and a testing sample set; constructing a 1DCNN network according to the characteristics of the single sample and setting corresponding super parameters; setting labels of whether all samples of the training data set belong to the high-speed rail seismic event or not; setting network training parameters, and sending a training sample set and a label into a 1DCNN network for classification training; applying the trained network to the classification of the actual test dataset; and determining the time sequence range of the high-speed rail earthquake event according to the prediction label and the sample serial number of the actual test data set, and completing the extraction of the high-speed rail earthquake event.
Description
Technical Field
The invention belongs to the technical field of exploration geophysics, and particularly relates to a high-speed rail seismic event extraction method, medium and equipment using a convolutional neural network.
Background
There has been a long history of monitoring studies for train operating conditions and geological environments. In 2004, li Li and the like initially proposed the possibility of train vibrations as a source for detecting underground structures. In 2015, li Zhimin and the like diagnose the train running state by utilizing the frequency characteristic of the vibration monitoring signal of the motor train, so that the working condition type of the train can be distinguished, and a data base of the train running state can be provided. In 2017, xu Shanhui and the like are carrying out a large number of seismic records on the Jinjin inter-city railway, which shows that the waveform of the vibration signal can be used as early warning for abnormal states of a track and an overhead structure. In 2019, liu Lei and the like construct a convolutional neural network model taking a time spectrum image corresponding to a high-speed rail seismic record as an input, train a classifier by artificially marking positive and negative event samples, and finally apply the classifier to screening a large number of real high-speed rail seismic event records. The future research of applying the high-speed rail earthquake to the geological environment has great potential, and the accurate and rapid acquisition of the real high-speed rail earthquake event record is a primary premise for the analysis and processing of the earthquake signals in the later period. The existing method for extracting the high-iron seismic event records mainly comprises the following steps:
prior art 1: high-iron seismic events are obtained by manual interception: the observation data is directly visualized to judge the high-speed rail signals, and then corresponding high-speed rail seismic events are obtained by manual interception, but the high-speed rail seismic events cannot be automatically extracted. In addition, this method requires excessive time and effort when the volume of seismic recording data is large, and a large number of high-speed rail seismic events cannot be obtained quickly.
Prior art 2: convolution neural network model with time spectrum image as input: firstly, artificially marking positive and negative event samples of a high-speed rail earthquake record, constructing a convolutional neural network model taking a time spectrum image of the samples as input, and screening out real high-speed rail earthquake events by using the model. The above method requires obtaining a time-frequency spectrum image of the seismic record, and thus it takes more time to extract a large number of high-iron seismic events when the seismic record data volume is excessive.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a high-speed railway seismic event extraction method, medium and equipment by utilizing a convolutional neural network, seismic data acquired by a plurality of detectors outside a high-speed railway line isolation area are used, and a large number of high-speed railway seismic events are extracted by utilizing a method for training a 1D convolutional neural network model, so that data is provided for subsequent geological information analysis.
The invention adopts the following technical scheme:
the high-speed rail seismic event extraction method using the convolutional neural network comprises the following steps:
s1, carrying out normalization processing on corresponding original seismic records according to statistical parameters of each detector to obtain normalized complete seismic records;
s2, setting a single sample time sequence length according to the time sequence length parameter of the high-speed rail seismic event;
s3, sectionally intercepting the complete seismic record normalized in the step S1 according to the single sample time sequence length determined in the step S2, and constructing a training sample set and a test sample set;
s4, setting labels for all samples of the training sample set in the step S3 to obtain label vectors of the training sample set;
s5, constructing a 1DCNN network according to the self characteristics of the single sample and setting corresponding super parameters;
s6, setting training parameters, and sending the training sample set and the label vector in the step S3 into the 1DCNN network constructed in the step S5 for classification training;
s7, applying the trained network in the step S6 to classification and discrimination of the high-speed rail seismic event of the test sample set to obtain a prediction label;
and S8, determining a time sequence range of the high-speed rail seismic event according to the prediction label and the sample sequence number obtained in the step S7, and completing the extraction of the high-speed rail seismic event.
Specifically, in step S1, the number of array elements of the detector array is assumed to be N, and the seismic record of the ith detector is assumed to be x i (n) corresponding to a mean value u i Maximum absolute value of m i Obtaining normalized complete seismic record y i (n)：
Where i denotes the detector number, and the values are i=1, 2, …, N.
Specifically, in step S3, the complete seismic record is x= [ X ] 1 … x N ]Containing components of E, N, Z with the greatest variance in three directions, adjacent samples being taken when the samples are taken from a complete seismic record segmentSetting 50% overlap length, namely overlapping the latter sample and the former sample by 0.5l, and intercepting seismic records by sections to obtain a single sample T of a data set k The method comprises the steps of carrying out a first treatment on the surface of the And then dividing all samples into a training set and a testing set in proportion, and obtaining a training sample set R and a testing sample set S on the assumption that the segmented and intercepted samples contain K samples in total.
Specifically, in step S4, whether all samples contain a high-speed rail seismic event is determined according to the complete seismic record waveform, if so, the samples belong to the high-speed rail seismic event, and the label is set to be 1; otherwise, the training data set does not belong to the high-speed rail seismic event, the label is set to be 0, and a label vector r of the training data set is obtained.
Specifically, in step S5, a 1DCNN network with 2 convolutional layers+relu layers, 2 pooling layers, 1 full connection layer and Softmax layer as structures is constructed; the output of the full connection layer is y 1 ,y 2 ,…,y n Obtaining an output Softmax (y) after Softmax regression treatment i The method comprises the steps of carrying out a first treatment on the surface of the And setting corresponding super parameters for each layer of the 1DCNN network according to the array element number of the seismic record detector, the time sequence length of the sample and the time domain waveform characteristic parameters.
Specifically, in step S6, the batch size and the network learning rate are selected according to the number of training samples, after Softmax regression processing, an error between the predicted output and the ideal output is calculated, and if p represents the correct probability output and q represents the predicted probability output, the cross entropy loss H (p, q) of the network training is obtained.
Specifically, in step S7, after model training is completed, the constructed actual test sample set is sent to a network to perform high-speed rail seismic event detection, so as to obtain a corresponding predictive label vector S:
s＝[s 1 s 2 … s K-p ]
wherein s is j And (3) representing a prediction label corresponding to the jth sample, wherein K is the total number of samples, and p is the number of samples of the training sample set.
Specifically, in step S8, after obtaining the prediction labels of the actual test sample set, according to the sample serial numbers of 0 appearing on both sides of the samples with the prediction labels of continuously 1, it is determined that a plurality of samples with the labels of 1 in the range are one complete high-speed rail seismic event, and all the high-speed rail seismic events are extracted from the original seismic record according to the time sequence range.
Another aspect of the invention is a computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computing device, cause the computing device to perform any of the methods.
Another aspect of the present invention is a computing device, including:
one or more processors, memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing any of the methods.
Compared with the prior art, the invention has at least the following beneficial effects:
the invention relates to a method for extracting a high-speed rail earthquake event by using a 1D convolutional neural network (1 DCNN) model, which can realize the extraction of the vibration event excited by a high-speed rail train by using a plurality of geophone data outside an isolation area. According to the method, firstly, the integral seismic records of the multiple detectors are normalized to obtain integral seismic records with original relative amplitude differences, then the time sequence length of a single sample is selected according to the statistics of the time sequence length of the high-speed rail seismic event, the integral seismic records of the multiple detectors are segmented into a series of samples, a part of the samples are provided with labels of whether the high-speed rail seismic event is, a 1DCNN network is built, super parameters are set to achieve feature extraction, finally event classification is completed through Softmax regression, and therefore the high-speed rail seismic event can be extracted according to the predicted labels and the sample positions of actual test data. Compared with the existing extraction method of the high-speed rail seismic events, the method can realize automatic extraction of a large number of high-speed rail seismic events by acquiring the classifier trained by only relying on a small number of seismic record samples.
Furthermore, the single sample time sequence length is set according to the time sequence length parameter of the high-speed rail seismic event, so that the subsequent classification error of the high-speed rail seismic event caused by the overlarge and undersize sample time sequence length can be avoided.
Furthermore, when multi-channel complete seismic record data are segmented, 50% of overlapping length is set between adjacent samples, so that the high-speed rail seismic event can be identified according to the classification labels and the sample serial numbers of the adjacent samples, and meanwhile, more accurate estimation of the high-speed rail seismic event position can be obtained.
Further, according to the seismic signal statistical characteristics of the geophone E, N, Z in three directions, the seismic record with the direction with the maximum variance is selected to construct a sample data set, so that the data volume participating in network training is reduced, and the characteristics of relevant redundancy are removed.
Further, a label whether the part of the samples intercepted by the segmentation belongs to the high-speed rail seismic event is set, and a sample classification basis is provided for training of a subsequent network model.
Further, in step S5, a 1DCNN network is constructed according to the characteristics of the sample itself and super parameters are set, so as to provide a mapping model for implementing sample data to classification labels, which is beneficial to completing the identification of the high-speed rail seismic event subsequently.
And further, based on the actual test sample intercepted in the step S3, the high-speed rail seismic event classification detection is carried out on the actual test sample, so that the sample belonging to the high-speed rail seismic event is determined, and the subsequent extraction of the high-speed rail seismic event is facilitated.
Further, the high-iron seismic event location is determined according to the prediction tag and the sample location in the step S7, so that all the detected high-iron seismic events are extracted accordingly.
In conclusion, the method can effectively and accurately extract the high-speed railway train seismic event by utilizing the data of the plurality of geophones, adopts a convolutional neural network deep learning model, has the characteristics of high accuracy, automation, rapidness and the like, and simultaneously provides a convenient method for extracting the high-speed railway seismic event without manual interception.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a three-component time domain waveform diagram of a single detector for a single high-iron seismic event;
FIG. 3 is a block diagram of a 1DCNN high-speed rail seismic event extraction model;
FIG. 4 is a graph showing the change in the loss function during a training session;
fig. 5 is a graph of 30 high-iron seismic event detections.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be understood that the terms "comprises" and "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It should be further understood that the term "and/or" as used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.
Various structural schematic diagrams according to the disclosed embodiments of the present invention are shown in the accompanying drawings. The figures are not drawn to scale, wherein certain details are exaggerated for clarity of presentation and may have been omitted. The shapes of the various regions, layers and their relative sizes, positional relationships shown in the drawings are merely exemplary, may in practice deviate due to manufacturing tolerances or technical limitations, and one skilled in the art may additionally design regions/layers having different shapes, sizes, relative positions as actually required.
The invention provides a high-speed rail seismic event extraction method by utilizing a convolutional neural network, which can realize the extraction of high-speed rail train vibration events by using a plurality of geophone data outside an isolation area. According to the method, firstly, the integral seismic records of the multiple detectors are normalized to obtain integral seismic records with original relative amplitude differences, then the time sequence length of a single sample is selected according to the statistics of the time sequence length of the high-speed rail seismic event, the integral seismic records of the multiple detectors are segmented into a series of samples, a part of the samples are provided with labels of whether the high-speed rail seismic event is, a 1DCNN network is built, super parameters are set to achieve feature extraction, finally event classification is completed through Softmax regression, and therefore the high-speed rail seismic event can be extracted according to the predicted labels and the sample positions of actual test data.
Referring to fig. 1, the method for extracting the high-speed rail seismic event by using the convolutional neural network of the invention comprises the following steps:
s1, normalizing corresponding original seismic records according to statistical parameters of each detector;
assuming that the array element number of the detector array is N, the seismic record of the ith detector is x i (n) corresponding to a mean value u i Maximum absolute value of m i Obtaining normalized complete seismic record y i (n)：
Where i denotes the detector number, and the values are i=1, 2, …, N.
S2, setting a single sample time sequence length according to the time sequence length parameter of the high-speed rail seismic event;
assuming a statistic of the high-speed rail seismic event time series length of L, a single sample time series length of l=0.6l is set when constructing the seismic record training and sample data set to be tested.
S3, sectionally intercepting the complete seismic record according to the single sample time sequence length to construct a training sample set and a test sample set;
let the complete seismic record be x= [ X 1 … x N ]When the samples are intercepted on the whole seismic record section, the overlapping length between the adjacent samples is set to be 50 percent, namely, the overlapping length between the later sample and the former sample is 0.5l, the seismic record is intercepted on the section, and a single sample T of the dataset is obtained k ：
Then dividing all samples into a training sample set R and a test sample set S according to a certain proportion, and obtaining the training sample set R and the test sample set S on the assumption that the segmented and intercepted samples contain K samples in total:
R＝[T 1 T 2 … T p ] (3)
S＝[T p+1 T p+2 … T K ] (4)
s4, setting labels of whether all samples of the training sample set R belong to the high-speed rail earthquake event or not;
judging whether all samples contain the high-speed rail seismic event according to the complete seismic record waveform, if so, belonging to the high-speed rail seismic event, and setting a label as '1' for the high-speed rail seismic event; otherwise, the training data set does not belong to the high-speed rail seismic event, and a label is set as 0 to obtain a label vector r of the training data set:
r＝[r 1 r 2 … r p ] (5)
wherein r is j The label corresponding to the j-th sample is represented, and the value of the label is 0 or 1.
S5, constructing a 1DCNN network according to the self characteristics of the single sample and setting network super parameters;
aiming at the characteristics of time sequence data of a sample, a 1DCNN network with a structure of 2 convolution layers, a Relu layer, 2 pooling layers, 1 full connection layer and a Softmax output layer is constructed.
Assuming that the output of the full link layer is y 1 ,y 2 ,…,y n Obtaining an output Softmax (y) after Softmax regression treatment i ：
And setting corresponding super parameters for each layer of the 1DCNN network according to parameters such as the array element number of the seismic record detector, the time sequence length of a sample, the time domain waveform characteristics and the like.
S6, setting network training parameters, and sending the training data set and the label into a 1DCNN network for classification training;
selecting training parameters such as batch processing size, network learning rate and the like according to the number of training samples, calculating an error between a predicted output and an ideal output after Softmax regression processing, wherein p represents a correct probability output, q represents a predicted probability output, and obtaining cross entropy loss H (p, q) of network training:
s7, applying the trained network to the classification and discrimination of the high-speed rail seismic event of the actual test data set;
after model training is completed, the constructed actual test sample set is sent into a network to carry out classification and discrimination on the high-speed rail seismic event, and a corresponding predictive label vector s is obtained:
s＝[s 1 s 2 … s K-p ] (8)
wherein s is j The prediction label corresponding to the jth sample is represented, and the value of the prediction label is '0' or '1'.
S8, determining a time sequence range of the high-speed rail earthquake event according to the prediction label and the sample serial number of the test data set, and completing high-speed rail earthquake event extraction.
After the prediction label of the actual test sample is obtained, according to the sample serial numbers of the samples with the prediction label of continuously 1 and the sample serial numbers of 0 at the two sides of the sample, a plurality of samples with the label of 1 in the range are judged to be one complete high-speed rail seismic event, so that all the high-speed rail seismic events can be extracted from the original seismic record according to the time sequence range.
In yet another embodiment of the present invention, a terminal device is provided, the terminal device including a processor and a memory, the memory for storing a computer program, the computer program including program instructions, the processor for executing the program instructions stored by the computer storage medium. The processor may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf Programmable gate arrays (FPGAs) or other Programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., which are the computational core and control core of the terminal adapted to implement one or more instructions, in particular adapted to load and execute one or more instructions to implement a corresponding method flow or a corresponding function; the processor of the embodiment of the invention can be used for the operation of a high-speed rail seismic event extraction method using a convolutional neural network, and comprises the following steps:
normalizing the corresponding original seismic records according to the statistical parameters of each detector; setting a single sample time sequence length according to the time sequence length parameter of the high-speed rail seismic event; intercepting normalized complete seismic records according to the time sequence length of a single sample to construct a training sample set and a test sample set; setting labels of whether all samples of the training sample set belong to the high-speed rail seismic event or not; constructing a 1DCNN network according to the characteristics of the single sample and setting corresponding super parameters; setting network training parameters, and sending a training sample set and a label into a 1DCNN network for classification training; applying the trained model to classification discrimination of the actual test sample set; and determining the time sequence range of the high-speed rail earthquake event according to the prediction label and the sample serial number of the actual test sample set, and completing the extraction of the high-speed rail earthquake event.
In a further embodiment of the present invention, the present invention also provides a storage medium, in particular, a computer readable storage medium (Memory), which is a Memory device in a terminal device, for storing programs and data. It will be appreciated that the computer readable storage medium herein may include both a built-in storage medium in the terminal device and an extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space storing an operating system of the terminal. Also stored in the memory space are one or more instructions, which may be one or more computer programs (including program code), adapted to be loaded and executed by the processor. The computer readable storage medium herein may be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory.
One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the corresponding steps of the method for high-speed rail seismic event extraction using a convolutional neural network in the above-described embodiments; one or more instructions in a computer-readable storage medium are loaded by a processor and perform the steps of:
normalizing the corresponding original seismic records according to the statistical parameters of each detector; setting a single sample time sequence length according to the time sequence length parameter of the high-speed rail seismic event; intercepting normalized complete seismic records according to the time sequence length of a single sample to construct a training sample set and a test sample set; setting labels of whether all samples of the training sample set belong to the high-speed rail seismic event or not; constructing a 1DCNN network according to the characteristics of the single sample and setting corresponding super parameters; setting network training parameters, and sending a training sample set and a label into a 1DCNN network for classification training; applying the trained model to classification discrimination of the actual test sample set; and determining the time sequence range of the high-speed rail earthquake event according to the prediction label and the sample serial number of the actual test sample set, and completing the extraction of the high-speed rail earthquake event.
Taking the signals received by 27 low-frequency detectors which are arranged at equal intervals along the rail direction of the high-speed rail line 15m when the high-speed rail passes through as an example.
Table 1 shows a sample data set of high-speed rail seismic records constructed in accordance with this embodiment
Table 2 shows the super parameter settings of the 1DCNN network in this embodiment
| Sequence number | Network layer | Input dimension | Output dimension | Super parameter setting |
| 1 | Conv1d+Relu | 27×1000 | 16×246 | kernel＝16×(20×27), |
| 2 | MaxPool1d | 16×246 | 16×123 | kernel＝2×1,stride＝2 |
| 3 | Conv1d+Relu | 16×123 | 32×60 | kernel＝32×(5×16), |
| 4 | MaxPool1d | 32×60 | 32×30 | kernel＝2×1,stride＝2 |
| 5 | Full connection | (32×30) | 2 | Without any means for |
| 6 | Softmax | 2 | 2 | Without any means for |
Referring to table 1, table 1 is a high-speed rail seismic record sample data set constructed in this embodiment, and a single sample is a seismic record containing 27 detectors and having a time sequence length of 1000, and thus its dimension is 27×1000. The training sample set contains 116 high-iron seismic event samples and 3934 non-high-iron seismic event samples; the test sample set contains 128 high-iron seismic event samples and 3921 non-high-iron seismic event samples. Accordingly, the high-speed rail seismic event sample classification label is "1", while the non-high-speed rail seismic event sample classification label is "0", and the specific division of the data set is detailed in table 1. Referring to table 2, table 2 shows the super parameter settings of the 1DCNN network in this embodiment, the input samples are seismic records with 27 detectors and a time sequence length of 1000, and the network structure is a convolution layer+relu layer, a pooling layer, a full connection layer and a Softmax output layer, and the main input channels, output channels and super parameter settings (convolution kernel/pooling kernel, step size, etc.) of the network are shown in table 2 in detail.
Referring to FIG. 2, FIG. 2 is a single-detector three-component time domain waveform of a single high-speed rail seismic event during a train pass; the sampling interval is 4ms, and there are 2008 sampling points. Referring to fig. 3 and 4, fig. 3 is a diagram of a model of 1DCNN high-speed rail seismic event extraction, and fig. 4 is a graph of a change in a loss function during a training process. It can be seen from fig. 4 that the 1DCNN network has converged after 20 iterations, so that it takes a short time to complete model training. The data received by 27 detectors when 30 trains pass through are detected and analyzed by the method, and the obtained 30 high-speed rail seismic event detection results are shown in fig. 5, are consistent with the number and the positions of the high-speed rail seismic events known in the actual seismic records, and can be sequentially extracted from the high-speed rail seismic events at the corresponding positions.
In summary, the method, medium and equipment for extracting the high-speed rail earthquake event by utilizing the convolutional neural network combine the time domain waveform characteristics of the high-speed rail earthquake event, firstly perform normalization processing on the original one-dimensional earthquake record, intercept the original one-dimensional earthquake record into a series of single samples in a sectioned manner, further construct a high-speed rail earthquake event data set, set a label of a training data set, realize characteristic extraction by training a 1DCNN network, and finally finish sample classification by Softmax regression, thereby extracting the high-speed rail earthquake event according to the classification label and time sequence range of the samples.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above is only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited by this, and any modification made on the basis of the technical scheme according to the technical idea of the present invention falls within the protection scope of the claims of the present invention.
Claims (6)
1. The high-speed rail seismic event extraction method using the convolutional neural network is characterized by comprising the following steps of:
s1, carrying out normalization processing on corresponding original seismic records according to statistical parameters of each detector to obtain normalized complete seismic records;
s2, setting a single sample time sequence length according to a time sequence length parameter of the high-speed rail seismic event, wherein the statistic value of the time sequence length of the high-speed rail seismic event is L, and setting the single sample time sequence length of l=0.6L;
s3, sectionally intercepting the complete seismic record normalized in the step S1 by using the single sample time sequence length determined in the step S2, and constructing a training sample set and a test sample set, wherein the complete seismic record is X= [ X ] 1 … x N ]When the samples are intercepted on the whole seismic record section, the overlapping length between the adjacent samples is set to be 50%, namely, the overlapping length between the later sample and the former sample is 0.5l, the seismic record is intercepted on the section, and a single sample T of the dataset is obtained k The method comprises the steps of carrying out a first treatment on the surface of the Then dividing all samples into a training set and a testing set in proportion, intercepting the samples in sections, and obtaining a training sample set R and a testing sample set S when the samples contain K samples in total;
wherein N is the array element number of the detector array;
s4, setting labels for all samples of the training sample set in the step S3 to obtain label vectors of the training sample set;
s5, constructing a 1DCNN network with a structure of a convolution layer+Relu layer, a pooling layer, a full connection layer and a Softmax layer according to the self characteristics of a single sample; the output of the full connection layer is y 1 ,y 2 ,…,y n The output after Softmax regression treatment was obtained as Softmax (y 1 )，Softmax(y 2 )，…，Softmax(y n ) The method comprises the steps of carrying out a first treatment on the surface of the Setting corresponding super parameters for each layer of the 1DCNN network according to the array element number of the seismic record detector, the time sequence length of the sample and the characteristic parameters of the time domain waveform;
s6, setting training parameters, and sending the training sample set and the label vector in the step S3 into the 1DCNN network constructed in the step S5 for classification training;
s7, applying the trained network in the step S6 to the classification and discrimination of the high-speed rail seismic event of the test sample set to obtain a prediction label, and after model training is completed, sending the constructed actual test sample set into the network to detect the high-speed rail seismic event to obtain a corresponding prediction label vector S:
s＝[s 1 s 2 …s j … s K-p ]
wherein s is j Representing a prediction label corresponding to the jth sample, wherein K is the total number of samples, and p is the number of samples of the training sample set;
s8, determining a time sequence range of the high-speed rail seismic event according to the prediction label and the sample serial number obtained in the step S7, completing high-speed rail seismic event extraction, judging that a plurality of samples with labels of 1 in the range are complete high-speed rail seismic events once according to sample serial numbers of 0 appearing on two sides of samples with the prediction label of 1 continuously after the prediction label of the actual test sample set is obtained, and extracting all the high-speed rail seismic events from the original seismic record according to the time sequence range.
2. The method of claim 1, wherein in step S1, the number of array elements of the detector array is set to N, and the seismic record of the ith detector is set to x i (n) corresponding to a mean value u i Maximum absolute value of m i Obtaining normalized complete seismic record y i (n)：
Where i denotes the detector number, and the values are i=1, 2, …, N.
3. The method according to claim 1, wherein in step S4, whether all samples contain a high-iron seismic event is determined according to the complete seismic record waveform, and if so, the samples belong to the high-iron seismic event, and a tag is set to be 1; otherwise, the training data set does not belong to the high-speed rail seismic event, the label is set to be 0, and a label vector r of the training data set is obtained.
4. The method according to claim 1, wherein in step S6, the batch size and the network learning rate are selected according to the number of training samples, and after Softmax regression processing, an error between the predicted output and the ideal output is calculated, and assuming that p represents the correct probability output and q represents the predicted probability output, the cross entropy loss H (p, q) of the network training is obtained.
5. A computer readable storage medium storing one or more programs, wherein the one or more programs comprise instructions, which when executed by a computing device, cause the computing device to perform any of the methods of claims 1-4.
6. A computing device, comprising:
one or more processors, memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing any of the methods of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110269510.9A CN113095364B (en) | 2021-03-12 | 2021-03-12 | High-speed rail seismic event extraction method, medium and equipment using convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110269510.9A CN113095364B (en) | 2021-03-12 | 2021-03-12 | High-speed rail seismic event extraction method, medium and equipment using convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113095364A CN113095364A (en) | 2021-07-09 |
| CN113095364B true CN113095364B (en) | 2023-12-19 |
Family
ID=76667066
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110269510.9A Active CN113095364B (en) | 2021-03-12 | 2021-03-12 | High-speed rail seismic event extraction method, medium and equipment using convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113095364B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113807305A (en) * | 2021-09-27 | 2021-12-17 | 宫雪峰 | Road destructive prediction automatic driving method based on naive Bayes classification algorithm |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1649289A (en) * | 2004-03-17 | 2005-08-03 | 威盛电子股份有限公司 | Differential phase detecting method and system |
| WO2019100724A1 (en) * | 2017-11-24 | 2019-05-31 | 华为技术有限公司 | Method and device for training multi-label classification model |
| CN111126471A (en) * | 2019-12-18 | 2020-05-08 | 中国石油大学(华东) | Microseism event detection method and system |
| CN111367961A (en) * | 2020-02-27 | 2020-07-03 | 西安交通大学 | Time sequence data event prediction method and system based on graph convolution neural network and application thereof |
| CN111505710A (en) * | 2020-05-06 | 2020-08-07 | 西安交通大学 | High-speed rail seismic source seismic signal extraction method based on equidistant spectrum characteristics |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107463907B (en) * | 2017-08-08 | 2021-06-25 | 东软集团股份有限公司 | Vehicle collision detection method and device, electronic equipment and vehicle |
-
2021
- 2021-03-12 CN CN202110269510.9A patent/CN113095364B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1649289A (en) * | 2004-03-17 | 2005-08-03 | 威盛电子股份有限公司 | Differential phase detecting method and system |
| WO2019100724A1 (en) * | 2017-11-24 | 2019-05-31 | 华为技术有限公司 | Method and device for training multi-label classification model |
| CN111126471A (en) * | 2019-12-18 | 2020-05-08 | 中国石油大学(华东) | Microseism event detection method and system |
| CN111367961A (en) * | 2020-02-27 | 2020-07-03 | 西安交通大学 | Time sequence data event prediction method and system based on graph convolution neural network and application thereof |
| CN111505710A (en) * | 2020-05-06 | 2020-08-07 | 西安交通大学 | High-speed rail seismic source seismic signal extraction method based on equidistant spectrum characteristics |
Non-Patent Citations (2)
| Title |
|---|
| Poststack Seismic Data Denoising Based on 3-D Convolutional Neural Network;Dawei Liu etal.;《IEEE Transactions on Geoscience and Remote Sensing》;第第58卷卷(第第3期期);全文 * |
| 大量高铁地震事件的属性体提取与特性分析;刘磊等;《地球物理学报》;20190630;第62卷(第6期);第1节第1段 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113095364A (en) | 2021-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20230333533A1 (en) | System and method for generating machine learning model with trace data | |
| CN112508105B (en) | Fault detection and retrieval method for oil extraction machine | |
| CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
| KR20160130422A (en) | Method and system for checking goods | |
| CN101614787A (en) | Analogical Electronics method for diagnosing faults based on M-ary textural classification device | |
| CN112633339A (en) | Bearing fault intelligent diagnosis method, bearing fault intelligent diagnosis system, computer equipment and medium | |
| CN106778908A (en) | A kind of novelty detection method and apparatus | |
| CN111964909A (en) | Rolling bearing operation state detection method, fault diagnosis method and system | |
| CN112766301B (en) | Oil extraction machine indicator diagram similarity judging method | |
| CN116012681A (en) | Method and system for diagnosing motor faults of pipeline robot based on sound vibration signal fusion | |
| CN110659682A (en) | Data classification method based on MCWD-KSMOTE-AdaBoost-DenseNet algorithm | |
| CN113095364B (en) | High-speed rail seismic event extraction method, medium and equipment using convolutional neural network | |
| CN110598999A (en) | Traffic travel analysis method, system and storage medium based on individual data | |
| CN114429152A (en) | Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption | |
| CN112147221B (en) | Steel rail screw hole crack identification method and system based on ultrasonic flaw detector data | |
| CN116466408B (en) | Artificial neural network superbedrock identification method based on aeromagnetic data | |
| CN112836719A (en) | Indicator diagram similarity detection method fusing two classifications and three groups | |
| CN114756420A (en) | Fault prediction method and related device | |
| CN106951924A (en) | Acceleration Algorithm in Seismic Coherence Cube image slices automatic identifying method and system based on AdaBoost algorithms | |
| CN116662800A (en) | Rolling bearing fault diagnosis method based on self-adaptive attention mechanism | |
| CN112215307B (en) | Method for automatically detecting signal abnormality of earthquake instrument by machine learning | |
| CN116168720A (en) | Motor sound abnormality fault diagnosis method, system and storable medium | |
| CN112801073B (en) | Acoustic emission first-break signal classification and identification method based on deep neural network | |
| CN115659271A (en) | Sensor abnormality detection method, model training method, system, device, and medium | |
| Deuschle et al. | Robust sensor spike detection method based on dynamic time warping |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |