CN113537636A - Earthquake destructive power prediction method and device based on bidirectional gating circulation unit - Google Patents
Earthquake destructive power prediction method and device based on bidirectional gating circulation unit Download PDFInfo
- Publication number
- CN113537636A CN113537636A CN202110927660.4A CN202110927660A CN113537636A CN 113537636 A CN113537636 A CN 113537636A CN 202110927660 A CN202110927660 A CN 202110927660A CN 113537636 A CN113537636 A CN 113537636A
- Authority
- CN
- China
- Prior art keywords
- seismic
- destructive power
- data
- earthquake
- time
- 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.)
- Pending
Links
- 230000001066 destructive effect Effects 0.000 title claims abstract description 71
- 238000000034 method Methods 0.000 title claims abstract description 24
- 230000002457 bidirectional effect Effects 0.000 title claims abstract description 19
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 25
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims abstract description 11
- 238000004458 analytical method Methods 0.000 claims description 20
- 238000004364 calculation method Methods 0.000 claims description 14
- 239000013598 vector Substances 0.000 claims description 9
- 230000005540 biological transmission Effects 0.000 claims description 8
- 238000012544 monitoring process Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 6
- 125000004122 cyclic group Chemical group 0.000 claims description 5
- 230000001902 propagating effect Effects 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- QBPFLULOKWLNNW-UHFFFAOYSA-N chrysazin Chemical compound O=C1C2=CC=CC(O)=C2C(=O)C2=C1C=CC=C2O QBPFLULOKWLNNW-UHFFFAOYSA-N 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000006870 function Effects 0.000 description 42
- 239000000306 component Substances 0.000 description 41
- 230000001133 acceleration Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000006073 displacement reaction Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000012216 screening Methods 0.000 description 5
- 238000011161 development Methods 0.000 description 4
- 230000008034 disappearance Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000004880 explosion Methods 0.000 description 3
- 241000288105 Grus Species 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 239000012533 medium component Substances 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000007873 sieving Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
- 238000012038 vulnerability analysis Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The invention relates to a seismic destructive power prediction method based on a bidirectional gating circulation unit, which comprises the following steps: acquiring seismic motion time-course data from an open source seismic motion database to obtain sample data and a sample label; decomposing various data included in the seismic motion time-course data into a plurality of intrinsic mode function components respectively through an empirical mode decomposition algorithm; taking each intrinsic mode function component obtained by decomposition as an input characteristic of the BiGRU model, and training by combining with a sample label to obtain an EMD-BiGRU model; and predicting the earthquake destructive force through the model. In the invention, the BiGRU model is used as a basic model to predict the earthquake destructive power level, so that the past and future seismic wave data can be linked with the current seismic wave prediction, and the model prediction is more facilitated; the EMD is combined with the BiGRU model, the new model comprises local characteristic signals of seismic data at different time scales, and high prediction accuracy can be achieved.
Description
Technical Field
The invention belongs to the technical field of earthquake destructive power prediction, and relates to an earthquake destructive power prediction method and device based on a bidirectional gating circulation unit.
Background
After an earthquake occurs, the earthquake damage condition of the disaster area is rapidly and accurately acquired, and the method is very important for timely and reasonably scheduling rescue force and reducing casualties and loss of the disaster area. The uncertainty of earthquake disaster occurrence and the high complexity of disaster area environment lead to the difficulty of accurately and timely predicting disaster area earthquake damage conditions. Besides seismic parameters, numerous factors such as fault development, structural environment, seismic distance and the like may influence the final prediction result. Traditional machine learning methods such as Support Vector machine (Support Vector machine), Decision Tree (Decision Tree), etc. are not suitable for processing high-dimensional time series data such as seismic destructive power.
Seismic destructive power based on numerical simulations there are now mainly two approaches. One is a vulnerability Analysis method (framework Analysis), which simplifies seismic motion intensity indexes and a region structure model, is difficult to comprehensively reflect complex time-domain and frequency-domain characteristics of seismic motion, and has high prediction speed but insufficient accuracy. A Nonlinear Time-history Analysis (NLTHA) has good retention on seismic parameters and building structures and high prediction accuracy, but the method has large calculation amount, needs a high-performance computer and has certain defect on prediction timeliness.
Disclosure of Invention
In view of the above, the present invention provides a method and an apparatus for predicting earthquake destructive power based on a bidirectional gated loop unit.
In order to achieve the purpose, the invention provides the following technical scheme:
a seismic destructive power prediction method based on a bidirectional gating cycle unit comprises the following steps:
s1, acquiring seismic motion time-course data from an open source seismic motion database, and establishing the database as sample data;
step S2, carrying out destructive power analysis on the sample data to divide destructive power grades, and taking the destructive power grades as sample labels;
step S3, decomposing the seismic motion time-course data into a plurality of intrinsic mode function components through an empirical mode decomposition algorithm;
step S4, defining a loss function of the BiGRU model, taking each intrinsic mode function component obtained by decomposition as an input characteristic of the BiGRU model, and training the intrinsic mode function component by combining a sample label;
step S5, when the value of the loss function of the BiGRU model converges to a fixed value and keeps unchanged, ending the training to obtain an EMD-BiGRU model;
and step S6, inputting seismic motion data acquired by a sensor or seismic motion data acquired through a seismic motion monitoring network into an EMD-BiGRU model, and predicting through the EMD-BiGRU model to obtain the seismic destructive power grade.
Further, the input features of the BiGRU model at any time are seismic motion time-course data or a vector formed by values of all eigenmode function components of each item of data included in the seismic motion data at the time.
Furthermore, in the BiGRU model, a hidden layer unit h of GRUtThe calculation formula of (2) is as follows:
rt＝σ(Wrxt+Urht-1)
zt＝σ(Wzxt+Uzht-1)
wherein, the sigma is a Sigmoid function; tan h is a hyperbolic tangent function; r istTo reset the gate; z is a radical oftTo update the door;
is a candidate activation state at time t; h ist-1Hidden layer unit at time (t-1); h istA hidden layer unit at the time t; wr、Ur、Wz、WzW, U is a training parameter matrix;
representing the multiplication of corresponding elements; x is the number oftRepresenting the input characteristics at time t.
Further, the calculation formula of the BiGRU model is as follows:
wherein x istAn input feature representing time t; (ii) a
Representing a backward propagation hidden state at time t;
represents a backward propagation hidden state at time (t + 1);
representing a forward propagation hidden state at time (t-1);
representing a forward propagation hidden state at time t; o istRepresenting a hidden layer state at time t; alpha is alphatRepresenting the hidden layer output weight of the forward transmission GRU unit at the time t; beta is atRepresenting the output weight of the hidden layer of the backward transmission GRU unit at the time t; btRepresenting the offset corresponding to the hidden layer state at the time t; w is atA weight coefficient representing a hidden layer state; y represents the final output prediction result.
Further, the loss function l (x, y) of the BiGRU model is defined as:
where N represents the number of samples, xiRepresenting the actual seismic destructive power level, y, of the ith sampleiRepresenting the predicted seismic destructive power level of the ith sample model.
An earthquake destructive power prediction device based on a bidirectional gating cycle unit comprises:
the earthquake motion data acquisition module is used for acquiring earthquake motion data in real time or acquiring the earthquake motion data through an earthquake motion monitoring network and transmitting the earthquake motion data to the analysis module;
the analysis module is used for analyzing the seismic motion data sent by the seismic motion data acquisition module and predicting the destructive power of the earthquake to obtain the destructive power grade of the earthquake; and
and the prediction data output module is used for outputting the destructive power level of the earthquake.
The early warning module is used for comparing the earthquake destructive power grade predicted by the analysis module with a preset earthquake destructive power grade threshold value, and sending out an early warning signal when the predicted earthquake destructive power grade is greater than or equal to the earthquake destructive power grade threshold value.
The early warning system further comprises a propagation module, wherein the propagation module is used for transmitting the early warning signal and the predicted earthquake destructive power level to specific equipment when the early warning module sends out the early warning signal.
Further, the analysis module comprises a data preprocessing unit, an EMD-BiGRU model and an output unit;
the data preprocessing unit is used for reading the seismic motion data acquired by the seismic motion data acquisition module and performing format conversion;
the EMD-BiGRU model comprises an empirical mode decomposition module and a BiGRU model, wherein the empirical mode decomposition module is used for decomposing seismic motion time-course data into a plurality of intrinsic mode function components through an empirical mode decomposition algorithm, and vectors formed by the intrinsic mode function components are used as input features of the BiGRU model; the BiGRU model is used for predicting the damage level of the earthquake according to the input characteristics;
and the output unit is used for sending the predicted earthquake destructive power grade to the predicted data output module and the early warning module.
According to the method, the EMD and the BiGRU model are combined to obtain the EMD-BiGRU model for predicting the earthquake destructive power level, and as GRUs capture long sequence semantic association, gradient disappearance or explosion can be effectively inhibited, the effect is better than that of the traditional RNN, and the calculation complexity is smaller than that of LSTM. Considering that the seismic wave development is associated with the seismic wave data at the previous moment, the current hidden layer state of the bidirectional gating circulation unit (BiGRU) is obtained by weighting and summing the forward hidden layer state and the reverse hidden layer state, the past seismic wave data, the future seismic wave data and the current seismic wave prediction can be linked, and the model prediction is facilitated. In addition, each IMF component decomposed by EMD contains local characteristic signals of seismic data at different time scales, so that the EMD-BiGRU model obtained by combining the EMD with the BiGRU model can achieve higher prediction accuracy. Moreover, the EMD algorithm and the BiGRU model have low requirements on hardware and low implementation cost.
Drawings
For the purposes of promoting a better understanding of the objects, aspects and advantages of the invention, reference will now be made to the following detailed description taken in conjunction with the accompanying drawings in which:
FIG. 1 is a flow chart of a preferred embodiment of a seismic destructive power prediction method based on a bidirectional gated cyclic unit according to the invention;
FIG. 2 is a schematic diagram of EMD decomposition;
FIG. 3 is a schematic diagram of the structure of the EMD-BiGRU model;
FIG. 4 is a schematic structural diagram of a single neuron architecture of a GRU;
fig. 5 is a block diagram of the earthquake destructive power prediction device based on the bidirectional gating cycle unit according to a preferred embodiment of the invention.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention in a schematic way, and the features in the following embodiments and examples may be combined with each other without conflict.
As shown in FIG. 1, a preferred embodiment of the seismic destructive power prediction method based on a bidirectional gated cyclic unit of the present invention comprises the following steps:
and step S1, acquiring seismic motion time-course data from the open-source seismic motion database, and establishing the database as sample data. For example: the Japanese K-NET database and the American PEER NGA database are open source earthquake motion databases and store massive earthquake motion data.
And step S2, carrying out destructive power analysis on the sample data, and dividing destructive power grades according to national standards to be used as sample labels. For example: the destructive power Analysis can be carried out on the sample data by a City-Scale online Time History Analysis (City-Scale NLTHA) method which is widely accepted in the field of civil engineering and has high precision.
Step S3, decomposing each item of data included in the seismic time-course data into a plurality of IMF (Intrinsic Mode Functions) components by an EMD (Empirical Mode Decomposition) algorithm. The earthquake motion time-range data comprises earthquake motion acceleration data, earthquake motion speed data, earthquake motion displacement data and other data.
The IMF component satisfies the following two conditions:
(1) in the whole time range of the function, the number of the extreme points is equal to or different from the number of the zero crossing points by 1;
(2) at any time point, the mean of the upper envelope and the lower envelope is 0.
The EMD algorithm is realized through a screening process; the non-linear data can be decomposed into a linear combination of a finite number of IMF components with frequencies from high to low by an EMD algorithm, and each decomposed IMF component contains local characteristic signals of different time scales of the original signal. The EMD algorithm specifically comprises:
step S11, obtaining an original signal x (t) according to one of the data in the earthquake time-course data, wherein t represents time; the original signal x (t) is a function of time t. For example, the seismic acceleration data may be decomposed first, and then the original signal x (t) may be obtained first according to the seismic acceleration data.
And step S32, screening the original signal x (t). The screening process is to subtract the average envelope function of the signal to obtain a new function; the method specifically comprises the following steps: finding out all maximum value points of the original signal x (t), and fitting an upper envelope line of the original signal x (t) by using a cubic spline function; finding out all minimum value points of the original signal x (t), and fitting a cubic spline function into a lower envelope curve of the original signal x (t); calculating the mean value of the upper envelope and the lower envelope to obtain a first mean envelope function m1(t); subtracting the first average envelope function m from the original signal x (t)1(t) obtaining a first intermediate component function d1,1(t)。
Step S33, determining the intermediate component function d1,1(t) whether two conditions for the IMF component are satisfied, and if so, d1,1(t) is denoted as the first IMF component IMF1(t) of the original signal; if not, continue to step S12 paird1,1(t) performing screening until the intermediate component function satisfies the condition of the IMF component. Assuming a medium component function d obtained after K screening1,k(t) if the IMF component is satisfied, d is1,k(t) is denoted as the first IMF component IMF1(t) of the original signal.
Step S34, subtracting the first IMF component IMF1(t) from the original signal x (t) to obtain a first residual component function r1(t); the first residual component function r1(t) continuing to decompose (i.e., decompose the IMF component from the signal by repeated sieving) according to steps S12 and S13 to obtain a second IMF component IMF2 (t); using a first residual component function r1(t) subtracting the second IMF component IMF2(t) to obtain a second residual component function r2(t) of (d). Continuing to apply the second residual component function r according to steps S12 and S132(t) carrying out decomposition; suppose that after n decompositions, the nth residual component function r is obtainedn(t) is a monotonic function, the decomposition is not continued, and the residual component function r is usedn(t) as residual amount RES; therefore, the decomposition of the seismic dynamic acceleration data is completed. The original signal can be represented as:
and S35, obtaining an original signal x (t) according to the earthquake vibration velocity data, the earthquake vibration displacement data and other earthquake motion time range data in sequence, returning to S12 to decompose the original signal x (t) formed by the data, and stopping decomposition until all the data included in the earthquake motion time range data are decomposed into IMF components and residual quantities RES. It should be noted that the time-lapse vibration time-course data may be decomposed from any one of the data included therein when decomposed, and a specific order is not required.
As shown in fig. 2, it is a schematic diagram of EMD decomposition of various items of data included in seismic motion time-course data. Compared with Fourier transform and wavelet decomposition, the EMD does not need to set a basis function, has self-adaptability, and can effectively eliminate the fluctuation of original data to make the data more stable.
And S4, defining a loss function of the BiGRU model, and training each IMF component obtained by decomposition as an input feature of the BiGRU model. The BiGRU, i.e., a two-way GRU (Gated recovery Units), is formed by combining two independent GRUs (i.e., a forward GRU and a backward GRU), has a function of capturing front and rear information characteristics, and is shown in fig. 3, which is a schematic structural diagram of an EMD-BiGRU model.
Wherein the loss function l (x, y) of the BiGRU model can be defined as:
where N represents the number of samples, xiRepresenting the actual seismic destructive power level, y, of the ith sampleiRepresenting the predicted seismic destructive power level of the ith sample model.
The input characteristics of the BiGRU model at any moment in time are that the values of each IMF component at that moment in time constitute a vector. Compared with the traditional RNN, the BiGRU does not clear previous information along with time, can effectively capture semantic association between long sequences, and relieves the phenomena of gradient disappearance or explosion. Considering that seismic wave development will have some correlation with seismic wave data at the last time, linking the current seismic wave data with the past seismic wave data in the prediction process will help the model to predict more. The working process of the BiGRU model comprises the following steps:
step S401, according to the input characteristic xtCalculating to obtain a hidden layer unit h of the GRUt. As shown in FIG. 4, the architecture of a single neural unit of a GRU includes a reset gate and an update gate, a hidden layer unit h of the GRUtCan be calculated by the following formula:
rt＝σ(Wrxt+Urht-1)
zt＝σ(Wzxt+Uzht-1)
wherein, the sigma is a Sigmoid function; tan h is a hyperbolic tangent function; r istTo reset the gate; z is a radical oftTo update the door;
candidate activation states for time t (i.e., the current time); h ist-1The hidden layer unit at the last moment (t-1 moment in forward transmission and t +1 moment in backward transmission); h istA hidden layer unit at the current time (namely, time t); wr、Ur、Wz、WzW, U is a training parameter matrix;
representing the multiplication of corresponding elements; x is the number oftAnd the input characteristic representing the time t moment is a vector formed by the values of all eigenmode function components of each item of data included in the seismic motion time-course data at the time t moment.
Assuming that when the seismic motion acceleration data is decomposed into four IMF components including IMF1(t), IMF2(t), IMF3(t) and IMF4(t), the seismic motion velocity data is decomposed into four IMF components including IMF1'(t), IMF2' (t), IMF3'(t) and IMF4' (t), and the seismic motion displacement data is decomposed into four IMF components including IMF1 "(t), IMF 2" (t), IMF3 "(t) and IMF 4" (t), … …; then at t0Time of day, x0Represented as a vector [ IMF1(t)0)，IMF2(t0)，IMF3(t0)，IMF4(t0)，IMF1'(t0)，IMF2'(t0)，IMF3'(t0)，IMF4'(t0)，IMF1”(t0)，IMF2”(t0)，IMF3”(t0)，IMF4”(t0)，……](ii) a At tiTime of day, xiRepresented as a vector [ IMF1(t)i)，IMF2(ti)，IMF3(ti)，IMF4(ti)，IMF1'(ti)，IMF2'(ti)，IMF3'(ti)，IMF4'(ti)，IMF1”(ti)，IMF2”(ti)，IMF3”(ti)，IMF4”(ti)，……]. Of course, the number of IMF components decomposed by the seismic displacement data, the seismic acceleration data and the seismic velocity data may be different.
Step S402, obtaining a forward propagation hidden layer state according to a hidden layer unit calculation formula of a GRU
The calculation formula is as follows:
wherein x istAn input feature representing time t;
representing a forward propagation hidden state at time (t-1);
representing the forward propagating hidden state at time t.
Step S403, obtaining a backward propagation hidden layer state according to a hidden layer unit calculation formula of GRU
The calculation formula is as follows:
wherein the content of the first and second substances,
representing a backward propagation hidden state at time t;
represents (t +1)The hidden state is propagated backwards in time.
Step S404, propagating the hidden layer state forward
And propagating hidden states backwards
Superposed and combined to obtain a hidden layer state Ot(ii) a The calculation formula is as follows:
wherein, OtRepresenting a hidden layer state at time t; alpha is alphatRepresenting the hidden layer output weight of the forward transmission GRU unit at the time t; beta is atRepresenting the output weight of the hidden layer of the backward transmission GRU unit at the time t; btAnd represents the offset corresponding to the hidden layer state at the time t.
Step S405, obtaining a prediction result Y according to the weight of each hidden layer state; the calculation formula is as follows:
wherein, wtA weight coefficient representing a hidden layer state; y represents the final output prediction result.
Step S5, in the training process of the BiGRU model, when the value of the loss function converges to a fixed value and keeps unchanged, the parameter of the BiGRU model at the moment is considered to be the optimal model parameter, and the model training is stopped.
And step S6, acquiring earthquake motion data in real time through a sensor or acquiring quasi-real-time earthquake motion data through an earthquake motion monitoring network, inputting the earthquake motion data into an EMD-BiGRU model, and predicting to obtain the earthquake destructive power grade through the EMD-BiGRU model.
In the embodiment, the BiGRU model is used as the basic model to predict the earthquake destructive power level, and the GRU can effectively inhibit gradient disappearance or explosion when capturing long sequence semantic association, so that the effect is better than that of the traditional RNN, and the calculation complexity is smaller than that of the LSTM. Considering that the seismic wave development is associated with the seismic wave data at the previous moment, the current hidden layer state of the bidirectional gating circulation unit (BiGRU) is obtained by weighting and summing the forward hidden layer state and the reverse hidden layer state, the past seismic wave data, the future seismic wave data and the current seismic wave prediction can be linked, and the model prediction is facilitated. In addition, each IMF component decomposed by EMD contains local characteristic signals of the original signal in different time scales, so that the EMD-BiGRU model obtained by combining the EMD and the BiGRU model can achieve higher prediction accuracy.
The invention also provides a seismic destructive power prediction device based on the bidirectional gating circulation unit, and as shown in fig. 5, a preferred embodiment of the seismic destructive power prediction device based on the bidirectional gating circulation unit comprises a seismic motion data acquisition module, an analysis module, a prediction data output module, an early warning module and a propagation module.
The earthquake motion data acquisition module can adopt a sensor module, and the sensor module is integrated with a sensor for acquiring data such as displacement, acceleration, speed and the like, so that earthquake motion data are acquired in real time; the seismic data acquisition module can also adopt a module which can be linked to a seismic monitoring network, so that quasi-real-time seismic data can be acquired through the seismic monitoring network.
The analysis module is used for analyzing the seismic motion data sent by the seismic motion data acquisition module and predicting the destructive power of the earthquake to obtain the destructive power grade of the earthquake. The analysis module comprises a data preprocessing unit, an EMD-BiGRU model and an output unit.
The data preprocessing unit is used for reading the earthquake motion data acquired by the earthquake motion data acquisition module and performing format conversion.
The EMD-BiGRU model comprises an empirical mode decomposition module and a BiGRU model, wherein the empirical mode decomposition module is used for decomposing seismic motion time-course data into a plurality of intrinsic mode function components through an empirical mode decomposition algorithm, and vectors formed by the intrinsic mode function components are used as input features of the BiGRU model; the BiGRU model is used for predicting the destructive power level of the earthquake according to the input features.
And the output unit is used for sending the predicted earthquake destructive power grade to the predicted data output module and the early warning module.
The prediction data output module is used for outputting the predicted earthquake destructive power grade and related data. The predictive data output module preferably employs a visualization module, such as a display, to visually output the seismic destructive power level and associated data.
The early warning module is used for comparing the earthquake destructive power grade predicted by the analysis module with a preset earthquake destructive power grade threshold value, and sending out an early warning signal when the predicted earthquake destructive power grade is greater than or equal to the earthquake destructive power grade threshold value. The propagation module is used for transmitting the early warning signal and the predicted earthquake destructive power grade to a rescue command center or other specific equipment in time when the early warning module sends the early warning signal, so that rescue work can be carried out in time conveniently.
In this embodiment, the analysis module adopts an EMD algorithm and a BiGRU model, and has low requirements on calculation and storage capabilities, low requirements on hardware, and low implementation cost.
Finally, the above embodiments are only intended to illustrate the technical solutions of the present invention and not to limit the present invention, and although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions, and all of them should be covered by the claims of the present invention.
Claims (9)
1. A seismic destructive power prediction method based on a bidirectional gating cycle unit is characterized by comprising the following steps:
s1, acquiring seismic motion time-course data from an open source seismic motion database, and establishing the database as sample data;
step S2, carrying out destructive power analysis on the sample data to divide destructive power grades, and taking the destructive power grades as sample labels;
step S3, decomposing the seismic motion time-course data into a plurality of intrinsic mode function components through an empirical mode decomposition algorithm;
step S4, defining a loss function of the BiGRU model, taking each intrinsic mode function component obtained by decomposition as an input characteristic of the BiGRU model, and training the intrinsic mode function component by combining a sample label;
step S5, when the value of the loss function of the BiGRU model converges to a fixed value and keeps unchanged, ending the training to obtain an EMD-BiGRU model;
and step S6, inputting seismic motion data acquired by a sensor or seismic motion data acquired through a seismic motion monitoring network into an EMD-BiGRU model, and predicting through the EMD-BiGRU model to obtain the seismic destructive power grade.
2. The method of claim 1, wherein the input features of the BiGRU model at any time are seismic time-course data or a vector formed by values of all eigenmode function components of each item of data included in the seismic data at that time.
3. The method for predicting seismic destructive power based on bidirectional gating cyclic unit of claim 1, wherein in the BiGRU model, the hidden layer unit h of GRUtThe calculation formula of (2) is as follows:
rt＝σ(Wrxt+Urht-1)
zt＝σ(Wzxt+Uzht-1)
wherein, the sigma is a Sigmoid function; tan h is a hyperbolic tangent function; r istTo reset the gate; z is a radical oftTo update the door;
is a candidate activation state at time t; h ist-1A hidden layer unit at the previous moment; h istA hidden layer unit at the current moment; wr、Ur、Wz、WzW, U is a training parameter matrix;
representing the multiplication of corresponding elements; x is the number oftRepresenting the input characteristics at time t.
4. The method for predicting seismic destructive power based on the bidirectional gating cyclic unit according to claim 3, wherein the calculation formula of the BiGRU model is as follows:
wherein x istAn input feature representing time t; (ii) a
To representBackward propagating hidden layer state at time t;
represents a backward propagation hidden state at time (t + 1);
representing a forward propagation hidden state at time (t-1);
representing a forward propagation hidden state at time t; o istRepresenting a hidden layer state at time t; alpha is alphatRepresenting the hidden layer output weight of the forward transmission GRU unit at the time t; beta is atRepresenting the output weight of the hidden layer of the backward transmission GRU unit at the time t; btRepresenting the offset corresponding to the hidden layer state at the time t; w is atA weight coefficient representing a hidden layer state; y represents the final output prediction result.
5. The method of claim 1, wherein the BiGRU model loss function i (x, y) is defined as:
where N represents the number of samples, xiRepresenting the actual seismic destructive power level, y, of the ith sampleiRepresenting the predicted seismic destructive power level of the ith sample model.
6. An earthquake destructive power prediction device based on a bidirectional gating cycle unit, comprising:
the earthquake motion data acquisition module is used for acquiring earthquake motion data in real time or acquiring the earthquake motion data through an earthquake motion monitoring network and transmitting the earthquake motion data to the analysis module;
the analysis module is used for analyzing the seismic motion data sent by the seismic motion data acquisition module and predicting the destructive power of the earthquake to obtain the destructive power grade of the earthquake; and
and the prediction data output module is used for outputting the destructive power level of the earthquake.
7. The earthquake destructive power prediction device based on the bidirectional gating cyclic unit as claimed in claim 6, further comprising an early warning module for comparing the level of the earthquake destructive power predicted by the analysis module with a preset earthquake destructive power level threshold value, and sending an early warning signal when the predicted earthquake destructive power level is greater than or equal to the earthquake destructive power level threshold value.
8. The apparatus of claim 7, further comprising a propagation module for transmitting the early warning signal and the predicted seismic destructive power level to a specific device when the early warning signal is emitted by the early warning module.
9. The apparatus of claim 6, wherein the analysis module comprises a data preprocessing unit, an EMD-BiGRU model and an output unit;
the data preprocessing unit is used for reading the seismic motion data acquired by the seismic motion data acquisition module and performing format conversion;
the EMD-BiGRU model comprises an empirical mode decomposition module and a BiGRU model, wherein the empirical mode decomposition module is used for decomposing seismic motion time-course data into a plurality of intrinsic mode function components through an empirical mode decomposition algorithm, and vectors formed by the intrinsic mode function components are used as input features of the BiGRU model; the BiGRU model is used for predicting the damage level of the earthquake according to the input characteristics;
and the output unit is used for sending the predicted earthquake destructive power grade to the predicted data output module and the early warning module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110927660.4A CN113537636A (en) | 2021-08-13 | 2021-08-13 | Earthquake destructive power prediction method and device based on bidirectional gating circulation unit |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110927660.4A CN113537636A (en) | 2021-08-13 | 2021-08-13 | Earthquake destructive power prediction method and device based on bidirectional gating circulation unit |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113537636A true CN113537636A (en) | 2021-10-22 |
Family
ID=78122421
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110927660.4A Pending CN113537636A (en) | 2021-08-13 | 2021-08-13 | Earthquake destructive power prediction method and device based on bidirectional gating circulation unit |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113537636A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108663712A (en) * | 2018-05-08 | 2018-10-16 | 四川大学 | A kind of Thin interbed method of side slope earthquake damage |
| CN110555784A (en) * | 2019-07-23 | 2019-12-10 | 中国地质大学（武汉） | wind power prediction method based on gated cyclic neural network wind speed correction |
| CN110780347A (en) * | 2019-11-22 | 2020-02-11 | 清华大学 | Earthquake destructive power prediction device and method based on cyclic neural network |
| CN111007569A (en) * | 2019-12-11 | 2020-04-14 | 中国人民解放军96901部队23分队 | Sample entropy threshold microseism signal noise reduction method integrated with empirical mode decomposition |
| CN111193256A (en) * | 2020-01-07 | 2020-05-22 | 上海电力大学 | Power load prediction method based on variational modal decomposition and gated cyclic unit |
| CN112949164A (en) * | 2021-01-27 | 2021-06-11 | 上海电机学院 | Lithium battery health state prediction method |
| CN113096378A (en) * | 2021-02-23 | 2021-07-09 | 北京航空航天大学 | Highway OD prediction method based on depth set empirical mode decomposition |
-
2021
- 2021-08-13 CN CN202110927660.4A patent/CN113537636A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108663712A (en) * | 2018-05-08 | 2018-10-16 | 四川大学 | A kind of Thin interbed method of side slope earthquake damage |
| CN110555784A (en) * | 2019-07-23 | 2019-12-10 | 中国地质大学（武汉） | wind power prediction method based on gated cyclic neural network wind speed correction |
| CN110780347A (en) * | 2019-11-22 | 2020-02-11 | 清华大学 | Earthquake destructive power prediction device and method based on cyclic neural network |
| CN111007569A (en) * | 2019-12-11 | 2020-04-14 | 中国人民解放军96901部队23分队 | Sample entropy threshold microseism signal noise reduction method integrated with empirical mode decomposition |
| CN111193256A (en) * | 2020-01-07 | 2020-05-22 | 上海电力大学 | Power load prediction method based on variational modal decomposition and gated cyclic unit |
| CN112949164A (en) * | 2021-01-27 | 2021-06-11 | 上海电机学院 | Lithium battery health state prediction method |
| CN113096378A (en) * | 2021-02-23 | 2021-07-09 | 北京航空航天大学 | Highway OD prediction method based on depth set empirical mode decomposition |
Non-Patent Citations (4)
| Title |
|---|
| BIAO ZHANG 等: "Network Security Situation Prediction Model Based on EMD and ELPSO Optimized BiGRU Neural Network", 《COMPUTATIONAL INTELLIGENCE AND NEUROSCIENCE》 * |
| QING ZHU等: "A hybrid VMD–BiGRU model for rubber futures time series forecasting", 《APPLIED SOFT COMPUTING》 * |
| 徐增光: "地震信号的时频分析方法研究及实际应用", 《中国优秀硕士学位论文全文数据库基础科学辑》 * |
| 石战战等: "一种基于EMD和GHT的储层识别方法", 《岩性油气藏》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mignan et al. | Neural network applications in earthquake prediction (1994–2019): Meta‐analytic and statistical insights on their limitations | |
| CN110738984A (en) | Artificial intelligence CNN, LSTM neural network speech recognition system | |
| Mousavi et al. | Developing deep neural network for damage detection of beam-like structures using dynamic response based on FE model and real healthy state | |
| CN111243682A (en) | Method, device, medium and apparatus for predicting toxicity of drug | |
| CN113988357A (en) | High-rise building wind-induced response prediction method and device based on deep learning | |
| Ren et al. | A $ T^{2} $-Tensor-Aided Multiscale Transformer for Remaining Useful Life Prediction in IIoT | |
| CN116451848A (en) | Satellite telemetry data prediction method and device based on space-time attention mechanism | |
| Sadhukhan et al. | Predicting the magnitude of an impending earthquake using deep learning techniques | |
| CN115795343A (en) | Method for analyzing state of unbalanced sample ammeter production equipment based on digital twin | |
| Zhang et al. | Deep convolutional neural network for microseismic signal detection and classification | |
| CN109190800A (en) | A kind of sea surface temperature prediction technique based on spark frame | |
| CN117216522A (en) | Bearing fault type diagnosis method, device and equipment based on machine learning | |
| CN113537637B (en) | Earthquake destructive power prediction method and device based on gated cycle unit | |
| CN113537636A (en) | Earthquake destructive power prediction method and device based on bidirectional gating circulation unit | |
| Fuentes et al. | Spatio-temporal seismicity prediction in Chile using a multi-column ConvLSTM | |
| Pei | Construction of a legal system of corporate social responsibility based on big data analysis technology | |
| CN114358024A (en) | Log analysis method, apparatus, device, medium, and program product | |
| CN113469504A (en) | Dynamic testing method, medium and system for water traffic risk | |
| Huang et al. | A hybrid FCN-BiGRU with transfer learning for low-velocity impact identification on aircraft structure | |
| CN111695989A (en) | Modeling method and platform of wind-control credit model | |
| CN111160715A (en) | BP neural network based new and old kinetic energy conversion performance evaluation method and device | |
| Kostomakhin et al. | The use of artificial intelligence algorithms to monitor the technical condition of energy-intensive tractors | |
| Kuo et al. | GNN-LSTM-based fusion model for structural dynamic responses prediction | |
| Крикун | Improving the Accuracy of the Neural Network Models Interpretation of Nonlinear Dynamic Objects | |
| Colace et al. | Unsupervised Learning Techniques for Vibration-Based Structural Health Monitoring Systems Driven by Data: A General Overview |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211022 |