CN112508124A - Gravel soil earthquake liquefaction discrimination method based on Bayesian network - Google Patents
Gravel soil earthquake liquefaction discrimination method based on Bayesian network Download PDFInfo
- Publication number
- CN112508124A CN112508124A CN202011529365.5A CN202011529365A CN112508124A CN 112508124 A CN112508124 A CN 112508124A CN 202011529365 A CN202011529365 A CN 202011529365A CN 112508124 A CN112508124 A CN 112508124A
- Authority
- CN
- China
- Prior art keywords
- liquefaction
- bayesian network
- gravel soil
- earthquake
- data
- 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
- 239000002689 soil Substances 0.000 title claims abstract description 91
- 238000012850 discrimination method Methods 0.000 title claims abstract description 12
- 238000000034 method Methods 0.000 claims abstract description 66
- 238000012360 testing method Methods 0.000 claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000011156 evaluation Methods 0.000 claims abstract description 15
- 230000035515 penetration Effects 0.000 claims description 12
- 230000001133 acceleration Effects 0.000 claims description 4
- 230000019771 cognition Effects 0.000 claims description 4
- 239000010419 fine particle Substances 0.000 claims description 4
- 239000003673 groundwater Substances 0.000 claims description 4
- 239000002245 particle Substances 0.000 claims description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 4
- 238000002790 cross-validation Methods 0.000 claims description 3
- 230000001502 supplementing effect Effects 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims 1
- 238000009933 burial Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 229920006395 saturated elastomer Polymers 0.000 description 3
- 239000004927 clay Substances 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000002265 prevention Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 206010012735 Diarrhoea Diseases 0.000 description 1
- 125000002015 acyclic group Chemical group 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004836 empirical method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011065 in-situ storage Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Algebra (AREA)
- Pure & Applied Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Mathematical Optimization (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Mathematical Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Investigation Of Foundation Soil And Reinforcement Of Foundation Soil By Compacting Or Drainage (AREA)
Abstract
The invention discloses a gravel soil earthquake liquefaction discrimination method based on a Bayesian network, which comprises the following steps: the method comprises the following steps: selecting key factors from a plurality of factors influencing the gravel soil earthquake liquefaction as a judgment index of the gravel soil earthquake liquefaction; step two: collecting historical data of the gravel soil earthquake liquefaction field according to the selected discrimination indexes, and dividing the data into a training set data set and a test set data set; step three: carrying out structure learning and parameter learning of the Bayesian network by using the training set data, generating a gravel soil earthquake liquefaction judging method, and constructing a dependency relationship between a gravel soil liquefaction judging index and a gravel soil liquefaction potential; step four: and verifying the obtained gravel soil earthquake liquefaction judging method based on the test set data, and obtaining the performance evaluation index of the method. The method solves the problems of insufficient precision, low applicability and the like of the existing discrimination method.
Description
Technical Field
The invention belongs to the technical field of geotechnical engineering seismic resistance, and particularly relates to a gravel soil liquefaction discrimination method based on a Bayesian network.
Background
The earthquake liquefaction is an important problem which is not negligible in the field of geotechnical engineering earthquake resistance, and means that saturated sandy soil, saturated light sub-clay with low clay content, saturated gravel soil with poor gradation and the like are likely to liquefy under the action of earthquake dynamic force, so that the soil strength is greatly suddenly lost. Earthquake liquefaction disasters occur in a plurality of historical earthquakes, such as Japanese New diarrhea earthquake in 1964, American Alaska earthquake, Wenchuan earthquake in 2008 in China, Indonesian Sulawesi earthquake in 2018 in China and the like, large-area earthquake liquefaction disasters occur, and serious life and property losses are caused. In view of this, the related research on the prevention and treatment of the earthquake liquefaction disaster is receiving continuous attention in the engineering field.
Earthquake liquefaction judgment is one of main liquefaction disaster prevention measures and is listed as a mandatory execution term of relevant specifications in the earthquake-resistant field. However, the factors influencing the seismic liquefaction are numerous, and extremely strong nonlinear relations exist among the factors, so that the complexity and uncertainty of seismic liquefaction judgment are determined. The traditional seismic liquefaction discrimination method is a semi-empirical method based on indoor tests and field in-situ tests, and the method is simple and easy to use, strong in interpretability, but insufficient in discrimination accuracy. With the development of machine learning technology, some experts and scholars propose to process the actual measurement data of the historical liquefaction disasters by means of a machine learning method so as to establish a corresponding earthquake liquefaction judging method. The method can obviously improve the discrimination precision, can consider the uncertainty in the discrimination process, and has wide application prospect. However, the advantages and disadvantages of various machine learning methods are different, and the establishment processes of some liquefaction discrimination methods are not scientific and transparent enough, so that further research and application are needed.
Traditional seismic liquefaction research focuses primarily on sandy soils, which have been considered impossible to undergo seismic liquefaction, and historical seismic liquefaction records relating to the sandy soils are rare. However, in 2008, the liquefaction phenomenon of large-area gravel soil appears in Wenchuan earthquake, and the inherent cognition that the gravel soil belongs to non-liquefiable safe soil is broken. At present, the method for discriminating the earthquake liquefaction of the gravel soil is still immature, and the triggering condition and the generation mechanism of the liquefaction of the gravel soil are also obviously different from those of the gravel soil, so that the method for discriminating the earthquake liquefaction of the gravel soil is urgent to propose and develop.
Disclosure of Invention
In order to respond to the actual requirement of liquefaction judgment of a gravel soil field in engineering, the problems of insufficient precision, low applicability and the like of the conventional judgment method are solved. The invention provides a Bayesian network model suitable for judging the seismic liquefaction of sandy soil, and details the establishment process and the theoretical basis of the method. The method achieves a better discrimination effect through verification.
In order to achieve the technical features, the invention is realized as follows: a gravel soil earthquake liquefaction discrimination method based on Bayesian network comprises the following steps:
the method comprises the following steps: selecting key factors from a plurality of factors influencing the gravel soil earthquake liquefaction as a judgment index of the gravel soil earthquake liquefaction;
step two: collecting historical data of the gravel soil earthquake liquefaction field according to the selected discrimination indexes, and dividing the data into a training set data set and a test set data set;
step three: carrying out structure learning and parameter learning of the Bayesian network by using the training set data, generating a gravel soil earthquake liquefaction judging method, and constructing a dependency relationship between a gravel soil liquefaction judging index and a gravel soil liquefaction potential;
step four: and verifying the obtained gravel soil earthquake liquefaction judging method based on the test set data, and obtaining the performance evaluation index of the method.
The Bayesian network modeling process comprises selection of key influence factors of gravel soil earthquake liquefaction, and structural learning and parameter learning of a Bayesian network model;
when a judgment index influencing the gravel soil earthquake liquefaction is selected, a plurality of key factors are preferably selected from three influence factors of earthquake information, soil information and site conditions.
The key factors are 13, including: magnitude MwEarthquake center distance R, earthquake duration t, peak acceleration PGA, gravel content GC, fine particle content FC, average particle diameter D50Correction ofNumber of power penetration hammering N1′20Correcting shear wave velocity Vs1Overlying effective soil pressure σv', depth of groundwater burial DwThickness H of overlying impervious layernThickness D of unsaturated soil layer between underground water level and overlying impervious layern。
Modified power penetration hammering number N among key factors1′20And correcting shear wave velocity Vs1And (3) not considering simultaneously, respectively adopting power penetration test data and shear wave velocity data to construct different 12-factor Bayesian network models.
Discretizing the training set data set and the test set data set collected in the step two by regions, retrieving data correlation among a plurality of selected key factor variables by using a Maximum Information Coefficient (MIC) method, and considering that a dependency relationship exists between the two variables when the mutual information value between the two variables reaches 0.9 times of the maximum mutual information value;
based on the detected variable dependency relationship, correcting and supplementing the dependency relationship among the input nodes by combining professional knowledge, so that the dependency relationship among all the influence factors expressed by the network structure is more consistent with professional cognition;
then constructing to obtain an initial structure of the Bayesian network as a prior node sequence of a K2 structural algorithm;
and on the basis of the obtained initial structure, learning from the training data set by using a K2 structural algorithm to obtain an optimal Bayesian network structure, thereby obtaining a final Bayesian network structure.
Aiming at the fact that a large amount of data are missing in a training set data set, parameter learning of the Bayesian network is conducted through a maximum Expectation (EM) algorithm, conditional probability table parameters of the Bayesian network are obtained, and then the data are divided into the training set and a testing set through a K-fold cross-validation method and used for repeated experiments.
And calculating 7 individual performance evaluation indexes of the model obtained by using the test set data, including the overall accuracy OA, the recall rate Rec, the accuracy Pre, the F1 value, the AUC value, the mean absolute error MAE and the root mean square error RMSE, and comparing the 7 individual performance evaluation indexes with the existing method.
The invention has the following beneficial effects:
1. the gravel soil earthquake liquefaction judging method provided by the invention better solves the problem that the existing gravel soil earthquake liquefaction judging method is insufficient, and further improves the effectiveness and accuracy of gravel soil earthquake liquefaction judgment.
2. The gravel soil earthquake liquefaction judging method provided by the invention adopts 15 historical gravel soil liquefaction survey data in different earthquakes, so that the application range of the method is greatly expanded, and the generalization capability of the method is enhanced.
3. The invention provides a method for constructing a gravel soil seismic liquefaction discrimination method by utilizing a Bayesian network, wherein the Bayesian network is a directed acyclic network graph model based on probability theory and graph theory, has a solid mathematical principle, can better handle uncertainty inference and complex nonlinear problems, and has obvious advantages in the field of seismic liquefaction discrimination.
4. The invention provides a Bayesian network structure for gravel soil seismic liquefaction judgment, which is constructed by utilizing a maximum information coefficient method, a K2 algorithm and professional knowledge, can well reduce subjective factors, and fully excavate objective relations in historical data.
Drawings
The invention is further illustrated by the following figures and examples.
FIG. 1 is a technical scheme of the present invention.
FIG. 2 is a Bayesian network structure for identifying the seismic liquefaction of the gravelly soil based on the maximum information coefficient method and the K2 algorithm.
FIG. 3 shows the performance evaluation index of the method for discriminating the seismic liquefaction of gravelly soil according to the present invention.
Detailed Description
Embodiments of the present invention will be further described with reference to the accompanying drawings.
Example 1:
as shown in fig. 1-3, a method for discriminating a gravel soil earthquake liquefaction based on a bayesian network comprises the following steps:
the method comprises the following steps: selecting key factors from a plurality of factors influencing the gravel soil earthquake liquefaction as a judgment index of the gravel soil earthquake liquefaction;
step two: collecting historical data of the gravel soil earthquake liquefaction field according to the selected discrimination indexes, and dividing the data into a training set data set and a test set data set;
step three: carrying out structure learning and parameter learning of the Bayesian network by using the training set data, generating a gravel soil earthquake liquefaction judging method, and constructing a dependency relationship between a gravel soil liquefaction judging index and a gravel soil liquefaction potential;
step four: and verifying the obtained gravel soil earthquake liquefaction judging method based on the test set data, and obtaining the performance evaluation index of the method.
Further, the Bayesian network modeling process comprises selection of key influence factors of gravel soil earthquake liquefaction, and structure learning and parameter learning of a Bayesian network model;
furthermore, when a judgment index influencing the gravel soil earthquake liquefaction is selected, a plurality of key factors are preferably selected from three types of influencing factors, namely earthquake information, soil information and site conditions.
Further, the key factors are 13, including: magnitude MwEarthquake center distance R, earthquake duration t, peak acceleration PGA, gravel content GC, fine particle content FC, average particle diameter D50Correcting the power penetration hammering number N1′20Correcting shear wave velocity Vs1Overlying effective soil pressure σv', depth of groundwater burial DwThickness H of overlying impervious layernThickness D of unsaturated soil layer between underground water level and overlying impervious layern。
Further, the modified power penetration hammering number N among the key factors1′20And correcting shear wave velocity Vs1And (3) not considering simultaneously, respectively adopting power penetration test data and shear wave velocity data to construct different 12-factor Bayesian network models.
Further, discretizing the training set data set and the test set data set collected in the step two by regions, retrieving data correlation among the selected multiple key factor variables by using a Maximum Information Coefficient (MIC) method, and considering that a dependency relationship exists between the two variables when the mutual information value between the two variables reaches 0.9 times of the maximum mutual information value;
furthermore, based on the detected variable dependency relationship, the dependency relationship between the input nodes is corrected and supplemented by combining with professional knowledge, so that the dependency relationship between all the influencing factors represented by the network structure is more consistent with professional cognition; then constructing to obtain an initial structure of the Bayesian network as a prior node sequence of a K2 structural algorithm; and on the basis of the obtained initial structure, learning from the training data set by using a K2 structural algorithm to obtain an optimal Bayesian network structure, thereby obtaining a final Bayesian network structure.
Further, aiming at more data loss in the training set data set, the maximum Expectation (EM) algorithm is utilized to learn parameters of the Bayesian network, conditional probability table parameters of the Bayesian network are obtained, and then the data are divided into the training set and the testing set by utilizing a K-fold cross-validation method for repeated experiments.
Further, 7 individual performance evaluation indexes of the model obtained by using the test set data are calculated, wherein the evaluation indexes comprise the overall precision OA, the recall rate Rec, the accuracy Pre, the F1 value, the AUC value, the mean absolute error MAE and the root mean square error RMSE, and the evaluation indexes are compared with the existing method.
Example 2:
the embodiment 1 of the invention provides a gravel soil earthquake liquefaction judging method based on a Bayesian network, which comprises the following specific steps as shown in a technical route diagram of FIG. 1:
205 groups of historical earthquake liquefaction survey data based on dynamic sounding test, and 12 gravel soil earthquake liquefaction discrimination indexes (seismic level M) are consideredwEarthquake center distance R, earthquake duration t, peak acceleration PGA, gravel content GC, fine particle content FC, average particle diameter D50Correcting the power penetration hammering number N1′20Overlying effective soil pressure σv', depth of groundwater burial DwThickness H of overlying impervious layernUnderground water level and overlying impervious layerThickness of unsaturated soil layer D betweenn) Constructing a Bayesian network method for judging the seismic liquefaction of the gravel soil by taking the 12 judgment indexes as input nodes and the seismic liquefaction potential LP of the gravel soil as output nodes;
and constructing an initial Bayesian network structure based on a maximum information coefficient method and professional knowledge. First, considering the network connection structure between the above-mentioned 12 input nodes, a mutual information value MIC (x, y) is calculated between every 2 input nodes and a mutual information value matrix is formed, and a maximum value maxmic (x) of each row and a maximum value maxmic (y) of each column in the matrix are marked. The larger the mutual information value, the stronger the correlation between the two variables, and when the mutual information value between 2 variables is greater than the maximum mutual information value of α times (i.e., MIC (x, y) ≧ α. maxMIC (X) or MIC (y, x) ≧ α. maxMIC (Y)), a dependency relationship is considered to exist between them, where α is 0.9. And correcting the dependency relationship between the supplementary variables by professional knowledge, determining the connection line between the network nodes, and indicating the direction of an arrow of the connection line, thereby preliminarily obtaining the network initial structure between the input nodes. And then determining a network connection structure between the input nodes and the output nodes, wherein although all the input nodes have dependency relations with the output nodes, part of the parent nodes and the output nodes obtained in the previous step may be indirectly dependent, so that a conditional mutual information value MIC (f, r | c) between the parent node f and the output node r under the condition of a given child node c is calculated by using a conditional maximum information coefficient method, and when the conditional mutual information value is greater than a beta value (namely MIC (f, r | c) ≧ beta), the network connection structure is considered to be directly dependent between the parent node and the output nodes, and the beta value is taken as an average value of all the conditional mutual information values. Based on the steps, an initial Bayesian network structure for judging the seismic liquefaction of the gravel soil is obtained, and prior node order, the maximum father node number and independent constraint among the nodes are provided for the K2 algorithm.
Since the K2 algorithm and the Bayesian network model need discrete data, the discrimination index and the sample data set are dispersed according to value intervals, wherein t, PGA, GC and N are1′20,Vs1,σ′V,Dw,Hn,DnIs divided into 4-5 intervals according to the equidistance standard, R and D50Then is based onDividing the test sample into 4 intervals according to professional knowledge;
and (3) fully mining implicit information in the data by using a K2 algorithm, and further optimizing the Bayesian network structure. The total number of nodes n is specified as 13, variable names represented by the nodes are defined, and then prior node orders and node independent constraints are provided for the K2 algorithm based on the initial bayesian network structure, wherein the nodes in the order that are arranged in front may be parents of the nodes that follow, but the nodes that follow cannot be parents of the nodes that follow. Next, the maximum connection number 10 of a single node in the initial bayesian network is designated as the maximum parent node number limit in the K2 algorithm, and finally, 205 sets of dynamic sounding historical seismic liquefaction survey data are read for network structure learning of the K2 algorithm. Finally, a Bayesian network structure of the gravel soil seismic liquefaction discrimination method is obtained, as shown in FIG. 2.
And after the Bayesian network structure is obtained, parameter learning is carried out to obtain a conditional probability table of the Bayesian network. And (4) adopting a maximum expectation algorithm to learn parameters because more data are missing in the data. In addition, the data set is fully utilized to train and test the model by adopting a 5-fold cross test method in consideration of the small amount of the gravel soil seismic liquefaction data. The 205 sets of data were equally divided into 5 arbitrary portions, and the number of liquefied samples and the number of non-liquefied samples in each data were set to be close to each other in order to reduce sampling variation. The experiment was then repeated 5 times, each time taking 1 different set of data as the test set and the other sets of data as the training set.
After the 5 times of tests, 5 models are obtained, and for the result of each model, the evaluation indexes of the models are calculated, such as the total accuracy (OA) represents the proportion of the number of samples predicted to be correct by the model to the total number of samples, the accuracy (Pre) represents the proportion of the number of samples predicted to be liquefied and predicted to be correct to the number of samples predicted to be liquefied, the recall rate (Rec) represents the proportion of the number of samples predicted to be liquefied and predicted to be correct to the actual number of liquefied samples, F1The values are the harmonic mean of Pre and Rec, and the area under the subject curve (AUC) can be used to determine the optimal liquefaction potential cut-off to discriminate liquefaction. In addition, two indexes for measuring the prediction deviation of the model are a prediction error average value MAE and a prediction error index respectivelyAlignment difference RMSE. And then, averaging the evaluation indexes of the 5 models to obtain a performance evaluation index Macro value (Macro-) of the gravel soil earthquake liquefaction discrimination method based on the dynamic penetration test. The prediction performance index of the method is shown in figure 3, the Macro total precision Macro-OA of the method is 92.2 percent, and the Macro accuracy Macro-Pre, the Macro recall Macro-Rec and the Macro-F1The values and the Macro-AUC values are high, which indicates that the classification performance of the model is good, the prediction error Macro-average value Macro-MAE and the prediction error Macro-standard deviation Macro-RMSE are small, and indicates that the prediction deviation of the model is small. In conclusion, the method has excellent performance and has wider application prospect in the field of judgment of the earthquake liquefaction of the gravel soil.
Example 3:
the embodiment 2 of the invention provides a gravel soil earthquake liquefaction judging method based on a Bayesian network, which is basically the same as the embodiment 1, except that the corrected power penetration hammering number N in the judging index1′20Substituted by modified shear wave velocity Vs1And structure and parameter learning of the model is performed using 205 sets of shear wave velocity historical survey data.
Claims (7)
1. A gravel soil earthquake liquefaction discrimination method based on Bayesian network is characterized by comprising the following steps:
the method comprises the following steps: selecting key factors from a plurality of factors influencing the gravel soil earthquake liquefaction as a judgment index of the gravel soil earthquake liquefaction;
step two: collecting historical data of the gravel soil earthquake liquefaction field according to the selected discrimination indexes, and dividing the data into a training set data set and a test set data set;
step three: carrying out structure learning and parameter learning of the Bayesian network by using the training set data, generating a gravel soil earthquake liquefaction judging method, and constructing a dependency relationship between a gravel soil liquefaction judging index and a gravel soil liquefaction potential;
step four: and verifying the obtained gravel soil earthquake liquefaction judging method based on the test set data, and obtaining the performance evaluation index of the method.
2. The method for discriminating the gravel soil earthquake liquefaction based on the Bayesian network as claimed in claim 1, wherein the Bayesian network modeling process comprises selection of key influence factors of the gravel soil earthquake liquefaction and structural learning and parameter learning of a Bayesian network model;
when a judgment index influencing the gravel soil earthquake liquefaction is selected, a plurality of key factors are preferably selected from three influence factors of earthquake information, soil information and site conditions.
3. The method for discriminating gravel soil seismic liquefaction based on the Bayesian network as claimed in claim 2, wherein the number of the key factors is 13, and the method comprises the following steps: magnitude MwEarthquake center distance R, earthquake duration t, peak acceleration PGA, gravel content GC, fine particle content FC, average particle diameter D50Corrected power penetration number N'120Correcting shear wave velocity Vs1Overlying effective soil pressure σ'vBuried depth of ground water DwThickness H of overlying impervious layernThickness D of unsaturated soil layer between underground water level and overlying impervious layern。
4. The method for discriminating gravel soil seismic liquefaction based on Bayes network as claimed in claim 3, wherein the corrected power penetration hammering number N 'is selected from the key factors'120And correcting shear wave velocity Vs1And (3) not considering simultaneously, respectively adopting power penetration test data and shear wave velocity data to construct different 12-factor Bayesian network models.
5. The method for discriminating sandy soil earthquake liquefaction based on the Bayesian network according to claim 1, wherein the training set data set and the test set data set collected in the second step are discretized in regions, and a Maximum Information Coefficient (MIC) method is used for retrieving data correlation among a plurality of selected key factor variables, and when the mutual information value between two variables reaches a maximum mutual information value which is 0.9 times that between the two variables, the two variables are considered to have a dependency relationship;
based on the detected variable dependency relationship, correcting and supplementing the dependency relationship among the input nodes by combining professional knowledge, so that the dependency relationship among all the influence factors expressed by the network structure is more consistent with professional cognition;
then constructing to obtain an initial structure of the Bayesian network as a prior node sequence of a K2 structural algorithm;
and on the basis of the obtained initial structure, learning from the training data set by using a K2 structural algorithm to obtain an optimal Bayesian network structure, thereby obtaining a final Bayesian network structure.
6. The method for discriminating the sandy soil earthquake liquefaction based on the Bayesian network as claimed in claim 5, wherein the parameters of the Bayesian network are learned by using a maximum Expectation (EM) algorithm aiming at more data loss in the training set data set, so as to obtain the conditional probability table parameters of the Bayesian network, and then the data are divided into the training set and the test set by using a K-fold cross-validation method for repeated tests.
7. The method for discriminating the sandy soil earthquake liquefaction based on the Bayesian network as claimed in claim 1, wherein 7 performance evaluation indexes of the model obtained by calculation of the test set data are compared with the existing method, wherein the evaluation indexes comprise an overall accuracy OA, a recall rate Rec, an accuracy Pre, an F1 value, an AUC value, an average absolute error MAE and a root mean square error RMSE.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011529365.5A CN112508124A (en) | 2020-12-22 | 2020-12-22 | Gravel soil earthquake liquefaction discrimination method based on Bayesian network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011529365.5A CN112508124A (en) | 2020-12-22 | 2020-12-22 | Gravel soil earthquake liquefaction discrimination method based on Bayesian network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112508124A true CN112508124A (en) | 2021-03-16 |
Family
ID=74922016
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011529365.5A Pending CN112508124A (en) | 2020-12-22 | 2020-12-22 | Gravel soil earthquake liquefaction discrimination method based on Bayesian network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112508124A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113379105A (en) * | 2021-05-21 | 2021-09-10 | 浙江大学 | Site earthquake liquefaction disaster rapid estimation method based on artificial intelligence |
| CN114154332A (en) * | 2021-12-06 | 2022-03-08 | 西南石油大学 | Gravel soil earthquake liquefaction evaluation method |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106980902A (en) * | 2017-04-05 | 2017-07-25 | 广州中国科学院工业技术研究院 | Liquefied gas storage accident prediction system and method |
-
2020
- 2020-12-22 CN CN202011529365.5A patent/CN112508124A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106980902A (en) * | 2017-04-05 | 2017-07-25 | 广州中国科学院工业技术研究院 | Liquefied gas storage accident prediction system and method |
Non-Patent Citations (3)
| Title |
|---|
| 曹振中: "汶川地震液化特征及砂砾土液化预测方法研究", 《中国优秀博硕士学位论文全文数据库(博士)基础科学辑》 * |
| 曾千千等: "基于最大信息系数的贝叶斯网络结构学习算法", 《计算机工程》 * |
| 胡记磊: "基于贝叶斯网络的地震液化风险分析模型研究", 《中国优秀博硕士学位论文全文数据库(博士)工程科技Ⅱ辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113379105A (en) * | 2021-05-21 | 2021-09-10 | 浙江大学 | Site earthquake liquefaction disaster rapid estimation method based on artificial intelligence |
| CN113379105B (en) * | 2021-05-21 | 2022-05-31 | 浙江大学 | Site earthquake liquefaction disaster rapid evaluation method based on artificial intelligence |
| CN114154332A (en) * | 2021-12-06 | 2022-03-08 | 西南石油大学 | Gravel soil earthquake liquefaction evaluation method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hanna et al. | Neural network model for liquefaction potential in soil deposits using Turkey and Taiwan earthquake data | |
| CN108256141B (en) | Main and aftershock joint vulnerability analysis method based on Copula theory | |
| CN101266299B (en) | Method for forecasting oil gas utilizing earthquake data object constructional features | |
| CN106407493A (en) | Multi-dimensional Gaussian cloud model-based rock burst grade evaluation method | |
| CN112508124A (en) | Gravel soil earthquake liquefaction discrimination method based on Bayesian network | |
| CN111476472A (en) | Sulfur-iron mine geological environment evaluation method | |
| CN114036841A (en) | Landslide incidence prediction method and system based on semi-supervised support vector machine model | |
| CN114821976B (en) | Intelligent forecasting system for multi-element karst collapse | |
| Shebalin et al. | Long-delayed aftershocks in New Zealand and the 2016 M7. 8 Kaikoura earthquake | |
| Licata et al. | A multi-level study for the seismic microzonation of the Western area of Naples (Italy) | |
| McBirney et al. | Volcanic and seismic hazards at a proposed nuclear power site in central Java | |
| Sandri et al. | The need to quantify hazard related to non-magmatic unrest: from BET_EF to BET_UNREST | |
| Handrianto et al. | C. 45 Algorithm for classification of causes of landslides | |
| Kamura et al. | Exploring the possibility of assessing the damage degree of liquefaction based only on seismic records by artificial neural networks | |
| Zeng et al. | Local coupled Markov chain model for simulating varied stratigraphy | |
| Bayraktarli | Application of Bayesian probabilistic networks for liquefaction of soil | |
| Torres et al. | An interpretable machine learning approach in understanding lateral spreading case histories | |
| Du et al. | Development of an automatic tracking model for seismic stratigraphic correlation based on information entropy theory | |
| Vinciguerra et al. | Identifying and discriminating seismic patterns leading flank eruptions at Mt. Etna Volcano during 1981–1996 | |
| Liu et al. | Liquefaction evaluation based on shear wave velocity using random forest | |
| CN116797755B (en) | Modeling method for multi-time-space three-dimensional geological structure of mixed rock zone | |
| Zhan | Data-Driven Assessment of Site Responses at Liquefiable Sites | |
| Al Bodour et al. | A model to predict the standard penetration test N60 value from cone penetration test data | |
| Belhadj et al. | Towards Seismic Damage Assessment Model for Soil Structure Interaction-based Building | |
| CN116502895B (en) | Open cut highway tunnel adjacent subway engineering collaborative construction risk coupling analysis method |
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 |
Application publication date: 20210316 |
|
| RJ01 | Rejection of invention patent application after publication |