Earthquake slope instability probability intelligent prediction method based on coupling of physical mechanism and neural network model

Abstract

The invention relates to the technical field of intelligent prediction of geological disasters, and particularly discloses an intelligent prediction method for an earthquake slope instability probability based on coupling of a physical mechanism and a neural network model. Seismic oscillation data, side slope geometric parameters and stability states are collected and integrated, and a database based on a physical mechanism is constructed; through a database, extracting seismic oscillation frequency spectrum features, pulse features, time domain / time frequency features of slope displacement, acceleration, stress and the like and accumulated displacement values; determining a slope instability index and a threshold value based on a database case; a neural network method based on Bayesian optimization is adopted to establish a slope instability probability intelligent prediction model, seismic slope instability probability intelligent prediction is realized, and a theoretical basis and technical support are provided for prevention and mitigation of seismic slope disasters.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent prediction of geological disasters, and in particular to an intelligent prediction method for earthquake slope instability probability based on coupling of physical mechanism and neural network model. Background Art

[0002] Landslides caused by earthquakes are a serious form of geological disasters, posing a major threat to human life safety, infrastructure and the ecological environment. Slope stability analysis is the core content of landslide hazard prediction and prevention, and its accuracy and timeliness directly affect the scientific nature of earthquake disaster emergency decision-making.

[0003] Traditional methods mainly evaluate stability based on physical models, but they have certain complexities and limitations, as follows: At present, most studies on slope stability are based on physical models (such as limit equilibrium method, finite element method, etc.) for mechanical analysis, which requires a large amount of detailed slope data (such as geometric characteristics, geotechnical parameters) to support. However, the dynamic response of landslide instability under strong earthquakes is complex, and these methods are not applicable enough when facing large-scale regional assessments or lack of high-resolution data; landslide prediction relies on multi-source data (seismic motion, topography, geology, monitoring, etc.), which have significant differences in temporal and spatial distribution. For example, seismic data are mainly based on time domain / frequency domain features, while topography and geometric parameters are static spatial data. This heterogeneity increases the difficulty of fusion and feature extraction of multidimensional information, and traditional prediction models are difficult to fully utilize this information.

[0004] With the development of artificial intelligence technology, machine learning models have begun to show significant advantages in slope hazard prediction, especially in the processing of large-scale data. However, these data-driven models usually lack an understanding of physical mechanisms and the prediction results are not interpretable enough, which affects their credibility and universality in practical applications.

[0005] Therefore, a new method suitable for intelligent monitoring, risk assessment and earthquake disaster prevention and control of slope hazards is urgently needed. Summary of the invention

[0006] To solve the problems existing in the prior art, the present invention provides an intelligent prediction method for the probability of earthquake slope instability based on the coupling of physical mechanism and neural network model. Through physical mechanism modeling, data-driven feature extraction and Bayesian optimized neural network modeling technology, it provides a theoretical basis and technical support for the prediction of earthquake-induced landslides. It is suitable for intelligent monitoring, risk assessment and earthquake disaster prevention and control of slope disasters, and solves the problems mentioned in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: an intelligent prediction method for earthquake slope instability probability based on physical mechanism and neural network model coupling, comprising the following steps:

[0008] S1. Database construction based on earthquake slope stability analysis;

[0009] S2, ground motion and slope response data feature extraction;

[0010] S3. Determination of earthquake slope stability evaluation index and instability threshold;

[0011] S4, data preprocessing and eigenvalue selection;

[0012] S5. Establishment of intelligent prediction model for earthquake slope instability probability;

[0013] S6. Intelligent prediction of earthquake slope instability probability.

[0014] Preferably, in step S1, the following is specifically included:

[0015] S11. Collect earthquake motion data worldwide, including acceleration time history, conduct earthquake motion characteristic analysis, and obtain information on earthquake motion peak acceleration and velocity, Arias intensity, frequency, epicenter distance, and focal depth;

[0016] S12. Collect relevant data for seismic slope stability analysis based on numerical simulation and shaking table model test methods, and obtain slope response characteristic data including slope geometry, geological conditions, load, acceleration, displacement, and stress distribution;

[0017] S13. Collect real-time monitoring data based on the on-site slope, including displacement and strain data;

[0018] S14. Integrate the above data to build a database for earthquake slope stability analysis, and mark the slope stability state corresponding to each set of data, that is, mark stable as 0 and unstable as 1.

[0019] Preferably, in step S2, data feature extraction is performed based on the constructed database, specifically including the following:

[0020] S21. Characteristics of seismic data: Fourier transform FFT is used to extract the frequency domain characteristics of seismic motion and slope response, and spectrum analysis is performed to calculate the energy distribution of different frequency bands. Based on the improved M&P wave method, the seismic pulse components are identified and extracted, and the pulse energy rate is calculated using the time-frequency characteristics of seismic motion, and the pulse effect characteristics are analyzed. The calculation formula for pulse seismic identification is:

[0021] E R =-(2.68+0.1416×PGVp -0.0162×ER p +0.0002852×PGV p 2

[0022] -0.005476×ER p ×PGV p +0.0005197×ER p 2 )

[0023]

[0024] Among them, E R is the pulse index; PGVp is the pulse peak velocity; ERp is the pulse energy rate; when the pulse index E R When it is greater than 0, the earthquake motion is a pulse earthquake motion; t s and t e and represent the start time and end time of the velocity pulse respectively; E(t) represents the accumulated energy of the earthquake motion at time t;

[0025] S22. Extraction of slope response characteristics from numerical simulation and model test: Extract time domain characteristics from displacement, acceleration and stress curves, including maximum value, root mean square (RMS) and standard deviation; extract the cumulative displacement of the slope under earthquake action;

[0026] S23. Monitoring data feature extraction: Calculate the change rate from the displacement and acceleration time series data, including the displacement change rate and the velocity change rate.

[0027] Preferably, in step S3, the following is specifically included:

[0028] Through statistical analysis of stable and unstable samples in the database and slope response characteristics, the displacement critical value under specific seismic conditions is calculated to determine the displacement threshold; the seismic acceleration time history is extracted to calculate the slope critical acceleration, the relationship between the acceleration value and slope instability is analyzed to determine the slope critical acceleration; the stress distribution is extracted to calculate the critical stress value;

[0029] Based on the index thresholds of displacement, acceleration and stress, a multidimensional index system of slope instability is defined. When any stability index reaches or exceeds the instability threshold, the slope is considered to have reached a critical state of instability; when all indicators are lower than their thresholds, the slope is in a stable state.

[0030] Preferably, the expression for judging the indicator threshold is as follows:

[0031]

[0032] Among them, State represents the slope stability state, which takes the value of "1" or "0"; D represents the cumulative displacement; A represents the maximum acceleration; S represents the maximum stress; D crit is the instability threshold of cumulative displacement; A crit is the acceleration instability threshold; S crit is the instability threshold of stress.

[0033] Preferably, in step S4, the following is specifically included:

[0034] S41, according to the database constructed in step S1 and the data features extracted in step S2, a large number of slope characteristic parameters and ground motion intensity parameters are obtained, and these parameters are used as influencing factors to perform standardized data preprocessing to ensure that the numerical ranges between the data features are consistent;

[0035] The slope characteristic parameters include slope height, slope gradient, and rock and soil strength; the seismic intensity parameters include peak value, Arias intensity, and pulse energy rate;

[0036] S42, using the slope instability multidimensional index system and threshold judgment expression established in step S3, determine the stability state, that is, stability is 0 and instability is 1, and use the stability state as the target value / label;

[0037] S43. Using correlation analysis, retain the factors whose correlation with other influencing factors is less than 0.8, and then evaluate the importance of these retained influencing factors based on the random forest algorithm. Remove the influencing factors with an importance less than 0.1 or the lowest importance relative to others, and use the remaining influencing factors as the eigenvalues ​​required for model establishment.

[0038] Preferably, the calculation formula of the impact factor importance is as follows:

[0039]

[0040] Among them, Importance(X i ) is the impact factor X i The importance of N is the number of influencing factors, G is split is the gain contributed by the factor, and n is the nth influencing factor.

[0041] Preferably, in step S5, the feature values ​​and target values / labels in step S4 are used as a data set, and the data set is randomly divided into 5 parts using 5-fold cross validation, 4 of which are selected as training sets and 1 as a test set, and the training set is brought into a convolutional neural network (CNN) algorithm to build a prediction model;

[0042] Introducing Bayesian optimization into the algorithm to automatically adjust the model's hyperparameters, including convolution kernel size, number of convolution kernels, pooling kernel size, pooling method, number of neurons, number of layers, learning rate, and L2 regularization parameter;

[0043] Model testing: Each test uses accuracy, recall, and AUC curves to evaluate the stability and generalization ability of the model on the test set. Finally, the average value of each test is taken to obtain the trained model.

[0044] Preferably, in step S6, the closer the slope instability probability value is to 1, the more likely the slope is to become unstable, and the closer it is to 0, the more stable the slope is.

[0045] The beneficial effects of the present invention are as follows: the present invention constructs a database driven by physical mechanisms through numerical simulation, indoor model tests and slope monitoring; extracts seismic spectrum characteristics, time domain / time-frequency characteristics, slope displacement and stress accumulation characteristics through database cases, and realizes accurate prediction of slope stability by combining a neural network model based on Bayesian optimization. While integrating the advantages of physical models with the flexibility of data-driven methods, the method of the present invention significantly improves the reliability and scientificity of landslide prediction, providing technical support for the prevention and mitigation of earthquake disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic flow chart of an intelligent prediction method for earthquake slope instability probability based on coupling of physical mechanism and neural network model in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] The present invention provides a technical solution: an intelligent prediction method for earthquake slope instability probability based on the coupling of physical mechanism and neural network model, such as Figure 1 As shown, the following steps are included:

[0049] S1. Construction of database based on earthquake slope stability analysis.

[0050] In step S1, the specific steps include:

[0051] S11. Collect earthquake data (acceleration time history) worldwide, analyze earthquake characteristics, and obtain information including peak value (earthquake peak acceleration and velocity), intensity (Arias intensity), frequency, epicenter distance, and focal depth;

[0052] S12. Collect relevant data for seismic slope stability analysis based on numerical simulation and shaking table model test methods, and obtain slope response characteristic data including slope geometry, geological conditions, load, acceleration, displacement, and stress distribution;

[0053] S13. Collect real-time monitoring data based on the on-site slope, including displacement and strain data;

[0054] S14. Integrate the above data to build a database for earthquake slope stability analysis, and mark the slope stability state corresponding to each set of data (stable is marked as 0, unstable is marked as 1).

[0055] S2, ground motion and slope response data feature extraction;

[0056] In step S2, data feature extraction is performed based on the constructed database, specifically including the following:

[0057] S21. Characteristics of seismic data: Fourier transform FFT is used to extract the frequency domain characteristics of seismic motion and slope response, and spectrum analysis is performed to calculate the energy distribution of different frequency bands. Based on the improved M&P wave method, the seismic pulse components are identified and extracted, and the pulse energy rate is calculated using the time-frequency characteristics of seismic motion, and the pulse effect characteristics are analyzed. The calculation formula for pulse seismic identification is:

[0058] E R =-(2.68+0.1416×PGV p -0.0162×ER p +0.0002852×PGV p 2

[0059] -0.005476×ER p ×PGV p +0.0005197×ER p 2 )

[0060]

[0061] Among them, E R is the pulse index; PGVp is the pulse peak velocity; ERp is the pulse energy rate; when the pulse index E R When it is greater than 0, the earthquake motion is a pulse earthquake motion; t s and t eand represent the start time and end time of the velocity pulse respectively; E(t) represents the accumulated energy of the earthquake motion at time t;

[0062] S22. Extraction of slope response characteristics from numerical simulation and model test: Extract time domain characteristics from displacement, acceleration and stress curves, including maximum value, root mean square (RMS) and standard deviation; extract the cumulative displacement of the slope under earthquake action;

[0063] S23. Monitoring data feature extraction: Calculate the change rate from the displacement and acceleration time series data, including the displacement change rate and the velocity change rate.

[0064] S3. Determination of earthquake slope stability evaluation index and instability threshold;

[0065] In step S3, the specific steps include:

[0066] Through the statistical analysis of stable and unstable samples in the database in step S1 and the slope response characteristics in step S2, the displacement critical value under specific earthquake conditions is calculated to determine the displacement threshold; the earthquake acceleration time history is extracted to calculate the slope critical acceleration, the relationship between the acceleration value and the slope instability is analyzed to determine the slope critical acceleration; the stress distribution is extracted to calculate the critical stress value;

[0067] Based on the index thresholds of displacement, acceleration and stress, a multidimensional index system of slope instability is defined. To ensure the safety of slope engineering, when any stability index reaches or exceeds the instability threshold, the slope is considered to have reached a critical state of instability; when all indicators are below their thresholds, the slope is in a stable state.

[0068] The expression for indicator threshold judgment is as follows:

[0069]

[0070] Among them, State represents the slope stability state, which takes the value of "1" or "0"; D represents the cumulative displacement; A represents the maximum acceleration; S represents the maximum stress; D crit is the instability threshold of cumulative displacement; A crit is the acceleration instability threshold; S crit is the instability threshold of stress.

[0071] S4, data preprocessing and eigenvalue selection;

[0072] In step S4, the specific steps include:

[0073] S41, according to the database constructed in step S1 and the data features extracted in step S2, a large number of slope characteristic parameters (slope height, slope gradient, rock and soil strength) and earthquake intensity parameters (peak value, Arias intensity, pulse energy rate) are obtained, and these parameters are used as influencing factors to perform standardized data preprocessing to ensure that the numerical ranges between the data features are consistent;

[0074] S42, based on step S3, obtaining slope stability evaluation indicators such as displacement, acceleration and stress value, using the slope instability multidimensional indicator system and threshold judgment expression established in step S3, determining the stability state (stable is 0, instability is 1), and using the stability state as the target value / label;

[0075] S43. Using correlation analysis, retain the factors whose correlation with other influencing factors is less than 0.8, and then evaluate the importance of these retained influencing factors based on the random forest algorithm. Remove the influencing factors with an importance less than 0.1 or the lowest importance relative to others, and use the remaining influencing factors as the eigenvalues ​​required for model establishment.

[0076] The calculation formula of the impact factor importance is as follows:

[0077]

[0078] Among them, Importance(X i ) is the impact factor X i The importance of N is the number of influencing factors, G is split is the gain contributed by the factor, and n is the nth influencing factor.

[0079] S5. Establishment of intelligent prediction model for earthquake slope instability probability;

[0080] In step S5, the feature values ​​and target values / labels in step S4 are used as a data set, and the data set is randomly divided into 5 parts using K-fold cross-validation. Four of them are selected as training sets and one as a test set. The training set is brought into the convolutional neural network (CNN) algorithm to build a prediction model.

[0081] Introducing Bayesian optimization into the algorithm to automatically adjust the model's hyperparameters, including convolution kernel size, number of convolution kernels, pooling kernel size, pooling method, number of neurons, number of layers, learning rate, and L2 regularization parameter;

[0082] Model testing: Each test uses accuracy, recall, and AUC curves to evaluate the stability and generalization ability of the model on the test set. Finally, the average of each test is taken to obtain the trained model. The final performance evaluation of the model is obtained.

[0083] S6. Intelligent prediction of earthquake slope instability probability.

[0084] When seismic data and slope characteristic data are obtained, the probability of slope instability can be obtained using the neural network model trained in S5. The closer the value is to 1, the more likely the slope is to become unstable, and the closer it is to 0, the more stable the slope is.

[0085] The present invention provides a theoretical basis and technical support for the prediction of earthquake-induced landslides through physical mechanism modeling, data-driven feature extraction and Bayesian optimized neural network modeling technology, and is suitable for intelligent monitoring of slope hazards, risk assessment and earthquake disaster prevention and control.

[0086] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0087] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0088] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0089] The word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.

[0090] The "first\second" mentioned in the embodiments is only to distinguish similar objects, and does not represent a specific order for the objects. It is understandable that the "first\second" can be interchanged with the specific order or sequence where permitted. It should be understood that the objects distinguished by "first\second" can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.

[0091] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent prediction method for earthquake slope instability probability based on the coupling of physical mechanism and neural network model, characterized in that: The steps include: S1. Database construction based on earthquake slope stability analysis; S2, ground motion and slope response data feature extraction; S3. Determination of earthquake slope stability evaluation index and instability threshold; S4, data preprocessing and eigenvalue selection; S5. Establishment of intelligent prediction model for earthquake slope instability probability; S6. Intelligent prediction of earthquake slope instability probability.

2. The intelligent prediction method for earthquake slope instability probability based on coupling of physical mechanism and neural network model according to claim 1 is characterized by: In step S1, the specific steps include: S11. Collect earthquake motion data worldwide, including acceleration time history, conduct earthquake motion characteristic analysis, and obtain information on earthquake motion peak acceleration and velocity, Arias intensity, frequency, epicenter distance, and focal depth; S12. Collect relevant data for seismic slope stability analysis based on numerical simulation and shaking table model test methods, and obtain slope response characteristic data including slope geometry, geological conditions, load, acceleration, displacement, and stress distribution; S13. Collect real-time monitoring data based on the on-site slope, including displacement and strain data; S14. Integrate the above data to build a database for earthquake slope stability analysis, and mark the slope stability state corresponding to each set of data, that is, mark stable as 0 and unstable as 1.

3. The intelligent prediction method for earthquake slope instability probability based on coupling of physical mechanism and neural network model according to claim 1 is characterized by: In step S2, data feature extraction is performed based on the constructed database, specifically including the following: S21. Characteristics of seismic data: Fourier transform FFT is used to extract the frequency domain characteristics of seismic motion and slope response, and spectrum analysis is performed to calculate the energy distribution of different frequency bands. Based on the improved M&P wave method, the seismic pulse components are identified and extracted, and the pulse energy rate is calculated using the time-frequency characteristics of seismic motion, and the pulse effect characteristics are analyzed. The calculation formula for pulse seismic identification is: AND R =-(2.68+0.1416×PGV p -0.0162×ER p +0.0002852×PGV p 2 -0.005476×ER p ×PGV p +0.0005197×ER p 2 ) Among them, E R is the pulse index; PGVp is the pulse peak velocity; ERp is the pulse energy rate; when the pulse index E R When it is greater than 0, the earthquake motion is a pulse earthquake motion; t s and t e and represent the start time and end time of the velocity pulse respectively; E(t) represents the accumulated energy of the earthquake motion at time t; S22. Extraction of slope response characteristics from numerical simulation and model test: Extract time domain characteristics from displacement, acceleration and stress curves, including maximum value, root mean square (RMS) and standard deviation; extract the cumulative displacement of the slope under earthquake action; S23. Monitoring data feature extraction: Calculate the change rate from the displacement and acceleration time series data, including the displacement change rate and the velocity change rate.

4. The intelligent prediction method for earthquake slope instability probability based on coupling of physical mechanism and neural network model according to claim 1 is characterized by: In step S3, the specific steps include: Through statistical analysis of stable and unstable samples in the database and slope response characteristics, the displacement critical value under specific seismic conditions is calculated and the displacement threshold is determined; the seismic acceleration time history is extracted, the critical acceleration of the slope is calculated, the relationship between the acceleration value and the slope instability is analyzed, and the critical acceleration of the slope is determined; Extract stress distribution and calculate critical stress value; Based on the index thresholds of displacement, acceleration and stress, a multidimensional index system of slope instability is defined. When any stability index reaches or exceeds the instability threshold, the slope is considered to have reached a critical state of instability; when all indicators are lower than their thresholds, the slope is in a stable state.

5. The intelligent prediction method for earthquake slope instability probability based on coupling of physical mechanism and neural network model according to claim 4 is characterized by: The expression for judging the indicator threshold is as follows: Among them, State represents the slope stability state, which takes the value of "1" or "0"; D represents the cumulative displacement; A represents the maximum acceleration; S represents the maximum stress; D crit is the instability threshold of cumulative displacement; A crit is the acceleration instability threshold; S crit is the instability threshold of stress.

6. The intelligent prediction method for earthquake slope instability probability based on coupling of physical mechanism and neural network model according to claim 1 is characterized by: In step S4, the specific steps include: S41, according to the database constructed in step S1 and the data features extracted in step S2, a large number of slope characteristic parameters and ground motion intensity parameters are obtained, and these parameters are used as influencing factors to perform standardized data preprocessing to ensure that the numerical ranges between the data features are consistent; The slope characteristic parameters include slope height, slope gradient, and rock and soil strength; the seismic intensity parameters include peak value, Arias intensity, and pulse energy rate; S42, using the slope instability multidimensional index system and threshold judgment expression established in step S3, determine the stability state, that is, stability is 0 and instability is 1, and use the stability state as the target value / label; S43. Using correlation analysis, retain the factors whose correlation with other influencing factors is less than 0.8, and then evaluate the importance of these retained influencing factors based on the random forest algorithm. Remove the influencing factors with an importance less than 0.1 or the lowest importance relative to others, and use the remaining influencing factors as the eigenvalues ​​required for model establishment.

7. The intelligent prediction method for earthquake slope instability probability based on coupling of physical mechanism and neural network model according to claim 6 is characterized by: The calculation formula of the impact factor importance is as follows: Among them, Importance(X i ) is the impact factor X i The importance of N is the number of influencing factors, G is split is the gain contributed by the factor, and n is the nth influencing factor.

8. The intelligent prediction method for earthquake slope instability probability based on coupling of physical mechanism and neural network model according to claim 1 is characterized by: In step S5, the feature values ​​and target values / labels in step S4 are used as a data set, and the data set is randomly divided into 5 parts using 5-fold cross validation. Four of them are selected as training sets and one as a test set. The training set is brought into the convolutional neural network (CNN) algorithm to build a prediction model. Introducing Bayesian optimization into the algorithm to automatically adjust the model's hyperparameters, including convolution kernel size, number of convolution kernels, pooling kernel size, pooling method, number of neurons, number of layers, learning rate, and L2 regularization parameter; Model testing: Each test uses accuracy, recall, and AUC curves to evaluate the stability and generalization ability of the model on the test set. Finally, the average value of each test is taken to obtain the trained model.

9. The intelligent prediction method for earthquake slope instability probability based on coupling of physical mechanism and neural network model according to claim 1 is characterized by: In step S6, the closer the slope instability probability value is to 1, the more likely the slope is to become unstable, and the closer it is to 0, the more stable the slope is.

Images