A method for predicting and evaluating the response of an integrated modeling shear wave vibroseis vibrator

Abstract

The application provides a kind of integrated modeling of shear wave vibroseis vibrator response prediction and evaluation method, belong to the field of deep, unconventional oil and gas seismic exploration technology, method includes: S1, construct the coupling vibration numerical simulation model of shear wave vibroseis vibrator-ground;S2, input characteristic data experiment simulation sample is obtained by using progressive Latin hypercube sampling method, obtain training data set and verification data set;S3, construct public time scale, mapping to unified time scale, divide to obtain time series simulation training data;S4, establish chaotic improved transverse cross particle swarm optimization BP neural network regression prediction model;S5, construct factor set and evaluation set, determine the adaptability evaluation result by fuzzy transformation and maximum membership degree principle.The application is aimed at the excitation characteristics of shear wave vibroseis vibrator and establishes the adaptability evaluation index of vibrator for different working conditions by combining the fuzzy comprehensive evaluation method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of deep, unconventional oil and gas seismic exploration, and particularly relates to a method for predicting and evaluating vibrator response of an integrated modeling shear wave vibroseis. BACKGROUND

[0002] With the development of China's economy, the consumption of oil and gas resources is increasing, and the shallow strata and traditional conventional oil and gas resources are gradually exhausted. As one of the key technologies for oil and gas resource exploration, seismic exploration, the controllable source has become an important equipment due to its high efficiency. By mechanical vibration excitation of seismic waves, the controllable source has the characteristics of continuous controllable signal and concentrated distribution of excitation energy, and has minimal impact on the environment during use, meeting the demand for green development. In the field of oil and gas exploration, shear waves have shown significant advantages in seismic exploration due to their low speed and short wavelength. The core equipment of shear wave seismic exploration is the shear wave controllable source, and the controllability of its excitation signal makes it have a unique position in multi-wave exploration. In order to further improve the ability of oil and gas exploration, it is urgent to develop various types of controllable sources in coordination, among which large-tonnage shear wave controllable sources have become a key point. However, the shear wave controllable source faces many technical challenges in signal excitation and energy transmission.

[0003] Research shows that complex surface conditions significantly restrict the excitation characteristics of the vibrator of the shear wave controllable source. Therefore, improving the adaptability of the vibrator under complex surface conditions and ensuring its high excitation strength in different geological environments are the key to improving the effect of seismic exploration.

[0004] At present, in the process of modeling the dynamic characteristics of shear wave controllable sources, the finite element simulation method is mostly used to construct the source-foundation coupling model and calculate the excitation output response of the source; however, due to the multiple parameter dimensions involved (such as excitation frequency, soil elastic modulus, shear strength, etc.), each simulation calculation needs to repeatedly solve multiple nonlinear coupling fields, resulting in high overall model complexity and high computing resource consumption. A single complete simulation often takes more than ten hours or even several days, which seriously restricts the application efficiency of the model in large-scale parameter condition evaluation; in order to improve efficiency, only methods based on neural networks or support vector machines are introduced to construct proxy prediction models; however, such methods usually use the overall response curve as the model Modeling targets, ignoring the non-stationary nature of the time series data of the shear-wave controllable source response in the time dimension and the coupling influence of cross-parameter factors, resulting in the overall prediction model having problems such as overfitting, poor generalization ability, and poor cross-frequency prediction accuracy, making it difficult to support hourly prediction and fitting of high-precision response curves; and in terms of sample generation and adaptability analysis, traditional Latin hypercube sampling is difficult to take into account both sample uniformity and multi-factor coupling characteristics, and the existing methods also lack a quantitative adaptability evaluation mechanism that integrates multiple indicators, resulting in limited model training quality and discrimination ability. Therefore, relying solely on existing models to predict the excitation characteristics of shear-wave controllable sources is inefficient and unreasonable, and there is an urgent need to explore more efficient prediction and optimization methods. Summary of the Invention

[0005] The present invention provides an integrated modeling method for predicting and evaluating the response of a shear wave controllable source vibrator, which predicts the excitation characteristics of the shear wave controllable source vibrator and establishes an adaptability evaluation index of the vibrator for different working conditions in combination with a fuzzy comprehensive evaluation method.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] An integrated modeling method for predicting and evaluating the response of a shear-wave vibrator includes:

[0008] S1. Based on the geometric model of the shear wave vibrator, a numerical simulation model of the coupled vibration of the shear wave vibrator and the earth is constructed using finite element simulation software;

[0009] S2. Determine the input feature data range and the number of sampling samples based on the shear wave vibrator-earth coupled vibration numerical simulation model. Based on the determined input feature data range and the determined number of sampling samples, use a progressive Latin hypercube sampling method to obtain experimental simulation samples of the input feature data, and input these samples into the shear wave vibrator-earth coupled vibration numerical simulation model to obtain a training data set and a validation data set of the vibrator output feature data under different input characteristics.

[0010] S3, a common time scale is constructed for the time series curve data of the training data set and the validation data set, mapped to a unified time scale by linear interpolation and normalized, and divided to obtain time series simulation training data;

[0011] S4, a chaotic improved transverse cross particle swarm optimization BP neural network regression prediction model is established using the time series simulation training data, and an integrated time series prediction curve is obtained by optimization and integration;

[0012] S5, based on the complete time series prediction curve, evaluation index data under different working conditions is obtained, a factor set and an evaluation set are constructed, and a single-factor evaluation matrix and an optimized dynamic weight matrix are calculated, and an adaptive evaluation result is determined by fuzzy transformation and the maximum membership principle.

[0013] In the specification, the specific steps of constructing the numerical simulation model of the vibrator-ground coupling vibration of the shear wave controllable source include: establishing and simplifying the vibrator geometric model based on three-dimensional drawing and finite element simulation software, dividing the grid by rigid element and hexahedral element, setting the vibrator material to 45 steel and the ground to clay, defining the hard contact and Coulomb friction characteristics between the plate and the ground; setting the six-degree-of-freedom constraint of the ground, the static load of the vibrator car weight and the dynamic load of the piston rod hydraulic pressure, using infinite element to process the boundary and solving the model through geostress analysis, statics analysis and implicit dynamics analysis.

[0014] In the specification, based on the numerical simulation model of the vibrator-ground coupling vibration of the shear wave controllable source, an evaluation index system is constructed by using contact surface deformation, excited stress wave amplitude, vibration output force and energy transfer rate. The standards of the evaluation index system are: the greater the vibration displacement and the smaller the plastic deformation in the contact surface deformation, the greater the excited stress wave amplitude and the fewer the harmonic components, the more uniform the vibration output force distribution, and the more effective energy and the higher the transfer rate in energy transfer.

[0015] In the specification, the input feature data range is: excitation frequency 5-120Hz, soil elastic modulus 20-70MPa, soil cohesion 42-94KPa, and driving force amplitude 40%-90%.

[0016] In the specification, the specific steps of the progressive Latin hypercube sampling method include: dividing the value range of each input dimension into n equal probability intervals, generating an initial sample subset, then calculating the pair correlation by Pearson matrix correlation coefficient, adding a new sample point each time according to the maximum pair correlation principle and optimizing the sample point position to maximize the stratification in each dimension, and combining the input feature data points of different dimensions to generate experimental simulation samples.

[0017] In the specification, the specific steps of constructing the public time scale include: calculating the average step length as the public time step according to the total number of points of the output characteristic data time sequence curve, creating an equidistant time sequence from the starting time to the ending time based on the public time step; mapping the original time sequence data to the time sequence by using linear interpolation, normalizing the data to the range of [0, 1] by the maximum and minimum normalization method, and dividing the time sequence simulation training data according to the characteristics of the time sequence.

[0018] In the specification, the specific steps of the chaos improved transverse crossover particle swarm optimization include: generating a chaotic sequence by using the Singer mapping, updating the particle speed through the inertia weight, self-cognition coefficient and social cognition coefficient, and adjusting the speed range by combining the dynamic maximum flight speed limiting mechanism; selecting adjacent particle pairs for dimensional crossover, generating new particle positions based on the combined particle centroid, the random number and the disturbance factor, and optimizing the number of neurons and the learning rate of the BP neural network based on the mean square error of the five-fold cross-validation.

[0019] In the specification, the construction steps of the BP neural network regression prediction model include: setting 4 neurons in the input layer corresponding to the excitation frequency, soil elastic modulus, soil cohesion and driving force amplitude, setting 4 neurons in the output layer corresponding to the vibration output force, energy transmission rate, excitation stress wave amplitude and contact surface strain, and determining the number of hidden layer neurons by an optimization algorithm; using the hyperbolic tangent function as the hidden layer activation function and the linear function as the output layer activation function, constructing an independent sub-model for each public time point and integrating to obtain a complete time sequence prediction curve.

[0020] In the specification, the specific steps of constructing the factor set and the evaluation set include: defining the factor set as {excitation frequency, driving force amplitude, soil elastic modulus, soil cohesion}, and defining the evaluation set as {vibration output force, energy transmission rate, excitation stress wave amplitude, contact surface strain}; calculating the single-factor evaluation matrix of the factor set to the evaluation set based on the membership function, and using the improved transverse crossover particle swarm optimization to optimize the dynamic weight matrix, wherein the weight matrix elements correspond to the importance of each factor under different working conditions.

[0021] In the specification, the specific steps of determining the adaptive evaluation result include: mapping the factor set fuzzy set to the evaluation set fuzzy set by fuzzy transformation; according to the principle of comprehensive maximum membership degree, selecting the evaluation grade corresponding to the maximum value of the membership degree of the evaluation object to the grade fuzzy subset as the adaptive evaluation result of the transverse wave controllable vibrator under the complex surface environment.

[0022] In summary, the present application has at least the following beneficial effects:

[0023] The application integrates finite element simulation and machine learning modeling, adopts a multi-submodel time sequence prediction mechanism, and combines a fuzzy comprehensive evaluation method to construct a response characteristic prediction model of a transverse wave vibrator under multiple factor excitation conditions of a controllable vibrator. By modeling the excitation frequency, driving force amplitude, soil elastic modulus, cohesion and other parameters, the dynamic prediction of the vibrator output force, stress wave, contact strain and energy transmission and other output characteristics is realized, the calculation efficiency and model accuracy are improved; at the same time, the fuzzy comprehensive evaluation method is used to quantitatively evaluate the vibrator output response under different working conditions, the uncertainty adaptability and adaptability analysis ability of the model in complex environment are enhanced; in addition, in order to further improve the search ability and convergence precision of the optimization process, a dynamic maximum flight speed regulation strategy is introduced on the basis of the transverse cross particle swarm optimization algorithm, the global exploration ability in the early stage and the local convergence performance in the later stage are considered; the method not only improves the modeling and analysis ability of the transverse wave vibrator excitation process in complex geological environment, but also provides theoretical support and technical basis for the structural optimization design and engineering application of the transverse wave vibrator. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0025] Figure 1 The schematic diagram of the integrated modeling transverse wave vibrator response prediction and evaluation method involved in the present application;

[0026] Figure 2 The flowchart of the integrated modeling transverse wave vibrator response prediction and evaluation method involved in the present application;

[0027] Figure 3 The schematic diagram of the transverse wave vibrator-ground coupling model involved in the present application;

[0028] Figure 4 The expanded schematic diagram of the BP neural network structure involved in the present application;

[0029] Figure 5 The fitting schematic diagram of the chaotic improved transverse cross particle swarm optimization BP neural network involved in the present application;

[0030] Figure 6 The comparative schematic diagram of one of the excitation characteristic indexes, the output force peak value time point submodel involved in the present application;

[0031] Figure 7A comparison diagram of output force time sequence data prediction results for one of the excitation characteristic indexes involved in the present application is shown. DETAILED DESCRIPTION

[0032] In the following, only certain exemplary embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the embodiments of the present application. Therefore, the drawings and the description are considered to be exemplary in nature rather than limiting.

[0033] The following disclosure provides many different embodiments, or examples, for implementing different structures of the embodiments of the present application. For the purpose of simplifying the disclosure of the embodiments of the present application, the components and settings of specific examples are described in the following. Of course, they are only examples, and the purpose is not to limit the embodiments of the present application. In addition, the embodiments of the present application can refer to the same reference numerals and / or reference letters in different examples, and such repetition is for the purpose of simplification and clarity, which does not indicate the relationship between the various embodiments and / or settings discussed.

[0034] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0035] As Figure 1 shown, the present embodiment provides a method for predicting and evaluating the response of an integrated modeling shear wave vibroseis vibrator, comprising:

[0036] S1, according to the geometric model of the shear wave vibroseis vibrator, using finite element simulation software, a coupled vibration numerical simulation model of the shear wave vibroseis vibrator-ground is constructed;

[0037] S2, according to the coupled vibration numerical simulation model of the shear wave vibroseis vibrator-ground, determine the input characteristic data range and the sampling sample number, based on the determined input characteristic data range and the determined sampling sample number, use the progressive Latin hypercube sampling method to obtain the input characteristic data experimental simulation sample, and input to the coupled vibration numerical simulation model of the shear wave vibroseis vibrator-ground, get the training data set and the verification data set of the vibrator output characteristic data under different input characteristics;

[0038] S3, construct a common time scale for the time sequence curve data of the training data set and the verification data set, map to a unified time scale by linear interpolation and normalize, and divide to get time sequence simulation training data;

[0039] S4, use the time sequence simulation training data to establish a BP neural network regression prediction model optimized by a chaotic improved transverse cross particle swarm, and optimize and integrate to get a complete time sequence prediction curve;

[0040] S5, obtaining evaluation index data under different working conditions based on the complete time sequence prediction curve, constructing a factor set and an evaluation set, and calculating a single-factor evaluation matrix and an optimized dynamic weight matrix, determining an adaptive evaluation result through fuzzy transformation and the maximum membership principle.

[0041] In some embodiments, the specific steps of constructing the vibrator-ground coupling vibration numerical simulation model of the shear wave vibroseis include: establishing and simplifying the vibrator geometric model based on three-dimensional drawing and finite element simulation software, dividing the grid by using rigid elements and hexahedral elements, setting the vibrator material as 45 steel and the ground as clay, defining the hard contact and Coulomb friction characteristics between the flat plate and the ground; setting the six-degree-of-freedom constraint of the ground, the static load of the vibrator car weight, and the dynamic load of the piston rod hydraulic pressure, processing the boundary by using infinite elements, and solving the model through geostress analysis, statics analysis, and implicit dynamics analysis.

[0042] In some embodiments, based on the vibrator-ground coupling vibration numerical simulation model of the shear wave vibroseis, an evaluation index system is constructed by using contact surface deformation, excited stress wave amplitude, vibration output force, and energy transfer rate. The standards of the evaluation index system are: the greater the vibration displacement and the smaller the plastic deformation in the contact surface deformation, the greater the excited stress wave amplitude and the fewer the harmonic components, the more uniform the vibration output force distribution, and the more effective energy and the higher the transfer rate in energy transfer.

[0043] In some embodiments, the input feature data range is: excitation frequency 5-120Hz, soil elastic modulus 20-70MPa, soil cohesion 42-94KPa, and driving force amplitude 40%-90%.

[0044] In some embodiments, the specific steps of the progressive Latin hypercube sampling method include: dividing the value range of each input dimension into n equal probability intervals, generating an initial sample subset, calculating the pair correlation by using the Pearson matrix correlation coefficient, adding a new sample point and optimizing the sample point position each time according to the maximum pair correlation principle, so that the maximum stratification is achieved in each dimension, and combining the input feature data points of different dimensions to generate experimental simulation samples.

[0045] In some embodiments, the specific steps of constructing the common time scale include: calculating the average step length as the common time step length according to the total number of output feature data time sequence curves, creating an equal interval time sequence from the starting time to the ending time based on the common time step length; mapping the original time sequence data to the time sequence by using linear interpolation, normalizing the data to the range of [0, 1] by using the maximum and minimum normalization method, and dividing the time sequence characteristics into time sequence simulation training data.

[0046] In some embodiments, the specific steps of the chaos improved transverse cross particle swarm optimization include: generating a chaos sequence by using a Singer mapping, updating particle speed by inertia weight, self-cognition coefficient and social cognition coefficient, and adjusting the speed range by combining a dynamic maximum flight speed limiting mechanism;Selecting adjacent particle pairs for dimensional crossover, generating new particle positions based on the combined particle centroid, random numbers and disturbance factors, and optimizing the number of neurons and learning rate of the BP neural network based on the mean square error of five-fold cross-validation.

[0047] In some embodiments, the construction steps of the BP neural network regression prediction model include: setting 4 neurons in the input layer corresponding to the excitation frequency, soil elastic modulus, soil cohesion and driving force amplitude, setting 4 neurons in the output layer corresponding to the vibration output force, energy transmission rate, excitation stress wave amplitude and contact surface strain, and determining the number of hidden layer neurons by an optimization algorithm;Using the hyperbolic tangent function as the hidden layer activation function and the linear function as the output layer activation function, constructing an independent sub-model for each common time point and integrating to obtain a complete time series prediction curve.

[0048] In some embodiments, the specific steps of constructing the factor set and the evaluation set include: defining the factor set as {excitation frequency, driving force amplitude, soil elastic modulus, soil cohesion}, and defining the evaluation set as {vibration output force, energy transmission rate, excitation stress wave amplitude, contact surface strain};Based on the membership function, a single-factor evaluation matrix of the factor set on the evaluation set is calculated, and an improved transverse cross particle swarm optimization dynamic weight matrix is used, wherein the weight matrix elements correspond to the importance of each factor under different working conditions.

[0049] In some embodiments, the specific steps of determining the adaptive evaluation result include: mapping the factor set fuzzy set to the evaluation set fuzzy set by fuzzy transformation;According to the principle of comprehensive maximum membership degree, the evaluation grade corresponding to the maximum value of the membership degree of the evaluation object to the grade fuzzy subset is selected as the adaptability evaluation result of the vibroseis vibrator of the controlled transverse wave source in the complex surface environment.

[0050] The technical idea of the application is as follows:

[0051] A kind of integrated modeling of transverse wave controlled source vibrator response prediction and evaluation method (flow as Figure 2 Shown), comprising:

[0052] S1, first, the model of the transverse wave controlled source vibrator is established, the coupling vibration numerical simulation model of transverse wave controlled source vibrator-ground is established by finite element simulation software, as follows:

[0053] As Figure 3As shown, based on three-dimensional drawing and finite element simulation software, the geometric model of the transverse wave controlled source vibrator is established and simplified, and the coupled vibration simulation model of the vibrator and the earth is constructed. In the mesh division of the model, the upper part of the vibrator uses rigid body element R3D4, and the flat plate structure uses hexahedral element C3D8R. Reasonable grid size is selected through mesh independence analysis. In terms of materials, the vibrator uses 45 steel, and the earth model is clay. Hard contact and Coulomb friction characteristics are set between the flat plate and the earth.

[0054] The boundary conditions of the transverse wave controlled source vibrator-earth coupled vibration model include six degree of freedom constraints of the earth, static load and dynamic load. The static load is the weight of the vibrator shell through the action of the air spring. The dynamic load is the hydraulic pressure acting on the piston rod. In addition, the absorbing boundary condition is set to handle the model boundary with infinite elements. The model solution is divided into three analysis steps: geostress analysis, statics analysis for static load, and implicit dynamics analysis for dynamic load.

[0055] The excitation frequency range of the transverse wave controlled source vibrator is 3-120 Hz. The transverse wave controlled source vibrator has more effective downlink energy in the low frequency stage, which is beneficial to the improvement of exploration depth. However, the energy transfer efficiency is low. The high frequency source can generate shorter wavelength, so that the source energy is more concentrated and higher ground vibration intensity can be generated. The effective energy transfer of the system is improved, but the transfer efficiency is still low.

[0056] The elastic modulus of soil and soil cohesion are also important factors affecting the output characteristics of the vibrator. The value range of the soil elastic modulus is selected as 20-70 MPa. It can be seen that the amplitude of the stress wave excited by the vibrator on the earth's surface increases with the increase of the soil elastic modulus, that is, the exploration ability of the vibrator is stronger. Soil cohesion is an index of soil shear strength. During the operation of the transverse wave controlled source vibrator, the flat plate and the earth are coupled to vibrate in the horizontal direction, and the soil is sheared and deformed. The material value range of the cohesion is selected as 42-94 Kpa.

[0057] Based on the transverse wave controlled source vibrator-earth coupled finite element simulation model, combined with the input characteristic analysis, it can be known that the excitation ability of the transverse wave source vibrator is closely related to the response of the vibrator. The evaluation index system is constructed by using the characteristics of contact surface deformation, excitation stress wave propagation, vibration output force and energy transfer.

[0058] The contact surface displacement directly reflects the interaction strength between the flat plate of the transverse wave source vibrator and the earth. The greater the displacement, the stronger the interaction, and the higher the output strength of the vibrator. However, the plastic deformation of the contact surface will affect the tightness of the contact between the flat plate and the earth. The greater the plastic deformation, the more likely it is to cause decoupling, thereby reducing the output strength and accuracy of the vibrator.

[0059] The variation law of the excited stress wave with the propagation depth directly reflects the propagation of the vibrator excitation signal in the earth, determines the intensity of the excitation signal, and directly reflects the degree of harmonic distortion of the vibrator output signal, and affects the exploration precision of the cross-wave controlled source;

[0060] The vibration output force between the vibrator plate and the earth directly reflects the output capacity of the vibrator, the greater the vibration output force, the stronger the coupling vibration capacity of the vibrator driving the earth; and the distribution of the contact pressure reflects the contact condition between the plate and the earth, the more uniform the vibration output force distribution, the smaller the signal distortion caused by the uneven force on the earth, and the higher the precision of the vibrator output signal;

[0061] The greater the effective energy transmission of the vibrator, the greater the energy obtained by the earth, which is beneficial to improve the exploration depth; the energy transmission rate of the system reflects the conversion rate of the consumed external energy into effective energy, the higher the energy transmission rate, the less the energy waste, the higher the energy utilization rate of the system, which is beneficial to improve the exploration efficiency;

[0062] The evaluation index standard is that the greater the contact surface deformation and the smaller the plastic deformation, the greater the stress wave amplitude and the less the harmonic component, the more uniform the vibration output force distribution, and the more effective energy transmission and the higher the energy transmission rate.

[0063] S2, based on the gradual Latin hypercube sampling method, the input feature data experiment simulation sample is obtained, first, according to the actual demand and target, the range and quantity of preprocessed data sample are determined; in the input parameter space, the initial sample of the experiment is obtained based on the gradual Latin hypercube sampling method, and the sufficient and necessary condition formula of the gradual Latin hypercube sampling compared with the original Latin hypercube sampling is as follows:

[0064] ;

[0065] Wherein, S is the total number of slices, p is the dimension number, is the number of points collected in the s stage, is a binary auxiliary variable; j is the dimension index, and q is the sample index point.

[0066] The above formula ensures that the gradually added samples in the gradual Latin hypercube can meet the Latin hypercube distribution, wherein the Latin hypercube distribution is a sample distribution method based on the principle of Latin hypercube design, which realizes uniform and representative sample distribution in high-dimensional space by dividing each input variable into equal probability sub-intervals and sampling only once in each dimension, and the specific is as follows:

[0067] According to the shear wave vibrator-earth coupling model, the input characteristic data range is determined, where the excitation frequency is 5-120 Hz, the soil elastic modulus range is 20-70 MPa, the soil cohesion range is 42-94 kPa, and the driving force amplitude range is 40%-90%. The number of sampling samples is determined to be n, and the value range of each dimension is divided into n intervals, and the value range of each interval is determined;

[0068] An initial sample subset U is generated based on Latin hypercube sampling, where the dimension of each sample point is d and the number of samples is n. The specific formula is as follows:

[0069] ;

[0070] for Middle sample points.

[0071] New subsets are gradually added based on the initial subsets. The maximum pairwise correlation is used as the acquisition function, and the Pearson matrix correlation coefficient is used to calculate the pairwise correlation items. The specific formula is:

[0072] ;

[0073] in, is the κth sample in The value of the dimension, is the κth sample in The value of the dimension, For the The average value of the dimension samples, For the The average value of the dimension sample, n is the number of samples, is the correlation coefficient;

[0074] Each time a sample point is added, the progressive Latin hypercube optimizes the position of the sampling point to achieve maximum stratification in each dimension. The different input feature data points obtained in each dimension are combined to generate the required experimental simulation samples.

[0075] Table 1. Sampling data of some samples

[0076]

[0077] The progressive Latin hypercube sampling samples are input into the shear wave vibrator-ground coupled vibration numerical simulation model to obtain the training data set and validation data set of the vibrator output characteristic data under different input characteristics.

[0078] S3, the data set obtained by progressive Latin hypercube sampling and simulation, multi-factor input feature time series curve data, analyzing data time series data features to construct a common time scale, and pre-processing the data by linear interpolation, specifically:

[0079] According to the obtained output feature data time series curve, combined with its specific characteristics, the data is obtained, that is, the average step length As a common time step is as follows:

[0080] ;

[0081] Where N is the total number of time data points; represents the first time node in the original time series;

[0082] According to the common time step combined with the time series data range, create a common time scale as follows:

[0083] ;

[0084] is the first time point in the original time series; is the last time point in the original time series;

[0085] The output characteristics are mapped to a unified common time scale using a linear interpolation method to obtain the input time series curve for training;

[0086] ;

[0087] Where t is any time point on the common time scale, and are the closest known data points in the original time series curve; is the value corresponding to any time point t on the unified common time scale obtained by linear interpolation.

[0088] Based on the maximum and minimum normalization method, the original data of the training data set is linearized and converted to the range of [0, 1] using a linear function, and the specific function is:

[0089] ;

[0090] Where, is the normalized range feature data, represents the original data set, represents the minimum value of the feature data, represents the maximum value of the feature data;

[0091] ​According to the characteristics of the time sequence, the converted training data set is divided to obtain time sequence simulation training data; according to the public time scale, a multi-sub-model integrated time sequence prediction method is used to divide each public time point data into several training data sets.

[0092] S4, a multi-factor chaotic improved transverse cross particle swarm optimization BP neural network regression prediction model is established, and complex subsurface transverse wave vibrator evaluation parameter data set is obtained, specifically as follows:

[0093] A singer map is used to generate a chaotic sequence and is introduced into an improved transverse cross particle swarm optimization, wherein the singer map model is:

[0094] ;

[0095] wherein, is a chaotic variable, and μ∈[3,4] is a control parameter;

[0096] The singer map is used to generate a chaotic sequence to update the speed and position of the particle, the chaotic mapping is set to the range allowed by the particle, the inertia weight ω is set to control the speed of the particle in the updating process, c1 defines the self-cognition coefficient, c2 defines the social cognition coefficient, and a dynamic maximum flight speed limiting mechanism is used, and the dynamic limiting value is adjusted with the number of iterations, and the specific definition is as follows:

[0097] ;

[0098] wherein, represents the search interval of the particle, k1=0.02 and k2=0.04 are respectively the minimum and maximum flight speed proportion coefficients; k is the initial maximum speed proportion coefficient; r is the normalized proportion corresponding to the current iteration number; is the maximum iteration number; e is a natural constant about 2.718; in the is the current iteration number.

[0099] Combined with the above speed limitation, the particle speed updating formula is defined as:

[0100] ;

[0101] wherein, is the speed of the particle i in the dth dimension, and the value at the iteration t, ω is the inertia weight, c1 is the self-cognition coefficient, c2 is the social cognition coefficient, r1 and r2 are random numbers in the interval [0,1], is the historical best position of the particle i in the dth dimension, g d is the global best position, and clip(a,b,c) represents limiting the value a in the interval [b,c]. max and min flight speed limits, respectively; to represent the i-th particle in the d-th dimension, the position of the i-th particle at iteration t;

[0102] The particle position update formula is as follows:

[0103] ;

[0104] wherein, is the new position of particle i in the d-th dimension, is the updated position of the particle;

[0105] A transverse crossover-based random search algorithm is introduced, which can effectively avoid local optimum by transverse crossover of dimensions, and effectively avoid local optimal trap by selecting adjacent particles in the population for pairwise crossover in the decision dimension. Then, based on the centroid of the combined particles, the particles are updated by transverse crossover to generate new particle positions. The specific generation formula is as follows:

[0106] ;

[0107] ;

[0108] wherein, r ∈ [0, 1] is a random number, c ∈ [-1, 1] is a random disturbance factor, and represent the values of particles F(i) and F(i+1) in the d-th dimension, respectively, and represent the values of the offspring generated by transverse crossover in the d-th dimension, respectively;

[0109] The improved transverse crossover particle swarm optimization objective function is constructed based on the mean square error of five-fold cross-validation, and the training data set of a single time point model is input. The improved transverse crossover particle swarm optimization is combined to find the best number of neurons and learning rate, and the individual optimal extreme value and global optimal extreme value are generated in the particle iteration process;

[0110] The BP neural network prediction model is constructed for the time point according to the public time scale, including the design of input layer, hidden layer and output layer. The number of neurons x_p is 4 according to the number of input features in the input layer, the number of neurons in the hidden layer is determined by the optimization algorithm, and the number of neurons in the output layer is selected as n_out is 4 according to the output feature;

[0111] The output feature parameter is , and the number of output layer nodes is n_hid, and the weighted sum z of the output layer is: k

[0112] ​ ;

[0113] wherein, is the weight of the hidden layer the first node output layer in the υ nodes, is the bias term, is the activation function; h is the number of hidden neurons.

[0114] The BP neural network adopts the forward propagation method, and the input data is transmitted layer by layer to the output layer through weighted summation of each layer combined with the activation function, wherein the hyperbolic tangent function is used as the activation function of the hidden layer, and the linear activation function is selected as the activation function of the output layer; the BP neural network structure diagram is shown in Figure 4 ;

[0115] According to the improved transverse cross particle swarm optimization, the optimal initial weight and bias are further searched, and the network structure parameters of the BP neural network are set as follows: the iteration number is 1000, and the error threshold is ;

[0116] When the multi-sub-model integrated time series prediction method is used, the whole time series modeling problem is divided into multiple sub-problems, and a specific sub-model is constructed for each time segment on the public time scale. The specific time series data is:

[0117] ;

[0118] For each time point , an independent sub-model is constructed, which has:

[0119] ;

[0120] ;

[0121] wherein, X is an input feature vector, is the predicted output data at time point , is the sub-model grid parameter data at the time point, represents a BP sub-model optimized by C-PSO.

[0122] The improved transverse cross particle algorithm is used to optimize the BP neural network model in the embodiment, and the evaluation prediction data and the original data are compared in the curve of the model of a single time point. Figure 5 and Figure 6 It can be seen that the error fluctuation between the chaotic improved transverse cross particle swarm optimization BP neural network and the true data is smaller.

[0123] Define the error evaluation index of BP neural network time series model, test and verify the accuracy of BP neural network time series prediction model. The smaller the value of error measurement index, the better the prediction effect of the model and the stronger the fitting ability. The mean absolute error (MAE) and determination coefficient (R 2 ) and mean absolute percentage error (MAPE), the specific calculation formula is as follows:

[0124] ;

[0125] ;

[0126] ;

[0127] in, For the True value, For the predicted ý values, n is the number of samples, is the true sample mean.

[0128] like Figure 7 As shown in the figure, based on the model of each time point in the common time scale, an integrated time series prediction model is established to verify the fitting ability of the model. Taking the vibration output force, one of the evaluation indicators, as an example, the mean absolute error (MAE) is 1259.538 and the determination coefficient (R 2 ) is 0.9822 and the mean absolute percentage error (MAPE) is 11.653%.

[0129] S5. Based on the established chaos-modified lateral cross particle swarm optimization BP neural network prediction model, a large amount of evaluation index data of shear wave controllable source vibrators under different working conditions is obtained. Fuzzy comprehensive evaluation is used to establish the adaptability analysis of shear wave controllable source vibrators in complex surface environments, specifically including:

[0130] According to the prediction results of the established chaotic improved lateral cross particle swarm optimization BP neural network prediction model, the excitation characteristics under different excitation frequencies, driving force amplitudes, soil elastic modulus and cohesion are analyzed to obtain the excitation characteristics under different conditions;

[0131] Based on the fuzzy comprehensive evaluation method, the influencing factors are decomposed to construct an index system at different levels, and these indicators are assigned values ​​and weights are determined;

[0132] The establishment of the shear wave vibrator factor set determines the combination of multiple factors that affect the output characteristics, namely:

[0133] G = {g1, g2, g3, g4} = {excitation frequency, driving force amplitude, soil elastic modulus, soil cohesion};

[0134] The establishment of the evaluation set and the collection of various evaluation structures of the evaluation object are expressed as follows:

[0135] V={v1,v2,v3,v4}={vibration output force, energy transfer rate, excited stress wave amplitude, contact surface strain};

[0136] Among them, g is a factor, v is an evaluation index, and fuzzy mathematics uses a membership function with a value of [0,1] for quantification;

[0137] Based on the membership function of the determined influencing factor set X to the evaluation set V, the single factor evaluation matrix R is obtained, that is:

[0138] ;

[0139] To indicate the Influencing factors For the first Evaluation indicators v ń The attributed value, is the influencing factor number, is the evaluation index number;

[0140] The importance of each factor level is characterized by weights, and an improved horizontal cross particle swarm optimization is introduced to refine the weight setting, where the position of the particle represents the weight set W={w1,w2,…,w ū}, each particle has a position vector and a velocity vector;

[0141] To make the weights more dynamic and adaptive, the weight set is combined with variables such as influencing factors to form a dynamic weight matrix W(ē), which is dynamically adjusted according to the optimization structure in each iteration:

[0142] ;

[0143] in, Indicates the weight of the evaluation index ī for the ūth working condition or moment at the ē moment;

[0144] Update the output optimal dynamic weight matrix W and perform fuzzy comprehensive evaluation by transforming a fuzzy set W on the factor set X into a fuzzy set Z on the evaluation set V through the fuzzy relation R. When the fuzzy weight vector A and the fuzzy matrix R are known, the fuzzy transformation is:

[0145] ;

[0146] in, is a fuzzy synthesis operation, Z is a graded fuzzy subset on the evaluation set V, To evaluate the overall membership of the subject to the fuzzy subset Z.

[0147] According to the principle of comprehensive maximum membership evaluation, the maximum corresponding to the evaluation grade is selected as the adaptive evaluation result.

[0148] In summary, in this integrated modeling method, S3, S4 and S5 respectively improve the effectiveness of the scheme from the three key links of data preprocessing, model optimization prediction and comprehensive evaluation, and their core roles or contributions are as follows:

[0149] S3: Construct a common time scale and process time series data

[0150] Core role: Solve the non-stationary and scale inconsistency problem of time series data, and provide standardized input for subsequent prediction models.

[0151] Specific contribution:

[0152] Unified time series reference: Create a common time scale by calculating the average step, map the time series curves under different input features to a unified time series, avoid model training bias caused by time scale differences, prevent the inconsistency of the number of points on the curve corresponding to some sample data, and after using the common time scale, these sample data are unified into consistent samples to facilitate subsequent model training.

[0153] Enhance data consistency: Use linear interpolation and normalization processing to eliminate the dimensional differences of the original data, make the output features under different working conditions (such as vibration output force, energy transfer rate) comparable, and improve the learning efficiency of the model for time series features.

[0154] Optimize the structure of training data: Divide the training data by time series to lay the foundation for subsequent multi-submodel integrated prediction, and ensure that the prediction at each time point is based on standardized input features.

[0155] S4: Chaotic improved particle swarm optimization BP neural network prediction model

[0156] Core role: Improve the accuracy, generalization ability and anti-noise ability of the prediction model, and solve the problems of traditional BP network such as easy overfitting and slow convergence.

[0157] Specific contribution:

[0158] Optimized algorithm enhances search capability: Introduce chaotic sequence (Singer mapping) and dynamic maximum flight speed limiting mechanism to make particle swarm optimization algorithm avoid falling into local optimum during global search, and enhance population diversity through horizontal crossover operation to improve the optimization efficiency of BP network parameters (such as the number of hidden layer neurons and learning rate).

[0159] Multi-submodel integration to improve time series prediction accuracy: For each public time point, an independent BP submodel is constructed, and after integration, a complete time series curve is formed, effectively dealing with the time-varying characteristics of the shear wave source response, especially in cross-frequency band (5-120Hz) prediction to reduce accuracy loss.

[0160] Dynamic adaptation to complex working conditions: Through five-fold cross-validation to optimize model parameters, the model has stronger robustness to the coupling influence of soil elastic modulus (20-70MPa), cohesion (42-94KPa) and other parameters, and the prediction error (such as MAE, MAPE) is significantly reduced.

[0161] S5: Fuzzy comprehensive evaluation mechanism to build adaptive index

[0162] Core role: Establish a multi-index quantitative evaluation system to solve the limitations of qualitative analysis of seismic source adaptability in complex surface environment.

[0163] Specific contributions:

[0164] Multi-dimensional index integration: Combine factors such as excitation frequency and driving force amplitude with evaluation sets such as vibration output force and energy transfer rate, and through membership functions and single-factor evaluation matrices, quantify the influence of each factor on the seismic source response.

[0165] Dynamic weight optimization: Use improved particle swarm optimization algorithm to optimize weight matrix, so that the importance of each factor is dynamically adjusted under different working conditions (such as high cohesion soil vs. low elastic modulus stratum), avoiding the one-sidedness of traditional fixed weights.

[0166] Uncertainty quantification analysis: Through fuzzy transformation and maximum membership principle, the multi-index evaluation results are converted into clear adaptability levels, such as "high adaptability" and "moderate adaptability", providing decision-making basis for seismic source matching in complex surface conditions, filling the gap of lack of quantitative evaluation mechanism in existing methods.

[0167] Summary: S3, S4, and S5 form a technical closed loop from three aspects of data standardization, model intelligence, and evaluation scientization:

[0168] S3 solves the problem of "data heterogeneity" and provides high-quality input for prediction;

[0169] S4 improves "prediction accuracy and generalization ability" through algorithm innovation, and realizes dynamic modeling of response characteristics;

[0170] S5 builds a "multi-index adaptability evaluation system" to make the model have decision support capability for working condition matching.

[0171] The synergistic effect of the three ultimately realizes the whole process optimization of the shear wave controllable seismic source from response prediction to engineering application.

[0172] The above-described embodiments are intended to illustrate the present application, and are not intended to limit the present application, so the change of example values or the substitution of equivalent elements should still belong to the scope of the present application.

[0173] From the above detailed description, it can be seen that the present application can achieve the aforementioned purpose, and has met the requirements of the Patent Law.

[0174] Although the preferred embodiments of the present application have been described, those skilled in the art who have the benefit of the basic inventive concept can make additional changes and modifications to the embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application. The above description is merely preferred embodiments of the present application and is not intended to limit the present application. It should be noted that any modification, equivalent replacement and improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

[0175] It should be noted that the above description of the flow is merely for example and illustration, and does not limit the scope of the present application; various modifications and changes can be made to the flow under the guidance of the present application by those skilled in the art; however, these modifications and changes are still within the scope of the present application.

[0176] The above has described the basic concept, and it is obvious that the above-mentioned invention disclosure is only as an example and does not constitute a limitation on the present application for those skilled in the art after reading this application; although it is not explicitly stated here, those skilled in the art can make various modifications, improvements and modifications to the present application; such modifications, improvements and modifications are suggested in the present application, so such modifications, improvements and modifications still belong to the spirit and scope of the exemplary embodiments of the present application.

[0177] Meanwhile, specific words are used in the present application to describe the embodiments of the present application. For example, "one embodiment", "an embodiment", and / or "some embodiments" means a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that the "an embodiment" or "one embodiment" or "an alternative embodiment" mentioned in different positions in the present specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of the present application can be properly combined.

[0178] Moreover, as will be appreciated by persons skilled in the art, the present application is capable of being embodied with several different types of categories or circumstances of patentable subject matter including any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof. Accordingly, the various aspects of the present application can be embodied in hardware alone, in software alone, or in a combination of hardware and software. The above described hardware and software can be referred to as a "unit", "module" or "system". Furthermore, the various aspects of the present application can take the form of a computer program product embodied in one or more computer readable medium(s) having computer readable program code embodied thereon.

[0179] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, conventional procedural programming languages, such as the C programming language, Visual Basic, Fortran 2103, Perl, COBOL 2102, PHP, ABAP, dynamic programming languages, such as Python, Ruby and Groovy, or another programming language. The program code can execute entirely on the user's computer, or it can be executed as a stand-alone software package, or it can execute partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any network, such as a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet) or within a cloud computing environment, or as a service, such as Software as a Service (SaaS).

[0180] Moreover, the order of execution or sequence of processing elements and sequences, unless otherwise specifically denoted, can not be limited to the description as set forth herein, nor specifically exemplified in the examples thereof, and the use of numbering or letters in the examples is not intended to limit the scope of the application. Although the above disclosure discusses several exemplary embodiments of the application, it should be apparent that various changes and modifications can be made by persons skilled in the art. For example, while the implementation of various components described above can be embodied in hardware, it can also be implemented as a software solution, for example, as an installation on an existing server or mobile device.

[0181] For similar reasons, it is to be appreciated that the teachings of the present application provided herein can be applied to any embodiment of the present application, and that actual claims applied for or patent granted can be broader than any single, featured embodiment. Accordingly, a patent applicant has constructed and filed examples to particularly point out and distinctly claim those aspects which are regarded as novel and those aspects specifically shown.

Claims

1. A method for predicting and evaluating the response of a shear wave vibrator based on an integrated modeling approach, characterized in that: include: S1. Based on the geometric model of the shear wave vibrator, a numerical simulation model of the coupled vibration of the shear wave vibrator and the earth is constructed using finite element simulation software; S2. Determine the input feature data range and the number of sampling samples based on the shear wave vibrator-earth coupled vibration numerical simulation model. Based on the determined input feature data range and the determined number of sampling samples, use a progressive Latin hypercube sampling method to obtain experimental simulation samples of the input feature data, and input these samples into the shear wave vibrator-earth coupled vibration numerical simulation model to obtain a training data set and a validation data set of the vibrator output feature data under different input characteristics. S3. Construct a common time scale for the time series curve data of the training dataset and the validation dataset, map them to a unified time scale through linear interpolation and normalize them, and divide them to obtain time series simulation training data; S4. Using the time series simulation training data, establish a BP neural network regression prediction model with chaos-improved horizontal cross particle swarm optimization, optimize and integrate to obtain a complete time series prediction curve; S5. Based on the complete time series prediction curve, obtain the evaluation index data under different working conditions, construct the factor set and evaluation set, calculate the single factor evaluation matrix and optimize the dynamic weight matrix, and determine the adaptability evaluation results through fuzzy transformation and maximum membership principle; The specific steps of the chaos-improved lateral crossover particle swarm optimization include: using Singer mapping to generate a chaotic sequence; updating particle velocities through inertia weights, self-cognition coefficients, and social cognition coefficients; and adjusting the velocity range by combining a dynamic maximum flight speed limit mechanism; selecting adjacent particle pairs for dimensional crossover; generating new particle positions based on the combined particle centroids through random numbers and perturbation factors; and optimizing the number of neurons and learning rate of the BP neural network based on the mean square error of five-fold cross-validation. The construction steps of the BP neural network regression prediction model include: setting 4 neurons in the input layer corresponding to the excitation frequency, soil elastic modulus, soil cohesion and driving force amplitude; setting 4 neurons in the output layer corresponding to the vibration output force, energy transfer rate, excitation stress wave amplitude and contact surface strain; the number of neurons in the hidden layer is determined by the optimization algorithm; using the hyperbolic tangent function as the hidden layer activation function and the linear function as the output layer activation function, constructing an independent sub-model for each common time point and integrating them to obtain a complete time series prediction curve.

2. The integrated modeling shear wave vibrator response prediction and evaluation method according to claim 1 is characterized in that: The specific steps of constructing a numerical simulation model of the coupled vibration of a shear-wave controllable source vibrator and the earth include: establishing and simplifying the geometric model of the vibrator based on three-dimensional drawing and finite element simulation software, dividing the mesh using rigid body units and hexahedral units, setting the vibrator material to 45 steel and the earth to clay, defining the hard contact and Coulomb friction characteristics between the plate and the earth; setting the six-degree-of-freedom constraints of the earth, the static load of the weight of the source vehicle, and the dynamic load of the piston rod hydraulic pressure, using infinite elements to process the boundaries, and solving the model through ground stress analysis, static analysis, and implicit dynamics analysis.

3. The integrated modeling shear wave vibrator response prediction and evaluation method according to claim 1, characterized in that: Based on the numerical simulation model of the coupled vibration of a shear-wave controllable source vibrator and the earth, an evaluation index system is constructed using contact surface deformation, excited stress wave amplitude, vibration output force, and energy transfer rate. The standards of the evaluation index system are as follows: the larger the vibration displacement and the smaller the plastic deformation in the contact surface deformation, the better; the larger the excited stress wave amplitude and the fewer harmonic components, the better; the more uniform the distribution of the vibration output force, the better; and the more effective energy and the higher the transfer rate in the energy transfer, the better.

4. The integrated modeling shear wave vibrator response prediction and evaluation method according to claim 1 is characterized in that: The input characteristic data range is: excitation frequency 5-120Hz, soil elastic modulus 20-70MPa, soil cohesion 42-94KPa, and driving force amplitude 40%-90%.

5. The integrated modeling shear wave vibrator response prediction and evaluation method according to claim 1, characterized in that: The specific steps of the progressive Latin hypercube sampling method include: dividing the value range of each input dimension into n equally probable intervals, generating an initial sample subset, calculating the pairwise correlation using the Pearson matrix correlation coefficient, adding new sample points each time according to the principle of maximum pairwise correlation and optimizing the sampling point positions to achieve maximum stratification in each dimension, and combining input feature data points of different dimensions to generate experimental simulation samples.

6. The integrated modeling shear wave vibrator response prediction and evaluation method according to claim 1, characterized in that: The specific steps of constructing a common time scale include: calculating the average step length as the common time step according to the total number of points of the output feature data time series curve, creating an equally spaced time series from the start time to the end time based on the common time step; using linear interpolation to map the original time series data to the time series, normalizing the data to the range of [0,1] through the maximum and minimum normalization method, and dividing it into time series simulation training data according to the characteristics of the time series.

7. The integrated modeling shear wave vibrator response prediction and evaluation method according to claim 1, characterized in that: The specific steps of constructing the factor set and evaluation set include: the factor set is defined as {excitation frequency, driving force amplitude, soil elastic modulus, soil cohesion}, and the evaluation set is defined as {vibration output force, energy transfer rate, excitation stress wave amplitude, contact surface strain}; the single factor evaluation matrix of the factor set to the evaluation set is calculated based on the membership function, and the improved horizontal cross particle swarm optimization dynamic weight matrix is ​​used, where the weight matrix elements correspond to the importance of each factor under different working conditions.

8. The integrated modeling shear wave vibrator response prediction and evaluation method according to claim 1, characterized in that: The specific steps for determining the adaptability evaluation results include: mapping the fuzzy set of the factor set to the fuzzy set of the evaluation set through fuzzy transformation; according to the principle of comprehensive maximum membership, selecting the evaluation level corresponding to the maximum membership of the evaluation object to the grade fuzzy subset as a whole as the adaptability evaluation result of the shear wave controllable source vibrator in a complex surface environment.

Images