A method and system for inversion of airborne electromagnetic data based on approximate Jacobian matrix
By using the approximate Jacqueline matrix and the improved Laplace operator in the inversion of aeronautical electromagnetic data, the interpolation operator is optimized, and the problem of time-consuming calculation of Jacqueline matrix is solved and the inversion speed is improved.
Patent Information
- Application Number
- CN202410576270.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-10
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-05-10
AI Technical Summary
In the prior art, the calculation of the Jacobian matrix takes a long time, resulting in slow inversion of aeronautical electromagnetic data and lack of effective acceleration methods.
Using the method based on the approximate Jacqueline matrix, we quickly solve the inverse problem by approximating the gradient information, reduce the calculation amount, build the approximate Jacqueline matrix and use the improved Laplace operator to calculate the weighted terms, optimize the interpolation operator to reduce the Calculation of iterative system of equations.
Without affecting the inversion accuracy, the inversion time is significantly shortened and the calculation speed is improved.
Smart Images

Figure CN118244368B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of geophysical exploration inversion, and in particular relates to an airborne electromagnetic data inversion method and system based on an approximate Jacobian matrix. Background Art
[0002] In geophysical inversion problems, the calculation of the Jacobian matrix is the most time-consuming, which is the biggest factor restricting the speed of inversion calculation. The amount of airborne electromagnetic data is large, and the inversion speed is an important factor restricting the practicality of the algorithm. In order to speed up the inversion speed, it is necessary to improve the inversion process and reduce the amount of calculation required in the inversion process as much as possible. At present, there is a lack of an effective acceleration method to address the problem of slow inversion speed of time-domain airborne electromagnetic method. Summary of the invention
[0003] The present invention aims to solve the deficiencies of the prior art and proposes an airborne electromagnetic data inversion method and system based on an approximate Jacobian matrix to quickly solve the inverse problem by approximating the gradient information.
[0004] To achieve the above object, the present invention provides the following solutions:
[0005] An airborne electromagnetic data inversion method based on an approximate Jacobian matrix comprises the following steps:
[0006] Divide the inversion target area into grids, and set the initial model of the underground space based on the divided grids;
[0007] Constructing an airborne transient electromagnetic three-dimensional inversion objective function based on the initial model of the underground space, and calculating a weighted term of the model based on the observed data and the inversion objective function;
[0008] constructing an approximate Jacobian matrix, calculating an iterative equation group based on the approximate Jacobian matrix, updating the inversion objective function based on the iterative equation group, and obtaining an inversion model;
[0009] The airborne transient electromagnetic observation data is input into the inversion model for inversion to obtain the inversion results of various induced polarization parameters.
[0010] Preferably, the inversion objective function includes:
[0011]
[0012]
[0013]
[0014]
[0015] Among them, λ1 and λ2 are regularization factors, Wd is the data weighting matrix, W m is the improved model weighting matrix, s is the data difference matrix, d obs is the observed data, F(m) is the transient electromagnetic response of the underground space model, m is the parameter vector of the underground space model, m0 is the underground space reference model vector, and T is the transpose of the matrix.
[0016] Preferably, the method for calculating the weighted term includes: calculating the weighted term using an improved Laplace operator;
[0017] The diagonal elements of the improved Laplace operator are:
[0018]
[0019] Where m(i,j,k) is the electromagnetic parameter value in the (i,j,k)th model grid, i is the grid number in the x direction, j is the grid number in the y direction, k is the grid number in the z direction, and α is a constant;
[0020] The off-diagonal elements are arranged in the form of central differences, and the difference coefficients in the three directions are c x , c y and c z :
[0021]
[0022]
[0023]
[0024] Among them, w is the weighting coefficient, Δx is the grid size of the initial model of the underground space in the x direction, Δy is the grid size of the initial model of the underground space in the y direction, and Δz is the grid size of the initial model of the underground space in the z direction.
[0025] Preferably, the method for constructing the approximate Jacobian matrix includes:
[0026] The model parameter vector is derived using adjoint forward modeling to obtain the Jacobian matrix:
[0027]
[0028] Where J is the Jacobian matrix, L is the interpolation operator that converts the electric field component into the magnetic field component, P is the background electric field value, K is the forward operator, and E s is the electric field value of the scattered field;
[0029] Solve for the transposed matrix of the Jacobian matrix:
[0030]
[0031] Optimize the interpolation operator, select 4 field value calculation points in the neighborhood of the measurement point in the divided grid, and obtain an approximate interpolation operator L';
[0032] The approximate interpolation operator is used to replace the interpolation operator in the transposed matrix to obtain the approximate Jacobian matrix.
[0033] Preferably, the method for obtaining the iterative equation system is:
[0034] Calculating a first partial derivative matrix of the observed data with respect to the model variables and a second partial derivative matrix of the data difference with respect to the model variables;
[0035] Constructing update variables of model variables based on the inversion objective function and the second partial derivative matrix:
[0036]
[0037]
[0038]
[0039] Where J is the Jacobian matrix of the forward operator, s is the difference evaluation operator, F(m0) is the forward numerical simulation value of the reference model vector, and B s is the second partial derivative matrix of s with respect to the model variables, n iter is the number of iterations, b1, b2, c1 and c2 are constants;
[0040] The iterative equations are obtained based on the update variables:
[0041] m k+1 m k +△m
[0042]
[0043] Among them, m k is the model variable after the kth iteration, J k is the Jacobian matrix of the forward operator after the kth iteration, s k is the difference evaluation operator after the kth iteration, B sk For k The matrix of second partial derivatives with respect to the model variables.
[0044] The present invention also provides an airborne electromagnetic data inversion system based on an approximate Jacobian matrix, wherein the inversion system applies any of the above-mentioned inversion methods and comprises: a grid division module, an objective function construction module, an inversion model construction module and an inversion module;
[0045] The grid division module is used to divide the inversion target area into grids and set the initial model of the underground space based on the divided grids;
[0046] The objective function construction module constructs an airborne transient electromagnetic three-dimensional inversion objective function based on the initial model of the underground space, and calculates a weighted term of the model based on the observed data and the inversion objective function;
[0047] The inversion model construction module is used to construct an approximate Jacobian matrix, calculate an iterative equation group based on the approximate Jacobian matrix, and update the inversion objective function based on the iterative equation group to obtain an inversion model;
[0048] The inversion module inputs the airborne transient electromagnetic observation data into the inversion model for inversion to obtain the inversion results of various induced polarization parameters.
[0049] Preferably, the inversion objective function includes:
[0050]
[0051]
[0052]
[0053]
[0054] Among them, λ1 and λ2 are regularization factors, W d is the data weighting matrix, W m is the improved model weighting matrix, s is the data difference matrix, d obs is the observed data, F(m) is the transient electromagnetic response of the underground space model, m is the parameter vector of the underground space model, m0 is the underground space reference model vector, and T is the transpose of the matrix.
[0055] Preferably, the method for calculating the weighted term includes: calculating the weighted term using an improved Laplace operator;
[0056] The diagonal elements of the improved Laplace operator are:
[0057]
[0058] Where m(i,j,k) is the electromagnetic parameter value in the (i,j,k)th model grid, i is the grid number in the x direction, j is the grid number in the y direction, k is the grid number in the z direction, and α is a constant;
[0059] The off-diagonal elements are arranged in the form of central differences, and the difference coefficients in the three directions are c x , cy and c z :
[0060]
[0061]
[0062]
[0063] Among them, w is the weighting coefficient, Δx is the grid size of the initial model of the underground space in the x direction, Δy is the grid size of the initial model of the underground space in the y direction, and Δz is the grid size of the initial model of the underground space in the z direction.
[0064] Preferably, the method for constructing the approximate Jacobian matrix includes:
[0065] The model parameter vector is derived using adjoint forward modeling to obtain the Jacobian matrix:
[0066]
[0067] Where J is the Jacobian matrix, L is the interpolation operator that converts the electric field component into the magnetic field component, P is the background electric field value, K is the forward operator, and E s is the electric field value of the scattered field;
[0068] Solve for the transposed matrix of the Jacobian matrix:
[0069]
[0070] Optimize the interpolation operator, select 4 field value calculation points in the neighborhood of the measurement point in the divided grid, and obtain an approximate interpolation operator L';
[0071] The approximate interpolation operator is used to replace the interpolation operator in the transposed matrix to obtain the approximate Jacobian matrix.
[0072] Preferably, the method for obtaining the iterative equation system is:
[0073] Calculating a first partial derivative matrix of the observed data with respect to the model variables and a second partial derivative matrix of the data difference with respect to the model variables;
[0074] Constructing update variables of model variables based on the inversion objective function and the second partial derivative matrix:
[0075]
[0076]
[0077]
[0078] Where J is the Jacobian matrix of the forward operator, s is the difference evaluation operator, F(m0) is the forward numerical simulation value of the reference model vector, and B s is the second partial derivative matrix of s with respect to the model variables, n iter is the number of iterations, b1, b2, c1 and c2 are constants;
[0079] The iterative equations are obtained based on the update variables:
[0080] m k+1 m k +△m
[0081]
[0082] Among them, m k is the model variable after the kth iteration, J k is the Jacobian matrix of the forward operator after the kth iteration, s k is the difference evaluation operator after the kth iteration, B sk For k The matrix of second partial derivatives with respect to the model variables.
[0083] Compared with the prior art, the present invention has the following beneficial effects:
[0084] The present invention solves the inverse problem quickly by approximating the gradient information. By means of this approximate calculation method, the number of required calculation equations is reduced without affecting the inversion accuracy, thus greatly shortening the time required for inversion. BRIEF DESCRIPTION OF THE DRAWINGS
[0085] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0086] Figure 1 A schematic diagram of a method flow of an embodiment of the present invention;
[0087] Figure 2 It is a schematic diagram of an XOY plane for approximately calculating the Jacobian matrix of multiple measurement points according to an embodiment of the present invention;
[0088] Figure 3 A schematic diagram of the system structure of an embodiment of the present invention;
[0089] Figure 4 A schematic diagram of an initial spatial model according to an embodiment of the present invention;
[0090] Figure 5Schematic diagram for comparing the inversion results of four induced polarization parameters according to an embodiment of the present invention;
[0091] Figure 6 Schematic diagram of the effect of Jacobian matrix calculation methods with the same approximation degree on calculation speed according to an embodiment of the present invention. DETAILED DESCRIPTION
[0092] 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.
[0093] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0094] Embodiment 1
[0095] In this embodiment, if Figure 1 As shown, an airborne electromagnetic data inversion method based on an approximate Jacobian matrix includes the following steps:
[0096] S1. Divide the inversion target area into grids and set the initial model of the underground space based on the divided grids.
[0097] In this embodiment, the inversion target area is divided into hexahedral grids, and an initial model of the underground space is set. If there is a priori model, it can be set as the priori model, and if there is no priori model, it can be set as a uniform half-space model. The initial model is forward calculated to obtain the transient electromagnetic response of the initial model of the underground space.
[0098] S2. Construct the airborne transient electromagnetic three-dimensional inversion objective function based on the initial model of the underground space, and calculate the weighted terms of the model based on the observed data and the inversion objective function.
[0099] The inversion objective function includes:
[0100]
[0101]
[0102]
[0103]
[0104] Among them, λ1 and λ2 are regularization factors, W d is the data weighting matrix, W mis the improved model weighting matrix, s is the data difference matrix, d obs is the observed data, F(m) is the transient electromagnetic response of the underground space model, m is the parameter vector of the underground space model, m0 is the underground space reference model vector, and T is the transpose of the matrix.
[0105] It should be noted that m here is the model parameter vector of the current iteration; if the prior model is not used, m0 is the reference model vector of the previous iteration of the underground space; if the prior model is used, m0 is the reference model vector of the prior model in each iteration.
[0106] The observed data (induced electromotive force attenuation curve) and the measuring point position are input into the inversion program, and the data weighting items are determined according to the data. The data weighting item diagonal matrix is expressed as follows:
[0107]
[0108] The method for calculating the weighted term includes: calculating the weighted term using an improved Laplace operator;
[0109] The diagonal elements of the modified Laplacian operator are:
[0110]
[0111] Where m(i,j,k) is the electromagnetic parameter value in the (i,j,k)th model grid, i is the grid number in the x direction, j is the grid number in the y direction, k is the grid number in the z direction, and α is a constant;
[0112] The off-diagonal elements are arranged in the form of central differences, and the difference coefficients in the three directions are c x , c y and c z :
[0113]
[0114]
[0115]
[0116] Among them, w is the weighting coefficient, Δx is the grid size of the initial model of the underground space in the x direction, Δy is the grid size of the initial model of the underground space in the y direction, and Δz is the grid size of the initial model of the underground space in the z direction.
[0117] S3. Construct an approximate Jacobian matrix, calculate an iterative equation group based on the approximate Jacobian matrix, update the inversion objective function based on the iterative equation group, and obtain an inversion model.
[0118] The construction methods of the approximate Jacobian matrix include:
[0119] The Jacobian matrix is obtained by using the adjoint forward modeling to derive the model parameter vector:
[0120]
[0121] Where J is the Jacobian matrix, L is the interpolation operator that converts the electric field component into the magnetic field component, P is the background electric field value, K is the forward operator, and E s is the electric field value of the scattered field;
[0122] Solve for the transpose of the Jacobian matrix:
[0123]
[0124] The expression of the designed intermediate variable v is:
[0125] v=K -1 L
[0126] The calculation equations can be written as:
[0127] Kv=L′
[0128] Optimize the interpolation operator, select 4 field value calculation points in the neighborhood of the measurement point in the divided grid, and obtain the approximate interpolation operator L': In this embodiment, Figure 2 As shown, the size of the interpolation operator L in the conventional Jacobian matrix calculation method is (Nx*N y ,3*Nx*N y *Nz), at least Nx*N needs to be calculated y The system of equations ( Figure 1 If four field value calculation points are taken in the neighborhood of the measurement point, the size of the approximate operator matrix L' is (4*N p ,3*N x *N y *N z ), only 4*N needs to be calculated p The equations are N, and the measuring point is p The number of N is much smaller than the number of field value calculation points x *N y If further approximation is made and only one field value calculation point is taken within the domain, the number of equations calculated can be reduced to N at most. p This greatly reduces the amount of calculation;
[0129] The approximate interpolation operator is replaced by the interpolation operator in the transposed matrix to obtain the approximate Jacobian matrix.
[0130] The method to obtain the iterative equation system is:
[0131] Calculate the first partial derivative matrix of the observed data with respect to the model variables, and the second partial derivative matrix of the data difference with respect to the model variables; in this embodiment, first calculate the frequency domain partial derivative matrix of the observed data with respect to the model variables, and then convert the frequency domain partial derivative matrix into a time domain partial derivative matrix by time-frequency transformation to obtain the first partial derivative matrix; calculate the data difference and the second partial derivative matrix of the data difference with respect to the model variables according to the multi-property parameter model;
[0132] Construct the update variables of the model variables based on the inversion objective function and the second partial derivative matrix:
[0133]
[0134]
[0135]
[0136] Where J is the Jacobian matrix of the forward operator, s is the difference evaluation operator, F(m0) is the forward numerical simulation value of the reference model vector, and B s is the second partial derivative matrix of s with respect to the model variables, n iter is the number of iterations, b1, b2, c1 and c2 are constants;
[0137] The iterative equation system is obtained based on the updated variables:
[0138] m k+1 =m k +△m
[0139]
[0140] Among them, m k is the model variable after the kth iteration, J k is the Jacobian matrix of the forward operator after the kth iteration, s k is the difference evaluation operator after the kth iteration, B sk For k The matrix of second partial derivatives with respect to the model variables.
[0141] The updated model is input into the forward modeling program to compare the data fitting difference between the inversion data and the observed data. If the fitting difference is small enough to meet the set threshold, the inversion results of different IP parameters are output, and the inversion results of IP parameters are plotted as the final inversion model; if the inversion iteration termination condition is not met, the first partial derivative matrix and the second partial derivative matrix are recalculated.
[0142] S4. Input the airborne transient electromagnetic observation data into the inversion model for inversion to obtain the inversion results of various induced polarization parameters.
[0143] Embodiment 2
[0144] In this embodiment, if Figure 3 As shown, an airborne electromagnetic data inversion system based on approximate Jacobian matrix includes: a grid division module, an objective function construction module, an inversion model construction module and an inversion module.
[0145] The grid division module is used to divide the inversion target area into grids and set the initial model of the underground space based on the divided grids.
[0146] In this embodiment, the inversion target area is divided into hexahedral grids, and an initial model of the underground space is set. If there is a priori model, it can be set as the priori model, and if there is no priori model, it can be set as a uniform half-space model. The initial model is forward calculated to obtain the transient electromagnetic response of the initial model of the underground space.
[0147] The objective function construction module constructs the airborne transient electromagnetic three-dimensional inversion objective function based on the initial model of the underground space, and calculates the weighted items of the model based on the observed data and the inversion objective function.
[0148] The inversion objective function includes:
[0149]
[0150]
[0151]
[0152]
[0153] Among them, λ1 and λ2 are regularization factors, W d is the data weighting matrix, W m is the improved model weighting matrix, s is the data difference matrix, d obs is the observed data, F(m) is the transient electromagnetic response of the underground space model, m is the parameter vector of the underground space model, m0 is the underground space reference model vector, and T is the transpose of the matrix.
[0154] It should be noted that m here is the model parameter vector of the current iteration; if the prior model is not used, m0 is the reference model vector of the previous iteration of the underground space; if the prior model is used, m0 is the reference model vector of the prior model in each iteration.
[0155] The observed data (induced electromotive force attenuation curve) and the measuring point position are input into the inversion program, and the data weighting items are determined according to the data. The data weighting item diagonal matrix is expressed as follows:
[0156]
[0157] The method for calculating the weighted term includes: calculating the weighted term using an improved Laplace operator;
[0158] The diagonal elements of the modified Laplacian operator are:
[0159]
[0160] Where m(i,j,k) is the electromagnetic parameter value in the (i,j,k)th model grid, i is the grid number in the x direction, j is the grid number in the y direction, k is the grid number in the z direction, and α is a constant;
[0161] The off-diagonal elements are arranged in the form of central differences, and the difference coefficients in the three directions are c x , c y and c z :
[0162]
[0163]
[0164]
[0165] Among them, w is the weighting coefficient, Δx is the grid size of the initial model of the underground space in the x direction, Δy is the grid size of the initial model of the underground space in the y direction, and Δz is the grid size of the initial model of the underground space in the z direction.
[0166] The inversion model building module is used to construct an approximate Jacobian matrix, calculate an iterative equation group based on the approximate Jacobian matrix, update the inversion objective function based on the iterative equation group, and obtain an inversion model.
[0167] The construction methods of the approximate Jacobian matrix include:
[0168] The Jacobian matrix is obtained by using the adjoint forward modeling to derive the model parameter vector:
[0169]
[0170] Where J is the Jacobian matrix, L is the interpolation operator that converts the electric field component into the magnetic field component, P is the background electric field value, K is the forward operator, and E s is the electric field value of the scattered field;
[0171] Solve for the transpose of the Jacobian matrix:
[0172]
[0173] Optimize the interpolation operator, select 4 field value calculation points in the neighborhood of the measurement point in the divided grid, and obtain the approximate interpolation operator L': In this embodiment, the size of the interpolation operator L in the conventional Jacobian matrix calculation method is (N x *N y ,3*N x *N y *N z ), at least N x *N y If four field value calculation points are taken in the neighborhood of the measurement point, the size of the approximate operator matrix L' is (4*N p ,3*N x *N y *N z ), only 4*N needs to be calculated p The equations are N, and the measuring point is p The number of N is much smaller than the number of field value calculation points x *N y If further approximation is made and only one field value calculation point is taken within the domain, the number of equations calculated can be reduced to N at most. p This greatly reduces the amount of calculation;
[0174] The approximate interpolation operator is replaced by the interpolation operator in the transposed matrix to obtain the approximate Jacobian matrix.
[0175] The method to obtain the iterative equation system is:
[0176] Calculate the first partial derivative matrix of the observed data with respect to the model variables, and the second partial derivative matrix of the data difference with respect to the model variables; in this embodiment, first calculate the frequency domain partial derivative matrix of the observed data with respect to the model variables, and then convert the frequency domain partial derivative matrix into a time domain partial derivative matrix by time-frequency transformation to obtain the first partial derivative matrix; calculate the data difference and the second partial derivative matrix of the data difference with respect to the model variables according to the multi-property parameter model;
[0177] Construct the update variables of the model variables based on the inversion objective function and the second partial derivative matrix:
[0178]
[0179]
[0180]
[0181] Where J is the Jacobian matrix of the forward operator, s is the difference evaluation operator, F(m0) is the forward numerical simulation value of the reference model vector, and B s is the second partial derivative matrix of s with respect to the model variables, n iteris the number of iterations, b1, b2, c1 and c2 are constants;
[0182] The iterative equation system is obtained based on the updated variables:
[0183] m k+1 =m k +△m
[0184]
[0185] Among them, m k is the model variable after the kth iteration, J k is the Jacobian matrix of the forward operator after the kth iteration, s k is the difference evaluation operator after the kth iteration, B sk For k The matrix of second partial derivatives with respect to the model variables.
[0186] The updated model is input into the forward modeling program to compare the data fitting difference between the inversion data and the observed data. If the fitting difference is small enough to meet the set threshold, the inversion results of different IP parameters are output, and the inversion results of IP parameters are plotted as the final inversion model; if the inversion iteration termination condition is not met, the first partial derivative matrix and the second partial derivative matrix are recalculated.
[0187] The inversion module inputs the airborne transient electromagnetic observation data into the inversion model for inversion to obtain the inversion results of various induced polarization parameters.
[0188] Embodiment 3
[0189] In this embodiment, a set of Figure 4The model is calculated and a 28*28*18 hexahedral grid is designed. The radius of the transmitting coil is 15m, the transmitting current is 100A, and the transmitting coil and the receiving coil are both located in the air at 30m, with a total of 121 measuring points. The calculation area is divided into 20 grids with equal intervals in the x and y directions, with a grid size of 20m. The grid size in the z direction gradually increases from the surface to the depth. The first grid size is 20m, and each grid size thereafter is 1.1 times the previous grid size; the boundary and air area include 8 expanded grids in the x and y directions and 8 grids in the z direction. The synthetic model is a uniform half-space buried with an inclined low-resistance polarizer, which gradually deepens from the negative direction of X and Y to the positive direction of X and Y. The resistivity of the surrounding rock is 200Ω·m, the polarizability is 0.01, the time constant is 0.0001, and the frequency correlation coefficient is 0.01; the resistivity of the target body is 50Ω·m, the polarizability is 0.5, the time constant is 0.005, and the frequency correlation coefficient is 0.5. The initial model of the inversion is designed as a uniform half-space with a resistivity of 100Ω·m, a polarizability of 0.1, a time constant of 0.001, and a frequency correlation coefficient of 0.1.
[0190] Figure 5 The inversion results obtained by the two Jacobian matrix calculation methods were compared. The approximate Jacobian matrix calculation method uses the method with the highest approximation, that is, only taking the value of one field value calculation point in the neighborhood to replace the value of its sampling point. From the results, there is no big difference between the inversion results obtained by the exact Jacobian matrix calculation method and the approximate Jacobian matrix calculation method, and both can restore the electrical characteristics of the underground medium well.
[0191] Figure 6 The time consumed for each iteration and the speedup ratio of different approximate Jacobian calculation methods are shown. It can be clearly seen from the figure that this Jacobian matrix approximation method can greatly reduce the time required for each iteration in the inversion. When the largest approximation ratio is selected for calculation, the calculation speed is 2.88 times that of the exact Jacobian matrix calculation method.
[0192] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.
Claims
1. A method for inversion of airborne electromagnetic data based on an approximate Jacobian matrix, characterized in that: The following steps are involved: Divide the inversion target area into grids, and set the initial model of the underground space based on the divided grids; Constructing an airborne transient electromagnetic three-dimensional inversion objective function based on the initial model of the underground space, and calculating a weighted term of the model based on the observed data and the inversion objective function; constructing an approximate Jacobian matrix, calculating an iterative equation group based on the approximate Jacobian matrix, updating the inversion objective function based on the iterative equation group, and obtaining an inversion model; Inputting the airborne transient electromagnetic observation data into the inversion model for inversion to obtain the inversion results of various induced polarization parameters; The inversion objective function includes: Among them, λ1 and λ2 are regularization factors, W d is the data weighting matrix, W m is the improved model weighting matrix, s is the data difference matrix, d obs is the observed data, F(m) is the transient electromagnetic response of the underground space model, m is the parameter vector of the underground space model, m0 is the underground space reference model vector, and T is the transpose of the matrix; The method for constructing the approximate Jacobian matrix includes: The Jacobian matrix is obtained by using the adjoint forward modeling to derive the parameter vector of the underground space model: Among them, J is the Jacobian matrix of the forward operator, L is the interpolation operator that converts the electric field component into the magnetic field component, P is the background field electric field value, K is the forward operator, and E s is the electric field value of the scattered field; Solve for the transposed matrix of the Jacobian matrix: Optimize the interpolation operator, select 4 field value calculation points in the neighborhood of the measurement point in the divided grid, and obtain an approximate interpolation operator L'; Replacing the interpolation operator in the transposed matrix of the Jacobian matrix with the approximate interpolation operator to obtain the approximate Jacobian matrix; The method for obtaining the iterative equation system is: Calculating a first partial derivative matrix of the observed data with respect to the model variables and a second partial derivative matrix of the data difference with respect to the model variables; Constructing update variables of model variables based on the inversion objective function and the second partial derivative matrix: Where J is the Jacobian matrix of the forward operator, s is the data difference matrix, F(m0) is the forward numerical simulation value of the reference model vector, and B s is the second partial derivative matrix of s with respect to the model variables, n iter is the number of iterations, b1, b2, c1 and c2 are constants; The iterative equations are obtained based on the update variables: m k+1 =m k +△m Among them, m k is the model variable after the kth iteration, J k is the Jacobian matrix of the forward operator after the kth iteration, s k is the data difference matrix after the kth iteration, B sk For k The second partial derivative matrix with respect to the model variables; The updated model is input into the forward modeling program to compare the data fitting error between the inversion data and the observation data. If the fitting error is small enough to meet the set threshold, the inversion results of different induced polarization parameters are output, and the inversion results of the induced polarization parameters are plotted as the final inversion model. If the inversion iteration termination condition is not met, the first partial derivative matrix and the second partial derivative matrix are recalculated.
2. The method for inversion of airborne electromagnetic data based on an approximate Jacobian matrix according to claim 1, characterized in that: The method for calculating the weighted term includes: calculating the weighted term using an improved Laplace operator; The diagonal elements of the improved Laplace operator are: Where m(i,j,k) is the electromagnetic parameter value in the (i,j,k)th model grid, i is the grid number in the x direction, j is the grid number in the y direction, k is the grid number in the z direction, and α is a constant; The off-diagonal elements are arranged in the form of central differences, and the difference coefficients in the three directions are c x , c y and c z : Among them, w is the weighting coefficient, Δx is the grid size of the initial model of the underground space in the x direction, Δy is the grid size of the initial model of the underground space in the y direction, and Δz is the grid size of the initial model of the underground space in the z direction.
3. An airborne electromagnetic data inversion system based on an approximate Jacobian matrix, wherein the inversion system applies the inversion method according to any one of claims 1 to 2, characterized in that: include: Meshing module, objective function building module, inversion model building module and inversion module; The grid division module is used to divide the inversion target area into grids and set the initial model of the underground space based on the divided grids; The objective function construction module constructs an airborne transient electromagnetic three-dimensional inversion objective function based on the initial model of the underground space, and calculates a weighted term of the model based on the observed data and the inversion objective function; The inversion model construction module is used to construct an approximate Jacobian matrix, calculate an iterative equation group based on the approximate Jacobian matrix, and update the inversion objective function based on the iterative equation group to obtain an inversion model; The inversion module inputs the airborne transient electromagnetic observation data into the inversion model for inversion to obtain the inversion results of various induced polarization parameters; The inversion objective function includes: Among them, λ1 and λ2 are regularization factors, W d is the data weighting matrix, W m is the improved model weighting matrix, s is the data difference matrix, d obs is the observed data, F(m) is the transient electromagnetic response of the underground space model, m is the parameter vector of the underground space model, m0 is the underground space reference model vector, and T is the transpose of the matrix; The method for constructing the approximate Jacobian matrix includes: The Jacobian matrix is obtained by using the adjoint forward modeling to derive the parameter vector of the underground space model: Among them, J is the Jacobian matrix of the forward operator, L is the interpolation operator that converts the electric field component into the magnetic field component, P is the background field electric field value, K is the forward operator, and E s is the electric field value of the scattered field; Solve for the transposed matrix of the Jacobian matrix: Optimize the interpolation operator, select 4 field value calculation points in the neighborhood of the measurement point in the divided grid, and obtain an approximate interpolation operator L'; Replacing the interpolation operator in the transposed matrix of the Jacobian matrix with the approximate interpolation operator to obtain the approximate Jacobian matrix; The method for obtaining the iterative equation system is: Calculating a first partial derivative matrix of the observed data with respect to the model variables and a second partial derivative matrix of the data difference with respect to the model variables; Constructing update variables of model variables based on the inversion objective function and the second partial derivative matrix: Where J is the Jacobian matrix of the forward operator, s is the data difference matrix, F(m0) is the forward numerical simulation value of the reference model vector, and B s is the second partial derivative matrix of s with respect to the model variables, n iter is the number of iterations, b1, b2, c1 and c2 are constants; The iterative equations are obtained based on the update variables: m k+1 =m k +Δm Among them, m k is the model variable after the kth iteration, J k is the Jacobian matrix of the forward operator after the kth iteration, s k is the data difference matrix after the kth iteration, B sk For k The second partial derivative matrix with respect to the model variables; The updated model is input into the forward modeling program to compare the data fitting error between the inversion data and the observation data. If the fitting error is small enough to meet the set threshold, the inversion results of different induced polarization parameters are output, and the inversion results of the induced polarization parameters are plotted as the final inversion model. If the inversion iteration termination condition is not met, the first partial derivative matrix and the second partial derivative matrix are recalculated.
4. The airborne electromagnetic data inversion system based on the approximate Jacobian matrix according to claim 3, characterized in that: The method for calculating the weighted term includes: calculating the weighted term using an improved Laplace operator; The diagonal elements of the improved Laplace operator are: Where m(i,j,k) is the electromagnetic parameter value in the (i,j,k)th model grid, i is the grid number in the x direction, j is the grid number in the y direction, k is the grid number in the z direction, and α is a constant; The off-diagonal elements are arranged in the form of central differences, and the difference coefficients in the three directions are c x , c y and c z : Among them, w is the weighting coefficient, Δx is the grid size of the initial model of the underground space in the x direction, Δy is the grid size of the initial model of the underground space in the y direction, and Δz is the grid size of the initial model of the underground space in the z direction.
