Ground surface settlement measurement data real-time fitting method based on evolutionary algorithm

Abstract

The invention discloses a ground surface settlement measurement data real-time fitting method based on an evolutionary algorithm. The method comprises the following steps: S1, acquiring a ground surface settlement measurement data set; s2, preprocessing the ground surface settlement measurement data set; s3, generating a ground surface settlement measurement data feature matrix; s4, constructing a radial basis function neural network model for ground surface settlement measurement data fitting; s5, optimizing parameters of the initial ground surface settlement measurement data fitting function by adopting an alpha evolutionary algorithm; and S6, adopting the optimized ground surface settlement measurement data fitting function to predict the input ground surface settlement measurement data. According to the method, the calculation efficiency, the nonlinear expression capability and the global optimization capability of ground surface settlement measurement data fitting are effectively improved, and the method can be widely applied to multiple scenes of ground surface settlement monitoring, infrastructure safety evaluation and mining area collapse prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of surface subsidence measurement, and particularly to a real-time fitting method for surface subsidence measurement data based on an evolutionary algorithm. Background Art

[0002] With the development of surface monitoring technology, surface subsidence measurement has become an important research content in multiple fields such as infrastructure safety assessment, mining area subsidence prediction, groundwater resource management, and geological disaster warning. Currently, surface subsidence measurement usually relies on multiple monitoring means such as global navigation satellite systems, synthetic aperture radars, and lidars to obtain surface subsidence data. However, surface subsidence data is often affected by environmental factors, measurement errors, and multi-source data fusion biases, resulting in large data noise, obvious nonlinear characteristics, and a large time-scale span, making it a challenge to accurately fit the surface subsidence trend.

[0003] At present, the fitting methods for surface subsidence measurement data mainly include traditional statistical regression methods, machine learning methods, and physical modeling methods. Among them, traditional statistical regression methods rely on fixed mathematical models and are difficult to adapt to the dynamic changes of complex nonlinear subsidence processes. The fitting accuracy is low in the case where the subsidence rate changes significantly with time. Machine learning methods can capture the nonlinear characteristics of surface subsidence data to a certain extent, but usually rely on large-scale labeled data for training and have a large computational overhead, and are prone to computational delays in high-frequency data stream processing. In addition, although physical modeling methods can deduce the subsidence trend based on geomechanics principles, they have a strong dependence on input parameters and are difficult to adapt to the subsidence characteristics of different regions and different environments in practical engineering applications.

[0004] Therefore, the existing technologies have problems such as low computational efficiency, the fitting accuracy being affected by dynamic changes, the optimization process being prone to falling into local optima, lack of self-adaptability, and insufficient ability to identify abnormal data in the real-time fitting of surface subsidence measurement data. There is an urgent need for a method that can improve computational efficiency, enhance the self-adaptability of the model, and optimize the fitting accuracy to meet the real-time monitoring requirements of surface subsidence measurement. Summary of the Invention

[0005] An object of the present invention is to propose a real-time fitting method for surface subsidence measurement data based on an evolutionary algorithm. The present invention effectively improves the computational efficiency, nonlinear expression ability, and global optimization ability of surface subsidence measurement data fitting, and can be widely applied to multiple scenarios such as surface subsidence monitoring, infrastructure safety assessment, and mining area subsidence prediction.

[0006] A real-time fitting method for surface subsidence measurement data based on an evolutionary algorithm according to an embodiment of the present invention includes the following steps: S1. Obtain a surface subsidence measurement data set;

[0007] S2. Preprocess the surface settlement measurement dataset to obtain a standardized surface settlement measurement dataset;

[0008] S3. Perform time series analysis on the standardized surface settlement measurement data to generate a surface settlement measurement data feature matrix;

[0009] S4. Construct a radial basis function neural network model for fitting surface settlement measurement data, and initialize the parameters of the radial basis function neural network model to obtain an initial surface settlement measurement data fitting function;

[0010] S5. Optimize the parameters of the initial surface settlement measurement data fitting function using the alpha evolution algorithm to obtain an optimized surface settlement measurement data fitting function;

[0011] S6. Use the optimized surface settlement measurement data fitting function to predict the input surface settlement measurement data, output a surface settlement trend fitting curve, and calculate fitting error, trend change rate, and prediction confidence interval indicators.

[0012] Optionally, S1 includes the following steps:

[0013] S11. Obtain a surface settlement measurement dataset, where the surface settlement measurement dataset includes global navigation satellite system measurement data, synthetic aperture radar measurement data, and lidar point cloud measurement data:

[0014] D GNSS ={(x i ,y i ,t i ,h i )|i = 1, 2,..., N GNSS};

[0015] D InSAR ={(x j ,y j ,t j ,h j )|j = 1, 2,..., N InSAR};

[0016] D LiDAR ={(x k ,y k ,t k ,h k )|k = 1, 2,..., N LiDAR};

[0017] where x, y represent the geographical coordinate positions, t represents the timestamp, h represents the surface height value, N GNSS ,N InSAR ,N LiDARare the number of sampling points of the global navigation satellite system measurement data, synthetic aperture radar measurement data, and lidar point cloud measurement data, respectively;

[0018] S12. Perform a spatial reference coordinate system conversion on the surface subsidence measurement data set to unify the coordinate reference of the surface subsidence measurement data set, and obtain the surface subsidence measurement data set in the unified spatial reference coordinate system:

[0020] D aligned = D' GNSS ∪D' InSAR ∪D' LiDAR ;

[0022] where D' GNSS 、D' InSAR and D' LiDAR represent the global navigation satellite system measurement data, synthetic aperture radar measurement data, and lidar point cloud measurement data in the unified spatial reference coordinate system, respectively.

[0023] Optionally, the S2 includes the following steps:

[0024] S21. Use the wavelet transform method to denoise the surface subsidence measurement data set D aligned in the unified spatial reference coordinate system, remove the random noise in the surface subsidence measurement data, and obtain the denoised surface subsidence measurement data set;

[0025] S22. Use the Mahalanobis distance analysis method to detect abnormal data in the denoised surface subsidence measurement data set, and calculate the abnormality degree of each data point:

[0026]

[0027] where M i is the Mahalanobis distance of the surface subsidence measurement data point, μ h is the mean vector of the surface subsidence measurement data height, and Σ h is the covariance matrix of the surface subsidence measurement data height;

[0028] Set a threshold M th . When M i > M th , determine the data point as an abnormal data point and mark the abnormal data point set;

[0029] S23. Perform interpolation correction on the abnormal data point set, use the spline interpolation method to calculate the correction value of the abnormal data point, and replace the original data point with the corrected abnormal data point to obtain the corrected surface subsidence measurement data set;

[0030] S24. Standardize the corrected surface settlement measurement dataset using the mean-standard deviation normalization method to finally obtain the standardized surface settlement measurement dataset D standardized 。

[0031] Optionally, the S3 includes the following steps:

[0032] S31. Based on the changing trend of time series data, calculate the surface settlement rate characteristics of each measurement point using the time difference method to reflect the degree of change in surface settlement within different time intervals:

[0033]

[0034] Among them, v i represents the surface settlement rate at the i-th moment, is the standardized surface height value of the i-th measurement point, t i -t i-1 is the time interval between two consecutive observations for calculating the instantaneous rate of surface settlement;

[0035] S32. Capture the surface settlement acceleration characteristics by calculating the rate of change of the settlement rate to reflect the intensification or mitigation of the surface settlement trend:

[0036]

[0037] Among them, a i represents the surface settlement acceleration of the i-th measurement point, reflecting the changing trend of the settlement rate at this point at time t i moment, v i is the surface settlement rate of the i-th measurement point, representing the intensity of the settlement change at the current moment, t i -t i-1 is the time interval;

[0038] S33. Use the polynomial fitting method to perform trend modeling on the surface settlement data to depict the global trend of the non-linear change of surface settlement:

[0039]

[0040] Among them, is the fitted settlement trend value of the i-th measurement point, c0, c1,..., c n are the fitting coefficients, t i is the timestamp;

[0041] S34. Use the fast Fourier transform method to transform the settlement data into the frequency domain to identify potential periodic characteristics of surface settlement:

[0042]

[0043] Among them, H(f) represents the spectral amplitude of the surface settlement data at the f-th frequency, reflecting the dominant frequency of the periodic settlement. j is the imaginary unit, and the main frequency f with the largest amplitude in the spectrum is selected max as the main periodic feature of the surface settlement:

[0044] f max = argmaxH(f);

[0045] Among them, f max represents the main periodic information of the settlement data, and is used to analyze the periodic settlement phenomenon caused by factors such as geological structure, climate change or underground engineering construction;

[0046] S35. Based on the comprehensively calculated surface settlement rate characteristics, surface settlement acceleration characteristics, surface settlement non-linear change trend characteristics and surface settlement periodic characteristics, construct a surface settlement measurement data feature matrix:

[0047]

[0048] Optionally, the S4 includes the following steps:

[0049] S41. Construct a radial basis function neural network model for fitting surface settlement measurement data, set the input layer of the radial basis function neural network model, and the input layer receives the surface settlement measurement data feature matrix F subsidence as the input. Let the number of neurons in the input layer be d, then the input layer neurons are represented as:

[0050] X = [x1, x2,..., x d ;

[0051] Among them, X is the input neuron vector, and x i represents the characteristic variable of the surface settlement measurement data, and d = 4 corresponds to the dimensions of the rate characteristic, acceleration characteristic, non-linear trend characteristic and periodic characteristic;

[0052] S42. Set the hidden layer of the radial basis function neural network model, and the hidden layer uses the Gaussian radial basis function as the activation function:

[0053]

[0054] Among them, H j is the output of the j-th hidden layer neuron, X is the input neuron vector, C j is the j-th center vector, σ j is the scale parameter of the j-th radial basis function, controlling the response range of the neuron, ∥X - C j∥ represents the Euclidean distance between the input neurons and the center vector. The total number of neurons in the hidden layer is set to m, and the output vector of the hidden layer is expressed as:

[0055] H = [H1, H2,..., H j ;

[0056] S44. Set the output layer of the radial basis function neural network model. The output layer is used to calculate the fitting result of the surface settlement measurement data and obtain the initial surface settlement measurement data fitting function:

[0057]

[0058] where y is the fitting output of the surface settlement measurement data, W is the weight matrix from the hidden layer to the output layer, b is the bias term of the output layer, represents the fitting surface settlement height value of the i-th measurement point.

[0059] Optionally, the S5 includes the following steps:

[0060] S51. Generate initial solutions for multiple fitting parameters to form an optimization population. Set the population size to P and initialize the population individual Θ 0 :

[0061]

[0062] where, is the parameter vector of the p-th individual, including the weight matrix W, the bias term b, and the hidden layer center vector C of the radial basis function neural network, that is:

[0063]

[0064] S52. Combine the surface settlement measurement data feature matrix F subsidence of the sample dimension N to calculate the fitness evaluation sample size:

[0065] P = αN;

[0066] where α is the population size coefficient;

[0067] S53. Use the mean square error and the model complexity penalty term as the optimization objective to calculate the fitness of the population individuals. Set the fitness function f(θ p ) to measure the fitting error of the individual:

[0068]

[0069] where, is the true surface settlement height value, is the fitting height value calculated using the parameters θ of the p-th individual p calculated, λ∥θp ∥ 2 is the regularization term;

[0070] S54. Optimize using the alpha evolutionary algorithm and adaptively adjust the evolutionary operator according to the evolutionary stage:

[0071] In the global exploration stage (G < G1), use the alpha hybrid strategy to combine genetic algorithm and differential evolution for wide-area search:

[0072]

[0073] where θ r g , θ s g , θ t g are randomly selected different individuals, and F is the scaling factor;

[0074] Adopt the elite retention mechanism to retain the current optimal individual:

[0075]

[0076] In the local development stage (G1 ≤ G < G2), use the alpha adaptive mutation strategy to select the optimal individual for local fine-tuning:

[0077]

[0078] where p best,p is the individual optimal solution, g best is the global optimal solution of the population, c1 and c2 are the adaptive weight coefficients, and r1 and r2 are random numbers;

[0079] Combine gradient descent optimization to adjust the parameters of the optimal individual:

[0080]

[0081] where η is the learning rate, is the gradient of the fitness function;

[0082] S55. Optimize the computational complexity using the dynamic weight adjustment strategy and set the weight adjustment factor λ g :

[0083]

[0084] where λ min and λ max are the lower and upper limits of the weight adjustment factor respectively;

[0085] S56. Determine whether the optimization has reached the convergence condition. If the convergence condition is met, output the optimal fitting parameters; otherwise, return to S53 to continue iterative optimization until the optimized fitting function for surface settlement measurement data is obtained.

[0086] Optionally, S56 includes the following steps:

[0087] S561. Based on the optimization objective of the fitting function for surface settlement measurement data, calculate the mean fitness of the optimization population in the current generation g:

[0088]

[0089] where, represents the average fitness value of the current population, P is the number of individuals in the optimization population, is the fitness function value of the p-th individual in the g-th generation, and the fitness function measures the error degree of the fitting function for surface settlement measurement data;

[0090] S562. Calculate the fitness variance based on the fitness distribution of the current population to measure the convergence degree of the population:

[0091]

[0092] where, represents the standard deviation of the fitness of the optimization population. If gradually converges to a preset value, it indicates that the fitness of the population individuals tends to be consistent;

[0093] S563. Determine whether the convergence condition is met. Set a convergence determination threshold. If the following conditions are met, it is considered that the optimization process has converged:

[0094] and

[0095] where, ∈1 is the fitness convergence threshold and ∈2 is the fitness variance threshold;

[0096] S564. Determine the optimal fitting parameters. When the convergence condition is met, select the optimal individual in the current generation as the optimal parameters of the final fitting function for surface settlement measurement data:

[0097]

[0098] where, θ best is the optimized fitting parameter for surface settlement measurement data, including the weight matrix, bias term, and hidden layer center vector of the radial basis function neural network model, that is:

[0099] θ best ={W best ,bbest ,C best};

[0100] S565. If the convergence condition is not met, continue to optimize and iterate, return to S53 to recalculate the fitness of the population individuals, and perform the next-generation iterative optimization based on the alpha evolutionary algorithm until the convergence condition is met;

[0101] S566. Obtain the optimized fitting function of the ground settlement measurement data:

[0102]

[0103] where is the fitted ground settlement height value of the i-th measurement point, W best is the optimal weight matrix from the hidden layer to the output layer optimized by the alpha evolutionary algorithm, H i is the output vector of the hidden layer of the radial basis function neural network, b best is the optimal bias term optimized by the alpha evolutionary algorithm.

[0104] The beneficial effects of the present invention are:

[0105] (1) The present invention proposes an adaptive hybrid optimization strategy. During the evolution process, the algorithm can adjust the optimization operator according to the dynamic changes of the ground settlement data. In the global search stage, genetic algorithm and differential evolution are combined for wide-area exploration. In the local optimization stage, particle swarm optimization combined with gradient descent is used for fine tuning, and an elite retention mechanism is adopted to ensure that the optimal individual is retained. The alpha evolutionary algorithm can adaptively adjust the strategy according to the search stage, reduce unnecessary calculations, improve the convergence speed, enable the fitting process to complete optimization in a short time, and enhance the real-time computing ability.

[0106] (2) The present invention uses a radial basis function neural network as the fitting model for the ground settlement measurement data, and optimizes the parameters of the neural network through the alpha evolutionary algorithm. The local response characteristics of the Gaussian radial basis function are used to enable the model to more accurately capture the nonlinear characteristics of the ground settlement and perform accurate fitting on different time scales. In addition, an adaptive weight adjustment mechanism is combined to ensure finding the optimal balance between computational complexity and fitting accuracy.

[0107] (3) The present invention proposes an adaptive convergence determination mechanism based on the mean and variance of fitness, that is, calculating the mean and variance of fitness of the current optimized population during each generation of evolution to ensure that the optimization process searches within the global scope. When the mean fitness of the population reaches the preset threshold and the fitness variance converges, the algorithm automatically terminates the optimization to ensure obtaining the global optimal solution. The convergence strategy can ensure the global optimum while reducing the computational cost and avoiding falling into the local optimum. Description of the Drawings

[0108] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the accompanying drawings:

[0109] Figure 1 It is a flowchart of a real-time fitting method for surface settlement measurement data based on an evolutionary algorithm proposed by the present invention. Detailed implementation manners

[0110] Now, the present invention will be further described in detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0111] Reference Figure 1 , a real-time fitting method for surface settlement measurement data based on an evolutionary algorithm, includes the following steps: S1. Obtain a surface settlement measurement data set;

[0112] S2. Preprocess the surface settlement measurement data set to obtain a standardized surface settlement measurement data set;

[0113] S3. Perform time series analysis on the standardized surface settlement measurement data to generate a surface settlement measurement data feature matrix;

[0114] S4. Construct a radial basis function neural network model for fitting surface settlement measurement data, and initialize the parameters of the radial basis function neural network model to obtain an initial surface settlement measurement data fitting function;

[0115] S5. Use the alpha evolutionary algorithm to optimize the parameters of the initial surface settlement measurement data fitting function to obtain an optimized surface settlement measurement data fitting function;

[0116] S6. Use the optimized surface settlement measurement data fitting function to predict the input surface settlement measurement data, output a surface settlement trend fitting curve, and calculate fitting error, trend change rate, and prediction confidence interval indicators.

[0117] In this embodiment, S1 includes the following steps:

[0118] S11. Obtain a surface settlement measurement data set, and the surface settlement measurement data set includes global navigation satellite system measurement data, synthetic aperture radar measurement data, and lidar point cloud measurement data:

[0119] D GNSS ={(x i ,y i ,t i ,h i)|i = 1, 2, ..., N GNSS};

[0120] D InSAR = {(x j , y j , t j , h j )|j = 1, 2, ..., N InSAR};

[0121] D LiDAR = {(x k , y k , t k , h k )|k = 1, 2, ..., N LiDAR};

[0122] Among them, x and y represent the geographical coordinate positions, t represents the timestamp, h represents the surface height value, N GNSS , N InSAR , N LiDAR are respectively the number of sampling points of the global navigation satellite system measurement data, synthetic aperture radar measurement data, and lidar point cloud measurement data;

[0123] S12. Perform a spatial reference coordinate system conversion on the surface subsidence measurement data set to unify the coordinate reference of the surface subsidence measurement data set, and obtain the surface subsidence measurement data set under the unified spatial reference coordinate system:

[0125] D aligned = D' GNSS ∪ D' InSAR ∪ D' LiDAR ;

[0127] Among them, D' GNSS , D' InSAR and D' LiDAR respectively represent the global navigation satellite system measurement data, synthetic aperture radar measurement data, and lidar point cloud measurement data under the unified spatial reference coordinate system.

[0128] In this embodiment, S2 includes the following steps:

[0129] S21. Use the wavelet transform method to denoise the surface subsidence measurement data set D aligned under the unified spatial reference coordinate system, remove the random noise in the surface subsidence measurement data, and obtain the denoised surface subsidence measurement data set;

[0130] S22. Use the Mahalanobis distance analysis method to detect abnormal data in the denoised surface subsidence measurement data set and calculate the abnormality degree of each data point:

[0131] M i =(h i ″ -μ h ) T Σ h -1 (h i ″ -μ h );

[0132] Wherein, M i is the Mahalanobis distance of the ground settlement measurement data points, μ h is the mean vector of the ground settlement measurement data heights, and Σ h is the covariance matrix of the ground settlement measurement data heights;

[0133] Set a threshold M th . When M i >M th , determine the data point as an abnormal data point and mark the abnormal data point set;

[0134] S23. Interpolate and correct the abnormal data point set, calculate the correction value of the abnormal data point by using the spline interpolation method, replace the original data point with the corrected abnormal data point, and obtain the corrected ground settlement measurement data set;

[0135] S24. Standardize the corrected ground settlement measurement data set by using the mean-standard deviation normalization method, and finally obtain the standardized ground settlement measurement data set D standardized .

[0136] In this embodiment, S3 includes the following steps:

[0137] S31. Based on the change trend of the time series data, calculate the ground settlement rate characteristics of each measurement point by using the time difference method to reflect the change degree of the ground settlement within different time intervals:

[0138]

[0139] Wherein, v i represents the ground settlement rate at the i-th moment, is the standardized ground height value of the i-th measurement point, and t i -t i-1 is the time interval between two consecutive observations for calculating the instantaneous rate of ground settlement;

[0140] S32. Capture the ground settlement acceleration characteristics by calculating the change rate of the settlement rate to reflect the intensification or slowdown of the ground settlement trend:

[0141]

[0142] Among them, a i represents the ground settlement acceleration at the i-th measurement point, reflecting the change trend of the settlement rate at this point at time t i moment, v i is the ground settlement rate of the i-th measurement point, representing the intensity of the settlement change at the current moment, t i -t i-1 is the time interval;

[0143] S33. Use the polynomial fitting method to perform trend modeling on the ground settlement data to depict the global trend of the non-linear change of the ground settlement:

[0144]

[0145] Among them, is the fitting settlement trend value of the i-th measurement point, c0, c1,..., c n are fitting coefficients, t i is the timestamp;

[0146] S34. Use the fast Fourier transform method to transform the settlement data into the frequency domain to identify potential periodic characteristics of the ground settlement:

[0147]

[0148] Among them, H(f) represents the spectral amplitude of the ground settlement data at the f-th frequency, reflecting the dominant frequency of the periodic settlement, j is the imaginary unit, and the main frequency f with the largest amplitude in the spectrum is selected max as the main periodic characteristic of the ground settlement:

[0149] f max = argmaxH(f);

[0150] Among them, f max represents the main period information of the settlement data, which is used to analyze the periodic settlement phenomenon caused by geological structure, climate change or underground engineering construction factors;

[0151] S35. Synthesize the obtained ground settlement rate characteristics, ground settlement acceleration characteristics, non-linear change trend characteristics of the ground settlement and periodic characteristics of the ground settlement, and construct a characteristic matrix of the ground settlement measurement data:

[0152]

[0153] In this embodiment, S4 includes the following steps:

[0154] S41. Construct a radial basis function neural network model for fitting surface subsidence measurement data, set the input layer of the radial basis function neural network model, and the input layer receives the feature matrix F of the surface subsidence measurement data subsidence As the input, let the number of neurons in the input layer be d, then the neurons in the input layer are represented as:

[0155] X = [x1, x2,..., x d ;

[0156] Among them, X is the input neuron vector, and x i represents the characteristic variable of the surface subsidence measurement data, and d = 4 corresponds to the dimensions of the rate characteristic, acceleration characteristic, nonlinear trend characteristic, and periodic characteristic;

[0157] S42. Set the hidden layer of the radial basis function neural network model, and the hidden layer uses the Gaussian radial basis function as the activation function:

[0158]

[0159] Among them, H j is the output of the jth hidden layer neuron, X is the input neuron vector, C j is the jth center vector, σ j is the scale parameter of the jth radial basis function, which controls the response range of the neuron, and ∥X - C j ∥ represents the Euclidean distance between the input neuron and the center vector. The total number of neurons in the hidden layer is set to m, and the hidden layer output vector is represented as:

[0160] H = [H1, H2,..., H j ;

[0161] S44. Set the output layer of the radial basis function neural network model. The output layer is used to calculate the fitting result of the surface subsidence measurement data and obtain the initial surface subsidence measurement data fitting function:

[0162]

[0163] Among them, y is the fitting output of the surface subsidence measurement data, W is the weight matrix from the hidden layer to the output layer, b is the bias term of the output layer, represents the fitting surface subsidence height value of the ith measurement point.

[0164] In this embodiment, S5 includes the following steps:

[0165] S51. Generate initial solutions of multiple fitting parameters to form an optimization population, set the population size to P, and initialize the population individual Θ 0 :

[0166]

[0167] Among them, is the parameter vector of the p-th individual, including the weight matrix W, the bias term b, and the center vector C of the hidden layer of the radial basis function neural network, that is:

[0168]

[0169] S52. Combine the surface settlement measurement data feature matrix F subsidence to calculate the fitness evaluation sample size according to the sample dimension N:

[0170] P = αN;

[0171] Among them, α is the population size coefficient;

[0172] S53. Use the mean square error and the model complexity penalty term as the optimization objective to calculate the fitness of the population individuals, and set the fitness function f(θ p ) to measure the fitting error of the individual:

[0173]

[0174] Among them, is the true surface settlement height value, is the fitting height value calculated using the parameters θ of the p-th individual p , and λ∥θ p ∥ 2 is the regularization term;

[0175] S54. Use the alpha evolutionary algorithm for optimization, and adaptively adjust the evolutionary operator according to the evolutionary stage:

[0176] In the global exploration stage (G < G1), use the alpha hybrid strategy to combine genetic algorithms and differential evolution for wide-area search:

[0177]

[0178] Among them, are randomly selected different individuals, and F is the scaling factor;

[0179] Use the elitist retention mechanism to retain the current optimal individual:

[0180]

[0181] In the local development stage (G1 ≤ G < G2), use the alpha adaptive mutation strategy to select the optimal individual for local fine-tuning:

[0182]

[0183] Among them, pbest,p is the individual optimal solution, g best is the global optimal solution of the population, c1 and c2 are adaptive weight coefficients, and r1 and r2 are random numbers;

[0184] Combine gradient descent optimization to adjust the parameters of the optimal individual:

[0185]

[0186] where η is the learning rate, is the gradient of the fitness function;

[0187] S55. Optimize the computational complexity using a dynamic weight adjustment strategy, and set the weight adjustment factor λ g :

[0188]

[0189] where λ min and λ max are the lower and upper limits of the weight adjustment factor, respectively;

[0190] S56. Determine whether the optimization reaches the convergence condition. If the convergence condition is met, output the optimal fitting parameters; otherwise, return to S53 to continue iterative optimization until the optimized fitting function for the ground settlement measurement data is obtained.

[0191] In this embodiment, S56 includes the following steps:

[0192] S561. Based on the optimization objective of the ground settlement measurement data fitting function, calculate the average fitness of the optimized population in the current generation g:

[0193]

[0194] where, represents the average fitness value of the current population, P is the number of individuals in the optimized population, is the fitness function value of the p-th individual in the g-th generation, and the fitness function measures the error degree of the ground settlement measurement data fitting function;

[0195] S562. Calculate the fitness variance based on the fitness distribution of the current population to measure the convergence degree of the population:

[0196]

[0197] where, represents the standard deviation of the fitness of the optimized population. If gradually converges to a preset value, it indicates that the fitness of the population individuals tends to be consistent;

[0198] S563. Determine whether the convergence condition is satisfied, set the convergence determination threshold. If the following conditions are met, it is considered that the optimization process has converged:

[0199] and

[0200] where ∈1 is the fitness convergence threshold and ∈2 is the fitness variance threshold;

[0201] S564. Determine the optimal fitting parameters. When the convergence condition is satisfied, select the current generation's best individual as the optimal parameters of the fitting function for the final surface settlement measurement data:

[0202]

[0203] where θ best is the optimized fitting parameter for the surface settlement measurement data, including the weight matrix, bias term, and hidden layer center vector of the radial basis function neural network model, that is:

[0204] θ best ={W best ,b best ,C best};

[0205] S565. If the convergence condition is not satisfied, continue the optimization iteration. Return to S53 to recalculate the fitness of the population individuals, and perform the next-generation iterative optimization based on the alpha evolutionary algorithm until the convergence condition is satisfied;

[0206] S566. Obtain the optimized fitting function for the surface settlement measurement data:

[0207]

[0208] where is the fitted surface settlement height value at the i-th measurement point, W best is the optimal weight matrix from the hidden layer to the output layer optimized by the alpha evolutionary algorithm, H i is the output vector of the hidden layer of the radial basis function neural network, and b best is the optimal bias term optimized by the alpha evolutionary algorithm.

[0209] Example 1:

[0210] On April 15, 2024, during the construction of a section of the subway Line 10 in City A, the monitoring system collected abnormal settlement data at the GNSS monitoring point GNSS-045 at 03:20. The settlement rate of this point increased significantly within the past 6 hours, accelerating from -0.0145 m / d to -0.0321 m / d, and the settlement acceleration changed from 0.0008 m / d 2Increased to 0.0024 m / d 2 , indicating that there may be potential non-uniform settlement of the formation in this area. At the same time, InSAR data also shows that the surface deformation amplitude in this area exceeds 15 mm at 03:45, exceeding the normal safety threshold of the subway tunnel.

[0211] At 04:00, the alpha evolutionary algorithm starts real-time data fitting. The system automatically retrieves the surface settlement measurement data within the past 24 hours and conducts data analysis in combination with the radial basis function neural network.

[0212] The initial data input is as shown in Table 1 below (data of some monitoring points):

[0213]

[0214] In the RBF-NN calculation, based on the non-linear characteristics of the data, the system uses 10 radial basis functions. The alpha evolutionary algorithm randomly initializes 200 candidate parameter sets and starts the optimization calculation. At 04:03:15, the system completes the 85th iteration and finds that the average fitness of the population has converged to the set threshold, finally determining the optimal fitting parameters and outputting the latest settlement trend prediction.

[0215] Real-time fitting result:

[0216]

[0217] The fitting curve shows that the predicted settlement of this monitoring point within the next 24 hours is -0.045 m, reaching the engineering warning value.

[0218] At 04:05, the system detects that the settlement trend exceeds the warning threshold, automatically generates a tunnel structure settlement warning report, and sends the report to the construction unit and the subway operation management center. The report content includes:

[0219] Warning time: 2024-04-15 04:05;

[0220] Monitoring point number: GNSS-045;

[0221] Settlement trend: The predicted settlement within the next 24 hours is -0.045 m, exceeding the safety threshold;

[0222] Abnormal detection result: The settlement rate increases sharply, and the regional settlement change amplitude is 15 mm;

[0223] Suggested measures: Immediately check the groundwater pumping situation around, and check the stability of the tunnel structure;

[0224] At the same time, at 04:06, the subway operation dispatching center of City A receives the system alarm and immediately dispatches engineering and technical personnel to the site for inspection, and suspends the construction work on this section.

[0225] To verify the effectiveness of the present invention, the system simultaneously uses polynomial regression and MLP neural network for fitting, and compares the computational efficiency and prediction error. The data is shown in Table 2 below:

[0226]

[0227]

[0228] According to the above table, the computational efficiency of the present invention is improved. This method completes fitting within 0.91 seconds, which is 3.1 times faster than the PR method and 1.95 times faster than the MLP. The fitting error of this method is reduced, and the mean square error is only 0.0035, which is 65.3% lower than the PR method and 51.4% lower than the MLP method. The prediction error of this method is reduced, and the 24-hour settlement prediction error is only 0.014m, which is 62.1% less than the PR method and 51.7% less than the MLP method, effectively improving the accuracy of long-term prediction.

[0229] At 04:40, the technical personnel of the subway management center arrived at the scene and found that the groundwater level in this area had dropped by 1.2m in the past 12 hours, which might lead to an intensification of local stratum settlement. Subsequently, the construction unit took measures to adjust the groundwater pumping volume at 05:15 and re-measured the settlement rate at 06:30, and found that the settlement trend tended to be stable.

[0230] When a new round of data analysis was carried out at 04:00 on April 16, the deviation between the 24-hour settlement trend predicted by the system and the actual measurement value was less than 0.012m, and the deviation was reduced by 58% compared with the traditional method, proving the high efficiency and reliability of this method in the real-time fitting of surface settlement measurement data.

[0231] In this embodiment, the radial basis function neural network optimized by the alpha evolutionary algorithm is successfully applied in the subway tunnel construction area of City A, and the advantages of this method in terms of calculation speed, fitting accuracy and prediction stability are verified through comparative experiments. At the same time, the real-time alarm function of the system enables the subway management party to respond in a timely manner, and finally successfully avoids the risk of tunnel structure settlement, proving the practical value of the present invention in the field of high-precision surface settlement monitoring.

[0232] The present invention proposes an adaptive hybrid optimization strategy. During the evolution process, the algorithm can adjust the optimization operator according to the dynamic changes of surface settlement data. In the global search stage, genetic algorithm and differential evolution are combined for wide-area exploration. In the local optimization stage, particle swarm optimization is combined with gradient descent for fine tuning, and an elite retention mechanism is adopted to ensure that the optimal individual is retained. The alpha evolutionary algorithm can adaptively adjust the strategy according to the search stage, reduce unnecessary calculations, and improve the convergence speed, enabling the fitting process to complete optimization in a short time and enhancing the real-time computing ability.

[0233] The present invention uses a radial basis function neural network as a fitting model for surface settlement measurement data, optimizes the parameters of the neural network through the alpha evolutionary algorithm, and utilizes the local response characteristics of the Gaussian radial basis function to enable the model to more accurately capture the non-linear characteristics of surface settlement and perform precise fitting on different time scales. In addition, an adaptive weight adjustment mechanism is combined to ensure finding the optimal balance between computational complexity and fitting accuracy.

[0234] The present invention proposes an adaptive convergence determination mechanism based on fitness mean and variance, that is, calculating the fitness mean and fitness variance of the current optimized population in each generation of evolution to ensure that the optimization process searches within the global scope. When the population fitness mean reaches the preset threshold and the fitness variance converges, the algorithm automatically terminates the optimization to ensure the acquisition of the global optimal solution. The convergence strategy can ensure the global optimum while reducing the computational cost and avoiding falling into the local optimum.

[0235] As described above, only the preferred specific embodiments of the present invention are given, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A real-time fitting method for surface subsidence measurement data based on evolutionary algorithm, characterized in that: The steps include: S1. Obtain surface subsidence measurement dataset; S2. preprocessing the surface settlement measurement data set to obtain a standardized surface settlement measurement data set; S3. Perform time series analysis on the standardized surface subsidence measurement data to generate a surface subsidence measurement data feature matrix; S4. constructing a radial basis function neural network model for fitting the surface settlement measurement data, and initializing the parameters of the radial basis function neural network model to obtain an initial surface settlement measurement data fitting function; S5. using an alpha evolution algorithm to optimize the parameters of the initial surface settlement measurement data fitting function to obtain an optimized surface settlement measurement data fitting function; S6. Use the optimized surface settlement measurement data fitting function to predict the input surface settlement measurement data, output the surface settlement trend fitting curve, and calculate the fitting error, trend change rate, and prediction confidence interval indicators.

2. The real-time fitting method for surface subsidence measurement data based on evolutionary algorithm according to claim 1 is characterized in that: The S1 comprises the following steps: S11. Acquire a surface subsidence measurement data set, wherein the surface subsidence measurement data set includes global navigation satellite system measurement data, synthetic aperture radar measurement data, and lidar point cloud measurement data: D GNSS ={(x i ,y i ,t i ,h i )|i=1,2,...,N GNSS }; D InSAR ={(x j ,y j ,t j ,h j )|j=1,2,...,N InSAR }; D LiDAR ={(x k ,y k ,t k ,h k )|k=1,2,...,N LiDAR }; Among them, x, y represent the geographic coordinates, t represents the timestamp, h represents the surface height, N GNSS ,N InSAR ,N LiDAR The number of sampling points for GNSS measurement data, SAR measurement data, and LiDAR point cloud measurement data, respectively; S12. Perform spatial reference coordinate system conversion on the surface settlement measurement dataset, unify the coordinate base of the surface settlement measurement dataset, and obtain the surface settlement measurement dataset under the unified spatial reference coordinate system: D aligned =D′ GNSS ∪D′ InSAR ∪D′ LiDAR ; Among them, D′ GNSS , D′ InSAR and D′ LiDAR They respectively represent global navigation satellite system measurement data, synthetic aperture radar measurement data and lidar point cloud measurement data in a unified spatial reference coordinate system.

3. The real-time fitting method for surface subsidence measurement data based on evolutionary algorithm according to claim 1 is characterized in that: The S2 comprises the following steps: S21. Using wavelet transform method to analyze the surface settlement measurement data set D in the unified spatial reference coordinate system aligned Perform denoising to remove random noise in the surface settlement measurement data and obtain a denoised surface settlement measurement data set; S22. Use the Mahalanobis distance analysis method to detect abnormal data on the denoised surface settlement measurement data set and calculate the abnormality of each data point: Among them, M i is the Mahalanobis distance of the surface settlement measurement data points, μ h is the mean vector of the height of the surface settlement measurement data, Σ h is the covariance matrix of the height of the surface subsidence measurement data; Set the threshold M th , when M i >M th , the data point is determined as an abnormal data point, and the abnormal data point set is marked; S23. Perform interpolation correction on the abnormal data point set, calculate the correction value of the abnormal data point using the spline interpolation method, replace the original data point with the corrected abnormal data point, and obtain a corrected surface settlement measurement data set; S24. The corrected surface settlement measurement data set is standardized using the mean-standard deviation normalization method, and finally the standardized surface settlement measurement data set D is obtained. standardized .

4. The real-time fitting method for surface subsidence measurement data based on evolutionary algorithm according to claim 1 is characterized in that: The S3 comprises the following steps: S31. Based on the changing trend of time series data, the time difference method is used to calculate the surface subsidence rate characteristics of each measuring point to reflect the degree of change of surface subsidence in different time intervals: Among them, v i represents the surface subsidence rate at the i-th moment, is the standardized surface height value of the i-th measurement point, t i -t i-1 is the time interval between two consecutive observations, which is used to calculate the instantaneous rate of surface subsidence; S32. By calculating the rate of change of the settlement velocity, the acceleration characteristics of the surface settlement are captured to reflect the intensification or slowdown of the surface settlement trend: Among them, a i Represents the surface settlement acceleration of the i-th measuring point, reflecting the point at time t i The sedimentation rate change trend at the moment, v i is the surface subsidence rate at the i-th measuring point, representing the intensity of subsidence change at the current moment, t i -t i-1 is the time interval; S33. Use polynomial fitting method to build trend model for surface subsidence data and describe the global trend of nonlinear change of surface subsidence: in, is the fitted settlement trend value of the i-th measurement point, c0, c1, ..., c n is the fitting coefficient, t i is the timestamp; S34. Use the fast Fourier transform method to convert the settlement data into the frequency domain and identify potential periodic characteristics of surface settlement: Among them, H(f) represents the spectrum amplitude of the surface settlement data at the fth frequency, reflecting the dominant frequency of periodic settlement, j is an imaginary unit, and the main frequency f with the largest amplitude in the spectrum is selected. max As the main periodic characteristics of surface subsidence: f max =argmaxH(f); Among them, f max Represents the main periodic information of settlement data, which is used to analyze the periodic settlement phenomenon caused by geological structure, climate change or underground engineering construction factors; S35. The surface settlement rate characteristics, surface settlement acceleration characteristics, surface settlement nonlinear change trend characteristics and surface settlement periodic characteristics obtained by comprehensive calculation are used to construct the surface settlement measurement data feature matrix:

5. The real-time fitting method for surface subsidence measurement data based on evolutionary algorithm according to claim 1 is characterized in that: The S4 comprises the following steps: S41. Construct a radial basis function neural network model for fitting surface settlement measurement data, set the input layer of the radial basis function neural network model, and the input layer receives the characteristic matrix F of the surface settlement measurement data. subsidence As input, let the number of neurons in the input layer be d, then the neurons in the input layer are represented as: X=[x1,x2,...,x d ]; Among them, X is the input neuron vector, x i It represents the characteristic variables of the surface settlement measurement data, d = 4 corresponds to the dimensions of rate characteristics, acceleration characteristics, nonlinear trend characteristics and periodic characteristics; S42. Set the hidden layer of the radial basis function neural network model, and the hidden layer uses the Gaussian radial basis function as the activation function: Among them, H j is the output of the jth hidden layer neuron, X is the input neuron vector, C j is the jth center vector, σ j is the scale parameter of the jth radial basis function, controlling the response range of the neuron, ∥XC j ∥ represents the Euclidean distance between the input neuron and the center vector. The total number of neurons in the hidden layer is set to m, and the output vector of the hidden layer is expressed as: H=[H1,H2,...,H j ]; S44. Set the output layer of the radial basis function neural network model, the output layer is used to calculate the fitting result of the surface settlement measurement data, and obtain the initial surface settlement measurement data fitting function: Among them, y is the fitting output of the surface settlement measurement data, W is the weight matrix from the hidden layer to the output layer, b is the bias term of the output layer, Represents the fitted surface settlement height value of the i-th measuring point.

6. The real-time fitting method for surface subsidence measurement data based on evolutionary algorithm according to claim 1 is characterized in that: The S5 comprises the following steps: S51. Generate initial solutions for multiple fitting parameters to form an optimized population, set the population size to P, and initialize the population individuals Θ 0 : in, is the parameter vector of the pth individual, including the weight matrix W, bias term b and hidden layer center vector C of the radial basis function neural network, that is: S52. Combined with surface settlement measurement data feature matrix F subsidence The sample dimension N of the fitness evaluation sample size is calculated as: P = αN; Among them, α is the population size coefficient; S53. Use mean square error and model complexity penalty as optimization targets to calculate the fitness of individuals in the population, and set the fitness function f(θ p ) measures the individual fitting error: in, is the true surface subsidence height value, To use the pth individual parameter θ p The calculated fitting height value, λ∥θ p ∥ 2 is the regularization term; S54. Optimize using the alpha evolutionary algorithm and adaptively adjust the evolutionary operator according to the evolutionary stage: In the global exploration stage (G < G1), use the alpha hybrid strategy to combine genetic algorithm and differential evolution for wide-area search: in, are different individuals randomly selected, and F is the scaling factor; Adopt the elitist retention mechanism to retain the current optimal individual: In the local development stage (G1 ≤ G < G2), use the alpha adaptive mutation strategy to select the optimal individual for local fine-tuning: Among them, p best,p is the individual optimal solution, g best is the global optimal solution of the population, c1, c2 are adaptive weight coefficients, r1, r2 are random numbers; Combine gradient descent optimization to adjust the parameters of the optimal individual: Where η is the learning rate, is the gradient of the fitness function; S55. Use dynamic weight adjustment strategy to optimize computational complexity and set weight adjustment factor λ g : Among them, λ min and λ max are the lower and upper limits of the weight adjustment factor, respectively; S56. Judge whether the optimization reaches the convergence condition. If the convergence condition is met, output the optimal fitting parameters; otherwise, return to S53 to continue iterative optimization until the optimized fitting function of the surface settlement measurement data is obtained.

7. The real-time fitting method for surface subsidence measurement data based on evolutionary algorithm according to claim 6 is characterized in that: The above S56 includes the following steps: S561. Based on the optimization objective of the fitting function of the surface settlement measurement data, calculate the mean fitness of the optimization population in the current generation g: in, represents the average fitness value of the current population, P is the number of individuals in the optimized population, is the fitness function value of the pth individual in the gth generation. The fitness function measures the error degree of the fitting function of the surface settlement measurement data; S562. Calculate the fitness variance based on the current population fitness distribution To measure the degree of convergence of the population: in, represents the standard deviation of the optimized population fitness. If Gradually converge to the preset value, indicating that the fitness of individuals in the population tends to be consistent; S563. Judge whether the convergence condition is met. Set the convergence determination threshold. If the following conditions are met, it is considered that the optimization process has converged: and where ∈1 is the fitness convergence threshold and ∈2 is the fitness variance threshold; S564. Determine the optimal fitting parameters. When the convergence condition is met, select the optimal individual in the current generation as the optimal parameters of the final fitting function of the surface settlement measurement data: Among them, θ best The optimized fitting parameters of the surface settlement measurement data include the weight matrix, bias term and hidden layer center vector of the radial basis function neural network model, namely: θ best ={W best ,b best ,C best }; S565. If the convergence condition is not met, continue the optimization iteration. Return to S53 to recalculate the fitness of the population individuals and perform the next-generation iterative optimization based on the alpha evolutionary algorithm until the convergence condition is met; S566. Obtain the optimized fitting function of the surface settlement measurement data: in, is the fitted surface settlement height value of the i-th measurement point, W best is the optimal hidden layer to output layer weight matrix after optimization by the Alpha evolution algorithm, H i is the output vector of the hidden layer of the radial basis function neural network, b best It is the optimal bias term after optimization by the Alpha evolutionary algorithm.

Images