Method, equipment and medium for robust estimation of parameters by probability integral method

Abstract

The invention discloses a method, equipment and medium for robust estimation of the parameters of the probability integral method based on CA-rPSO. Firstly, according to the observation data of the surface movement in the mining area, the initial parameters of the probability integral method are obtained by combining the truncated least squares LTS method and CA-rPSO. Estimation and the error in the initial unit weight; according to the error in the initial unit weight, the weight function IGGIII is used to obtain the initial weight of each observation data; Search and iterative calculation to obtain the optimal solution of the parameters of the robust and high-precision probability integration method; the present invention can invert the parameters of the probability integration method with high precision in the case of gross errors in the observation data, and is useful for making full use of various surface mobile observations It is of great scientific value to collect data, study geological hazards caused by underground coal mining, predict and evaluate building / structure damage, and optimize underground coal mining schemes.

Description

technical field [0001] The invention relates to the fields of geodetic surveying and data mining, in particular to a method, device and medium for robust estimation of parameters of a probability integral method based on CA-rPSO. Background technique [0002] The prediction of surface movement and deformation caused by underground coal mining is one of the core contents of mining damage and protection, and it is of great significance to the theoretical research and production practice of mining subsidence. The probability integration method is one of the most commonly used mining subsidence prediction methods in my country. The parameters of the probability integration method (including the subsidence coefficient q , horizontal movement coefficient b, main influence angle β, main influence propagation angle θ and inflection point offset s) determine the accuracy of mining subsidence prediction results. [0003] The parameters of the probability integration method are usuall...

AI Technical Summary

Problems solved by technology

However, the above methods all have certain shortcomings: the feature point method only uses individual key points to solve the parameters of the probability integration method. Although the efficiency is high, the solution process lacks stability and has large errors; The increase is exponential, which is not convenient for computer implementation; the total least squares method needs to linearize and invert the probability integral method model, but the coefficient matrix of the normal equation after linearization is ill-conditioned, resulting in unstable inversion; Gauss-Newton method, Iterative algorithms such as McQuarter's method and modulus method are sensitive to the initial value of iteration, which may easily cause divergence in the iterative process;

Among them, the weight selection iterative least squares method based on Huber and the minimum norm assigns a small weight value to the observation data containing gross errors, which can reduce its impact on the solution results, but these two weight functions do not have an elimination section, so it is difficult to Effective resistance to gross errors; the robust total least squares method based on total least squares and weight selection iteration method, because its iterative initial value is easily affected by gross errors, it is easy to cause iterative distortion

Images