CN-121997734-A - Mining subsidence prediction parameter solving method based on improved center collision optimization algorithm

CN121997734ACN 121997734 ACN121997734 ACN 121997734ACN-121997734-A

Abstract

The invention discloses a mining subsidence prediction parameter solving method based on an improved central collision optimization algorithm, and belongs to the field of mine deformation monitoring data processing. Aiming at the defects that a basic center collision optimization algorithm is easy to premature convergence and insufficient in optimizing precision when solving the mining subsidence parameter inversion problem, an adaptability evaluation model is built by adopting a Huber loss function to enhance the algorithm robust capability, a cube map is utilized to generate random numbers to enhance the stability of the searching process, and a Cauchy variation mechanism guided by a self-adaptive elite is introduced to enhance the global exploration capability. The method comprises the steps of firstly constructing a probability integration method parameter inversion problem into an optimization model and initializing a population, then carrying out double-space collaborative search in an original space and a decorrelation space constructed based on principal component analysis, combining a dynamic space allocation strategy, iteratively updating the population, and finally outputting an optimal parameter solution. The method effectively improves convergence accuracy, stability and robustness of the algorithm in complex nonlinear parameter inversion.

Inventors

XU LIANGJI
WANG GUOHUI
Han Jiazhang
ZHANG KUN
CHEN YONGCHUN
CHEN SHUAIPENG

Assignees

安徽理工大学
淮南矿业(集团)有限责任公司

Dates

Publication Date: 20260508
Application Date: 20260121

Claims (9)

1. The mining subsidence prediction parameter solving method based on the improved center collision optimization algorithm is characterized by comprising the following steps of: Determining a parameter vector to be solved by a probability integration method, presetting corresponding value ranges for each parameter, randomly generating an initial pedestrian population based on the value ranges, wherein each pedestrian individual represents a parameter vector; acquiring position information of each surface monitoring point on the mining working surface, and calculating and acquiring an estimated sinking value of each monitoring point by a probability integration method based on a parameter vector represented by each pedestrian; obtaining measured sinking values of all surface monitoring points on the mining working surface, and calculating the fitness of each pedestrian by combining the predicted sinking values; Presetting a termination condition, and iteratively updating the pedestrian population with the aim of minimizing the fitness, wherein the iterative updating process comprises the steps of generating pedestrian individuals representing new parameter vectors through original space searching, decorrelation space searching and self-adaptive elite mutation based on the fitness, calculating the fitness of corresponding pedestrian individuals, and selecting a next generation population from the current population by a preset selection mechanism; And when the iteration termination condition is met, outputting a parameter vector represented by the individual minimum fitness pedestrian in the next generation population, and taking the parameter vector as a final mining subsidence prediction parameter.
2. The mining subsidence prediction parameter solving method based on the improved center collision optimization algorithm according to claim 1, wherein the original space search and the decorrelation space search in the iterative updating process specifically comprise the following steps: Constructing a decorrelation space conversion matrix according to individuals with poor adaptability in the current population; dynamically distributing a certain proportion of pedestrian individuals to update in an original space, and mapping the rest individuals to a decorrelation space to update; in the original space, the updating position of the pedestrian individual is determined by the individual with better adaptability, the randomly selected individual and the dynamic parameters; In the decorrelation space, the updating position of the pedestrian individual is determined by the same kind rule after the space transformation, and the updating is reversely transformed back to the original space.
3. The mining subsidence prediction parameter solving method based on the modified center collision optimizing algorithm according to claim 2, wherein the location update formula in the original space or the decorrelation space is as follows: ; Wherein, the The method comprises the steps of selecting a reference individual from an individual set with adaptability better than that of a current individual; is a method comprising random number and dynamic parameter Comparing the generated binary vectors for controlling the updating mode of each parameter dimension; the mixed vector is generated by linear combination of a plurality of other reference individuals and random individuals, and the specific expression is as follows: ; Wherein, the Other reference individuals selected from the optimal fitness individuals; is an individual randomly selected from a population; is the current individual; is a dynamic parameter; Is a random number within the interval of [0,1 ].
4. The improved center collision optimization algorithm-based mining subsidence prediction parameter solving method according to claim 3, wherein the dynamic parameters The calculation formula of (2) is as follows: ; Wherein, the The current iteration number; is the total number of iterations.
5. The method for solving the estimated parameters of mining subsidence based on the improved central collision optimization algorithm according to claim 1, wherein the fitness of each pedestrian is calculated by a robust estimation method based on a Huber loss function, and the specific calculation formula is as follows: ; Wherein, the Residual errors of the predicted sinking value and the actually measured sinking value; Is a preset threshold value parameter, individual fitness The sum of Huber losses for all monitoring point residuals.
6. The method for solving the estimated parameters of mining subsidence based on the improved central collision optimization algorithm according to claim 1, wherein the iterative updating process adopts a cube mapping to generate random numbers for initializing population and randomly selecting in the updating process, and the mapping formula is as follows: ; Wherein, the Is a random number uniformly distributed over the interval [0,1 ]; is a random number after being mapped and transformed by the cube.
7. The mining subsidence prediction parameter solving method based on the improved center collision optimization algorithm according to claim 1, wherein the adaptive elite variation specifically comprises the following steps: presetting a variation probability attenuated with the number of iterations Variation scale ; After each position update, probability is used Performing elite-guided cauchy variance on the new individual generated, the variance formula being: ; Wherein, the The method is a current global optimal individual; 、 is a parameter boundary vector; Is a uniform random number in the interval of 0, 1; And comparing the fitness of the mutated individual with the fitness of the original updated individual, and preferentially keeping.
8. The method for solving the estimated parameters of mining subsidence based on the improved central collision optimization algorithm according to claim 1, wherein the ratio of the dynamic allocation original space to the decorrelated space searching individuals adopts an adaptive adjustment strategy based on the searching success rate, and is specifically as follows: calculating the individual proportion of successful updates in the Decorrelation Space (DS) in the current iteration And the individual proportion of successful updates in the Original Space (OS) ; Individual scale of next generation allocation to decorrelated space The updating is as follows: And is constrained at preset upper and lower limits An inner part; The ratio allocated to the original space is 。
9. The mining subsidence prediction parameter solving method based on the improved center collision optimization algorithm according to claim 1, wherein the probability integration method is used for solving a parameter vector Wherein Is a sinking coefficient; To influence the tangent; Influencing the propagation angle for mining; Is an inflection point offset distance parameter.

Description

Mining subsidence prediction parameter solving method based on improved center collision optimization algorithm Technical Field The invention belongs to the field of mine deformation monitoring data processing, and particularly relates to a mining subsidence prediction parameter solving method based on an improved center collision optimization algorithm. Background The mining subsidence prediction theory plays a vital role in guiding mining under buildings, railways and water bodies, evaluating potential geological disasters and analyzing mining area subsidence mechanisms, the probability integration method is a mining subsidence prediction method appointed by Chinese authorities, the traditional probability integration method prediction parameter inversion method mainly comprises a characteristic point method, a linear approximation method and an orthogonal experiment method, the characteristic point method is large in parameter solving error due to the fact that characteristic points of curves are difficult to accurately determine, the linear approximation method mainly comprises a least square iteration method, a Gauss Newton method, a steepest descent method and the like, the linear approximation method is strict in parameter solving theory and high in parameter solving accuracy, but the parameter solving model belongs to a complex nonlinear function, and part of probability integration parameters have correlation, so that the linear approximation method has extremely high requirements on the layout form of a surface movement observation station and the accuracy of initial values of the parameter solving model (the initial values are low in accuracy and easily cause the parameter solving model to diverge), and the engineering application of the method is difficult; In recent years, intelligent optimization methods such as genetic algorithm, particle swarm optimization and the like are introduced into the field, so that the solving effect is improved to a certain extent, however, due to the fact that the probability integral parameter searching space is complex, a plurality of local extrema exist, the algorithms are easy to generate premature convergence, sink to local optimum and the like in the solving process, the fluctuation of parameter inversion results is large, and stable solving precision is difficult to guarantee, so that the problems of insufficient convergence precision and robustness in mining subsidence parameter inversion exist in the prior art. Disclosure of Invention Aiming at the problems that the mining subsidence parameter inversion method in the prior art is easy to converge in premature, falls into local optimum and has insufficient solving precision and robustness, the invention aims to provide the mining subsidence predicted parameter solving method based on the improved central collision optimization algorithm. The aim of the invention can be achieved by the following technical scheme: A mining subsidence prediction parameter solving method based on an improved center collision optimization algorithm comprises the following steps: S1, determining a parameter vector to be solved by a probability integration method, presetting corresponding value ranges for each parameter, randomly generating an initial pedestrian population based on the value ranges, wherein each pedestrian individual represents a parameter vector; S2, acquiring position information of each surface monitoring point on the mining working surface, and calculating and acquiring an expected sinking value of each monitoring point through a probability integration method based on a parameter vector represented by each pedestrian; S3, obtaining measured sinking values of all earth surface monitoring points on the mining working surface, and calculating the fitness of each pedestrian by combining the predicted sinking values; S4, presetting a termination condition, and iteratively updating pedestrian population with the aim of minimizing fitness, wherein the iterative updating process comprises the steps of generating pedestrian individuals representing new parameter vectors by three stages of constructing a double search space, executing central collision interaction and self-adaptive elite variation based on population interaction behaviors of the pedestrian population, and calculating the fitness of corresponding pedestrian individuals; And S5, outputting a parameter vector represented by the individual minimum fitness pedestrian in the next generation population when the iteration termination condition is met, and taking the parameter vector as a final mining subsidence prediction parameter. Further, the step S3 of calculating the fitness adopts a robust estimation method based on Huber loss function, and the specific formula is as follows: For the ith pedestrian, the fitness calculation formula is as follows: Wherein, the Residual errors of the j-th monitoring point; is the Huber loss function. The definition is as