基于稳健遗传算法的矿山开采沉陷预计参数反演

Mining Subsidence Prediction Parameter Inversion Based on Robust Genetic Algorithm

概率积分法是我国矿山开采沉陷预计的主要方法,其预计的精度直接取决于参数的准确性,如何利用科学的方法准确地求取预计参数是实际应用中的关键问题。本文针对利用遗传算法反演概率积分法参数时抗粗差能力差的问题,提出将稳健估计与遗传算法相融合的方法,提高反演参数过程中对异值点的抗干扰能力,保证参数求取结果稳健、精确。通过实验得出:单纯采用遗传算法进行参数的反演时,当粗差出现在下沉盆地边缘时,其抗差能力较强;而当粗差出现在拐点处和下沉盆地中心时,其抗差能力较差,反演结果严重偏离理论值,而采用稳健遗传算法可以获得较为稳定、精确的参数。

Probability integral method is the main method of mining subsidence prediction in China,and its prediction accuracy directly depends on the accuracy of parameters.How to use scientific method to accurately calculate the predicted parameters is the key problem in practical application.Aiming at the problem of poor anti-gross error ability when using genetic algorithm to invert the parameters of probability integral method,a method combining robust estimation and genetic algorithm is proposed to improve the anti-interference ability to outliers in the process of parameter inversion,and ensure the robustness and accuracy of parameter obtaining results.The results show that when the gross error occurs at the edge of subsidence,the genetic algorithm has a strong resistance to the error.However,when the gross error occurs at the inflection point and the center of the sinking basin,its resistance to error is poor,and the inversion results deviate seriously from the theoretical values,while the robust genetic algorithm can obtain more stable and accurate parameters.