基于RSM和GA-ANN的侵蚀沟煤矸石复垦优化设计

文章基于沟道复垦后导排水能力最大原则研究侵蚀沟煤矸石填埋复垦方法,采用煤矸石作为侵蚀沟复垦材料,各层填埋材料最优厚度为变量,出流时间为响应值,依次建立中心组合设计响应曲面法(RSM)模型和遗传算法优化神经网络遗传算法—人工神经网络模型(GA-ANN),验证两种模型优异度及最终预测结果,分析试验结果全局灵敏性。研究表明,响应曲面法模型(RSM)和遗传算法—人工神经网络(GA-ANN)模型决定系数R~2分别为0.91810和0.94497,与响应曲面法(RSM)模型相比,遗传算法—人工神经网络模型(GA-ANN)建模能力、模型精度更优。响应曲面法(RSM)模型和遗传算法—人工神经网络模型(GA-ANN)确定的厚度最优阙值及预测出流时间分别为53.42、38.51、90 cm,145.129 s和50、41、90 cm,140.542 s。模型验证偏差率分别为2.85%和1.43%。改进全局敏感性分析结果表明,覆土厚度X_1、混合煤矸石层厚度X_2和大粒径煤矸石层厚度X_3对于全局敏感性为X_3X1X_2,大粒径煤矸石厚度单位变化对于复垦沟道导排水能力影响最大。在实际生产中,精度要求较小情况下,采用响应曲面法设计煤矸石填埋复垦方案简洁高效。

Optimal design of erosion gully reclamation using coal gangue based on RSM and GA-ANN

A method of gully landfill reclamation was presented in this paper using coal gangue based on the principle of maximum drainage capacity after gully reclamation. Using coal gangue as erosion gully reclamation materials, the optimal thickness of each layer of landfill materials for variables,outflow time for the response values, the response surface method(RSM) model of central composite design and the neural network optimized by genetic algorithm model(GA-ANN) were established successively, the superiority of the two models and the final prediction results were verified. Finally, the global sensitivity of the experimental results was analyzed. The results showed that R~2 of the model based on the response surface method(RSM) and the genetic algorithm-artificial neural network model(GA-ANN) were 0.91810 and 0.94497. Compared with the model based on the response surface method(RSM), the model based on the the genetic algorithm-artificial neural network model(GA-ANN)had better modeling ability and precision. The optimal thickness threshold and outflow time predicted by the response surface method(RSM) model and the genetic algorithm——artificial neural network model(GA-ANN) were 53.42, 38.51, 90 cm, 145.129 s and 50, 41, 90 cm 140.542 s. The deviation rates of model validation were 2.85% and 1.43%. The improved global sensitivity analysis results showed that the soil layer thickness X_1, mixed size gangue layer thickness X_2 and large size gangue layer thickness X_3 had the highest global sensitivity to X_3> X_1> X_2, and the unit change of large size gangue thickness had the largest impact on the drainage capacity of reclamation channels. In actual production and life, it was a simple and efficient method to design the reclamation scheme of coal gangue landfill by response surface method when the precision was not required.