基于正交设计的BP神经网络上覆岩层移动研究

Research on Overlying Strata Movement Based on Orthogonal Design BP Neural Network

采用数值模拟对正交实验设计的4个影响因素16种组合进行模拟,并且构建相应BP神经网络预测方法,同时对覆岩移动进行预测并对数值模拟结果进行对比分析,研究结果表明:正交实验可大大降低实验次数,提高工作效率,第12组实验对煤层开采上覆岩层移动的影响最大,在正交实验设计的数值模拟结果的基础上,运用神经网络对煤层开采引起的覆岩移动预测结果误差小,具有较高的可信度。

Numerical simulation of four factors orthogonal experiment designed to simulate the 16 combinations, and to construct the corresponding BP neural network forecasting method, while the overlying strata movement and predict simulation results were analyzed, the results showed that： orthogonal experiment can greatly reduce the number of experiments, improve work efficiency, maximum, based on the impact of the first set of experiments, 12 overburden movement on coal mining numerical simulation results in orthogonal experimental design on the use of neural networks caused to the coal mining overburden movement prediction error is small, has high credibility.