为避免传统评价方法中主观因素造成的影响并弥补神经网络在风险预测中的不足,将统计学中的主成分分析法(PCA)与 BP 神经网络模型进行有机结合,并根据大量有关矿山充填管道系统资料,建立充填管道失效风险评价模型。研究发现,将原始数据经过主成分分析法处理,不仅可以有效地减少模型输入维数,便于消除各指标间的相关性,而且与未经 PCA 处理的 BP 神经网络相比,训练收敛速度明显加快,预测结果更加准确。针对某矿山充填管道系统实际情况,利用该模型进行模拟预测的结果与实际情况相符合,证明模型合理。%For the sake of covering the shortages for neural network in risk evaluation and eliminating human error and subjective grounds,principal component analysis in statistic and BP neural network were combined and used to constructing the invalidate risk evaluation model of backfill pipe,which couple with a large amount of relative data of mine’s backfill pipe system.The investigations found that the input dimension of neural network were reduced and the relationship of all the indexes were also eliminated through dealing with the original data by means of PCA method,and the contrast of optimized BP neural network and standard BP neural network without principal components analysis turned out the former has outstanding merits of rapid analysis and high accuracy in predicting,meanwhile,the rationality of the model was certified according to the results from simulation test.
展开▼