突变理论与BP神经网络相结合的堤防安全综合评价

Catastrophe theory and BP neural network-jointed comprehensive evaluation on levee safety

针对堤防工程失事后果严重,难以进行全线监控等问题,从致灾因子、孕灾环境和承灾体3个方面考虑,选取恰当的评价指标,构建了多层次多目标半结构性的堤防工程安全评价指标体系。在堤防安全评价指标体系的基础上,提出基于突变理论与BP神经网络相结合的堤防安全综合评价模型,即将突变理论归一公式处理后的指标作为BP神经网络的输入,通过试错法确定中间层,输出为堤防安全综合评价值。利用该模型对黄河宁夏段堤防工程进行堤防安全综合评价,并将综合评价模型评价结果分别与主成分分析法和层次分析法的评价结果进行对比分析。结果表明,模型评价结果合理,既考虑了系统内部各影响指标间的矛盾关系,也具有较强的鲁棒性,简单实用,对同类水利工程安全评价具有借鉴意义。

Aiming at the problem of that the failure consequence of levee project is quite serious and difficult to be monitored along the whole line,a multi-level,multi-target and semi-structural levee project safety evaluation system is established by selecting appropriate evaluation indexes under the consideration from the aspects of disaster-inducing factor,disaster-pregnant environment and disaster-bearing body. On the basis of the levee safety evaluation index system,a catastrophe theory and BP neural network-jointed comprehensive evaluation model for levee safety is put forward herein,i. e. taking the indexes processed by the normalization formula of catastrophe theory as the inputs for BP neural network,which are outputted as the comprehensive values by determining the mid-layers through the trial-and-error method. The model is applied to the comprehensive evaluation made on the levee project of Ningxia Section of the Yellow River,from which the evaluation results are comparatively analyzed with those from the methods of principal components analysis and analytic hierarchy process. The result shows that the evaluation result from the model is reasonable,which not only considers the contradiction relations in-between all the impacting indexes inside of the system,but also has stronger robustness,thus is simple and practicable and has a referential significance for the safety evaluation on the similar water conservancy project.