摘要
为有效解决某些复杂系统安全评价中涉及的评价指标数目较多、复杂度较高、难以运用传统建立数学解析式的方法对其进行综合评价等问题,提出基于Relief F和BP神经网络的安全评价指标体系(SAI)精简化建模方法。首先,用Relief F算法对SAI进行特征选择,剔除对安全评价主导指标影响较小的辅助指标;然后用BP神经网络对精简后的指标体系进行建模,建立SAI的精简化模型;最后,根据钢板缺陷评价试验数据将这种建模方法与其他方法对比。结果表明,用该方法能够简化安全评价系统模型、降低运算复杂度、提高系统效率。
In some complex safety assessment systems, there are many safety assessment indexes and high complexity. It is difficult to apply traditional mathematical methods for their comprehensive assess- ment. In this paper, a simplifying modeling method for safety assessment index system based on Relief F and neural network was described. Firstly, Relief F algorithm was used to select safety assessment index system features on the basis of empirical data, and exclude auxiliary indexes having little influence on key indexes were excluded. Secondly, neural network was used for modeling of safety assessment index sys- tem, and a simplified model for safety assessments index was built. Finally, a series of experiments on steel plates faults assessment were conducted to this approach and other approaches. The experimental results show the merits of model built by us, and that proposed method can simplify models for some com- plex safety assessment systems, reduce computational complexity, and improve efficiency.
出处
《中国安全科学学报》
CAS
CSCD
北大核心
2013年第10期15-20,共6页
China Safety Science Journal
基金
国家自然科学基金资助(5107418)
重庆市自然科学基金资助(CSTC2012JJA40018)
重庆市教委科研项目(KJ131409)
