According to the nonlinear links between blasting-vibration-induced damage degree and its influencing factors, a rough set-based fuzzy-neural network model is proposed to seek a method that can overcome the limitations in the single-factor case and improve the damage prediction precision. In the proposed prediction model, there are 10 factors to be taken into account, which include particle vibration velocity (PPV) , dominant frequency, dominant frequency duration and dynamic characteristics of structures. The prediction model is trained and tested by a series of data from the observations of blasting vibration and damage degree of houses in Tongliishan Copper Mine. The training results are in agreement with the field observations. The rough set-based fuzzy-neural network can reduce data indexes and simplify input variables, and minify the decision table size and accelerate the approach to the minimal rules. The proposed method considering the manifold factors can improve the prediction precision of damage degree induced by blasting vibration.%为了探索一种能克服单因素预测的局限性,提高爆破振动危害预测精度的方法,基于粗糙集模糊神经网络理论,建立了综合考虑爆破振动幅值、主频率、主频率持续时间及结构动力特性等10个因素的民房破坏程度预测模型;用铜绿山矿爆破振动和民房破坏情况观测数据,对该模型进行了训练和测试,测试结果与现场观测结果具有良好的一致性.研究表明,粗糙集理论可将现场数据进行属性约简,简化输入变量,缩小神经网络的搜索空问,改善爆破振动的预测性能;基于粗糙集模糊神经网络理论的爆破振动危害预测模型,能更好地考虑各种因素对危害程度的综合影响,避免了单因素预测的局限性.
展开▼
