爆破振动特征参量的粗糙集模糊神经网络预测

摘 要 爆破振动特征参量对爆破振动危害效应有重要影响。首次用粗糙集模糊神经网络方法对振幅、主频率及主频持续时间进行预测。首先介绍了粗糙集模糊神经网络的基本思想，其次，分析了印象爆破振动特征参量的主要因素，建立了基于粗糙集模糊神经网络的爆破振动特征参量预测模型；最后用某边坡开挖爆破中的振动观测指标对模型进行了训练，并对15组指标进行了测试。结果表明：粗糙集模糊神经网络预测模型能反映了影响因素与特征量之间的非线性关系，适用于爆破振动特征参量预测。一次预测1个指标的精度高于同时预测3个指标的精度。

A fuzzy neural network prediction model based on rough set for characteristic variables of blasting vibration

Characteristic variables of blasting vibration have great effects on its damage level. The prediction of characteristic variables caused by blasting vibration is helpful to study blasting vibration effect. Here, the prediction of the amplitude and the first dominant frequency band and its time duration were achieved on the basis of rough set and fuzzy neural network (FNN) theory. The purpose of this study was to explore a method which could avoid the limitation of the prediction with only one index and to improve the prediction precision. Firstly, the drawback of the prediction of the am-plitude based on Sadov's vibration formula was analyzed. Secondly, rough set and fuzzy neural network (RSFNN) theory were introduced briefly. Thirdly, a rough set-based FNN prediction model for characteristic variables of blasting vibration was established based on analysis of factors affecting blasting vibration characteristic variables. Finally, the model was trained with data come from Tonglvshan Copper Mine and was tested by 15 groups of data. The results showed that therough set and FNN prediction model reflected the nonlinear relationship between factors and characteristic variables and could be used to predict characteristic variables of blasting vibration. It was also found that the precision of predicting sin-gle index a time was higher than that of predicting three indexes at the same time.