摘要
预测隧道工程中TBM掘进速度,主要有完全经验的、半理论半经验的模型和人工智能等方法,所用参数均为确定性的,未考虑参数存在的随机性,故导致预测结果的不准确性。基于此,提出了Monte Carlo-BP神经网络TBM掘进速度预测模型,着重考虑了一些重要输入参数的随机性,其中输入参数重要性的大小通过粗糙集进行计算排序。采用Monte Carlo产生随机数时,由于参量的样本数据的有限,分布函数均采用阶梯形经验分布函数。如果采用的数据是来自不同类型的TBM,则应当考虑机器性能参数,并重新对参数重要性进行排序。实例计算表明,Monte Carlo-BP神经网络模型预测结果和实测值总体趋势和均值比较一致。
Penetration rate of TBM is mainly predicted by fully empirical, semi-theoretical, semi-empirical models and artificial intelligence in engineering. The parameters used in these models are all deterministic and their uncertainties are neglected, so it leads to inaccuracy of the results. Because of this, a Monte Carlo-BP network model is proposed and the uncertainty of some important parameters are considered in this model. The importance of each parameter is calculated by rough set. Due to the limit of sample data, stepped empirical distributed function is used when Monte Carlo is used to produce random numbers. If sample data are not from the same type TBM, the TBM performance parameters should be considered and the importance of parameters should be calculated again. It is proved that the calculated results of proposed model are in accordance with measured results.
出处
《岩土力学》
EI
CAS
CSCD
北大核心
2009年第10期3127-3132,共6页
Rock and Soil Mechanics
基金
国家十一五科技支撑项目(No.2006BAB04A06)