摘要
提出了一种基于应力波反射波法判别基桩桩身完整性的BP型神经网络模型。该模型以桩身应力波波形曲线、桩的几何尺寸和桩身混凝土波速等为网络的输入信息,预测作为网络输出信息的桩身完整性特征,如正常、缩径、扩径、离析和开裂等。通过采用多种运算改进技术,提高了网络的可行性和计算速度。对天津市软土地基中的数十根灌注桩的实测波形等资料学习和预测,取得了令人满意的精度。
A neural network model for diagnosing pile integrity based on reflection wave method is suggested. The model predicts the basic features of pile integrity such as its normality and various defects including partially enlargement or reduction in pile diameter, concrete segregation, cracks and breakages in pile stem and so on. The service ability and efficiency of the neural computing are enhanced by implementation of some elaborated operation technologies. The application of the neural network model in the integrity check of dozens of concrete piles, in soft soils in Tianjin region, tested with reflection wave method yielded a fairly high accuracy of prediction.
出处
《岩土力学》
EI
CAS
CSCD
北大核心
2003年第6期952-956,共5页
Rock and Soil Mechanics
