基于BP神经网络的新浇筑混凝土爆破安全震动速度预测  

Forecasting for safety vibration velocity of freshly-made concrete based on BP neural network

作　　者：刘敦文[1] 崔朋波 

机构地区：[1]中南大学资源与安全工程学院,长沙410083

出　　处：《安全与环境学报》2014年第1期43-46,共4页Journal of Safety and Environment

基　　金：国家自然科学基金项目(50490270)

摘　　要：岩体开挖爆破与局部浇筑混凝土交叉施工时,爆破产生的地震波对新浇筑混凝土的力学性能会产生不利影响,预测控制混凝土质点峰值震动速度对确保新浇筑混凝土工程质量至关重要。以新浇筑混凝土结构质点峰值震动速度为研究对象,基于BP神经网络理论,选取混凝土极限拉应变、混凝土的弹性模量和泊松比、基岩的弹性模量和泊松比5个影响因素,建立了新浇筑混凝土爆破安全震动速度BP神经网络预测模型。并对龄期为1.5 d和40 d的混凝土在两种不同基岩条件下的爆破安全震动速度进行了预测,预测结果相对误差分别为0.025、0.011、0.004和0.002,满足工程需要。研究表明,该方法计算量小,预测性能好,便于工程人员掌握。This paper is aimed at introducing our novel forecasting model of the safety vibration velocity of freshly-made concrete based on the back propagation （BP） neural network. As is known, the peak particle vibration velocity （PPV） is usually taken as the bench- mark in deciding whether the mechanical properties of the concrete have been changed or not due to the blast vibration. Therefore, we would like to take the peak particle vibration velocity of fresh concrete as our study topic. While considering both the mechanical properties of the concrete and solidness of the supporting rock-pebbles, we have chosen five factors of the concrete as chief decisive ones that may af- fect the concrete＇s eventual tensile strain and stress, such as, the e- lastic modulus, the Poisson＇ s ratio, the bedrock and elastic modulus. Then we have worked out a forecasting means based on the back propagation neural network, with its functions on predicting the safety vibration velocity of the newly prepared concrete. Specifically speak- ing, the model consists of five input layer nodes, thirteen hidden lay- er nodes and one output layer node. Furthermore, the number of the hidden layer nodes, which can be represented by a few formulae and subject to ten training tests. In the said model, the hidden layers known as ＇tansig＇ layers, actually function as the output layer ＇pure- lin＇ and the training ＇trainlm＇ function. The model can then be used to forecast the safety vibration velocity of the concrete in a period from 1.5 d to 40 d mending age under the condition of two kinds of bedrock. The results have shown that the relative errors can be illus- trated as 2.5%, 1.1%, 0.4% and 0.2%, symbolizing that the model can be expected to satisfy the demands of general engineering programs and projects. The results also indicate that the said BP neu- ral network forecasting model can be used by means of the LM algo- rithm for forecasting the safety vibration velocity of the concrete. Thus, it can be seen that the model we have

关 键 词：安全工程 安全震动速度 爆破震动 新浇筑混凝土 BP神经网络 

分 类 号：X932[环境科学与工程—安全科学]