基于神经网络的混凝土强度预测研究

发布时间：2018-01-08 16:26

本文关键词：基于神经网络的混凝土强度预测研究　出处：《辽宁工程技术大学》2013年硕士论文　论文类型：学位论文

【摘要】：近年来在役土木工程结构的安全性越来越受到重视，而混凝土材料的强度在土木工程结构的安全性及耐久性评估中扮演着重要角色。在土木工程结构的长期服役中，混凝土的强度会慢慢下降。准确地预测出混凝土材料在长期使用过程的强度退化对于正确评估结构安全性有着重要的意义。本文在收集整理混凝土试验数据的基础上，利用神经网络模型和MATLAB软件中的神经网络工具箱，，对混凝土强度进行了预测研究。首先，本文论述了神经网络的基本原理，BP网络的结构、算法原理，BP网络建立过程以及BP网络的程序实现过程，利用收集整理到的国内外试验资料，建立BP网络预测模型预测了混凝土强度值，并分析了运用人工神经网络进行混凝土强度预测的可行性。其次，分析总结了影响混凝土强度的主要因素，并利用收集到的试验数据，运用MATLAB神经网络工具箱创建了BP网络模型，根据敏感性分析方法，对几个主要的影响因素进行了混凝土强度敏感性分析。最后，本文通过试验研究了混凝土的力学性能，并将试验数据应用到混凝土强度预测之中。测试结果表明，经过训练的人工神经网络能够准确地预测混凝土材料的强度，其误差控制在7%以内，可以满足工程的需要。
[Abstract]:In recent years, more and more attention has been paid to the safety of civil engineering structures in service. The strength of concrete plays an important role in the evaluation of safety and durability of civil engineering structures and plays an important role in the long-term service of civil engineering structures. The strength of concrete will decrease slowly. It is very important to predict the strength degradation of concrete material in long-term use for evaluating the structural safety correctly. On the basis of collecting and finishing the test data of concrete, this paper makes use of neural network model and neural network toolbox in MATLAB software to predict the strength of concrete. First of all, this paper discusses the basic principle of neural network and the structure of BP network, the algorithm principle of BP network establishment process and BP network program implementation process, using the collected domestic and foreign experimental data. The prediction model of BP neural network is established to predict the strength of concrete, and the feasibility of using artificial neural network to predict the strength of concrete is analyzed. Secondly, the main factors affecting the strength of concrete are analyzed and summarized, and the BP neural network model is created by using the MATLAB neural network toolbox, and the sensitivity analysis method is used. The sensitivity analysis of concrete strength is carried out for several main influencing factors. Finally, this paper studies the mechanical properties of concrete through experiments, and applies the test data to the strength prediction of concrete. The test results show that. The trained artificial neural network can accurately predict the strength of concrete material, and its error is controlled within 7%, which can meet the need of engineering.
【学位授予单位】：辽宁工程技术大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TU528

【参考文献】