海洋环境下混凝土强度演化预测评价系统

发布时间：2019-06-26 22:13

【摘要】：海洋环境下腐蚀离子对水泥基材料的破坏严重。为解决多因素耦合作用下混凝土强度的预测评价,从大数据挖掘分析的角度出发,通过建立干湿循环盐结晶造成的应变计算模型,对影响因素进行降维,最终采用Matlab GUI平台建立了海洋环境下混凝土强度预测评价系统。开发的海洋混凝土性能劣化评价系统1.0软件包,为今后混凝土材料性能的预测评估提供了便利。围绕建立海洋环境下混凝土强度的预测评价系统这一中心问题,开展了大量计算模拟和试验验证工作。(1)搜集整理了大量试验数据,综合考虑了海洋环境下14个影响因素,建立了 BP神经网络、支持向量机和决策树模型,比较了三种模型的预测评价效果。研究表明,支持向量机在数据处理后的精确度较高,训练集的平均相对误差为5.87%,测试集的平均相对误差为8.33%,优于BP神经网络和决策树。(2)从干湿循环盐结晶机理出发,将部分干湿循环影响参数转化为反映混凝土缺陷的参数一应变,建立了干湿循环盐结晶过程造成的混凝土应变计算模型。提出Comsol有限元法求解水分传输方程,推导了盐结晶应变计算公式。结果表明,Comsol求解的结果与试验吻合度较好,最终建立的干湿循环盐结晶应变计算结果可以较好地预测混凝土的应变发展趋势。(3)引入缺陷参数(应变)后,对多因素进行降维处理,利用支持向量机建立了降维后的混凝土强度预测模型,并与全因素所建模型的分析结果进行了比较。结果表明,虽然测试集的预测精度比降维后略有下降,但将应变参数引入模型可以很好的反映每次干湿循环后样本的缺陷度,且降低了影响因素的维度,减少了运算量。(4)通过研究海洋混凝土性能劣化预测方法,建立了预测精度较高、适应性较好的预测方法及预测模型。然后从界面优化、GUI之间的相互调用和快捷方式对主界面进行设计,利用MATLAB GUI平台实现了人机交互界面的建立。最后通过MatlabGUI平台开发了“海洋混凝土强度演化预测评价系统VI.0”软件包,为相关领域人员研究混凝土在海洋环境下服役过程中强度性能的演化提供了便利,为混凝土材料的设计和维护提供了参考。
[Abstract]:The damage of corrosion ions to cement-based materials is serious in marine environment. In order to solve the prediction and evaluation of concrete strength under the action of multi-factor coupling, from the point of view of big data mining analysis, the strain calculation model caused by dry and wet circulating salt crystallization is established, and the influencing factors are reduced. Finally, the concrete strength prediction and evaluation system in marine environment is established by using Matlab GUI platform. The 1.0 software package of marine concrete performance deterioration evaluation system is developed, which provides convenience for the prediction and evaluation of concrete material properties in the future. Around the central problem of establishing the prediction and evaluation system of concrete strength in marine environment, a large number of calculation simulation and experimental verification work have been carried out. (1) A large number of experimental data have been collected and sorted out, 14 influencing factors in marine environment have been comprehensively considered, BP neural network, support vector machine and decision tree model have been established, and the prediction and evaluation effects of the three models have been compared. The results show that the accuracy of support vector machine is high after data processing, the average relative error of training set is 5.87%, and the average relative error of test set is 8.33%, which is better than BP neural network and decision tree. (2) based on the crystallization mechanism of dry and wet cycle salt, the influence parameters of partial dry and wet cycle are transformed into parameter strain reflecting concrete defects. The strain calculation model of concrete caused by dry and wet cyclic salt crystallization process is established. The Comsol finite element method is proposed to solve the water transfer equation, and the formula for calculating the crystallization strain of salt is derived. The results show that the results of Comsol solution are in good agreement with the experimental results, and the final calculation results of dry and wet cyclic salt crystallization strain can predict the strain development trend of concrete. (3) after introducing defect parameters (strain), the dimension reduction treatment of multi-factors is carried out, and the strength prediction model of concrete after dimension reduction is established by support vector machine, and the results are compared with those of the whole factor model. The results show that although the prediction accuracy of the test set is slightly lower than that after dimension reduction, the introduction of strain parameters into the model can well reflect the defect degree of samples after each dry and wet cycle, reduce the dimension of influencing factors and reduce the amount of calculation. (4) by studying the prediction method of marine concrete performance deterioration, a prediction method and prediction model with high prediction accuracy and good adaptability are established. Then the main interface is designed from the interface optimization, the mutual call between GUI and the shortcut, and the human-computer interaction interface is established by using MATLAB GUI platform. Finally, the software package VI.0 is developed through MatlabGUI platform, which provides convenience for people in related fields to study the evolution of strength performance of concrete in the process of service in marine environment, and provides a reference for the design and maintenance of concrete materials.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TU528

【参考文献】