混凝土徐变模型随机性及参数敏感性分析

发布时间：2018-03-26 21:23

本文选题：徐变模型　切入点：参数相关性　出处：《北京交通大学》2013年硕士论文

【摘要】：目前,混凝土徐变的理论预测值与实际值存在较大偏差,这主要是由于徐变的不确定性所致。这种基于徐变模型的不确定性,由徐变影响因素的不确定性、徐变模型本身的不确定性两部分构成。徐变模型本身的不确定性源于对徐变机理认识不足、试验数据的拟合误差等因素所致,此方面的研究已经有了一定的结果。而徐变影响因素的随机性对徐变预测值变异性的影响还需进一步的研究,开展这方面的工作在研究徐变不确定性、提高徐变预测精度方面具有重要意义。本文基于数理统计理论,研究了四种徐变模型对9种徐变影响因素的敏感性。基于典型徐变数据库,得到了徐变影响因素的秩相关系数矩阵。在考虑徐变参数之间相关性的基础上,采用改进的傅里叶幅度灵敏度检验法,计算了徐变模型对计算参数的全局敏感性。结果表明,混凝土干燥龄期、混凝土强度、弹性模量、加载龄期以及环境湿度对徐变度变异性影响较大,骨料水泥比对徐变度预测值变异性影响最小。各模型的预测精度随龄期增大而减小,四种模型的预测精度从大到小依次为GL2000、RILEM B3、CEB MC90-99、ACI209R-92。参数敏感度中包含了其他参数相关性对徐变预测值不确定性的影响,采用神经网络模型进行非线性回归,将各参数的总敏感度分解为参数自身的独立贡献和与其他参数相关部分的相关贡献两部分。结果表明,当混凝土龄期大于30d时,四种模型的混凝土强度、弹性模量的相关贡献占总贡献的90%左右,这两个因素可以综合反映混凝土材料特性的对徐变预测不确定性的贡献。GL2000模型在实用角度比RILEM B3模型更具优势。
[Abstract]:At present, there is a big deviation between the theoretical prediction value of concrete creep and the actual value, which is mainly caused by the uncertainty of creep, which is caused by the uncertainty of creep model, which is influenced by the uncertainty of creep factors. The uncertainty of creep model itself is caused by the lack of understanding of creep mechanism, the fitting error of test data, and so on. Some results have been obtained in this field, and the influence of randomness of creep factors on creep prediction variability needs to be further studied. It is of great significance to improve the accuracy of creep prediction. Based on mathematical statistical theory, the sensitivity of four creep models to 9 creep factors is studied. The rank correlation coefficient matrix of creep influencing factors is obtained. On the basis of considering the correlation between creep parameters, an improved Fourier amplitude sensitivity test method is used to calculate the global sensitivity of creep model to the calculated parameters. The dry age of concrete, the strength of concrete, the modulus of elasticity, the age of loading and the humidity of environment have great influence on creep variability. The prediction accuracy of each model decreases with the increase of age, and the prediction accuracy of the four models is GL2000RILEM B3CEB MC90-99 and ACI209R-92. The parameter sensitivity includes the influence of the correlation of other parameters on the uncertainty of creep prediction value. The neural network model is used to carry out nonlinear regression. The total sensitivity of each parameter is decomposed into two parts: the independent contribution of the parameter itself and the related contribution of the other parameters. The results show that when the concrete age is more than 30 days, the concrete strength of the four models is obtained. The relative contribution of elastic modulus is about 90% of the total contribution. The contribution of these two factors to the uncertainty of creep prediction can reflect the properties of concrete materials. GL2000 model is superior to RILEM B3 model in practical terms.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TU528

【相似文献】