Study Of Uniform Experiment Design Method Applying To Civil Engineering

In order to solve some problems in the field of civil engineering, a lot of physical tests, numerical calculations or simulation tests are usually required. These tests may cause considerable experimental cost since they could involve some related expensive materials and equipment. Meanwhile, these massive numerical tests have large calculation efforts, which could consume plenty of time of researcher and computer. And even in some cases, to complete these tests may be an impossible task within effective time. After obtaining experimental results, a regression analysis is usually made on them to get some objective laws or relations. If the experiment involves many parameters (e.g.>3), conventional regression method, e.g. parameter regression may not get the best result, even fail to get a reasonable result. Therefore, an efficient experimental design method, i.e. uniform experiment design method (UD) and a non-parameter regression method, i.e. ACE (Alternating Conditional Expectations) are introduced. The former one can select some representative experiment samples to reduce required number of experiments, and the latter one can lead to an appropriate regression result with high precision. It is shown that they can simplify the process of experiments and solve the foregoing problems at a low cost of expense, manpower and time.The following projects are studied using uniform design and (or) ACE:(1) Damage detection of girder structure based on curvature mode. The modal analysis is made on 30 structural model samples with single damage which are defined by UD respectively. An equation of the relationship between the variation of curvature mode, damage position and damage degree of the single sample is obtained by making regression analysis on the experimental data, which can be used to detect the real single damage accurately and easily. During the analysis of sensibility of multiple damages, it is found that multiple damages just affect variation of curvature mode in the local areas, i.e. scattered damages can be seen as independent damages. Therefore, the equation of the single damage can be also used to detect the multiple damage degree.(2) Analysis of response spectrum of vibration induced by human walking. A few representative samples of time histories of human walking loads are simulated using UD. Then the dynamic analysis is made on many beam structures with different natural frequencies, spans and damping ratios under the excitement of these samples respectively. Some different response spectrums, including response spectrums of peak acceleration, RMS acceleration, eVDV acceleration and VDV acceleration, having a certain percentile are obtained by reliability analysis based on response surface determined by regression analysis on the input parameters and the related vibration response index. The general formula of each acceleration response spectrums are deduced from the sensitivity analysis of damping ratio and span. According to these formulae, the corresponding acceleration index of response can be figured out easily and directly. Comparing the results of the method and the results based on Monte Carlo simulation, it is proved that the method proposed in this thesis has a high precision and it can cuts down the required calculation work remarkably and it is an efficient way to simplify the calculation of dynamic response of random excitement.(3) Experimental research of concrete corrosion by sulfuric acid. Eight experimental parameters are defined by UD. And then the experiments of concrete corrosion by sulfuric acid are done in the lab. The general function between the apparent diffusion coefficient, water-cement ratio and pH value of the soak solution is obtained by ACE regression analysis of the experimental results, which can directly reflect the relationship between these parameters.(4) Reliability analysis of building fire. Using the UD and ACE regression method to improve the existing reliability analysis method based on response surface, an efficient reliability analysis method is proposed. In the case of building fire, the cumulative probability functions(CPF) of the two most important parameters, i.e. the temperature and the thickness of smoke layer, are obtained respectively just by some small amount of calculations. The results show that the proposed method is efficient and just costs little calculation in reliability analysis.(5) Reliability analysis of foundation pit. The reliability of a foundation pit case is analyzed adopting the method similar to the method in the reliability analysis of building fire, i.e. reliability analysis based on improved response surface. In the calculation case, owing to the use of ACE, a response surface relationship with high precision is obtained just by small amounts of calculation. Therefore, subsequent finite element analysis can be replaced by direct simple interpolation algorithm. In this way, a direct Mont Carlo simulation to obtain the probability density function of response parameters is carried out by small calculation work, which means that the method can make the reliability analysis of complex engineering more efficient and time-saving.From the foregoing problems that are solved successfully, it is shown that the UD, especially combining with ACE regression, is feasible and applicable in addressing many related problems in civil engineering. It can decrease the number of experiments remarkably for both physical experiment and numerical experiment, which can cut down the expense costs, calculation work and calculation time. Therefore, the UD can be adopted to solve similar problems in the fields of civil engineering and other.This thesis is partly sponsored by the Natural Science Foundation of China(Grant No.:51078175) and the Natural Science Foundation of Yunnan Province(Grant 2008E038M).