STATISTICAL MODELLING OF CONCRETE PROPERTIES

Abstract

Many of the present revised codes are based ou semi-probabilistic approach. It is generally known that the nature of loads and material resistance are probabilistic in nature. The present study is aimed to systematically generate data (experimentally and numerically) on time dependent properties of concrete, namely, strength, shrinkage, creep and creep recovery. These data were subjected to statistical analysis and probabilistic modelling. The specimens for various properties were prepared of plain concrete while for shrinkage data both plain and reinforced concrete were used. Strength and creep recovery were studied upto the age of 240-day and 28-day respectively. Shrinkage and creep were studied upto 180-day. The present investigation shows that the variance in the concrete properties (mentioned above) due to batching is insignificant at 1 percent level of significance. The time-invariant probability modelling indicates that most of the generated data, for all the concrete properties follow the Gaussian and lognormal distribution, with 1 percent level of significance. The time-variant probabilistic modelling of all the four properties studied has been done using `chi-square minimization' method. The mean function of each of these properties is found to be linearly related with the logarithm of time. Further most of the observed data lie inside the 99 percent confidence envelopes. Results of numerically generated data reauthenticate the results of experimental data. Keywords Creep, Creep Recovery, Numerical Simulation, Probability Modelling, Regression Analysis, Shrinkage, Statistical Analysis, Strength