Research And Prediction Of Time-varying Law Of Compressive Strength Of Existing Building Concrete

Concrete is the most widely used building material with excellent mechanics performances and durability.Compressive strength is an important mechanical property of concrete,which changes with time under the influence of load,environment and other factors.At present,there are rebound method,drill core method,pullout method and ultrasonic rebound method to test the compressive strength of concrete,among which the rebound method and drill core method are more common.It is necessary to know the compressive strength of concrete when evaluating,reinforcing and remodeling the existing buildings.Usually,the measured strength value is used to calculate the resistance of the structure,and pay little attention to the time variation of strength,so it is impossible to predict the safety performance index of the building in the remaining service life.Therefore,the rule of concrete strength variation is of great significance in the structural assessment,reliability analysis,determination of future service life and reinforcement of existing buildings and so on.This thesis summarized the existing theories of concrete strength varying with time at home and abroad.According to the appraisal report of buildings in Shandong,73,840 pieces of strength data measured by the rebound method and 4,149 pieces of strength data measured by the drill core method were collected and sorted out.The strength was divided by the mean value of axial compressive strength corresponding to different strength grades.Therefore,the influence of different design strengths can be eliminated and the applicability of time-dependent variation rule for concrete compressive strength of existing buildings can be improved.Based on drill core method and rebound method of actual measured data,using the hypothesis test of skewness and kurtosis test and?~2 fitting test and the probability density function of the relative strength of concrete,proved that the existing buildings of concrete compressive strength is normal distribution.In addition,due to different testing process and sample sizes,the mean value obtained by the rebound method is smaller than that of the drill core method,while the standard deviation of the drill core method is greater than that of the rebound method.In view of the lack of data over 30 years,there is a large deviation in the fitting of data when the service life is higher,so the neural network model is used to predict the concrete strength.On the basis of the original data and the Grubbs test in GB/T 4883-2008"Statistical Processing and Interpretation of Data Judgment and Processing of Normal Sample Outliers",the outliers were found and eliminated.Then,50 subsamples measured by the rebound method and 24 subsamples measured by the drill core method were obtained by averaging the data for each year.The quadratic function,exponential and logarithmic complex function,exponential and linear complex function are used to fit the data,and the results show that the quadratic function is more suitable to express the time-dependent variation rule for concrete strength of existing building.Finally,the time-varying law model of this study is compared with other models in China.The results show that the variation trend is basically the same,but due to different normalization methods and regions,the maximum value and the instant when the maximum value occurs are different.The prediction deviation between the ordinary BP neural network model and Genetic Algorithm BP neural network model is compared and analyzed.The results show that the prediction deviation of Genetic Algorithm BP neural network model is relatively small and stable,so it is used to predict the partial strength of the existing buildings with longer service life.Then GA-BP neural network model was used to predict relative strength of the existing building with the service life of 40?45 years by drilling core method and the relative strength of the existing building with the service life of 46?55 years by rebound method.By comparing the predicted values obtained by GA-BP neural network model and quadratic function time-dependent variation model,it is found that the difference between the two prediction models is less than 10%,so both models can be adopted.Finally,the main results of this thesis are summarized and the future research is prospected.