Based on a long-term exposure test under natural tidal environment,3,150 groups of data on free chloride ion mass fraction were obtained and used to established the chloride ion mass fraction prediction model by the support vector regression(SVR)method.We proposed a prediction model for the chloride ion mass fraction in fly ash concrete.The established SVR model analyzed the influences of data preprocessing,kernel function and hyperparameter optimization on the accuracy of prediction results.Besides,the effects of four input parameters,including water-cement ratio,fly ash content,exposure time and permeation depth,on the prediction results were explored.Meanwhile,predictions of free chloride ion mass fraction based on unmeasured parameters were also conducted.Results showed that the best prediction results for free chloride ion mass fraction were obtained by using a normalized data preprocessing method,RBF kerne function and Bayesian optimization hyperparameter optimization method.When the free chloride ion mass fraction was less than 0.1%,there were an obvious difference between prediction values obtained by the SVR method and the measured free chloride ion mass fraction.
free chloride ion mass fractionsupport vector regressionfly ash concreteprediction
维普
万方数据
