Prediction of Expansion Behavior of Cement Mortar Containing Different Steel Slag

• Main Authors: Juang, Chuen-Ul, 莊椿微
• Other Authors: Kuo, Wen-Ten
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/939tsg

碩士 === 國立高雄應用科技大學 === 土木工程與防災科技研究所 === 104 === Industrial byproducts arc furnace oxidation ballast (Electric arc furnace oxidizing slag, EOS) generated during the steel-making, reducing ballast (electric arc furnace reductive slag, ERS), Stone converter (basic oxygen furnace slag, BOF), desulfurization ballast ( desulfurization slag, DS), and byproduct lime (circulating fluidized bed boiler, CFB), is widely used in civil industries, but because circumstances have found long-term use of expanded production, leading to concerns about the security of the application. Therefore, this study will aggregate a series of laboratory steel furnace cement mortar expansion amount of data using multivariate non-linear regression forecast model and back propagation neural network, the establishment of the amount of expansion forecast model with different types of steel furnace cement mortar, Discussion different W / C, curing temperature, the amount of substitution, age resulting steel furnace cement mortar expansion behavior. The results showed that the use of multivariate nonlinear regression prediction model established by the amount of expansion forecast model prediction mode will be different according to the different coefficients of determination (coefficient of determination, R2), of which the exponential function equation, the exponential equation performed results for the best, R2 are higher than 0.8; and the use of back-propagation neural network forecasting model to explore the relationship of different w predict more than 0.9 R2. Ensure the viability of the prediction model, the present study and then take a group to carry out stabilization of steel ballast material, respectively, reducing ballast, ballast oxidative desulfurization ballast carried cement mortar bar expansion test values were changing the way the interpolation and extrapolation age and curing temperature of the variables, to predict model validation, analysis showed that the prediction model multivariate non-linear regression of a security alert in the range of 1.14-3.16. Prediction Model of the back-propagation neural network compared with the prediction of individual cases carried out, so verify the results of a small number of measured value will be greater than the predicted value. The prediction model established in this study is reliable and can be used as a reference for future projects on the application, that the early expansion of the material value, in order to clarify the applicability of the material.