Optimization strategy of model B4TW-SCC in predicting shrinkage of self-compacting concrete

Abstract

Shrinkage strain directly reduces the dimensions of concrete, leading to potential intrusion of deleterious substances into the structure, affecting the durability in the long term, especially self-compacting concrete (SCC), which differs from conventional concrete in its composition, featuring a higher paste content and a lower amount of coarse aggregate, and consequential reduction in restraining capacity of aggregate. Most available models were designed to estimate the shrinkage strain of conventional concrete and lack the efficient prediction models of SCC. The enhancement in predicting could be applied to minimize the drawback of shrinkage on the building quality by preventing cracks and surface defects. Therefore, the new model B4TW-SCC was adjusted from the model B4, encompassing composition and compressive strength, to closely fit the newly collected database comprising 1316 test curves of shrinkage, assembled from numerous published papers. The paper presented a detailed optimal process based on both the minimum error of statistical indicators and the weighting scheme of the database to determine the important factors of the model B4TW-SCC for predicting the shrinkage of SCC. The evaluation results indicated that the developed model exhibited the highest efficiency with NRMSE values of only 52.3% and 50.5% in estimating autogenous shrinkage and total shrinkage among the current models for conventional concrete of B4, fib, and ACI, as well as the models for SCC, including CEB90-SCC (Poppe) and JSCE-SCC (Aslani).