Data science approach to find an outlier in the group of cement dispersants

Abstract

Polycarboxylate ether (PCE) is an essential cement dispersant to be included in high-range water-reducing admixtures for concrete. The polymeric structure of PCE is easily modified to satisfy its performance requirement for a specific concrete produced in sites. A PCE developer polymerizes hundreds of PCE samples to find the final product to be applied and evaluating the performance of each sample is cumbersome considering the variability of concrete-constituting materials. A data-driven algorithm is proposed to probe a PCE showing superior performance. Principal component analysis is adopted to regulate the change in the viscosity curves of mortar samples incorporating the reference PCE. An outlier PCE is then found showing the highest dissimilarity, defined by the eigen-decomposition error and the changing pattern of the viscosity curves, compared to the reference PCE. As a result, the outlier PCE had the unique distinction of providing low viscosity for concrete.