Prediction of rheological properties of filler modified cement paste from constituent properties, flow measurements and modelling

The world is facing a sand crisis due to depletion of natural deposits of good quality construction sand. Replacement of natural sand is crucial to curb the impact of the crisis. Though several alternatives exist, this thesis solely focus on crushed sand as replacement for natural sand. However, unless the crushed sand is properly processed, it is more angular and contains more fines than natural sand, which have a negative impact on concrete rheology. Measuring concrete rheology requires a lot of materials and labour, however, small scale rheology testing on equivalent cement pastes is found to correlate to concrete rheology. The thesis investigates the rheology of cement pastes with fillers from crushed sand, denoted matrices. The scope of the thesis is to investigate methods for quantification of matrix rheology, with emphasis on characteristics of the ingoing materials. Special focus is on the possibilities of establishing a one-point measurement of matrix rheology, i.e., a method describing the rheological behaviour of matrix by only one
parameter. Limitations of and possibilities for further developing the flow viscometer FlowCyl are studied. Additionally, the accuracy and efficiency of artificial neural network predictions of the cement paste rheology are investigated and compared to semi-empirical suspension models. The rheological properties of the matrices have been quantified experimentally by four parameters; the flow resistance ratio (obtained from the FlowCyl), the mini slump flow, the Bingham yield stress and the Bingham plastic viscosity. From the work of the thesis it is found that it is possible to predict rheology of matrices based on constituent materials when all dry materials (except silica fume) are characterized by the volumetric specific surface area per volume of matrix, as well as their mix proportions. An empirical equation is developed, which correlates especially good to the plastic viscosity of the matrices. The main limitation of the FlowCyl is its weak correlation to yield stress, however, it is found that this correlation can be increased by changing the geometry of the FlowCyl. Based on numerical simulations a prototype of a modified geometry of the FlowCyl has been produced, called the FlowFunnel, which provides better correlations to the yield stress than the FlowCyl. Lastly, artificial neural network predictions are found to be an efficient and accurate prediction tool for matrix rheology.