The utility of proper orthogonal decomposition for dimensionality reduction in understanding behavior of concrete

Properties of wet and set concrete are influenced by a wide range of variables. With new formulations being tried and adopted, understanding workability, strength and durability characteristics of these formulations is of utmost importance. From among the wide range of variables that affect properties of concrete, identification of the most vital, interplay between variables, quantification of influence, for judicious manipulation of mix proportioning, placement, compaction and curing, to get the desired and targeted end results can vastly be improved by employing the state of the art data handling tools. Group method of data handling (GMDH), a set of mathematical algorithms, is of great usage potential in multi-variable data modeling, optimization and pattern recognition. Proper Orthogonal Decomposition (POD) a subset of GMDH, a technique for systematic dimensionality reduction and pattern recognition, is of great importance in studying complex datasets. This paper presents the need for adoption of GMDH techniques in concrete technology with an account of trends in this direction and also provides an illustration of POD's utility as a valid decision-making tool in dimensionality reduction and projection of behavior of concrete subjected to elevated temperature.