Optimization of aggregate testing using polynomial regression

Abstract

Recycled Aggregate (RA) has only been used as sub-grade applications, roadwork and unbound materials because it is believed that RA has lower strength and poorer quality than normal aggregate. This paper assesses RA quality by proposing a new testing procedure. The new procedure is developed by using correlation information among the standard twenty three tests commonly employed in the construction industry. Samples of RA from ten demolition sites were obtained with service life ranging from ten to forty years. One additional set of samples was specifically collected from the Tuen Mun Area 38 Recycling Plant, Hong Kong for testing. The characteristics of these eleven sets of samples were then compared with normal aggregate samples. A Vandermonde matrix for interpolation polynomials coefficient estimation is employed to yield detailed mathematical relationships between results of two particular tests, from that redundant tests can be identified. Different orders of interpolation polynomials are used for comparison, hence the best-fit equations with lowest fitting errors from different orders of polynomials can be found. This paper shows that four out of twenty three tests are required in the new testing procedure.