Application of cluster analysis and development of a lifecycle environmental performance indicator to categorise construction materials
Auteurs
A. Marvuglia, P. Hild, B. Schmitt, and E. Benetto
Référence
Journal of Environmental Accounting and Management, vol. 4, no. 1, pp. 1-11, 2016
Description
This paper presents a two-step analysis of a set of construction materials: (1) a preliminary classification of the materials according to their environmental performances; (2) the definition of an aggregated indicator (called UIL5) of the lifecycle environmental performance of each material. Two different clustering techniques were applied and the results obtained shown a very good level of agreement. The first is an agglomerative clustering technique and the second is a non-supervised neural network (called self-organizing map). Each material was represented by a vector of six elements: five midpoint indicator values of the potential environmental impact of the material and, in addition, its cumulated primary energy (non-renewable), all normalized by cubic meter and by a function of the conductivity of the material. This kind of normalization was chosen to take into account the function with respect to which we compare the materials to each other in this application, i.e. their thermal insulation capacity, which depends on the thickness of the material's layer and on its thermal conductivity. The two clustering techniques yielded coherent results and the visual observation of the obtained clusters allowed the identification of potential explanatory variables that could be used to determine threshold values to distinguish classes of materials on the basis of their environmental performances.
