The shrinkage of oak predicted from its anatomical pattern:: validation of a cognitive model

This paper deals with the ability of a scientific strategy to compute the shrinkage behaviour of any oak sample, regardless of its origin, density, growth ring, etc. This approach uses the description of the actual oak structure at the annual ring level (i.e. the spatial organisation of the radial ray-cells, fibre, parenchyma areas and large vessels), to evaluate its shrinkage/swelling and elastic properties in the transverse directions. For the shrinkage properties, computed results were compared with experimental values measured on a set of samples depicting a very large diversity of anatomical patterns. The accuracy of our prediction is about 5% in the tangential direction and 20% in the radial direction, which is much better than statistical models over a wide range of variables. These results are discussed and a few microscopic observations with ESEM allow explanation of anomaly points of behaviour to be formulated. Such good results could allow this approach to be used to study the influence of growing conditions or of global changes upon physical wood properties.