Accurate modelling of PET multiaxial behaviour is fundamental for blow moulding process optimisation. In this paper, we present the identification of a strain induced crystallisation model on uniaxial and biaxial tension tests, undertaken on PET amorphous samples above T g at various draw rates and tension speeds [Marco, Y., 2003. Caractérisation multi-axiale du comportement et de la microstructure d’un semi-cristallin: application au cas du PET. Thèse de doctorat de l’ENS de Cachan]. Using an accurate equivalent deformation, a non-Newtonian viscous model coupled with induced crystallisation seems to give a good representation of strain hardening for different kinds of solicitations. Moreover, a WLF-like parameter is identified and allows to be a representative of the range of the process temperatures with a single test at 90°C.As temperature is a crucial parameter for both process and mechanical tests, we focus on the study of the influence of the temperature regulation on the dispersion of mechanical results. We use here a probabilistic approach giving convincing results and leading to a direct industrial application: knowing the heating regulation default of the process, one can predict the variation of the mechanical behaviour of the perform during blow moulding and therefore the variability of final product properties like thickness or crystallinity ratio.