The numerical method developed for thermogravimetric analysis (TGA) of one-
stage processes is generalised to systems with multistage decomposition. Ei
ghteen different physicochemical models are used to describe kinetics at ea
ch of the stages. In a two-stage decomposition the algorithm realises seque
ntial analysis of 18 x 18 = 324 variants and characterises a point separati
ng the two stages on the TG curves. Thus, it provides the minimum of the av
erage square of deviation between the computed and the experimental curves
on the scale of the logarithm of the reduced time that is expressed as the
integral of the Arrhenius exponential. The same approach is used for the ap
proximation of three-stage decomposition. Three-stage decomposition of foam
ed polyurethane in air is considered as an illustrative example. The first
and second stages both correspond to bimolecular reactions, while the third
one has been found to be a diffusion controlled process.