We carried out deconvolution of the molecular weight distribution curves from gel permeation chromatography for polyolefins into individual active sites considering Flory distribution by an evolutionary-computing-based real-coded genetic algorithm, a nonlinear multivariate optimization algorithm. We applied the deconvolution to homo-polymers of 1-octene synthesized using heterogeneous Ziegler-Natta catalysts with different amounts of hydrogen. The molecular weight distribution was deconvoluted in to five Flory distributions, which showed a sensitivity to hydrogen amounts. With no hydrogen presence, the peaks corresponding to high-molecular-weight fractions were intense. As the amount of hydrogen was increased, not only did the intensities of the high-molecular-weight peaks decrease, but also peaks corresponding to low-molecular-weight fractions were observed. The method allowed us to determine the active site distribution of the polymer molecular weight distribution obtained from gel permeation chromatography. (C) 2010 Wiley Periodicals, Inc. J Appl Polym Sci 117: 3379-3385, 2010