Theoretical models of grain growth predict dust properties to change as a function of protoplanetary disk radius, mass, age and other physical conditions. We lay down the methodology for a multi-wavelength analysis of (sub-)mm and cm continuum interferometric observations to constrain self-consistently the disk structure and the radial variation of the dust properties. The computational architecture is massively parallel and highly modular. The analysis is based on the simultaneous fit in the uv-plane of observations at several wavelengths with a model for the disk thermal emission and for the dust opacity. The observed flux density at the different wavelengths is fitted by posing constraints on the disk structure and on the radial variation of the grain size distribution. We apply the analysis to observations of three protoplanetary disks (AS 209, FT Tau, DR Tau) for which a combination of spatially resolved observations in the range ~0.88mm to ~10mm is available (from SMA, CARMA, and VLA), finding evidence of a decreasing maximum dust grain size (a_max) with radius. We derive large a_max values up to 1 cm in the inner disk between 15 and 30 AU and smaller grains with a_max~1 mm in the outer disk (R > 80AU). In this paper we develop a multi-wavelength analysis that will allow this missing quantity to be constrained for statistically relevant samples of disks and to investigate possible correlations with disk or stellar parameters.