Figures & data
FIG. 1 The logarithmic sensitivities of the mean volume densities in bins 4 and 8 at final time with respect to the initial distribution bin averages (left) and bin slopes (right) for coagulation and growth.
TABLE 1 The logarithmic sensitivities of the mean volume densities in bins 4 and 8 with respect to the values of σ and β for coagulation and growth
FIG. 2 Results for recovering the initial distribution for coagulation and growth from a complete set of observations.
FIG. 3 Results for recovering β and σ for coagulation and growth from a complete set of observations.
FIG. 4 Results for recovering the initial distribution, β, and σ for coagulation and growth from a complete set of observations.
FIG. 5 Results for recovering the initial distribution without observations at bins 1, 2, and 3 for both coagulation and growth.
FIG. 6 Results for recovering the initial distribution with observations at only even bins for both coagulation and growth.
FIG. 7 Results for recovering β and σ with observations at bin 1 for both coagulation and growth.
FIG. 8 Results for recovering the initial distribution from a complete set of observations with different frequencies. RMS error on the left, and cost function value on the right.
FIG. 9 Recovering β and σ from observations of total surface density.
FIG. 10 Recovering the initial distribution from observations of total surface density.
FIG. 11 Recovering the initial distribution, β, and σ from observations of total surface density.
FIG. 12 Recovering the initial distribution and σ for growth using continuous and discrete adjoint gradients. The continuous formulation leads to a better performance of the optimization process.
FIG. 13 Recovering the initial distribution for coagulation using two implementations of the discrete adjoint gradients. The automatic differentiation and the direct implementation lead to similar performance.
