Aiming at introducing flexibility for starting the SA algorithm from different positions
in the search space, this work introduced a variation in the original cost formula, where,
instead of being treated as constants in the original algorithm, the α, β and γ elements are
turned into PFIH parameters. The new values of α, β and γ change at each execution being
captured by a normal distribution N(μ,σ), with average values of αμ = 0.7; βμ = 0.1;
γμ = 0.2 (suggested as optimal by Solomon), and standard deviation (σ ) = 1. With this
variation of the heuristic orders, the customers being inserted by the PFIH algorithm maintain
a good arrangement because the averages are centered on the optimal values obtained
empirically and the small variations cause perturbations that create different initial solutions
for the SA. This modification in the PFIH algorithm was inspired on the work of de Oliveira
et al. (2004) that implements the change of the heuristic through taking the values of α, β
and γ from a uniform distribution changing from 0 to 1.