4.2.2. Computation of upper bound
At each node, two feasible solutions are constructed by allocating the unassigned parts into dedicated classes
and a random class. The better (i.e., lower) objective value of these two solutions, Ud or Ur gives an upper
bound for node nði; jÞ at stage c 1.
Ud of node n(i, j): The unassigned parts i þ 1; . . . ; P are assigned to dedicated classes c; c þ 1; c þ 2 and so
on. Let us call this solution (part and allocation assignments) as SOLd. The objective function is computed
as