| Summary: | 碩士 === 國立臺灣大學 === 資訊管理學研究所 === 92 === This study proposes a heuristic algorithm to solve a
general master-planning problem of a supply chain network with
multiple final products. The objective of this planning algorithm
is (1)To minimize the processing, transportation , and inventory
costs under the constraints of the capacity limits of all the
nodes in a given supply chain network graph and the quantity and
due day requirements of all the orders. (2)To lower the impact of
fairness problem of greedy capacity allocation.
This study assumed that multi-finished items are made and shipped
on the given supply chain which results in common parts on common
nodes for different finished items. Three different ways are
proposed to solve the sharing capacity problem caused by common
components: greedy, average capacity, and proportional capacity.
All the three algorithms are basically composed of five steps.
(1). They split nodes in the supply chain network graph by
different functions the nodes perform, and set the initial
capacities of all nodes. (2).They transform the capacity units
shown on the graph, based on the unit of the final finished
product. (3).They sort all the orders by adopting a rule-based
sorting method to decide the scheduling sequence.(4).They extract
sub-networks from original networks according to final product
structure of orders. (5).Finally, for each order, the algorithms
find a minimum cost production tree under the constraints of the
order''s due day. They then compute the maximum available capacity
of this combination and arranges the suitable quantities of
production and transportation. If the demand cannot be fulfilled
before the due day, the order will have to be postponed. Repeating
the process above until the demand is completely fulfilled. The
difference is average capacity and proportional capacity are under
constraints of capacity using quota.
The three algorithms result in the same optimal solution as the
one by "Linear Programming" in eight different dimensions of
scenarios when no delayed orders present. In the four cases with
delayed orders, the three orders will still work out a
near-optimum solution in a shorter time.
|
|---|
