The effectiveness of logistic network design andmanagement for complex and geographically distributedproduction systems can be measured in terms of directlogistic costs and in terms of supply chain productionperformance. The management of transportation logistics,for instance, involves difficult trade-offs among capacityutilization, transportation costs, and productionvariability often leading to the identification of multiplelogistic solutions. This paper defines and compares threedifferent modeling approaches to systematically assesseach identified logistic alternative in terms of actualtransportation costs and expected production losses. Thefirst modeling approach examined in the paper is amathematical model which provides the statistical basisfor estimating costs and risks of production losses insimple application cases. The second model is astochastic,discrete event simulation model of bulkmaritime transportation specifically designed to capturethe dynamic interactions between the logistic network andthe production facilities. The third one is an AI-basedmodel implemented as a modular architecture of ArtificialNeural Networks (ANNs). In such an architecture eachnetwork establishes a correlation between the logisticvariables relevant to a specific sub-problem and thecorresponding supply chain costs. Preliminary testing ofthe three models shows the relative effectiveness andflexibility of the ANN-based model; it also shows thatgood approximation levels may be attained when eitherthe mathematical model or the simulation model are usedto generate accurate ANN training data sets for eachtransportation/production sub-problem.
展开▼
