A multi-modeling approach using simulation and optimization for supply-chain network systems.

摘要

Enabling two large complex models that use different types of execution algorithms to work together is a difficult problem. The semantics between the models must be matched and the execution of the algorithms have to be coordinated in a way that the data I/O timing is correct, the syntax and semantics of the data I/O to each algorithm is accurate, the synchronization of control is maintained, and the algorithms concurrent execution is managed.;In this work, an approach is developed that uses a Knowledge Interchange Broker (KIB) to enable composition of general optimization and discrete process models consistent with their respective execution algorithms. The KIB provides a model specification that matches data and control semantics across two distinct classes of optimization and discrete process models. The KIB has a sequential execution algorithm in which two independent execution algorithms are synchronized.;A domain where this kind of problem is seen is in the development of computational models of real world discrete manufacturing supply chain network systems. Modeling these systems require that the planning systems, manufacturing process flows, and their interactions to be concisely described. Planning systems generally use optimization algorithms for calculating future instructions to command what the manufacturing processes build. Manufacturing processes are commonly modeled using discrete event simulations. For supply chain network systems, the planning and manufacturing models can be very large. The KIB is an enabler for correctly combining these models into semantically consistent multi-models.;Two common types of models used for planning and manufacturing are Linear Programming (LP) and Discrete Event Simulation (DES). In this work, a KIB has been developed and demonstrated on a set of theoretical representative semiconductor supply-chain network problems using LP and DES. The approach has then been successfully applied to integrating planning and simulation models of real-world problems seen at Intel Corporation's multi-billon dollar supply-chain network.