DDDAS-based multi-fidelity simulation for online preventive maintenance scheduling in semiconductor supply chain

This research intends to augment the validity of simulation models in the most economic way using the DDDAS (Dynamic Data Driven Application Systems) paradigm. Implementation of DDDAS requires automated switching of model fidelity and incorporating selective, dynamic data into the executing simulation model. Comprehensive system architecture and methodologies are proposed, where the components include 1) RT (Real Time) DDDAS simulation, 2) grid computing modules, 3) Web Service communication server, 4) database, 5) various sensors, and 6) real system. Four algorithms are developed to facilitate integration of the various components in the DDDAS system. They are 1) data filtering algorithm using control charts, 2) preliminary fidelity selection algorithm using Bayesian belief network, 3) fidelity assignment algorithm using integer programming and 4) simulation model reconstruction algorithm using multiple linear regression. A prototype DDDAS simulation was successfully implemented for preventive maintenance scheduling in a semiconductor supply chain. The initial results look quite promising.