As the complexity of engineering systems increases, the cost of computer simulations, experimentation and physical tests for design and development also gets expensive. Due to the significant resources required by these physical and computational experiments on a new design, only sparse data are generally available and are often not sufficient enough to build accurate meta-models. A common approach to reduce the cost and the time cycle of a design process is to use meta-modeling techniques to build simpler and faster mathematical models to carry out the decision-making process for a new design. If datasets are available from legacy systems which belong to a similar family as the new design, one can leverage that information and knowledge to improve the accuracy of models built for the new system. In this work, a Bayesian multi-source modeling technique has been developed which combines models built on data from legacy systems with sparse data for a new design to improve the predictive capability of meta-model for the new design.
展开▼