Design and Modeling for Computer Experiments (Computer Science & Data Analysis)
Design and Modeling for Computer Experiments (Computer Science & Data Analysis)
| Hi-index | 0.00 |
As engineering analyses rely more on computer simulation, it is important to develop a methodology that accounts for inaccuracies present in all simulation models. Typically, simulation models are validated against the physical system they try to explain. This process is normally achieved in a qualitative manner. Thus, the comparison falls short of quantifying the agreement or disagreement between simulation model and the physical system. A better approach to validation considers all sources of error and uncertainties in both the simulation model and the physical system. Quantitative model validation is a means of achieving confidence in simulation models. This work will serve to outline a methodology for achieving the operating points required for validation and apply it to an electrical engineering problem. The study proposes using classical design of experiments, i.e. central composite design, in tandem with computer experiments to arrive at a small amount of physical experiments. These experiments allow for model comparison at multiple tiers. The system under study is a representative 400 V direct current (DC) distribution system suitable for evaluating the impact of a proposed grounding scheme [1]. Both system architecture and grounding method studied are under consideration for future naval vessels.