Methods for Early Model Validation : Applied on Simulation Models of Aircraft Vehicle Systems

Abstract: Simulation  models of physical systems, with or without control software, are widely used in the aeronautic industry in applications ranging from system development to verification and end-user training. With the main drivers of reducing the cost of physical testing and in general enhancing the ability to take early model-based design decisions, there is an ongoing trend of further increasing the portion of modeling and simulation.The work presented in this thesis is focused on development of methodology for model validation, which is a key enabler for successfully reducing the amount of physical testing without compromising safety. Reducing the amount of physical testing is especially interesting in the aeronautic industry, where each physical test commonly represents a significant cost. Besides the cost aspect, it may also be difficult or hazardous to carry out physical testing. Specific to the aeronautic industry are also the relatively long development cycles, implying long periods of uncertainty during product development. In both industry and academia a common viewpoint is that verification, validation, and uncertainty quantification of simulation models are critical activities for a successful deployment of model-based systems engineering. However, quantification of simulation results uncertainty commonly requires a large amount of certain information, and for industrial applications available methods often seem too detailed or tedious to even try. This in total constitutes more than sufficient reason to invest in research on methodology for model validation, with special focus on simplified methods for use in early development phases when system measurement data are scarce.Results from the work include a method supporting early model validation. When sufficient system level measurement data for validation purposes is unavailable, this method provides a means to use knowledge of component level uncertainty for assessment of model top level uncertainty. Also, the common situation of lacking data for characterization of parameter uncertainties is to some degree mitigated. A novel concept has been developed for integrating uncertainty information obtained from component level validation directly into components, enabling assessment of model level uncertainty. In this way, the level of abstraction is raised from uncertainty of component input parameters to uncertainty of component output  characteristics. The method is integrated in a Modelica component library for modeling and simulation of aircraft vehicle systems, and is evaluated in both deterministic and probabilistic frameworks using an industrial application example. Results also include an industrial applicable process for model development, validation, and export, and the concept of virtual testing and virtual certification is discussed.