Abstract

Aeronautical industry faces the increase of system complexity and reliability-based constraints, and in the meantime tries to reduce costs, to maximize the system performance and to improve the safety. The recent development of MBSE facilitates the transfer of models and enables to simulate the system behavior early in the development phase. In this context, the thesis aims to shape a collaborative and adaptive software environment to carry out uncertainty management on multidisciplinary aeronautical systems. Case Studies examine the implementation of a systematic CPM throughout the design process of a new commercial aircraft. This work raises the point of uncertainty-based optimization complexity and investigates different solutions to face this issue.

更多

Most of the analytical models tackled in the thesis derive from a set of regressions suited to an Airbus commercial aircraft. While Cameo Systems Modeler supports a modular modeling, ModelCenter bridges the gap between descriptive and analytical models while ensuring a great traceability. The multi-levels simulation enabled by ModelCenter helps identifying the critical parameters from the early steps of the design process. This holistic and data-driven approach drives the product development process by eliminating non-value-added activities. The variety of sensitivity analysis tools suits any type of system complexity.

The software environment supports the implementation of an uncertainty-based multidisciplinary optimization. Non-dominated Sorting Genetic Algorithm NSGA-II highlights the tradeoff between performance optimization and cost reduction and its influence on the optimal design. Reliability-based constraints reduce the solution space and affects the final design of the aircraft by shifting the Pareto-front away from the best objective values. ModelCenter provides effective tools to face the high level of complexity of optimization under uncertainty. While the parallelization of simulations on virtual machines enhances the computational performance, DOE screening enables reducing the design space by eliminating irrelevant inputs. The conversion of multi-objective into single objective function focuses the search for optimal on a part of the global Pareto-front and significantly shortens the computing time. However, this solution requires setting up a hierarchy between the objectives and thus leaves behind non-dominated design solutions.

Although the results show the ability of this software environment to design complex systems under uncertainty, it is difficult to extrapolate a general uncertainty-based multidisciplinary design optimization workflow for various aeronautical systems at Airbus. Each design under uncertainty depends on the model complexity, the size of the design space as well as the available computational resources. Improvements of the Case Studies models are possible by refining both performance and cost functions. While models linking them to the design parameters are difficult to set up, a precise definition may capture the complete product life cycle in the design process under uncertainty.

收起