Physically-Based, Real-Time Visualization and Constraint Analysis in Multidisciplinary Design Optimization

Abstract

As computational tools becomes a valuable part of the engineering process, multidisciplinary design optimization (MDO) has become a popular approach for the design of complex engineering systems. MDO has had considerable impact by improving the performance, lowering the lifecycle cost and shortening product design time for complex systems; however, lack of knowledge on the design process is often expressed by the engineering community. This thesis addresses this issue by proposing a novel approach that brings visualization into the MDO framework and delivers a physically-based real-time constraint analysis and visualization.A framework and methodology are presented for effective, intuitive visualization of design optimization data. The visualization is effected on a Computer-Aided-Design (CAD)-based, physical representation of the system being designed. The use of a parametric CAD model allow real-time regeneration by using the Computational Analysis PRogramming Interface (CAPRI). CAPRI is used to link a general optimization framework to the CAD model. An example is presented for multidisciplinary design optimization of an aircraft.The new methodology is used to visualize the path of the optimizer through the design space. Visualizing the optimization process is also of interest for optimization health monitoring. By detecting flaws in the optimization definition, useless computations and time can be saved. Visualization of the optimization process enables the designer to rapidly gain physical understanding of the design tradeoffs made by the optimizer. The visualization framework is also used to investigate constraint behavior. Active constraints are displayed on the CAD model and the participation of design variables in a given constraint is represented in a physically intuitive manner.This novel visualization approach serves to dramatically increase the amount of learning that can be gained from design optimization tools and also proves useful as a diagnostic tool for identifying formulation errors. The ACDL has been a wonderful place to work, but also to socialize and meet people.Garrett, on the other side of the wall, has been great teaching me Latex ("One day to learn, a lifetime to master") and giving me so many insights on the American culture. Life in the lab would have been different without the French speaking crew (David, Jérome, Hector and Matthieu) and the many special people that made each day unexpected (Mark, Victor and Vivian). Thank you to Joe, for sharing these many hockey moments that helped me stay fit and sane ! Other people at MIT also deserve a particular thank: Franck, Manu and JB.Of course, I must acknowledged my family across the Atlantic. A lot of emails, Yahoo messenger discussions and phone calls can easily bring a piece of France to the US. And yes, Laetitia, I will graduate after you, but at least, I am finally graduating ! Finally, I could not end this section without talking about Carole. I wanted to deeply thank her for listening to me day after day, for supporting me ...