An Object-Oriented Framework for Multidisciplinary Design Optimization

Abstract

We present πMDO, an object-oriented framework that facilitates the use of algorithms for multidisciplinary optimization (MDO). The resulting implementation of the MDO architectures is efficient, scalable, and portable. The main advantage of the proposed framework is that it is flexible, with a strong emphasis on object-oriented classes and operator overloading and it is thus ideal for the rapid development and evaluation of new MDO architectures, as well as the benchmarking of existing ones. The top layer interface is programmed in Python and it allows for the layers blow the interface to be programmed in C, C++, Fortran and other languages. We describe the implementation of πMDO and demonstrate that we can take advantage of object-oriented programming and operator overloading to obtain intuitive, easy-to-read, and easy-to-develop codes that are at the same time efficient. This allows developers to focus on the new algorithms they are developing and testing, rather than on implementation details. Several numerical experiments verify that the various MDO architectures yield the correct solutions and allow the evaluation of their relative performance.