Multidisciplinary design optimization for expendable launch vehicles

Since the dawn of the space age, development, production and operations of launch vehicles and other Space Transportation Systems (STS) have required huge investments. After the Moon landings, drastically reduced budgets forced engineers to start including in the design cost and programmatic aspects. In the current international context, which shows several attempts to reduce the cost for the access to space, the quality of the design process in general and of the initial design decisions in particular assume a key role, being the major drivers of the total life cycle cost (LCC) of launch systems. Specifically, it has been shown that about 80% of the LCC is determined in conceptual design, while optimization efforts in later design phases only result in minor improvements. For these reasons, Multidisciplinary Design Optimization (MDO) was chosen as the main topic of this doctoral research, in light of its potential benefits on both the efficiency of the design process (i.e. required time and effort) and the quality of the design solutions. In the European scenario, ESA’s Future Launchers Preparatory Programme (FLPP) and other national projects are paving the way - through technology demonstrators and system studies - for the transition in the 2025 timeframe from the current fleet of launchers (Ariane 5, Soyuz and Vega) to a more flexible and cost effective, modular family of Expendable Launch Vehicles (ELV). Hence, through collaboration with ESA, the primary application of the present research was identified in ELVs, although the developed environment was kept as flexible as possible to allow for easy extendibility to other STSs, such as re-entry vehicles, crewed and reusable systems. The main objective of the research was therefore identified in the development and quantitative assessment of a MDO environment capable of tackling the early design phases of ELVs. The focus was primarily on the engineering modelling aspects, whereas the necessary global and local optimization infrastructure was only considered as a mathematical tool enabling MDO, in light of the criticality of the models definition for the applicability of MDO methodology to real-world design. In fact, in spite of the rather impressive list of achievements promised by MDO and of a growing interest in the research community, successful industrial applications are still extremely rare. This is in part due to a certain resistance of design offices, motivated by the large required initial investments, to the introduction of MDO against the traditional fixed-point iterations process. But the largest obstacle is certainly constituted by the fact that MDO’s industrial applicability is subject to the assessment of the confidence which can be placed in the achieved design solutions. In particular, the definition and calibration of the multidisciplinary models are particularly challenging, the main obstacle being the identification of a reasonable compromise between simplicity and accuracy. With this problem in mind, a quantitative assessment of the engineering models was therefore carried out, including a detailed analysis of the accuracy of all developed analysis codes, both in terms of disciplinary errors and of system level sensitivities and results. Although a rather significant number of developments documented in recent years have been focused on the application of MDO to the design of different types of STS, the evaluation of accuracy and reliability of such models to the extent described in the present research represents an innovative and necessary endeavour, to the author’s knowledge. The main question proposed for the research - whether it is possible to develop relatively simple models permitting fast design cycles while still ensuring sufficient accuracy to place confidence in the achieved design solutions – does not hold a straightforward answer. The computational effort required with the implemented models surely matches the original target (i.e. <2 s for a complete multidiscipinary analysis), but it is not nearly as easy to measure the accuracy requirement. It can however be said that the decided two-steps development process proved to be of key importance for the incremental improvement of the models, which ensured a sensible enhancement in accuracy at a manageable price in computational effort from Version 1 to Version 2. As a result, the final models are characterized by accuracies on the global performance which should be sufficient in most of the cases to fairly compare two significantly different design solutions through MDO. Throughout lengthy details on the modelling and validation as well as on few relevant applicative cases, this PhD research highlights how the development and tuning of a reliable MDO environments is a very complex task, requiring large efforts in most engineering areas, besides computer science and mathematics. However, reasonable accuracies and physically sound design modifications can be obtained through the MDO approach, even when exploiting only fast engineering level models to avoid resorting to high performance computing. With today’s computational resources, which even allow conceiving the introduction of high fidelity information in the design cycle, MDO guided by human expertise is therefore a powerful approach for the initial design phases of launchers and other STS. Although the huge initial investment in terms of development and personnel training is a major obstacle to widespread industrial application, it is the opinion of the author that the resulting benefit in terms of design quality is well worth the effort, with the potential of contributing to the long term goal of achieving low cost access to space.

La tesi tratta il tema dell'ottimizzazione multidisciplinare per lanciatori spendibili, in particolare nell'ottica di una applicazione industriale e quindi con un forte accento sull'analisi quantitativa dell'accuratezza e affidabilità dei modelli ingegneristici multidisciplinari sviluppati.