On the algorithmic synthesis of fluid and thermofluid systems

Individual technical components are usually well optimized. However, the design process of entire technical systems, particularly in its early stages, is still dominated by the practical experience of system designers. Especially the interaction of the individual system components at varying conditions leads to a high level of design complexity. In this context, the application of modern algorithmic optimization methods offers the potential to reduce costs and at the same time increase energy efficiency. The overarching vision is therefore the widespread use of these methods to support the human-driven design process of technical systems. As a contribution to this, the algorithmic system synthesis of fluid and thermofluid systems using mathematical programming techniques is examined in this thesis. For this task, the objective is to select and combine individual components with defined characteristics from a set of available options in order to obtain an optimal overall system capable of covering a given collective of varying loads at minimum cost.

The starting point is the presentation of an optimization model for the synthesis of fluid systems. Based on this, model extensions and algorithmic methods are developed that aim at an optimal synthesis of systems on a practice-oriented scale. For fluid systems, an approach is proposed that exploits domain-specific properties and is able to outperform commercial standard solvers. Simulated Annealing is used to generate good initial solutions for the approach, whereupon a domain-specific relaxation procedure is used to identify strong bounds for evaluating the obtained solutions. In a subsequent step, globally optimal solutions can be obtained using the Branch-and-Bound method. This approach is then further analyzed for the application example of booster stations. Furthermore, it is demonstrated how the resilience of existing systems can be increased with respect to the failure of individual components by using quantified programming.

Moreover, a model extension is presented to extend the consideration to thermofluid systems. These systems are assumed to comprise fluid systems with superimposed heat transfer, which ensures compatibility with the basic fluid system model. However, while the previous considerations exclude dynamic behavior that results, inter alia, from the coupling of adjacent points in time by storage components, two time representations suitable for this application are presented and discussed. One of these two representations is a more traditional discrete-time approach and the other is a novel continuous-time approach based on the consideration of variable time intervals.

Overall, this thesis is intended to provide a foundation for further research in the field of algorithmic synthesis of technical systems and its transfer to engineering practice.