This thesis presents an innovative methodology for System Design Optimisation (SDO) through the framework of Model-Based System Engineering (MBSE) that bridges system modelling, Constrained Global Optimisation (CGO), Uncertainty Quantification (UQ), System Dynamics (SD) and other mathematical tools for the design of Complex Engineered and Engineering Systems (CEdgSs) under epistemic uncertainty. The problem under analysis has analogies with what is nowadays studied as Generative Design under Uncertainty. The method is finally applied to the design of Space Systems which are Complex Engineered Systems (CEdSs) composed of multiple interconnected sub-systems. A critical aspect in the design of Space Systems is the uncertainty involved. Much of the uncertainty is epistemic and is here modelled with Dempster Shafer Theory (DST). Designing space systems is a complex task that involves the coordination of different disciplines and problems. The thesis then proposes a set of building blocks, that is a toolbox of methodologies for the solution of problems which are of interest also if considered independently. It proposes then a holistic framework that couples these building blocks to form a SDO procedure. With regard to the building blocks, the thesis includes a network-based modelling procedure for CEdSs and a generalisation for CEdgSs where the system and the whole design process are both taken into account. Then, it presents a constraint min-max solver as an algorithmic procedures for the solution of the general Optimisation Under Uncertainty (OUU) problem. An extension of the method for the Multi-Objective Problems (MOP) is also proposed in Appendix as a minor result. A side contribution for the optimisation part refers to the extension of the global optimiser Multi Population Adaptive Inflationary Differential Evolution Algorithm (MP-AIDEA) with the introduction of constraint handling and multiple objective functions. The Constraint Multi-Objective Problem (CMOP) solver is however a preliminary result and it is reported in Appendix. Furthermore, the thesis proposes a decomposition methodology for the computational reduction of UQ with DST. As a partial contribution, a second approach based on a Binary Tree decomposition is also reported in Appendix. With regard to the holistic approach, instead, the thesis gives a new dentition and proposes a framework for system network robustness and for system network resilience. It finally presents the framework for the optimisation of the whole design process through the use of a multi-layer network model.