Generative design of robust modular system architectures

  • Giota Paparistodimou

Student thesis: Doctoral Thesis


Modern engineering systems are frequently formed from complex networks of interwoven technological solutions, whose functions combine to enable key functions in the society: the defence of nations, the transport of people, the transmission of energy. The creation of such systems requires a detailed understanding of how the components interact both physically and functionally as a whole. However, while physical and functional connections (e.g. lines of power and control) can be defined, characterising a system's level of robustness when subjected to different disruptions is more challenging.;Essentially although many system architectures aim to create a degree of redundancy, that ensures robust operation under disruptions, it's difficult to quantify the level of robustness and so to evaluate and select the right system architecture option during the initial stage of the design.;This problem can be observed in many different domains of distributed engineering systems i.e. systems dictated by their configuration of source and sink components, structured in such a way as to provide a specific set of functions. Because of this, architects often do not consider, how the definition of the modular configuration (i.e. the degree of modularity) affects the overall robustness of the system.;Indeed, the choice of new system architectures is often dictated by subject matter experts and previous designs during the initial design phase. This leads to limited exploration, analysis and evaluation of potential system architecture design options, and selection of the system architecture that proves to be balanced between robustness and modularity at the beginning of system development.;Motivated by this need this thesis proposes a 'Robust Modular Generation and Assessment' (RoMoGA) methodology that combines a network tool (used to create alternate system architecture options) with a robustness and modularity evaluation metrics (that quantifies the robustness and modularity of each candidate system architecture option). The methodology has been used in case studies of three naval ship distributed systems and an explorative application was performed though experiments.;The industrial evaluation was based on semi-structured interviews and industrial design practices. The findings of the evaluation highlighted the existence of trade-offs between redundancy, modularity and robustness, the influence of the type of redundancy, the effects of the type of disruptions and effects of patterns and topological features. The evaluation part of the study led to redesign and update proposals for the original naval system designs, which the experts have assessed as rational and improved design solutions. The thesis reflects on the limitations of the proposed methodology and makes recommendations for future work.
Date of Award13 Aug 2020
Original languageEnglish
Awarding Institution
  • University Of Strathclyde
SponsorsUniversity of Strathclyde
SupervisorAlex Duffy (Supervisor) & Ian Whitfield (Supervisor)

Cite this