Guenov, Marin D.Molina-Cristobal, ArturoBile, Yogesh Hanumant2023-04-062023-04-062018-08https://dspace.lib.cranfield.ac.uk/handle/1826/19418During the conceptual design of complex systems, architects study a number of different options, which comprise the architectural design space. Usually, new system architectures (SAs) are created by modifying existing ones, e.g., by deleting existing and/or adding new elements. Once the concept is synthesised, the architect wishes to swiftly find the effect of the proposed architectural changes at system level. This would involve sizing of the modified sub-systems, and then, obtaining the system level performance. In turn, this involves time-consuming activities, such as re-arrangement (orchestration) of computational tasks and models. Also, depending on the results, the architect may undertake further modifications. When doing this, a means to navigate across the RFLP (Requirements-Functional-Logical-Physical) views of the SA may be required, in order to trace elements affected by these modifications. Generally, several iterations are involved between architecting and sizing during conceptual design, which, if manually performed, result in a tedious and time-consuming process. There are existing methods which address this problem, but these have significant limitations in that they are usually system specific and often involve an excessive amount of time-consuming manual tasks. Within this context, the research aim is to improve the efficiency of the architectural design space exploration (ADSE) process, by automating repetitive computational tasks, thus enabling the designer to swiftly and interactively explore multiple SA options. A novel method, comprised of two parts, has been developed to achieve the aim. In the first part, a graph-theoretic approach is employed to enable architectural element dependency analysis. Here, the relationships between the architectural elements are stored as a graph. Algorithms, such as ‘Depth First Search’ and ‘Transitive Closure’ are then applied to assist the architect in tracing the dependencies between elements that might be affected by a proposed change to other elements of the SA. In the second part, the architecture is assessed to find the system level performance. The inputs needed for rapid assessment include the functional and logical views of the SA, and the requisite steady-state computational models associated with each of the ‘logical’ components. The assessment process itself consists of three steps. In the first step, the sequence of the sub-systems is automatically generated by extracting a sub-systems source-sink ‘Dependency Structure Matrix (DSM)’ from the logical view, followed by the application of an algorithm which determines the systems’ sizing sequence. In the second step, the individual sub-systems and system level workflows are constructed. Here, the computational workflow (a network of computational models) is represented as a bipartite graph. A maximum matching enumeration algorithm is used to find all possible workflows for a given model set, and another algorithm, to choose from these the most computationally efficient one, i.e., the workflow with the lowest number of reversed variables. In the third step, the workflows produced in the second step and subsystems ’sizing sequence obtained in the first step are combined to produce a complete workflow. To demonstrate and evaluate the proposed enablers, the author developed a prototype object-oriented architecting tool. The enablers were individually and collectively verified on representative test-cases. Comparison with the existing methods confirmed the claimed advantages of the proposed approach, namely, reducing the number of manual activities, which results in swifter and interactive ADSE process. Feedback obtained from experts in the aircraft industry during an initial qualitative evaluation session confirmed the usefulness of the proposed method.en© Cranfield University, 2015. All rights reserved. No part of this publication may be reproduced without the written permission of the copyright holder.Architectural design space exploration (ADSE)system architecture (SA)architectural element dependency analysisfunctional viewlogical viewdependency structure matix (DSM)Component-driven computational design of complex engineering systems.Thesis