January 2025
This doctoral thesis explores the design process for manufacturing systems, which encompasses different architectural design decisions (ADDs), such as deploying software to hardware components, determining shop-floor topology, and planning production tasks for stations. Searching for valid or optimized ADDs is conducted through design space exploration (DSE), a methodology dedicated to finding solutions within a predefined and constrained design space. Frequent production changes in evolving manufacturing systems, which are in the focus of the fourth industrial revolution (Industry 4.0), necessitate frequent re-evaluation and re-synthesis of various ADDs. This increases the already high engineering effort and introduces errors, highlighting the need for automating ADDs. The increasing significance of DSE in currently evolving industrial systems, next to a large landscape of DSE solutions in different domains, stresses the importance of establishing a clear and detailed overview of DSE in the context of Industry 4.0. Moreover, the complex interactions between ADDs call for a multi-dimensional DSE approach to synthesize diverse ADDs together. Integrating these ADDs within industrial automation remains challenging, given an applicability shortage of existing DSE solutions from other domains in the Industry 4.0 domain. To address this challenge, this doctoral thesis investigates ADDs and their interplay within a multi-dimensional DSE for evolving software-intensive manufacturing systems, organized into three parts, each addressing one research question. Following the identified problem, the thesis’ first part presents a literature review of existing DSE approaches in Industry 4.0. The review yields a taxonomy of DSE possibilities across three sub-topics: architectural models, ADDs, and optimization methods. The taxonomy clarifies and classifies DSE in Industry 4.0, improving domain understanding, guiding research positioning, and identifying research gaps. It also helps practitioners by providing a common ground for communication and initial guidance in identifying relevant ADDs, finding correlated ADDs, and selecting suitable architectures and techniques. The second part establishes a novel multi-dimensional DSE approach that uses domain-specific input models and a satisfiability modulo theories (SMT) solver to automatically generate valid and optimal deployment, service composition, shop-floor topology, and production planning ADDs. The applicability of the proposed DSE approach is evaluated on different domain-relevant demonstrators, showcasing reduced engineering effort and improved design quality. The third part introduces mitigation constraints to investigate the synergy and ensure the compatibility of ADDs. This enables the exploration of different sequential and hybrid DSE workflows in addition to a joint DSE workflow. Findings reveal that starting with production planning ADDs yields the best overall performance, though specific optimization goals may favor other workflows (e.g., starting with deployment ADDs to prioritize communication coupling). Sequential workflows starting with topology ADDs are the fastest, while the joint workflow offers the highest solution quality and is easier to implement. This analysis provides additional guidance for software architects on enhancing the quality and efficiency of ADDs. This thesis contributes to manufacturing system design and practice by providing a comprehensive framework to understand and optimize the interplay of various ADDs. This framework encompasses i) an SMT-based multi-dimensional DSE approach for ADD synthesis and optimization, ii) an evaluation of DSE workflows and their performances, iii) a taxonomy of ADDs in Industry 4.0 to guide research and practice, and iv) recommendations for practitioners on optimizing ADDs. Overall, the thesis provides practitioners and researchers in industrial automation with tools to support more effective decision-making in designing and optimizing manufacturing systems.