Ehsan Mahmoodi
School of Engineering Science
Ehsan Mahmoodi defends his thesis "Simulation-based knowledge-driven decision support for manufacturing systems design".

In today’s globalized manufacturing environment, companies face major challenges such as rapidly changing customer demands, short product life cycles, increasing customization requirements, and intense competition. To remain competitive, many manufacturers are adopting next-generation manufacturing systems (NGMSs) that use Industry 4.0 technologies to improve flexibility and intelligence. However, the successful implementation of these advanced systems depends heavily on effective system design. Manufacturing systems design involves determining key parameters such as buffer sizes, production rates, the number of stationary and mobile resources, system configuration, and control policies. Because these parameters are highly interdependent and often need to be optimized simultaneously across multiple objectives, the decision space becomes large and complex. This complexity, together with uncertainty in manufacturing environments, creates a difficult optimization challenge known as the manufacturing systems design problem (MSDP).
MSDP is characterized by the stochastic and dynamic behavior of manufacturing systems (MSs), conflicting objectives, and high-dimensional decision spaces. To address this problem, this thesis proposes a computer-based decision support approach that uses discrete-event simulation to capture the complexity and uncertainty of MSs. To deal with the multi-objective nature of MSDP, simulation-based multi-objective optimization (SBMOO) is employed to identify Pareto-optimal designs that balance trade-offs among conflicting objectives such as throughput, work-in-progress, and resource utilization. However, SBMOO is computationally expensive because it requires a large number of simulation evaluations. To improve its performance, this thesis investigates knowledge-driven optimization (KDO) approaches. In particular, knowledge about bottlenecks, resource utilization, and buffer utilization patterns is used to guide the search process more effectively. This thesis aims to advance SBMOO for MSDP, with a particular focus on the joint buffer and server allocation problem under uncertainty, which is a relatively understudied variant of the problem.
To achieve this overall aim, the thesis formulates three research questions. The findings, presented in seven appended papers, lead to three main conclusions. First, simulation-based optimization is an effective and practical approach for identifying near-optimal MS designs under uncertainty, provided that the optimization process is carefully adapted to the computational and structural characteristics of the problem. Second, the current design of an MS should not be viewed merely as a baseline for improvement, but also as a structured source of useful knowledge. When this knowledge is systematically extracted and represented, it can support continuous improvement at strategic, tactical, and operational levels. Third, when such extracted knowledge is integrated into the optimization process, both the efficiency of the search and the quality of the resulting designs improve substantially. Across the studies, improvements in key performance measures were demonstrated in industrially grounded contexts. Overall, the thesis shows that the progression from SBMOO to knowledge extraction and ultimately to KDO represents a coherent and transferable scientific contribution toward more effective design and operation of NGMSs.
2026. Doctoral thesis.
School of Engineering Science