This thesis proposes sophisticated simulation-based optimization approaches for the design of unreliable serial production lines. Two challenging problems are considered: the buffer allocation problem, and the joint buffer and service time allocation problem. Allocating optimal buffer capacities and the assignment of appropriate service time to each machine is fundamental to enhance the production rate and to reduce investment costs (either in the form of work-in-process inventory costs, floor space use or capital investment), and are then among the most essential insights needed for designing production lines. Therefore, the design of effective optimization techniques is essential for the competitiveness of enterprises.

The thesis is divided into four chapters. Chapter 1 and 2 aim to develop metaheuristics based on the particle swarm optimization. The algorithms proposed are developed to solve theoretical function, real optimization engineering, and the buffer allocation problem. Chapter 3 presents a robust hybrid technique, combining finite perturbation analysis with genetic algorithm, to solve the buffer allocation problem. The hybrid algorithm is extended in chapter 4 for allocating simultaneously service times and buffer capacities in production lines with unreliable machines.