Multi-objective worker allocation in U-shaped assembly lines

Abstract

This dissertation studies multi-objective U-shaped assembly line worker allocation problems with symmetrical and rectangular layouts having manually operated machines. The objective is to assign tasks into a U-line and allocate tasks to workers hierarchically. The primary objective function of a number of workers is minimized. Then, the deviation of operation times of workers and the walking time are minimized simultaneously. Several products are assembled in 7-task to 297-task benchmarked problems with given cycle times. This problem, a combinatorial optimization problem, is modeled with mathematical formulation initially. To produce a good quality solution and time, the development of evolutionary algorithms is used for a large-sized problem. With the Pareto dominance relationship that is used to solve a problem instead of relative preference of multiple objectives, four algorithms have been developed as follows: Non-dominated Sorting Genetic Algorithm-II (NSGA-II), Memetic Algorithm (MA), COINcidence algorithm (COIN), and Particle Swarm Optimization with Negative Knowledge (PSONK). The quality solutions of Convergence to the Pareto-optimal set, Spread, and Ratio of non-dominated solutions are compared in each problem for all multi-objective algorithms. The results indicate that MA is the best and the rest of algorithms are indifferent. For the performance of Central Processing Unit time (CPU time), the computational results show that PSONK is almost identical to COIN, but they are faster than NSGA-II and MA by at least 10 times for all problems.