GENETIC ALGORITHM WITH CYCLE TIME UPPER BOUND PARAMETER DEVELOPMENT ON HUMAN-ROBOT COLLABORATION ASSEMBLY LINE

• Main Author: Budhiarti, Diniarie
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/80625
• Institution: Institut Teknologi Bandung

Due to the decreasing product life cycle, the assembly line conditions must be increasingly flexible towards changes. Collaborative robots, or cobots, are designed to interact directly with humans in a production environment. Cobot possesses the flexibility to perform various tasks by exchanging end-effectors (grips), thereby influencing setup time in determining cross-assembly design solutions. This research aims to develop a genetic algorithm to minimize cycle time as an alternative solution to collaborative human-robot cross-assembly problems, considering setup time. The genetic algorithm is developed and divided into two main processes: initial solution generation and solution improvement. The experimental data is from secondary data from various references with different data sizes and precedence variations. The initial solution generation phase involves proposing a specific procedure. This procedure determines a task sequence that does not violate precedence constraints, allocates resources with feasible random times, and places them at stations that do not infringe the predetermined upper bound cycle time. A tournament selection procedure is implemented in this research. The crossover methods utilized include one-point and partially mapped crossovers. The proposed mutations include swap and scramble mutations used for the ALBP-HRC problem. Parameter determination is carried out through the experiment (DOE) design using full factorial design to identify the precise values of parameters used in the genetic algorithm. The development of the cycle time upper bound parameter (?) was added in this study and contributed to the algorithm's efficiency by 39%. The computational results of the development of genetic algorithms in this research case using secondary data show that the developed algorithm can achieve analytically appropriate solutions with the number of tasks reaching 35 and produce better solutions on data with the number of tasks more than 35. Genetic algorithms can produce optimal solutions on secondary data with an average difference of 2.13% and faster computation time with an average difference of 64.66%. The development of the genetic algorithm can obtain better solutions with fast computation times, thus aiding in improving efficiency and effectiveness in decision-making related to assembly line balancing problems.