MODEL AND PERMUTATION-BASED GENETIC ALGORITHM DEVELOPMENT OF MTSSDRC SCHEDULING IN UNRELATED PARALLEL MACHINE TO MINIMIZE MAKESPAN

• Main Author: Akbar Rugova Krisna P., M.
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/86528
• Institution: Institut Teknologi Bandung

The manufacturing industry uses semi-automatic production tools to minimize operational costs. The machine makes the operator's work lighter because he only needs to be involved in setup and unloading activities. This makes the operator more flexible, can be moved from machine to machine to supervise machining activities simultaneously. As a result, the manufacturing industry is able to assign fewer operators than the number of existing machines. Multi-task simultaneous supervision dual-resources constrained (MTSSDRC) is a scheduling term that describes the condition of the production floor. In this study, MTSSDRC is implemented in an unrelated parallel machine environment commonly referred to as the Unrelated Parallel Machine Scheduling Problem (UPMSP). This means that the machining time taken by a machine to process a particular job varies without any particular correlation between machines and jobs. Scheduling MTSSDRC on UPMSP is a study that has not been done before. The MTSSDRC scheduling problem on UPMSP can be modeled with mixed-integer linear programming (MILP) to minimize makespan. Therefore, the MILP solution model can be tested with hypothetical data input. The MTSSDRC problem in UPMSP is included in the NP-hard problem which is difficult to solve analytically due to its complexity. Therefore, the search for a solution is carried out using the metaheuristic method. Testing is carried out by grouping the size of the sub-case (jobs × machines × operators) into small, medium, and large cases. Testing is carried out using Gurobi software as a solver. Gurobi is able to obtain optimal solutions up to small cases (6 jobs, 4 machines, and 3 operators). However, the same cannot be done in medium cases (does not produce optimal solutions) and large cases (cannot produce solutions because they are 'out of memory'). Verification and validation are carried out by looking at the resulting gantt chart and comparing the test results of the reference model and the proposed model. The MILP model is then adjusted using the Permutation-based Genetic Algorithm (PGA). This PGA can produce the same optimal solution as Gurobi with consistent time the larger the size of the sub-case input index (jobs × machines × operators). PGA can produce better solutions in some sub-cases on the case size when the index input size is getting bigger with more effective computation time. For large cases, PGA can produce solutions while Gurobi cannot. Verification and validation on PGA can be done by visualizing the solution with a chart and comparing the resulting solution with the Gurobi solution. This research also brings MTSSDRC scheduling closer to the real situation in many manufacturing industries that have the same machine environment and constraints.