PENGEMBANGAN METODE PENJADWALAN PRODUKSI BERBASIS ALGORITMA VARIABLE NEIGHBORHOOD SEARCH PADA PROSES HEAT TREATMENT DI PT DIRGANTARA INDONESIA

• Main Author: Sekarrani Zharfan, Fiona
• Format: Final Project
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/74500
• Institution: Institut Teknologi Bandung

PT Dirgantara Indonesia (PTDI) is an indigenous aerospace company in Asia that produces various types of aircraft, such as CN235 aircraft, NC212i aircraft, and NAS330 Puma helicopter. PTDI also produces various parts and equipment for aircraft. In PTDI, the production process consists of several work areas, one of which is sheet metal work area to fabricate, assemble, and modify components of aircraft. The sheet metal area consists of five workstations, i.e., cutting workstation, press forming workstation, hydraulic workstation, stretch forming, and heat treatment. Press forming and heat treatment workstation process 70% – 80% of the aircraft parts that produced at sheet metal work area. All production process at heat treatment workstation uses the job shop concept. At this current condition, PTDI uses priority method of components to set production scheduling order. Priority method classified aircraft components into several groups, thus unfinished parts will continue to pile up. Hence, the delay for delivering goods to customer are inevitable. Based on these problems, design of production scheduling at heat treatment workstation in PTDI will be proposed to minimize production make span. Design of production scheduling at heat treatment workstation uses variable neighborhood search (VNS) algorithm and the calculation was done using Python. The proposed design removes the priority rule when clustering aircraft components into several groups. The experiment was done by using dummies dataset that consists of 20 groups of components. The VNS algorithm was experimented with full factorial design to determine the best combination of algorithm parameters and resulted in 30 times of outer loop iterations (N), 4 times of shaking iterations (s), and 3 times of local search iterations (l). The output from algorithm computation gave better results with total make span is 33.24% shorter than result from scheduling using priority method.