Pengembangan Model Penjadwalan Produksi Berbasis Hybrid MILP-Heuristic di Industri Pengolahan Kedelai

Putra, Dyan Kusuma Ristya (2026) Pengembangan Model Penjadwalan Produksi Berbasis Hybrid MILP-Heuristic di Industri Pengolahan Kedelai. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032232093-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Meningkatnya kompleksitas sistem manufaktur kontrak mendorong kebutuhan optimasi penjadwalan produksi yang lebih efektif. PT XYZ menghadapi tantangan signifikan dalam penjadwalan produksi dengan utilisasi kapasitas hanya 37-40% dari maksimum 260 ton/shift, fluktuasi permintaan harian 300-679 ton, dan waktu setup ±40 menit untuk pergantian antar grade. Kondisi ini menyebabkan inefisiensi operasional dan peningkatan biaya produksi yang substansial. Penelitian ini mengembangkan framework hibrida MILP-Palmer Heuristic yang dimodifikasi untuk mengoptimalkan penjadwalan produksi dengan mempertimbangkan fluktuasi permintaan dan alokasi bahan baku berbasis grade. Metodologi penelitian menggunakan pendekatan tiga fase terintegrasi: fase prediktif dengan peramalan Moving Average + Seasonal Index, fase strategis dengan Mixed Integer Linear Programming (MILP) untuk perencanaan jangka menengah, dan fase operasional dengan Palmer Heuristic yang dimodifikasi untuk penjadwalan harian responsif. Model MILP diimplementasikan menggunakan LINGO 21.0 dengan 1.258 variabel keputusan dan 2.211, menghasilkan solusi optimal global dengan waktu komputasi 0,32 detik pada kondisi baseline. Analisis sensitivitas menunjukkan model stabil pada variasi permintaan ±20%, namun mengalami peningkatan waktu komputasi signifikan hingga 308 detik pada skenario demand tertentu, memvalidasi kebutuhan komponen heuristik untuk responsivitas operasional. Palmer Heuristic yang dimodifikasi dengan Modified Priority Index (MPI) berhasil menangani rescheduling real-time dengan waktu solve 0,07 detik, memberikan solusi praktis untuk penyesuaian jadwal harian. Validasi terhadap data aktual 27 hari kerja menunjukkan sistem usulan mencapai pengurangan biaya operasional sebesar 21,52% atau Rp 629.630 per hari (setara Rp 204 juta per tahun, p=0,0285 one-tailed test, Cohen's d=0,47). Framework hybrid berhasil mereduksi operasi dua shift sebesar 75% (16→4 hari), menghilangkan switching events sepenuhnya (1→0 kejadian), dan meningkatkan safety stock compliance dari 70,4% menjadi 100%. Sistem usulan juga menunjukkan stabilitas operasional superior dengan reduksi variabilitas biaya sebesar 51,5% dibanding sistem eksisting.
Penelitian ini memberikan kontribusi teoritis dalam pengembangan framework optimasi hibrida untuk sistem manufaktur kontrak dengan membuktikan bahwa integrasi MILP dan heuristik mampu merealisasikan optimasi matematis menjadi peningkatan performa operasional yang tangible. Studi ini dapat menjadi referensi bagi industri agribisnis dalam mengelola kompleksitas penjadwalan produksi dengan karakteristik permintaan fluktuatif dan keterbatasan kapasitas.
============================================================
The increasing complexity of contract manufacturing systems drives the need for more effective production scheduling optimization. PT XYZ faces significant production scheduling challenges with capacity utilization of only 37-40% of maximum 260 tons/shift, daily demand fluctuations between 300-679 tons, and setup time of ±40 minutes for grade switching. These conditions cause substantial operational inefficiencies and increased production costs. This research develops a modified hybrid MILP-Palmer Heuristic framework to optimize production scheduling considering demand fluctuations and grade-based raw material allocation. The research methodology employs an integrated three-phase approach: predictive phase using Moving Average + Seasonal Index forecasting, strategic phase with Mixed Integer Linear Programming (MILP) for medium-term planning, and operational phase with modified Palmer Heuristic for responsive daily scheduling. The MILP model was implemented using LINGO 21.0 with 1,258 decision variables and 2,211 constraints, generating globally optimal solutions with computational time of 0.32 seconds under baseline conditions. Sensitivity analysis demonstrates model stability under ±20% demand variation but experiences significant computational time increases up to 308 seconds in certain demand scenarios, validating the necessity of the heuristic component for operational responsiveness. The modified Palmer Heuristic with Modified Priority Index (MPI) successfully handles real-time rescheduling with solve time of 0.07 seconds, providing practical solutions for daily schedule adjustments. Validation against 27 working days of actual data demonstrates the proposed system achieves operational cost reduction of 21.52% or Rp 629,630 per day (equivalent to Rp 204 million annually, p=0.0285 one-tailed test, Cohen's d=0.47). The hybrid framework successfully reduces two-shift operations by 75% (16→4 days), completely eliminates switching events (1→0 occurrences), and increases safety stock compliance from 70.4% to 100%. The proposed system also demonstrates superior operational stability with 51.5% reduction in cost variability compared to the existing system. This research provides theoretical contributions in developing hybrid optimization frameworks for contract manufacturing systems by proving that MILP-heuristic integration can translate mathematical optimization into tangible operational performance improvements. This study can serve as a reference for agribusiness industries in managing production scheduling complexity with fluctuating demand characteristics and capacity constraints.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Hybrid MILP–Heuristic, Penjadwalan Produksi, Mixed Integer Linear Programming, Palmer Heuristic, Manufaktur Kontrak Hybrid MILP–Heuristic, Production Scheduling, Mixed Integer Linear Programming, Palmer Heuristic, Contract Manufacturing
Subjects:	T Technology > TS Manufactures > TS157.5 Production scheduling
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Dyan Kusuma Ristya Putra
Date Deposited:	26 Jan 2026 08:39
Last Modified:	26 Jan 2026 08:39
URI:	http://repository.its.ac.id/id/eprint/130332

Actions (login required)

View Item