Flow Shop Production Scheduling Optimization Using Genetic Algorithms in the Garment Manufacturing Industry

Abstract

The garment industry is a labor-intensive, make-to-order sector characterized by complex flow shop production, though many small and medium-sized manufacturers still rely on experience-based rather than optimization-based scheduling. This study applies a Genetic Algorithm (GA) to real production scheduling data from a garment manufacturing flow shop comprising 30 jobs processed across five workstations (Cutting, Sewing 1, Sewing 2, Finishing, and Packing), comparing the resulting schedule with the company's existing conventional practice using makespan, waiting time, machine utilization, and tardiness as performance indicators. The baseline schedule produced a makespan of 1,996.57 minutes, average utilization of only 58.65 percent, and total tardiness of 2,036.00 minutes across seven delayed jobs. Following a parameter sensitivity analysis and validation across 30 independent GA runs, the best GA-optimized schedule achieved a makespan of 1,888.54 minutes, a 5.41 percent reduction, together with a 3.41 percent decrease in waiting time, a 5.72-percentage-point rise in utilization, and a 49.92 percent reduction in tardiness, cutting delayed jobs from seven to four. A one-sample t-test confirmed the makespan improvement was statistically significant (t = -107.51, p < 0.001) and highly reproducible, whereas the paired t-test on job-level tardiness was not significant (t = 1.11, p = 0.275) due to the skewed distribution of delays. These findings demonstrate that GA-based rescheduling, validated through repeated runs and formal statistical testing, offers a statistically robust and practically implementable alternative to conventional scheduling in garment manufacturing flow shops.