Adaptive large neighborhood search for a production planning problem arising in pig farming

• Published in: Journal of open innovation Vol. 5; no. 2; pp. 1 - 21
• Main Authors: Praseeratasang, Nat, Pitakaso, Rapeepan, Sethanan, Kanchana, Kaewman, Sasitorn, Golinska-Dawson, Paulina
• Format: Journal Article
• Language: English
• Published: Basel MDPI 01.06.2019; Elsevier Ltd; Elsevier Limited; Elsevier
• Subjects: adaptive large neighborhood search; Adaptive search techniques; Agriculture; Algorithms; Ant colony optimization; Assignment problem; assignment problems; Computing time; Datasets; Farms; Farmworkers; Heuristic; Heuristic methods; Hogs; Job rotation; Literature reviews; Mathematical models; Methods; Operations research; Operators (mathematics); Optimization; Production lines; Production planning; Profitability; Profits; Repair; scheduling problems; Skills; Sugarcane; Teaching assistants; Upper bounds; Workers; Thailand; adaptive large neighborhood search; assignment problems; ant colony optimization; scheduling problems
• ISSN: 2199-8531; 2199-8531
• DOI: 10.3390/joitmc5020026

This article aims to resolve a particular production planning and workforce assignment problem. Many production lines may have different production capacities while producing the same product. Each production line is composed of three production stages, and each stage requires different periods of times and numbers of workers. Moreover, the workers will have different skill levels which can affect the number of workers required for production line. The number of workers required in each farm also depends on the amount of pigs that it is producing. Production planning must fulfill all the demands and can only make use of the workers available. A production plan aims to generate maximal profit for the company. A mathematical model has been developed to solve the proposed problem, when the size of problem increases, the model is unable to resolve large issues within a reasonable timeframe. A metaheuristic method called adaptive large-scale neighborhood search (ALNS) has been developed to solve the case study. Eight destroy and four repair operators (including ant colony optimization based destroy and repair methods) have been presented. Moreover, three formulas which are used to make decisions for acceptance of the newly generated solution have been proposed. The present study tested 16 data sets, including the case study. From the computational results of the small size of test instances, ALNS should be able to find optimal solutions for all the random data sets in much less computational time compared to commercial optimization software. For medium and larger test instance sizes, the findings of the heuristics were 0.48% to 0.92% away from the upper bound and generated within 480-620 h, in comparison to the 1 h required for the proposed method. The Ant Colony Optimization-based destroy and repair method found solutions that were 0.98 to 1.03% better than the original ALNS.