Large-scale capacitated part-routing in the presence of process and routing flexibilities and setup costs

• Published in: The Journal of the Operational Research Society Vol. 57; no. 9; pp. 1100 - 1112
• Main Authors: Diaby, M, Nsakanda, A L
• Format: Journal Article
• Language: English
• Published: Basingstoke Taylor & Francis 01.09.2006; Palgrave Macmillan Press; Palgrave; Taylor & Francis Ltd
• Subjects: Applied sciences; Capital costs; cellular manufacturing systems; dynamic programming; Exact sciences and technology; Explicit costs; Heuristic; Integer programming; Lagrange multiplier; Lagrangean Relaxation; large-scale optimization; Linear programming; Machinery; Manufacturing; Materials handling; Mathematical programming; Mathematical vectors; Minimization of cost; Operational research and scientific management; Operational research. Management science; Operations research; Outsourcing; Problem solving; Production costs; Production management; Routing; Studies; Theoretical Papers; Unit costs; Production system; Process planning; Materials handling; Subcontracting; Flexibility; cellular manufacturing systems; Optimization; large-scale optimization; Heuristic method; Cellular manufacturing; Production cost; Outsourcing; Dynamic programming; Lagrangean Relaxation
• ISSN: 0160-5682; 1476-9360
• DOI: 10.1057/palgrave.jors.2602072

We develop a Lagrangean relaxation-based heuristic procedure to generate a near-optimal solution to large-scale capacitated part-routing problems through a cellular manufacturing system with both routing flexibilities and setup times. Several alternate process plans exist for each product. Any given operation can be performed on alternate machines at different costs. The part demands can be satisfied from internal production or through outsourcing. The objective is to minimize the total material handling, production, outsourcing, and setup costs, subject to satisfying all the part demands and not exceeding any of the machine capacity limits. Our computational experiments show that large problems involving several thousand products and decision variables can be solved in a reasonable amount of computer time to within 1% of their optimal solutions. The proposed procedure is general enough to be applied directly or with slight modifications to real-life, industrial-sized problems.