abstract: In this thesis, a single-level, multi-item capacitated lot sizing problem with setup carryover, setup splitting and backlogging is investigated. This problem is typically used in the tactical and operational planning stage, determining the optimal production quantities and sequencing for all the products in the planning horizon. Although the capacitated lot sizing problems have been investigated with many different features from researchers, the simultaneous consideration of setup carryover and setup splitting is relatively new. This consideration is beneficial to reduce costs and produce feasible production schedule. Setup carryover allows the production setup to be continued between two adjacent periods without incurring extra setup costs and setup times. Setup splitting permits the setup to be partially finished in one period and continued in the next period, utilizing the capacity more efficiently and remove infeasibility of production schedule. The main approaches are that first the simple plant location formulation is adopted to reformulate the original model. Furthermore, an extended formulation by redefining the idle period constraints is developed to make the formulation tighter. Then for the purpose of evaluating the solution quality from heuristic, three types of valid inequalities are added to the model. A fix-and-optimize heuristic with two-stage product decomposition and period decomposition strategies is proposed to solve the formulation. This generic heuristic solves a small portion of binary variables and all the continuous variables rapidly in each subproblem. In addition, the case with demand backlogging is also incorporated to demonstrate that making additional assumptions to the basic formulation does not require to completely altering the heuristic. The contribution of this thesis includes several aspects: the computational results show the capability, flexibility and effectiveness of the approaches. The average optimality gap is 6% for data without backlogging and 8% for data with backlogging, respectively. In addition, when backlogging is not allowed, the performance of fix-and-optimize heuristic is stable regardless of period length. This gives advantage of using such approach to plan longer production schedule. Furthermore, the performance of the proposed solution approaches is analyzed so that later research on similar topics could compare the result with different solution strategies. / Dissertation/Thesis / Masters Thesis Industrial Engineering 2015

Submitted by JACKELINE DEL CARMEN HUACCHA NEYRA null (jacky_157_93@hotmail.com) on 2017-03-24T15:38:11Z No. of bitstreams: 1 dissertação jackeline.pdf: 1416143 bytes, checksum: 3865afc18803fe4e45d315a9ee3afaf9 (MD5) / Approved for entry into archive by Luiz Galeffi (luizgaleffi@gmail.com) on 2017-03-24T18:02:08Z (GMT) No. of bitstreams: 1 huacchaneyra_jc_me_sjrp.pdf: 1416143 bytes, checksum: 3865afc18803fe4e45d315a9ee3afaf9 (MD5) / Made available in DSpace on 2017-03-24T18:02:08Z (GMT). No. of bitstreams: 1 huacchaneyra_jc_me_sjrp.pdf: 1416143 bytes, checksum: 3865afc18803fe4e45d315a9ee3afaf9 (MD5) Previous issue date: 2017-03-13 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Os problemas de dimensionamento de lotes consistem em determinar a quantidade de itens que devem ser produzidos em todos os períodos de um horizonte de planejamento. Em geral, são considerados custos de produção, preparação de máquina e de manutenção de estoque. Neste trabalho estuda-se uma extensão do problema de dimensionamento de lotes com restrição de capacidade que considera tempos de preparação, preparação carryover e crossover, em que se tem uma única máquina, único estágio, multi-itens e big-bucket (CLSP-SCC). Novas formulações para o CLSP-SCC são apresentadas e evitam a necessidade de definir novas variáveis binárias para modelar a preparação crossover. Também são propostas restrições de quebra de simetria para formulações propostas na literatura. São provadas as relações teóricas que existem entre cada uma destas formulações estudadas. Além disso, é proposta uma heurística híbrida que combina as heurísticas Relax-and-Fix e Fix-and-Optimize (RF-FO), em que a heurística Relax-and-Fix é usada para obter uma solução inicial e a heurística Fix-and-Optimize melhora essa solução. Por fim, apresentam-se os resultados computacionais e conclui-se que os resultados obtidos melhoram significativamente quando comparam-se a formulação clássica com as formulações sem preparação carryover. Compara-se também os resultados da heurística com os do pacote computacional CPLEX e, quando ambos são limitados ao mesmo tempo computacional, a heurística RF-FO obtém melhores resultados. / Lot-Sizing Problems consist of determining the quantity of items to be produced in each period of a planning horizon. In general, production, setup and inventory costs are considered. In this work an extension of the Capacitated Lot-Sizing Problem is studied, which considers setup times, Setup Carryover and Setup Crossover, single machine, single level, multi items, multi periods and big-bucket (CLSP-SCC). New formulations to the CLSP-SCC are presented and avoid the necessity of defining new extra binary variables to model the setup crossover. Furthermore, symmetry breaking constraints are proposed for formulations from the literature. The theoretical relations between the studied formulations are proved. A Relax-and-Fix and Fixand-Optimize (RF-FO) hybrid heuristic is proposed, in which the Relax-and-Fix helps to find an initial solution and the Fix-and-Optimize improves it. Computational results are presented and the obtained results improve significantly when comparing the classical formulation with the formulation without setup carryover. Finally, the results obtained by the RF-FO heuristic and the computational package CPLEX are compared and, when they both are limited to the same computational time, the RF-FO heuristic obtains better results.