The most recent revolution in industry (Industrial Revolution 4.0) requires increased flexibility, agility and efficiency in the use of production equipment. Dynamic Cellular Manufacturing System (DCMS) is one of the best production systems to meet such requirements. In addition, the increasing importance of sustainable development forces manufacturers and managers to take account of the environmental and social issues in the design and configuration of manufacturing systems. This thesis focuses on the sustainable configuration of DCMS by proposing three mathematical models. The main challenge of this study is to (i) choose appropriate social and environmental criteria, (ii) integrate them in mathematical models, and (iii) study the impact of these criteria on DCMS. The first model is bi-objective in order to make a trade-off between some social (job opportunity, potential machine hazards, etc.) and economic (various costs related to cell formation) criteria. To get closer to real-life situations, some parameters such as demand, machine-related costs and time capacity of the machines are considered as uncertain. To solve this problem, a robust optimization method is applied to cope with this uncertainty. In the second model, all dimensions of sustainable development are taken into account in a new bi-objective mathematical model. The first objective function models economic criteria (costs) and the second one environmental aspects (production waste), while social issues (mainly Daily Noise Dosage because of computational complexity) are modeled as constraints. Due to the NP-hardness of the problem, a new innovative approach called NSGA II-MOSA is proposed. The last model has three objective functions, one for each dimension of the sustainable development : environmental, social and economic. In order to be close to real life, some parameters of the model are expressed in terms of fuzzy value. We propose a hybridized possibilistic method to deal with uncertainty and an interactive fuzzy approach is considered to solve an auxiliary crisp multi-objective model in order to find trade-off solutions. Finally, the last part of the thesis studies the possibility to apply the three proposed models to the industry thanks to an easier method. A novel optimization-simulation approach is introduced to deal with the configuration of DCMS : (i) the optimization phase operates as scenario fraction method in order to reduce the number of alternative configurations by focusing on strategic and tactical levels ; (ii) next, a simulation tool investigates the operational level by studying the performance of each alternative and the interaction between several components of the cells.