Synthesis Of The Separation Processes For Azeotropic Mixtures

The separation of azeotropic mixtures is quite common in chemical industry. However, the optimal design of the separation processes still faces many challenges. Compared with traditional separation sequence integration problem, the synthesis of separation processes for azeotropic mixtures not only has difficulty with the combinational explosion effect when the number of components rises. What's more, the strong deviation from the ideal mixtures results in the formulation of distillation boundaries, which limit distillation operation in certain distillation region or compartment. Consequently, split feasibility and flowsheet feasibility have become the first problem.The work described in the thesis aims at developing a systematic framework for the synthesis of separation processes for azeotropic mixtures. The proposed framework, which is not based on traditional geometric analysis methods, is applicable to systems consisting of arbitrary number of components. The framework is mainly divided into two phases:System Analysis Phase:The main task of this phase is to identify the thermodynamic characteristic of the given system. Taking advantage of a sequence of equation-oriented algorithms, the structure of the composition space, including distillation regions and compartments, is explored. What's more important, the existence of unchangeable points is detected, which will serve as the targeting information of the overall flowsheet. As a result, in order to debottleneck the separation limitation, some methods for breaking these unchangeable points, i.e., decantation, extractive distillation, membrane process, has to be chosen.State-Space Superstructure based Optimization Phase:On the basis of the system analysis phase, the separation network can be represented by distribution network operator, OP-RCM and OP-AO. And then the state-space superstructure can be modeled as a mixed-integer nonlinear programming (MINLP). The MINLP problem can be solved in GAMS environment.Compared with former works, the proposed state-space superstructure in this thesis, featuring multi-stream mixing and stream splitting, is superior to previous ones because it significantly expands the feasible area. Moreover, detailed design parameters such as number of stages and reflux ratio are derived. In addition, flowsheet feasibility test rules are constructed to facilitate the analysis of the process, and are able to be used as heuristic methods to guide the design of ternary or quaternary systems.In order to overcome the difficulty in solving the MINLP model, reduced model is introduced in this work. By solving the reduced model, feasible solutions of the original model can be derived, which improves the robustness of the algorithm. Moreover, combined with random factors imbedded in the reduced model, stochastic starting point strategy is used to strengthen the capability to find the global optimal solution.Finally, three industrial cases are presented to illustrate the effectiveness of the proposed framework.