Detailed modelling and optmization of crystallization process / Modelagem detalhada e otimização de processos de cristalização

AUTOR(ES)
DATA DE PUBLICAÇÃO

2006

RESUMO

This work is focused on crystallization, a process widely used in industry, especially for the production of high added-value particles in pharmaceutical and fine chemistry industries. Although it is a process of established utilization, its mechanisms, modeling and the real control of its operation still require research and study. This thesis presents considerations and developments on the detailed deterministic modeling area and the process optimization with both deterministic and stochastic methods. The modeling is discussed in detail and the literature developed numerical methods for the population balance solution, which is part of the modeling, are presented focusing on crystallization processes and on the main advantages and drawbacks. Preliminary studies on batch cooling crystallization processes improvement drive to the need of cooling operating policy optimization. Since the Sequential Quadratic Programming deterministic method of optimization is inefficient for the optimization problem, the use of Genetic Algorithm (GA), a stochastic optimization method well established in literature, is evaluated in the global optimum search for this process, in a pioneering literature study of GA application in crystallization processes. Since the GA requires that many runs, with different values for its parameters, are executed, in order to increase the probability of global optimum (or its neighborhood) achievement, an original, general and relatively simple procedure for the detection of the parameters set with significant influence on the optimization response is developed and proposed. The proposed methodology is applied to general case studies, with different complexities and is very useful in the preliminary studies via GA. The procedure is, then, applied to the cooling profile optimization problem in a batch cooling optimization process. The results of the study presented in this thesis indicate that the deterministic optimization methods do not deal well with high dimensionality problems, leading to achievement of local optima. The evolutionary methods are able to detect the region of the global optimum but, on the other hand, are not fast codes. The developed procedure for the significant GA parameters detection is a relevant contribution of the thesis and can be applied to any optimization problem (of any complexity and of any dimensionality)

ASSUNTO(S)

crystallization mathematical optimization modelagem otimização matematica cristalização genetic algorithms modelling algoritmos geneticos

ACESSO AO ARTIGO

Documentos Relacionados