Development of block-oriented model for reactive distillation column: MTBE synthesis
Reactive distillation is nonlinear in nature and hence, the development of suitable nonlinear models to reactive distillation poses challenging problem to the industry. A good and robust nonlinear model is necessary to study the dynamics of reactive distillation and also to achieve better controller...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2012
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/3177/ http://umpir.ump.edu.my/id/eprint/3177/1/CD4173_LIANG_YONG_YEOW.pdf |
| Summary: | Reactive distillation is nonlinear in nature and hence, the development of suitable nonlinear models to reactive distillation poses challenging problem to the industry. A good and robust nonlinear model is necessary to study the dynamics of reactive distillation and also to achieve better controller performance using model-based control strategies. A first principle model of reactive distillation column was used as a platform in this research and the model equations are solved in MATLAB environment. The first principle model was validated using plant data. Then, the nonlinear empirical models were developed using the system identification toolbox in MATLAB. The data generated from the validated first principle model was used for the identification of nonlinear empirical block-oriented models namely, Wiener and Hammerstein models. Wiener model consists of linear dynamic block followed by nonlinear static block while Hammerstein model is the reverse connection order of Wiener model. In this research, a comparative study of different block-oriented models namely wavelet based Wiener model, wavelet based Hammerstein model and sigmoidnet based Wiener model was performed. In wavelet based Wiener and Hammerstein models, wavelet nonlinear function was used to describe the nonlinear static block and Output Error (OE) model was used to describe the linear dynamic block. Conversely, in sigmoidnet based Wiener model, sigmoidnet nonlinear function was used to describe the nonlinear static block and Output Error (OE) model was used to describe the linear dynamic block. The selection of input sequence plays an important role in nonlinear model identification. In this research, two types of input sequences namely random Gaussian and random uniform were implemented for the identification of each model and the results were compared. The parameters of the models were estimated using iterative prediction-error minimization method. Sigmoidnet based Wiener model using random Gaussian input sequence was chosen for modeling the reactive distillation column as it shown better agreement with first principle model results compared to other block-oriented models. The model analysis results proved the stability and suitability of sigmoidnet based Wiener model in capturing the dynamics of the reactive distillation column.
|
|---|