Identification of High-Purity Distillation Columns : Dynamic Neural Networks Model Development

• Main Authors: Ming Hsien Chung, 鍾明憲
• Other Authors: Shu-Woei Yu;I-Lung Chien
• Format: Others
• Language: zh-TW
• Published: 1995
• Online Access: http://ndltd.ncl.edu.tw/handle/17274952892096456857

碩士 === 國立中央大學 === 化學工程研究所 === 83 === Effective control of high-purity distillation columns is one of most challenging topics in the field of process control over the years. The unit is commonly employed to separate the final products and would normally consume vast amount of energy. Maintaining satisfactory separation of high-purity distillation columns is one of the most important concerns in in the chemical industry. Due to nonlinearity and loop inter- ation characteristics of high-purity distillation columns, it is difficult to describe dynamic behavior of such columns using simple linear mathematical models. A realistic dynamic simulation of a dual composition and temperature column is used in this study and the process is identified by using artificial neural network(ANN). Because ANN is capable of learning essential process nonlinearity from plant data , this ANN model can provide another means to describe the dynamic behavior of high-purity distillation column. Also in this work, different manipulated input excitation methods will be used to investigate the best input/output data as training set. The nonlinear model obtaioned via ANN will be compared to another nonlinear model using nonlinear ARX model.