| Autor: |
Vampola, Milan, Šoštarić, Igor, Gosak, Darko |
| Přispěvatelé: |
Budin, Leo, Ribarić, Slobodan |
| Jazyk: |
angličtina |
| Rok vydání: |
2004 |
| Předmět: |
|
| Popis: |
Distillation process is very common in chemical industry as the most appropriate method for separation of liquid mixtures. Main distillation disadvantage is high energy consumption in comparison with other separation techniques. Therefore, optimal operation with tight composition control is essential for any high capacity distillation column. This task is difficult because of process nonlinearity and inability to measure concentration directly without expensive equipment. In this paper, concentration soft sensor is developed for observing on line feed composition together with distillate and bottom composition in binary distillation process. Soft sensor uses temperature measurements along the column as a cheap and reliable measuring quantities. Observer model is based on externally recurrent dynamic neural net coupled with static net for describing temperature composition relationship on several column positions. Dynamic neural net learning is performed off line using data gathered on sets of experiments on pilot distillation column. Soft sensor on line calculation method is tested in sets of experiments on pilot column with methanol - water and acetone - water binary mixtures. Results of experiments has shown applicability of developed method and its suitability to be used as a base for non-linear composition control of distillation process. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
