Soft-sensing Modeling Based on MLS-SVM Inversion for L-lysine Fermentation Processes
A modeling approach 63 based on multiple output variables least squares support vector machine (MLS-SVM) inversion is presented by a combination of inverse system and support vector machine theory. Firstly, a dynamic system model is developed based on material balance relation of a fed-batch ferment...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Bulgarian Academy of Sciences
2015-06-01
|
| Series: | International Journal Bioautomation |
| Subjects: | |
| Online Access: | http://www.biomed.bas.bg/bioautomation/2015/vol_19.2/files/19.2_07.pdf |
| Summary: | A modeling approach 63 based on multiple output variables least squares support vector machine (MLS-SVM) inversion is presented by a combination of inverse system and support vector machine theory. Firstly, a dynamic system model is developed based on material balance relation of a fed-batch fermentation process, with which it is analyzed whether an inverse system exists or not, and into which characteristic information of a fermentation process is introduced to set up an extended inversion model. Secondly, an initial extended inversion model is developed off-line by the use of the fitting capacity of MLS-SVM; on-line correction is made by the use of a differential evolution (DE) algorithm on the basis of deviation information. Finally, a combined pseudo-linear system is formed by means of a serial connection of a corrected extended inversion model behind the L-lysine fermentation processes; thereby crucial biochemical parameters of a fermentation process could be predicted on-line. The simulation experiment shows that this soft-sensing modeling method features very high prediction precision and can predict crucial biochemical parameters of L-lysine fermentation process very well. |
|---|---|
| ISSN: | 1314-1902 1314-2321 |