| Revista: | Brazilian journal of chemical engineering |
| Base de datos: | PERIÓDICA |
| Número de sistema: | 000308954 |
| ISSN: | 0104-6632 |
| Autors: | Dondo, R1 |
| Institucions: | 1Universidad Nacional del Litoral, Instituto de Desarrollo Tecnológico para la Industria Química, Santa Fe. Argentina |
| Any: | 2004 |
| Període: | Sep |
| Volum: | 21 |
| Número: | 3 |
| Paginació: | 367-392 |
| País: | Brasil |
| Idioma: | Inglés |
| Tipo de documento: | Artículo |
| Enfoque: | Experimental, aplicado |
| Resumen en inglés | This work describes an application of maximum likelihood identification and statistical detection techniques for determining the presence and nature of abnormal behaviors in batch fermentations. By appropriately organizing these established techniques, a novel algorithm that is able to detect and isolate faults in nonlinear and uncertain processes was developed. The technique processes residuals from a nonlinear filter based on the assumed model of fermentation. This information is combined with mass balances to conduct statistical tests that are used as the core of the detection procedure. The approach uses a sliding window to capture the present statistical properties of filtering and mass-balance residuals. In order to avoid divergence of the nonlinear monitor filter, the maximum likelihood states and parameters are periodically estimated. The maximum likelihood parameters are used to update the kinetic parameter values of the monitor filter. If the occurrence of a fault is detected, alternative faulty model structures are evaluated statistically through the use of log-likelihood function values and chi2 detection tests. Simulation obtained for xanthan gum batch fermentations are encouraging |
| Disciplines | Química |
| Paraules clau: | Fermentaciones, Biotecnología, Fermentación por lotes, Detección de fallas, Probabilidad máxima, Modelos estocásticos |
| Keyword: | Chemistry, Fermentation, Biotechnology, Batch fermentation, Fault detection, Maximum likelihood, Stochastic models |
| Text complet: | Texto completo (Ver HTML) |
