Tesis doctoral de Rodolfo Vicente Tona Vasquez
This thesis presents a set of new methodologies that tries to exploit the information embedded in process historical data and effectively support process analysis and supervision tasks. The data rectification problem was considered in first place. The adequacy of some type of wavelet for univariate filtering of different signal patterns was studied. Then, a strategy to determine the best decomposition level was proposed and consequently, an initial step to improve current wavelet filtering approaches was found. The obtained results expand the applicability and reliability of existing filtering schemes with wavelets for on-line applications without losing of accuracy on signal estimation. Additionally, an alternative strategy was proposed to solve the problem of which wavelet to choose. This last strategy consist on a weighted combination of different wavelets functions with only one output. The data rectification problem was also studied through a data reconciliation (dr) approach. The focus was set on dr developments for dynamics and linear systems. The proposed strategy consists on first applying a trend extraction step, to identify measured process variables trends and then reconciling these trends to make them consistent with the dynamic process model studied. For the trend extraction step, filtering using wavelets was adopted. To reconcile the estimated variables trends, a extended polynomial approach was used. The comparison with existing rd approaches shows promising results in terms of accuracy and computing efficiency. Further extensions that contemplate nonlinear cases were also introduced, showing also satisfactory results. Process supervision problems were considered in second place. Primarily, principal components analysis (pca) based monitoring strategies for treatment of processes frequently affected by slowly appearing disturbances or small relative shifts were compared. This comparison included some new proposals combining wavelets filtering approac
Datos académicos de la tesis doctoral «Estrategias de analisis y exploracion de datos como soporte a la operacion y supervision deprocesos quimicos«
- Título de la tesis: Estrategias de analisis y exploracion de datos como soporte a la operacion y supervision deprocesos quimicos
- Autor: Rodolfo Vicente Tona Vasquez
- Universidad: Politécnica de catalunya
- Fecha de lectura de la tesis: 12/12/2006
Dirección y tribunal
- Director de la tesis
- Luis Puijaner Corbella
- Tribunal
- Presidente del tribunal: joseba-jokin Quevedo casin
- joan Colomer llinas (vocal)
- laureano Jiménez esteller (vocal)
- Piera i eroles miquel (vocal)