(84c) Online Dynamic Parameter Estimation, Data Reconciliation and Transition Planning for a Packed Distillation Column within a Model-Centric Process Support System

Today's world is experiencing events that demand changes in the goals and methodologies of research for the development and production of goods and commodities. The current globalization and tightening of environmental and safety laws, for example, require industries to have efficient production processes at a cost sufficiently low to stay afloat in this competitive market. Therefore, supporting tools and techniques are required so that process engineers are able to fulfill these demands. Computer-aided process engineering (CAPE) is the application of a modeling approach to the study of a process as an integrated whole (Mayer and Schoenmakers, 1998). Its objective is not only to promote the development of modeling and simulation tools, but also to assist in the integration of the process with its operating systems during design, as well as improve the mechanisms for data transfer between phases of the process life cycle.

Even though the central role of process models in cutting-edge technologies for advanced manufacturing operations cannot be ignored, rigorous mechanistic process models are just one of the many components of any sophisticated software tool targeting industrial applications. In order to bring the advances in modeling, simulation and optimization environments and open-software architectures closer to process industries, a series of novel mechanisms and advanced software tools must be devised so that the definition of complex model-based problems is simplified. Additionally, synergistic interactions between complementary model-based software tools must be refined in order to fully unlock the potential of model-centric technologies in the industrial workplace.

In previous works (Aragon et. al, 2008; Rolandi y Romagnoli, 2010) we have presented a framework for model-centric framework for support of process systems, integrating and easing the definition of offline simulation, parameter estimation, data reconciliation, and optimization problems. In this work, we extend the capabilities of the framework to formulate and solve online activities related to parameter estimation, data reconciliation and transition planning.

During process operation assisted by model-based applications, online parameter estimation and data reconciliation are indispensable because they provide the means to maintain an accurate model and to adjust measurements in accordance to such model. For online activities, such as advanced process control and real-time optimization, maintaining a model that represents the process behavior accurately and reducing the magnitude of errors is essential as new trajectories are computed based on information provided by both process model and plant measurements. Additionally, transition planning (i.e. a special case of dynamic optimization) is extremely important for processes which require changes either in product specification or productivity.

A multicomponent packed distillation unit has been selected to evaluate the performance of such online activities within the proposed framework. Modeling of tray and packed distillation columns is a common practice today. The use of fundamental modeling in model-based methodologies to packed distillation processes has been made possible due to advances in numerical methods and software applications. Temperature and flow measurements collected through the distributed control system (DCS) during the normal operation of a distillation unit at plant scale may serve for many different purposes such as process control, parameter estimation, data reconciliation and optimization, among others. However, the majority of studies in the area show the use of such information for control and optimization purposes (Bezzo et. al, 2005; Espinosa & Marchetti, 2007). Few authors report data reconciliation in distillation columns (Al-Arfaj, 2006; Bhat & Saraf, 2004) and even less report online dynamic parameter estimation and data reconciliation for these processes (Bai et al., 2007).

The framework for integrated model-centric process support formulated and solved different complex online dynamic parameter estimation, dynamic data reconciliation and transition planning problems when coupled to an OPC server for a packed distillation unit. This demonstrated the excellent capabilities of the framework for more advanced real-time applications.

Keywords: model-centric technologies, dynamic parameter estimation, dynamic data reconciliation, dynamic optimization, packed distillation

Al-Arfaj, M. Shortcut data reconciliation technique: Development and industrial application, AICHE Journal, 2006, 52, 141-417.

Aragon, D.; Rolandi, P.A. & Romagnoli, J. Implementation and validation of a model-centric support system within a pilot-plant scaled packed distillation column. Proceedings of the AIChE national spring meeting. New Orleans, U.S.A. April 6-11, 2008.

Bai, S.; McLean, D. & Thibault, J. Autoassociative Neural Networks for Robust Dynamic Data Reconciliation, AIChE Journal, 2007, 52, 438-448.

Bezzo, F.; Muradore, R. & Barolo, M. Using structured and unstructured estimators for distillation units: a critical comparison, Computer Aided Chemical Engineering, 2005, 20, 1201-1206.

Bhat, S. A. & Saraf, D. N. Steady-state identification, gross error detection, and data reconciliation for industrial process units, Ind. Eng. Chem. Res., 2004, 43, 4323-433.

Espinosa, J. & Marchetti, J. Conceptual Modeling and Referential Control Applied to Batch Distillations. Ind. Eng. Chem. Res. 2007, 46, 6000-6009.

Rolandi, P.A. & Romagnoli, J.A. Integrated model-centric framework for support of manufacturing operations. Part I: The framework. Computers & Chemical Engineering, 2010, 34, 17-35.