(110g) Managing Multi-Scale Modeling Issues in Chemical Engineering – a Computer-Aided Framework

Innovative product-process engineering involves the modeling and simulation of systems that require the consideration and integration of multiple time and/or length scales. However, the development of these multi-scale scenarios and their models can be a very challenging, time-consuming and therefore cost-intensive task requiring a lot of expertise. From this arises the necessity to systematize and structure the multi-scale modeling process and thereby increasing its efficiency. This motivates the development of a general methodology to manage the multi-scale modeling issues. The methodology needs to incorporate the analysis of a problem and the required scales for a desired degree of detail that fits the purpose of the modelling, the development and linking of the models and the application for simulation and optimization problems.

The objective for this work is to develop a computer-aided modeling framework for multi-scale model development, analysis and application which is based on the above mentioned general methodology. The framework is structured by providing general work-flows for the systematic development and analysis of a multi-scale scenario and the corresponding models as well as for the application in multi-scale simulation and optimization problems. Optimization problems are divided into model identification problems and design optimization problems. In each step of the work-flow the computer-aided framework provides the required tools and database connections to solve the problem. Furthermore, the framework offers expertise in each step to support the user with the decision making process by showing possible alter-natives and their consequences without restricting the freedom of decision making and flexibility of the modeler.

Important components of the multi-scale modeling framework and the issues that need to be managed are, among others:

? The re-use of the developed models and multi-scale scenarios is of utmost importance in order to increase the efficiency of the modeler by avoiding repetition of the modeling work. For that reason, the multi-scale modeling framework needs to allow the collection and storage of the different developed multi-scale scenarios for a system in an easily accessible and well-arranged way. At the same time model de¬composition and aggregation are important features. This means that the components, that is the models for the different scales, of one scenario need to be stored in a decomposed form to make them easily applicable in a different scenario for a different system by aggregation and combination with other models. In order to prevent errors in model re-use it is important that the modeling framework allows and encourages model documentation.

? With respect to the systematic generation of a multi-scale scenario the framework offers strategies the modeler can follow. Possible strategies in this context are bottom-up and top-down. Also, the framework handles the case where the modeler already knows how the multi-scale scenario looks like and constructs and analyzes the models for the different scales and communicates the data-flow between the scales and the update-scheme. Based on the selected strategy the framework guides the modeler through the development process of the scenario. This includes the systematic development and analysis of the models for the different scales as well as the performance of a multi-scale analysis after the development and incorporation of a new scale model. The multi-scale analysis supports the modeler in establishing and updating the data-flow between the different scales in the scenario and evaluating if an additional scale is required for the scenario. Further, the multi-scale modeling framework needs to support the selection of the applicable linking scheme. Linking scheme types depend on the types of equations on the different scales. Time scale issues arise for example if two or more models contain differential equations. The framework offers strategies to approach the problem of different time scales.

? For the application of a multi-scale scenario the framework needs to offer appropriate solvers and strategies to solve all types of ODE, AE and PDE systems and their combinations. In the case of dynamic simulations, options for analysis of the possible solutions of the steady state model and their stability are provided.

? The sensitivity analysis option is also provided for optimization problems with multi-scale models and in model identification problems where identifiability analysis and uncertainty analysis are performed.

A computer-aided multi-scale modeling framework that offers solutions to deal with the above mentioned problems and issues will be presented. The framework has been developed based on different multi-scale modeling examples with an application background in chemical and biochemical engineering. The development, analysis, solution and evaluation of different multi-scale scenarios of the well studied fluidized bed reactor will be presented in order to highlight the multi-scale modeling framework and its features and strategies.