Inhalt des Dokuments
Model Development
[1]
- © d|b|t|a 2017
Objective of this research area is the systematic and efficient development of robust steady state and dynamic models, which are the basis for optimal design and efficient operation of processes. Core task is the systematic development and implementation of models to mitigate numerical issues and decrease plant-model mismatch, improve documentation, and data management.
Contact Persons: Christian Hoffmann [2], Dr.-Ing. Erik Esche [3]
[4]
Model Development for Complex,
Non-standard Process units: Standard flowsheet simulators
typically have a large number of standard process unit models that can
be used for simulation purposes. However, models for new (intensified)
and non-standard process units are not available, yet. We develop
first-principle models to overcome these issues. The models are
generically implemented in our own modeling environment MOSAICmodeling
and exported for use in process simulators, such as Aspen Plus,
CHEMCAD, etc.
Multiscale Modeling for Process Design and
Optimization: Process development takes place on several
scales - from Molecular Dynamics and CFD simulations up to Planning
and Scheduling. However, instead of treating these models separately,
we combine them to gain a better understanding of basic physical
phenomena and their impact on large-scale processes. Typically,
information gained from smaller scales are included in larger scale
models via surrogate submodels.
Validation Based on Experimental Investigation in Test Rigs and Mini-plants: While simulation studies are certainly the future of process development, experiments will always be necessary to some extent - especially when developing new processes. An important aspect of our ongoing research is the planning and construction of mini-plants to validate our models, test and optimize new process designs and operation strategies, and to estimate unknown model parameters.
[5]
- © d|b|t|a 2018
Adaptive Model Reformulation and Convergence Analysis: The solution of highly nonlinear systems can be troublesome. However, their solution can be notably simplified by decomposing the model into appropriate submodels. Moreover, knowing the iteration variables that are most crucial for the solution of the problem is helpful. We study these aspects to suggest model (re-)formulations and initialization strategies to reduce convergence issues when using Newton-type solvers for simulation or optimization.
[6]
- © d|b|t|a 2014, J. Brehmer
Grey box and Black box models: While first-principle models are certainly the best approach to understand the behavior of a process unit, it can be quite difficult to solve them. Hence, grey box models are more frequently used to achieve a more robust process model while reducing the complexity of the system. These models may stem from a rigorous first-principle model or from process/plant data.
[7]
- © d|b|t|a 2018
Model Discretization: Many applications in process engineering demand a certain discretization - this may be regarding time and/or space. A well-known approach are finite differences. Other techniques include orthogonal collocation or Galerkin methods. Choosing the appropriate scheme is however not easy. We study different discretization approaches to find the appropriate ones for specific applications.
Tool Development - MOSAICmodeling Environment
[8]
- © d|b|t|a 2018
MOSAICmodeling – A Fully Equation-oriented, Collaborative Tool for Modeling, Simulation, and Optimization in Chemical Engineering
MOSAICmodeling is a free, web-based modeling, simulation, and optimization environment, which has been under development at the Process Dynamics and Operations Group at TU Berlin (Prof. Wozny / Prof. Repke) for the past 9 years. Based on a LaTeX-style entry method for algebraic and differential equations, inequality constraints, function calls, etc., equation systems can be built and subsequently used for simulation and optimization. This typical modeling procedure is carried out on the “documentation level”, i.e. a notation has to be specified and all model elements are associated with a description and keywords.
All model elements are stored on an online database and model elements as well as whole models can be easily shared between users. Hence, collaborative work on large-scale flowsheets is implemented by design and a great facilitator for speeding up engineering projects.
For more information, please visit mosaic-modeling.de [9]
rschung/Model_development.png
ssistants/christian_hoffmann_msc/parameter/de/font1/min
hilfe/
en_post_docs/dr_ing_erik_esche/parameter/de/font1/minhi
lfe/
ssistants/christian_hoffmann_msc/parameter/de/font1/min
hilfe/
rschung/Model_Development/BBTF.png
rschung/Model_Development/ROM.PNG
rschung/Model_Development/Discretization.png
rschung/Model_Development/Mosaic.png
orschung/Model_Development/MOSAICmodeling-Overview.png