Creating ANNs to study hydrodynamics of lab-scale bubbling fluidized bed gasifier using CFD data

Fluidized beds have been studied for multiple decades as alternative reactors for energy conversion and fuel generation from a wide variety of feedstock. There are various kinds of physics that are relevant in a fluidized bed, namely hydrodynamics, heat & mass transfer, and chemical kinetics. Several experimental and computational models have been defined to correlate the interplay of these physics. However, despite several correlations being employed, there is still a need to provide simplistic models to study the various interactions between the physics and chemistry involved for industrial-scale reactors. This is especially the case for spouted bubbling fluidized beds.

The main aim of this Master thesis is to develop a neural network to study the hydrodynamics of a lab-scale spouted bubbling fluidized by using information from physical conservation equations with artificial neural networks (ANNs). The idea of this model is to enhance the speed of calculations instead of employing full-scale 3D CFD calculations, which is useful for engineering calculations and the design of control systems for industry-scale applications. The results from the neural network will then be validated using experiments (with the CFD simulations being validated with experimental results earlier).

To generate training data for the ANNs, a two-fluid CFD model will be used, which will then be validated using lab-scale experiments. The CFD model will be developed by the PhD student. Thus, the student will assist in creating an ANN from CFD data. The experimental data that will be used for validation will be provided by another MSc student.

Chair:
Complex Fluid Processing

Involved People:
ir. Ravi Ramesh 
Prof. Johan Padding
Prof. Wiebren de Jong