The modeling and analysis of the transient characteristics of fluidization phenomena by using advanced instrumentation and industrial engineering related techniques

The particle transient velocity analysis in fluidized bed combustor (FBC) riser is critical for FBC's efficient and reliable operation. An experimental cold model of a FBC was designed and fabricated. Transient particle velocities in the FBC cold model riser were measured and analyzed with the assistance of an advanced laser-based Particle Image Velocimetry (PIV) instrumentation. Factorial design/response surface, regression analysis, and neural network analysis were applied to generate empirical models of the 2-dimensional (2-D) FBC particle transient velocity with the consideration of four (4) independent variables---x coordinate, y coordinate, fluidizing air velocity and time. The Design of Experiments played a dominant role in designing and conducting the experiments for this research.; The 2K central composite design and the full factorial design were used to analyze the experimental data. Both analyses created the models with very poor data fitting accuracies. The polynomial regression approach has a data fitting accuracy of 85%. The knowledge-based regression approach has a data fitting accuracy of 88%. These two models have good validation performances at both same and different experimental setups.; The two hidden layer neural network with 32 neurons at each hidden layer has a data fitting accuracy of 95%. However, the validation performance of this model is poor. The single hidden layer neural network with 8 neurons has a data fitting accuracy of 88%. The validation performance of this model is excellent in terms of data fitting accuracy.; The knowledge-based regression model and the single hidden layer neural network model (8 hidden neurons) are the best models to predict the particle transient velocity in the FBC cold model. The FBC hot model testing proves these two models' validities with above 80% accuracies.; The knowledge-based regression and the single hidden layer neural network modeling methodologies along with the laser-based PIV instrumentation are a unique approach to the research of particle movement in FBC risers. The research work proves that the industrial engineering related technologies (design of experiment, regression, and neural network) are very effective and practical for evaluating the experimental conditions and analyzing experimental results in FBC systems.