This contribution presents a new set of petroleum coke dry gasification tests performed on pilot-scale gasifier. Dry gas composition and flow rate, temperature distribution, conversion, and pollutant formation taken from the experimental tests and respective calculations were used to validate the prediction capabilities of a reduced order model (ROM) developed for the same gasifier. The ROM predicted the experimental observations for conversion in the range of 48−90%. This study confirms that a systematically developed ROM (with a fixed framework) can predict the behavior of a gasifier under different operating conditions with reasonable accuracy. Moreover, this study investigates the variability in the ROM's key outputs in the presence of uncertainty in the feed and model parameters, i.e., the volatile percentage of the fuel, solid particle diameters, angle of multiphase flow jet, and recirculation ratio. These parameters affect the feedstock's properties and the mixing/laminar flows within different zones of the gasifier. Insights gained from the uncertainty quantification study revealed significant variability in the conversion, peak temperature, and steam percentage in the syngas; while the dry syngas composition does not seem to be significantly affected by the uncertainty of the parameters considered.
Usage
of reduced order models (ROMs) and reactor networks are becoming
widely accepted tools for the modeling of complex reactors, such as
entrained-flow gasifiers. The approximations made in a ROM reduce
the required computational costs compared to computational fluid dynamic
(CFD) models; however; the capabilities of the model in predicting
the outputs for a range of operating conditions in the gasification
unit face challenges. The following contribution presents a comparison
between a ROM and the corresponding CFD model of a short-residence-time
gasifier under different operating conditions and kinetic parameters.
Although the framework of the proposed ROM was fixed and developed
on the basis of CFD simulations generated at a base-case condition,
the results showed reasonable agreement between the two models in
predicting syngas composition, carbon conversion, and temperature
profile in the gasification system. Sensitivity analysis of the inputs
of the ROM (including test condition and reactor network parameters)
has also been performed. This analysis has shown that the recirculation
ratio and oxygen flow rate have a greater effect on the outputs compared
to model geometry and kinetic parameters.
A classification approach is proposed for finding ranges of process inputs that result in corresponding ranges of a process profit function using deep learning. Two deep learning tools are used to formulate models for use in classification, based on either supervised learning or unsupervised learning approaches. The supervised learning models are based on long shortterm memory networks and multilayer perceptron networks while the unsupervised learning model consists of an autoencoder neural network connected to a support vector machine classifier. An algorithm referred to as sequential layer-wise relevance propagation for pruning (SLRPFP) is proposed and applied to the aforementioned models for selecting relevant inputs and for pruning the neural networks and its inputs such that the test accuracy at every step of the proposed sequential algorithm is maintained or even improved. It is also shown that the selected inputs from the proposed algorithm (SLRPFP) provide important process insights on the productivity, that is, the profit-based objective function. The approaches are illustrated for the Tennessee Eastman Process (TEP) and for an industrial vaccine manufacturing process (industrial process). The efficacy of the proposed supervised and unsupervised deep learning approaches over linear model-based classification methods that are based on linear dynamic principal component analysis combined with multiclass support vector machines classification is shown by comparing the performances of both a TEP and a vaccine manufacturing process.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.