Data-Driven Models and Correlation Strategies for Thermal Cracking of Polymeric Feedstocks in Dual Fluidized Beds

Licentiate thesis, 2024

The escalating global production and consumption of plastics pose a significant environmental threat, demanding innovative waste management solutions. Among the different kinds of recycling technologies, steam cracking is a promising alternative to mechanical recycling that allows the processing of highly heterogeneous plastic waste streams to recover the carbon into monomeric species that can be used to produce polymeric materials of virgin quality.

The product species obtained from the steam cracking comprise different kinds of molecules which include syngas, aliphatics, aromatics and soot. The distribution of species is intrinsically linked to both the reactor's operational conditions and the chemical characteristics of the feedstock. From a data analysis perspective, this distribution holds valuable information that can be leveraged for instrument validation, prediction of unmeasured species, and estimation of relevant process variables. However, this aspect often receives insufficient attention in the technology’s related studies. This thesis aims to contribute to the theoretical and practical understanding of steam pyrolysis, offering validated data-driven models that can serve as tools for optimizing pyrolysis processes and for gaining insights into the relationship between feedstock characteristics and pyrolysis outcomes. The research encompasses the development and validation of two data-driven models, aimed at enhancing data quality and improving the understanding of product species distribution from steam cracking processes.

Experiments were performed with different types of plastic feedstocks in a dual fluidized bed (DFB) plant with a semi-industrial scale bubbling fluidized bed cracking reactor that was coupled with a circulating fluidized bed combustor running on biomass. The analytical setup included a Solid Phase Adsorption (SPA) method for aromatic fraction collection and a High-Temperature Reactor (HTR) for complete reforming of the hydrocarbons to syngas, which allowed estimation of the char yield. The characterization of the produced chemical species was performed with gas chromatography using thermal conductivity (TCD), flame ionization (FID) and vacuum ultraviolet (VUV).

The first part of the work encompasses the introduction of a Parametric System Model (PSM), designed for data quality assessment. This model leverages the constraints imposed by the conservation laws and the chemical nature of steam pyrolysis to ensure physically and statistically meaningful results for the product species distribution obtained from the cracking process. Special focus is placed on the conceptualization, mathematical foundation, and experimental validation of the model. In the second part of the work, the influence of feedstock polymeric composition on the product distribution is examined through the species distribution data obtained from the cracking in the DFB system of heterogeneous mixtures that originated from the rejected materials of different industrial recycling processes. The study presents a novel carbon classification framework that aids in identifying correlations between the feedstock chemical structures and the cracking products.

This study encapsulates efforts towards creating generalizable data analysis frameworks that can be used as tools for predictive analysis between the composition of polymeric mixture feedstocks and the resulting product species from steam cracking in dual fluidized beds.