PROGRAM | Chemical Engineering

By: Yuqing Luo Chair: Marianthi Ierapetritou

ABSTRACT

The chemical industry is experiencing the transition into a more sustainable and flexible paradigm due to global competition and environmental regulations. Apart from using renewable energy, developing new chemical technologies that convert unconventional feedstocks such as biomass and plastic waste into value-added products is also essential to enable this transition. Tremendous research efforts have stimulated the advancement of new catalysts and reaction routes that valorize individual components of biomass (e.g., cellulose, hemicellulose, and lignin) or mono-stream plastic waste (e.g., polyolefin and polyester). However, many emerging sustainable chemical production technologies fail to live up to the expectations despite promising performance on a small scale, emphasizing the need for early-stage cost and emission analysis on a system level. This thesis addresses the sustainability of chemical production by applying process system engineering tools to analyze the potentials of renewable feedstocks (i.e., biomass and plastic waste), improve conversion technologies, and optimize process or supply chain designs.

The first focus of this thesis is to perform techno-economic analysis (TEA) and life-cycle assessment (LCA) on biomass conversion technologies and guide their improvement. Three main components of common lignocellulosic biomass feedstocks are cellulose, hemicellulose, and lignin. We start our discussion on cost- and emission-oriented process improvement involving two platform chemicals readily available from cellulose and hemicellulose – furfural and 5- hydroxymethylfurfural (HMF). First, a renewable furfural-based route of polyester intermediate (dimethylbiphenyl, DMBP) synthesis was simulated based on preliminary experiments. We utilized TEA to reveal the process bottleneck and suggested shorter reaction times, new solvents, and more stable catalysts to make this process cost-competitive with the petroleum-based routes. We further illustrated the TEA/LCA-guided reaction discovery by leveraging the Bayesian

optimization framework and applied it to a case study of the HMF hydrodeoxygenation reaction. By exploring the trade-off between product yield and solvent loss, the Bayesian optimization achieved a 26% reduction in production costs and a 15% decrease in global warming potentials.

Next, two potential improvements of lignin valorization were assessed to achieve full utilization of all biomass compounds. We first demonstrated the advantages of glycerin as the solvent for reductive catalytic fractionation (RCF) over the traditional methanol solvent. Using glycerin as the solvent in the RCF process enabled a continuous reaction distillation operation under ambient pressure, leading to a 60% reduction in the production cost for softwood Kraft lignin. Building upon this work, we examined the ‘lignin-first’ biorefinery configuration that integrated RCF units with hydrolysis operations, which further processed the remaining cellulose and hemicellulose components. LCA with expanded biorefinery system boundary was used to screen different yellow poplar forest residuals for greenhouse gas reduction effects.

The third part of this dissertation aims to develop process and supply chain optimization models for better integration of biomass feedstocks, conversion technologies, and products. We first developed a superstructure model to design a centralized biorefinery with profit and emission objectives. The price and supply uncertainties are captured as scenarios in two-stage stochastic programming. We then evaluated the process flexibility under demand and reaction yield variabilities using a max-min-max multi-level optimization formulation. The computational complexity of the integrated optimization model with flexibility constraints was reduced by a data- driven neural network surrogate model.

When we extend the system boundary beyond the production facilities, supply chain logistics become a crucial consideration in biomass-based chemical production. Consequently, we applied a modular production strategy to design distributed facilities in the biomass supply chain with the flexibility to handle uncertainties in spatial-temporal variabilities of biomass supply, facility disruption, and product demand. We demonstrated the effectiveness of rolling horizon planning and generalized Benders decomposition algorithm in handling computational challenges brought by high-dimensional and integer variables. The stochastic supply chain optimization problem results were used to provide quantitative insights into LCA uncertainties, which was not achievable in the traditional Pedigree-based LCA uncertainty analysis framework.

Finally, we moved our attention to another promising feedstock with high environmental impacts – plastic waste. Currently, the majority of plastic waste is landfilled or combusted and runs the risk of mismanagement. Polyethylene terephthalate (PET) is one of the top-consuming plastic materials; however, only 20% of it is recycled. We performed process simulation, TEA and LCA to evaluate a novel PET glycolysis process with heterogeneous catalyst and demonstrated its potential for modular production. The results suggested that this microwave-assisted glycolysis process could produce the bis(2-hydroxyethyl) terephthalate (BHET) monomer with a 13% less cost and an 85% less GWP, even at a much smaller scale owing to its high reaction efficiencies and easy product separation.

Back >