(283a) Predicting the Performance of Mesophase Formation and Properties of Mesophase Pitch Based on Experimental Investigation and Machine Learning | AIChE

(283a) Predicting the Performance of Mesophase Formation and Properties of Mesophase Pitch Based on Experimental Investigation and Machine Learning

Authors 

Wang, W. - Presenter, University of Utah
Cooley, M., University of Utah
Jolley, K., University of Utah
Kirby, R. M., The University of Utah
Eddings, E., University of Utah
Producing value-added mesophase pitch (MP) as the intermediate of high-performance carbon materials (carbon fibers, carbon foams, and graphite electrodes) via thermal polymerization is a promising process to utilize low-rank carbonaceous resources and waste-like bituminous coals, coal tar pitch, asphaltene, biomass, and waste plastics [1]. In this study, an orthogonal experimental design was applied to determine the effects of composition and operational conditions on the production of MP. However, effective optimizing operational conditions for desired product production through experiments can be energy- and time-consuming and cannot readily identify the interaction between variables[2]. Therefore, to accelerate the optimization process in an efficient, sustainable, and economical way, we created a systematically-compiled dataset of thermal polymerization to produce MP from the data mining of the literature and the results of our own experiments and employed machine learning (ML) tools to bridge the inputs and outputs, predict formation performance and properties of MP, and develop ML-based optimization for producing the desired MP. Several ML algorithms were evaluated with ten-fold cross-validation, and the algorithm with the greatest accuracy was selected. Moreover, additional detailed information behind the models was extracted. Partial dependence analysis determined the modes of each variable affecting the yield, mesophase content, and softening points of targeted MP. Further optimal solutions from reverse optimization were experimentally verified to produce MP with high mesophase content and suitable softening points. This research can provide a reference for value-added utilization of low-rank hydrocarbon resources and wastes and extend the knowledge of MP production.

References

[1] Wang WJ, Preciado I, Malzahn J, Eddings E. Mild Solvolysis Liquefaction of Low-Rank Coal into a Feedstock of Value-Added Carbon Materials. 2021 AIChE Annual Meeting. Boston, MA: AIChE; 2021.

[2] Wang W, Preciado I, Eddings E. Thermochemical Co-Conversion of Waste Polyolefins with Low-Rank Aromatic-Rich Hydrocarbons into an Intermediate of High-Quality Anisotropic Pitch. 2021 AIChE Annual Meeting. AIChE; 2021.