A Machine Learning Model for Predicting Composition of Catalytic Coprocessing Products from Molecular Beam Mass Spectra

Demand for the development of an automated and integratedrefiningprocess for biofuels has increased in recent years due to the lackof generalized process inspection tools. In bio-oil upgrading processes,all process variables are maintained based on the offline specificationof intermediates and products. A lack of real-time product specificationsin batch-wise monitoring can cause process failure and wasted resources.Therefore, there is a need for a fast and accurate intermediates/productspecification tool that can be used for real-time specification toreduce waste and mitigate the risk of process failure. To addressthis gap, we developed a machine learning (ML) model for predictingspeciated bio-oil composition, including paraffin, iso-paraffins, olefins, naphthene, and aromatics. The model is trainedusing the mass spectra from upgraded products collected in the vaporphase before condensation and predicts the composition of the condensedproduct. Training ML models using raw mass spectra is challengingdue to numerous overlapped peaks originating from different parentcompounds. With this in mind, we propose a protocol that (i) transformsraw mass spectra to chemistry-inspired predefined features and (ii)trains decision tree-based models using these features. Our resultsshow that the random forest model was robust against overfitting andhad the highest accuracy compared to other models. Moreover, a stochasticablation method determined the eight most significant features whilemaximizing the accuracy. Our protocol facilitates real-time compositionalanalysis of upgraded bio-oils and thus real-time process monitoring.Additionally, this protocol enables the rational design of efficientcatalysts and the determination of optimal process conditions. Machine learning models were developedto aid process monitoringin biomass upgrading by predicting product compositions from massspectra.