A data-driven approach to generate pseudo-reaction sequences for the thermal conversion of Athabasca bitumen

Sivaramakrishnan, Kaushik; Puliyanda, Anjana; Klerk, Arno de; Prasad, Vinay

doi:10.1039/d0re00321b

Cited by 10 publications

(10 citation statements)

References 77 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

Section: Methodsmentioning

confidence: 99%

“…However, when dealing with noisy experimental data, the ratio of derivatives of the above empirical metric was found to be more sensitive in gleaning the optimal rank, as explained in our previous works. 36,37 Bayesian Hierarchical Clustering (BHC). Clustering is a non-supervised machine learning technique that combines or clusters data points that are similar into a group, based on a similarity metric.…”

Section: ■ Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Data Fusion-Based Approach for the Investigation of Reaction Networks in Hydrous Pyrolysis of Biomass

Sattari

Srinivasan

Puliyanda

et al. 2023

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

In this work, we present and validate a methodology for generating reaction networks from spectroscopic data using data-driven methods by applying it to the hydrothermal liquefaction (HTL) of Monterrey pine biomass and its constituents, viz., cellulose and lignin. This work is presented as a step toward automated inference of chemistry of the hydrothermal liquefaction process, thus limiting the need for human expertise. Bayesian hierarchical clustering of spectra and selfmodeling multivariate spectral curve resolution are used to generate groups of chemically similar species, the reaction networks among which have been developed using Bayesian networks. Fourier transform infrared spectroscopy and proton nuclear magnetic resonance spectroscopy-based measurements are used as input data. The data-driven reaction network includes pathways representing decomposition of the biomass components, large molecule hydrolysis, and reformation of produced molecules and is consistent with the literature. Furthermore, the comparison of the networks generated for biomass and its components (levoglucosan, representing cellulose, and 2-phenoxy-ethyl benzene, representing lignin) reveals the relationship between the biomass HTL reaction network and the reaction networks of the components. The data-driven approach provides a diagnostic tool to identify the most probable reaction chemistry for complex biomass feedstocks and can be used for process understanding, design, and control.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: ■ Introductionmentioning

confidence: 99%

Data Fusion-Based Approach for the Investigation of Reaction Networks in Hydrous Pyrolysis of Biomass

Sattari

Srinivasan

Puliyanda

et al. 2023

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

show abstract

“…The discussion of the tensor decomposition and the subsequent interpretation of the Bayesian networks constructed from the pseudo-component spectra are provided together with the results for each case so as to have an easier interpretation. The reaction pathways hypothesized from the Bayesian networks have been validated against the literature pertaining to conversion chemistry in bitumen that has been investigated using quantitative metrics reflecting composition changes of model compounds, representative of the complex reactive system. − , It must be noted that the pseudo-component signatures from the tensor decomposition do not point to a single molecular structure but a class of compounds. Suitable model compounds with structures representative of the pseudo-component spectra have been used to indicate plausible conversion pathways in line with the Bayesian networks.…”

Section: Results and Discussionmentioning

confidence: 99%

Structure-Preserving Joint Non-negative Tensor Factorization to Identify Reaction Pathways Using Bayesian Networks

Puliyanda

Sivaramakrishnan

et al. 2021

J. Chem. Inf. Model.

Self Cite

View full text Add to dashboard Cite

Extracting meaningful information from spectroscopic data is key to species identification as a first step to monitoring chemical reactions in unknown complex mixtures. Spectroscopic data obtained over multiple process modes (temperature, residence time) from different sensors [Fourier transform infrared (FTIR), proton nuclear magnetic resonance ( 1 H NMR)] comprise hidden complementary information of the underlying chemical system. This work proposes an approach to jointly capture these hidden patterns in a structure-preserving and interpretable manner using coupled non-negative tensor factorization to achieve uniqueness in decomposition. Projections onto the modes of spectral channels, specific to each sensor, are interpreted as pseudo-component spectra, while projections onto the shared process modes are interpreted as the corresponding pseudo-component concentrations across temperature and residence times. Causal structure inference among these pseudo-component spectra (using Bayesian networks) is then used to identify plausible reaction pathways among the identified species representing each pseudo-component. Tensor decomposition of the FTIR data enables the development of reaction sequences based on the identified functional groups, while that of 1 H NMR by itself is lacking in mechanism development as it solely reveals the proton environments in a pseudocomponent. However, jointly parsing spectra from both the sensors is seen to capture complementary information, wherein insights into the proton environment from 1 H NMR disambiguate pseudo-components that have similar FTIR peaks. A scalable method of parallelizing tensor decomposition to handle high-dimensional modes in process data by using grid tensor factorization, while being robust to process data artifacts like outliers, noise, and missing data, has been developed.

show abstract

“…In the past 2 decades, RF, SVR, and GBR techniques have become attractive tools for various chemical engineering applications, such as quantitative structure-property relationship development, CO 2 capture, gas chromatography, olefin oligomerization, and development of groundwater potential maps . Though other chemometric methods such as self-modeling curve resolution (SMCR) − and Bayesian learning ,− have been used previously to extract unknown components from bitumen, decision trees and SVR methods have not been applied to TGA data from DAO in the past. RF trees are based on ensemble learning theory and require minimal hypertuning of the parameters as opposed to the choosing and tuning of the weights and number of layers in ANNs.…”

Section: Introductionmentioning

confidence: 99%

Prediction of Thermogravimetric Data for Asphaltenes Extracted from Deasphalted Oil Using Machine Learning Techniques

Sivaramakrishnan,

Tannous,

Chandrasekaran

2023

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

Thermogravimetric analysis (TGA) has been extensively used in the bitumen literature to investigate its thermal stability and various stages of thermal decomposition. The primary aim of these studies has been to calculate the kinetic parameters, such as activation energy and the pre-exponential factor of each thermal event. However, in our current paper, we explore the application of three machine learning (ML) techniques, namely, support vector regression (SVR), random forest (RF), and gradient booster regression (GBR), to predict the TGA data for the asphaltenes extracted from the feed and products of visbreaking of three types of materials: (i) deasphalted oil (DAO), (ii) DAO doped with 5.55 wt % indene, and (iii) DAO doped with 11.11 wt % indene. The addition of indene was shown to significantly affect the free-radical chemistry of DAO in a previous work, and the key contribution of our work in this paper was to minimize the requirement of the TGA instrument to obtain the mass loss curves by employing ML techniques on available experimental data. This will reduce the human errors involved in sample preparation and data collection as well as decrease the process time in obtaining the TGA data as compared to experimentation. We observed that the regression techniques based on decision trees, i.e., RF and GBR, showed the best performance and highest prediction accuracy of >0.99 for predicting the TGA data of the asphaltenes extracted from the feed and products obtained by reacting the feedstocks at visbreaking reaction times of 30, 45, and 60 min. A number of inputs were considered for the ML models, such as the temperature of the TGA chamber and sample, heat supplied to the sample, visbreaking time, and time spent inside the TGA chamber. The novelty of our work lies in the fact that no previous study has reproduced the TGA data for asphaltenes extracted from DAO and indene-added DAO and their visbroken products through ML approaches, and we believe that the results of this work will help in fastening the process times in the heavy oil industry by eliminating the need for offline measuring instruments.

show abstract

A data-driven approach to generate pseudo-reaction sequences for the thermal conversion of Athabasca bitumen

Abstract: We use self-modeling multivariate curve resolution to identify pseudo-components and chemical transformations in thermal conversion of Athabasca bitumen.

Cited by 10 publications

References 77 publications

Data Fusion-Based Approach for the Investigation of Reaction Networks in Hydrous Pyrolysis of Biomass

Data Fusion-Based Approach for the Investigation of Reaction Networks in Hydrous Pyrolysis of Biomass

Structure-Preserving Joint Non-negative Tensor Factorization to Identify Reaction Pathways Using Bayesian Networks

Prediction of Thermogravimetric Data for Asphaltenes Extracted from Deasphalted Oil Using Machine Learning Techniques

Contact Info

Product

Resources

About