Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Prediction of excess enthalpy in binary mixtures through probabilistic matrix completion
Hermanus, G. R. 2025. Prediction of excess enthalpy in binary mixtures through probabilistic matrix completion. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/62ffa814-3bef-4ef0-a940-56a51ed624eb
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Stellenbosch : Stellenbosch University
2025
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613902423982080 |
|---|---|
| access_status_str | Open Access |
| author | Hermanus, Garren Ryan |
| author2 | Cripwell, Jamie |
| author_browse | Cripwell, Jamie Hermanus, Garren Ryan |
| author_facet | Cripwell, Jamie Hermanus, Garren Ryan |
| author_sort | Hermanus, Garren Ryan |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Hermanus, G. R. 2025. Prediction of excess enthalpy in binary mixtures through
probabilistic matrix completion. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/62ffa814-3bef-4ef0-a940-56a51ed624eb |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/132418 |
| institution | Stellenbosch University (South Africa) |
| last_indexed | 2026-06-10T12:43:31.605Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2025 |
| publishDateRange | 2025 |
| publishDateSort | 2025 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/132418 Prediction of excess enthalpy in binary mixtures through probabilistic matrix completion Hermanus, Garren Ryan Cripwell, Jamie Louw, Tobi Stellenbosch University. Faculty of Engineering. Dept. of Chemical Engineering. Enthalpy Mixture distributions (Probability theory) Thermodynamics -- Mathematical models Chemical processes UCTD Hermanus, G. R. 2025. Prediction of excess enthalpy in binary mixtures through probabilistic matrix completion. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/62ffa814-3bef-4ef0-a940-56a51ed624eb Thesis (MEng)--Stellenbosch University, 2025. ENGLISH ABSTRACT: This study investigated the use of matrix completion (MC) methods for the generation of pseudo excess enthalpy data in binary mixtures across temperatures and composition. The motivation for such a study is to aid in experimental design and/or thermodynamic model development. The choice of excess enthalpy was motivated by the significant variations exhibited across mixtures and temperatures and is a notoriously difficult property to predict. The MC methods are generalizable to any composition dependent property. MC methods rely solely on the known experimental data to impute the data for unseen mixtures. Two MC models were proposed for the prediction of the excess enthalpy, the Pure MC and Hybrid MC models. The Pure MC model was aimed at performing interpolation of experimental data at discrete temperatures and compositions, with the addition of smoothness constraints. The Hybrid MC model focused on using the NRTL model and imputing the NRTL parameters of unknown mixtures which in turn informed the shape of the excess enthalpy. These two models were compared to UNIFAC (Dortmund) as a baseline. To increase the predictive accuracy and robustness of the MC models, additional regularization terms were incorporated. The most significant of which was a clustering framework, enforcing the feature matrices of similar compounds to be similar and hence enforcing the predictions of the excess enthalpy of similar mixtures to be similar. The introduction of the clustering framework yielded a significant increase in the predictive capabilities of both MC models. This indicated that the introduction of thermodynamic considerations in the form of regularization terms can lead to improved model capabilities. The Pure MC model outperformed UNIFAC on 56.1% of the testing mixtures, whilst the Hybrid MC model outperformed UNIFAC on 49.5% of the testing mixtures, speaking to the efficacy of MC approaches. The Pure MC model outperformed the Hybrid MC model on 64.1% of the testing mixtures. It was noted that in the majority of cases the MC and UNIFAC predictions were similar. Where the predictions of the Pure MC model and UNIFAC differ significantly, data is required to accurately infer the excess enthalpies, guiding future experiments. The Pure MC model maintained smoothness of the excess enthalpy predictions across both composition and temperature for known and unseen mixtures. The Hybrid MC model suffered from non-identifiability issues resulting from the use of the NRTL model. This made inference difficult with the results indicating local mode(s) were found, resulting in suboptimal predictions. A Bayesian model was developed for the Hybrid MC model. Markov Chain Monte Carlo (MCMC) sampling methods were used to draw samples from the posterior model. These samples were used to obtain a biased estimate of the uncertainty in the Hybrid MC model predictions. Large uncertainties were obtained for Acid-Aldehyde, Acid-Amine and Aldehyde-Alcohol mixtures, which aligns with some of the large differences in the predictions between UNIFAC and the Pure MC model. It was concluded that overall MC approaches can be applied for the generation of pseudo excess enthalpy data. AFRIKAANSE OPSOMMING: Geen opsomming beskikbaar. Masters 2025-06-06T09:36:50Z 2025-06-06T09:36:50Z 2025-03 Thesis https://scholar.sun.ac.za/handle/10019.1/132418 Stellenbosch University viii, 212 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Enthalpy Mixture distributions (Probability theory) Thermodynamics -- Mathematical models Chemical processes UCTD Hermanus, Garren Ryan Prediction of excess enthalpy in binary mixtures through probabilistic matrix completion |
| title | Prediction of excess enthalpy in binary mixtures through probabilistic matrix completion |
| title_full | Prediction of excess enthalpy in binary mixtures through probabilistic matrix completion |
| title_fullStr | Prediction of excess enthalpy in binary mixtures through probabilistic matrix completion |
| title_full_unstemmed | Prediction of excess enthalpy in binary mixtures through probabilistic matrix completion |
| title_short | Prediction of excess enthalpy in binary mixtures through probabilistic matrix completion |
| title_sort | prediction of excess enthalpy in binary mixtures through probabilistic matrix completion |
| topic | Enthalpy Mixture distributions (Probability theory) Thermodynamics -- Mathematical models Chemical processes UCTD |
| url | https://scholar.sun.ac.za/handle/10019.1/132418 |
| work_keys_str_mv | AT hermanusgarrenryan predictionofexcessenthalpyinbinarymixturesthroughprobabilisticmatrixcompletion |