BIOFUELS.PTML: AI-Driven Design of Eco-friendly Gasoline or Biofuels blends
Sustainable Gasoline Blends Design by AI-Driven Addition of Eco-friendly Chemical Compounds
Authors and affiliations Harbil Bediaga a, Isabel Moreno-Benítezb, Sonia Arrasate b, José Luis Vilas-Vilela a, Lucía Orbe c, Juan Pedro Gómez-Martín d, Elías Unzueta c, and Humberto González-Díaz b,e,f,*
a Department of Physical Chemistry, University of Basque Country UPV/EHU, 48940, Leioa, Spain.
b Department of Organic and Inorganic Chemistry, University of Basque Country UPV/EHU, 48940, Leioa, Spain.
c Petronor Innovación S.L., 48550, Muskiz, Spain.
d Universidad Rey Juan Carlos (URJC), 28933, Madrid, Spain
eBasque Center for Biophysics, CSIC-UPV/EHU, 48940, Leioa, Spain.
fIKERBASQUE, Basque Foundation for Science, 48011, Bilbao, Spain
- Correspondence: humberto.gonzalezdiaz@ehu.eus
Designing sustainable fuel blends by adding eco-friendly pre-mixtures of chemical compounds (gasoline compounding) may help to reduce environmentalimpact. However, experimental testing of all possible pre-mixtures is time and resources consuming. Artificial Intelligence / Machine Learning (AI/ML) models may help to reduce experimentation cost but there are not publicly available datasets with detailed chemical composition of gasoline blends. Different IFPTML models were trained/validatedusing the Multivariate Linear Regression (MLR) compared to different Artificial Neural Networks (ANN) algorithms such Multi-Layer Perceptron (MLP), Deep Learning Network (DLN), Radial Basis Function (RBF). The best IFPTML models found were able topredict the RON and MON values of these blends with R2 in the range 0.89 - 0.99 in training and validation series. We release here the C++ code of the ANN models trainined in this work. The new model may useful eco-friendly gasoline blends design with lower environmental impact. The data used to train/validate the models was released for public research purposes in the data repository Figshare: https://doi.org/10.6084/m9.figshare.21206174.v1
Supporting information files contain the following information:
Supporting information files ANN-MLP 1 4 12 12 1 1.c and ANN-MLP 1 4 15 8 1 1.c contain the code in c of the ANN models trained and validated in this paper. The code may be used to implement the model in other software ever following the licence specifications by Statsoft inc. specified inside these files. The code of the software used to generate the models is part of the STATISTICA package an is not propietary of the authors for release. Please in case you want to obtain similar ANN models use this package or other ANN algorithms package implemented in Phyton, WEKA, or other algorithms. In any case, we would like to point out that the linear models presented in the paper are simpler and have similar to better statistical performance than the ANN models released here. If your objective is using our models to predict new fuel blends we recommend the linear models instead of the ANN models released here. This linear models can be run in Excel or in any other script using the linear equations presented in the paper. You can reproduce the models using the data released at follows in the Supporting information files.
Supporting Information file SI01.xlsx Names, general formula, and code of 863 compounds usually find in byproducts of industrial processes which can be interesting to repurpose/reuse in more eco-friendly fuel blends in order to reduce carbon fingerprint. https://github.com/glezdiazh/BIOFUELS.PTML/blob/main/Supporting%20Information%20file%20SI01.xlsx
Supporting Information file SI02.xlsx Identifiers, RON, MON, PVR(kPa), and S values (Vm,k), Code Property, input variables: function of reference f(vm,j)ref and PT operators: ∆Vm,25,j ∆Vm,29,j ∆Vm,30,j ∆Vm,32,j ∆Vm,36,j along with calculated/predicted values f(vm,j)calc, and their Residuals, Standard Predicted, Standard Residual, Std Error Predicted, Mahananobis Distance, Deleted Residual, and CookDistance, for 1404 fuel blends used to train (Set = t) or validate (Set = v) the models. https://github.com/glezdiazh/BIOFUELS.PTML/blob/main/Supporting%20Information%20file%20SI02.xlsx
Supporting Information file SI03.xlsx RON and MON values along with Names and composition of the 10 main components of 5000 fuel blends https://github.com/glezdiazh/BIOFUELS.PTML/blob/main/Supporting%20Information%20file%20SI03.xlsx
Additionally, the data is also released for free in data repository Figshare and indexed with doi number: https://doi.org/10.6084/m9.figshare.21206174.v1.
H.B.: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software,Validation,Visualization,Writing original draft, Writing review and editing. M.I.M.: Conceptualization, Formal analysis, Investigation, Methodology, Resources, Software, Supervision,Validation,Visualization,Writing original draft, Writing review and editing. S.A.: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Supervision,Validation,Visualization,Writing original draft, Writing review and editing. J.L.V.: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision,Validation,Visualization,Writing original draft, Writing review and editing. L.O.: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision,Validation,Visualization,Writing original draft, Writing review and editing. E.U.: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision,Validation,Visualization,Writing original draft, Writing review and editing. J.P.M.: Conceptualization, Formal analysis, Investigation, Methodology, Supervision,Validation,Writing review and editing. H.G.-D.: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing original draft, Writing review and editing.
The authors acknowledge financial support from Basque government SPRI ELKARTEK program grant (KK-2019/00037). The authors are very especially grateful for the scientific lobbing and networking activity developed by F. Temprano towards the genesis of this project. The authors greatly appreciate and thanks expert consulting services from F. Temprano, J. Pérez-Mercader, andJ.I. Goicolea. It included state-of-art discussions and technical recommendations made on brain-storm meetings sponsored by the present grant. The authors are grateful and thank the help of the refinery Laboratory in the experimental tests that have been carried out in this work. The authors also acknowledge partial financial support from research grants of Ministry of Economy and Competitiveness, MICIIN (PID2019-104148GB-I00) - 2020 – 2022) and Basque Government (EuskoJaurlaritza) consolidation groups grant (IT1045-16). G.D.H. personally acknowledges the support of IKERBASQUE, Basque Foundation for Science. J.P.-M. thanks Petronor S.A. for their support.
The manuscript is under consideration. As soon as the manuscript is accepted, we'll include citation information in this repository.