Pith. sign in

REVIEW 3 major objections 5 minor 84 references

Fuelprop: Fuel property prediction from ATR-FTIR spectroscopic data

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Fuelprop claims that augmenting a surrogate-fuel spectral dataset with synthetic spectra, pseudo-labels, and consistency training lets a CNN predict real-fuel RON, MON, and DCN with a 23.9% error reduction and pooled R-squared of 0.984.

desk verdict The dataset is the real contribution; the 23.9% OOD gain is not interpretable because hyperparameters were tuned on the test set. read the letter →

arxiv 2506.01601 v1 pith:43SA54FJ submitted 2025-06-02 physics.chem-ph

Mohammed Almomtan , Emad Al Ibrahim , Aamir Farooq This is my paper
classification physics.chem-ph
keywords ATR-FTIRspectroscopyfuelpropertypredictionoctanenumberderivedcetanedataimputationpseudo-labelingconsistencytrainingout-of-distributiongeneralization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that a convolutional neural network can predict research octane number, motor octane number, and derived cetane number from ATR-FTIR spectra of real fuels even when trained only on pure components and surrogate blends, provided the training set is enhanced with synthetic spectra, pseudo-labels, and consistency-enforcing perturbations. The composite model, Fuelprop, reports a 23.9% lower mean absolute error than an unenhanced baseline CNN on 135 real fuels held out from training, with pooled $R^2=0.984$; individual strategies contribute reductions between 5.3% and 16.7%. A sympathetic reading is that the paper's contribution is a recipe for making small spectral datasets usable for out-of-distribution fuel-property prediction, which matters because synthetic fuel candidates are often available only in milliliter quantities while standard octane and cetane tests are slow, costly, and volume-hungry.

What carries the argument

The load-bearing machinery is a multi-head convolutional neural network (a network that scans the spectrum with learned local filters) with three convolutional layers and separate output heads for RON, MON, and DCN, plus four enhancement mechanisms: Eqn. 6, the non-linear blending equation that corrects the linear Beer-Lambert sum by adding calibration-curve-weighted terms for non-ideal components, used to impute missing spectra; pseudo-labeling, where a model trained on labeled data predicts labels for spectra without them; semi-supervised synthetic data generation, which creates random mole-fraction blends and pseudo-labels them; and consistency training, which penalizes disagreement between predictions on original and augmented spectra. The out-of-distribution protocol, reserving all 135 real fuels for testing, is what makes the claim about generalization rather than interpolation.

What would settle it

Fix the augmentation hyperparameters (s=50, m=250, c=0.001, n=0.001, h=10, v=0.05) and the synthetic-data count before any real-fuel spectrum is examined, retrain Fuelprop on the surrogate-only split, and compare out-of-distribution MAE on the 135 real fuels; if the reduction over baseline falls below the strongest individual strategy (pseudo-labeling, 16.7%), the headline 23.9% is partly a test-set artifact.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that four data-enhancement strategies—imputation of missing spectra by non-linear component blending, pseudo-labeling of samples with missing labels, semi-supervised generation of synthetic blends with pseudo-labels, and consistency enforcement via unsupervised data augmentation—each improve out-of-distribution accuracy, and their combination compounds the gains. Trained on pure components and surrogate blends and tested on 135 real fuels never used in training, Fuelprop achieves a pooled $R^2$ of 0.984 across RON, MON, and DCN and a 23.9% lower mean absolute error than the same CNN architecture trained without enhancement. The paper reads this as evidence that chemometric models can generalize from carefully chosen surrogates to genuine fuels, addressing the data-scarcity bottleneck that has limited ATR-FTIR-based fuel screening.

Load-bearing premise

The central claim assumes the augmentation parameters and synthetic-data amounts were chosen without using the 135 real-fuel test spectra, and that no real-fuel spectra or labels leaked into training through imputation or pseudo-labeling; if either fails, the reported 23.9% improvement is inflated.

Editorial extensions

If this is right

  • If Fuelprop's results hold, a fuel developer can screen candidate synthetic fuels from a single ATR-FTIR measurement in minutes, reserving CFR-engine and IQT tests for the few candidates that pass the spectral screening.
  • The released dataset of 757 spectra (104 pure components, 518 surrogate blends, 135 real fuels) gives the community a common benchmark for RON, MON, and DCN prediction under out-of-distribution evaluation.
  • The compounding of imputation, pseudo-labeling, and consistency training suggests these strategies address different failure modes, so combined gains can exceed the best single strategy.
  • The residual analysis identifies ether-heavy blends of light naphtha as the main failure region, implying that expanding the surrogate library with ether components and their DCN labels is the next concrete step to close the gap.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • This reader's extension: pre-registering augmentation hyperparameters on a surrogate-only validation split would separate genuine transfer from test-set selection and quantify the unbiased version of the 23.9% figure.
  • Going beyond the paper, the same consistency-training recipe is directly testable on near-infrared or Raman spectra of fuels, where the same small-dataset problem appears.
  • The outlier pattern suggests a screening rule for practitioners: flag any prediction whose composition lies in a chemical class absent from the training set, such as dipropyl ether, because pseudo-labels cannot teach a property the model never saw.
  • A stricter validation of the imputation step would compare synthetic Eqn. 6 spectra against measured spectra for a diverse set of oxygenated real fuels, not just two gasoline samples.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper introduces Fuelprop, a CNN-based model for predicting RON, MON, and DCN from ATR-FTIR spectra. The authors curate a dataset of 757 entries (104 pure components, 518 surrogate blends, 135 real fuels), measure spectra for surrogate blends and real fuels, and combine four data-enhancement strategies: synthetic spectra blending, pseudo-labeling, semi-supervised synthetic data generation, and consistency enforcement via unsupervised data augmentation. Models are trained on pure components and surrogate blends and evaluated out-of-distribution on the 135 real fuels. The headline result is a 23.9% reduction in mean absolute error relative to a baseline CNN, with a pooled R2 of 0.984, attributed to the combined data-enhancement pipeline. The paper also provides the dataset and reports per-strategy ablations over 20 random seeds.

Significance. If the reported gain were unbiased, this would be a valuable contribution to chemometric fuel characterization: the dataset is substantial, the out-of-distribution evaluation target (real fuels) is the right one, the ablations are run over 20 seeds, and the augmentation procedures are described in algorithmic detail. The public release of the ATR-FTIR dataset would be useful to the community. However, the quantitative headline is compromised by the manuscript's own protocol: augmentation strengths and synthetic-data counts are selected by sweeping the out-of-distribution test set, which inflates the reported improvement by an unknown amount. The significance of the work therefore depends on whether the protocol can be corrected and the gain reassessed on a properly separated validation set.

major comments (3)
  1. [Section 3.3.3, Figures 14-27] The augmentation strengths for multiplicative scaling (s), masking (m), spectral interference (c), additive noise (n), horizontal shift (h), and vertical shift (v) are selected by their out-of-distribution MAE on the real-fuel test set, as shown in Figures 14-25; Section 3.3.2 likewise selects the number of synthetic data points by out-of-distribution MAE in Figures 12-13. Because the test set is used as the model-selection criterion, the reported 23.9% reduction in MAE for Fuelprop (Figure 29) is optimistically biased and cannot be interpreted as an unbiased estimate of generalization. Please reselect all such hyperparameters on a validation split that is disjoint from the 135 real fuels, or use a nested procedure, and report the resulting test-set performance.
  2. [Section 2.4.1 and Section 2.5.2] Section 2.5.2 states that real fuels are reserved for testing, but Section 2.4.1 describes synthetic spectra blending 'for real fuels' and pseudo-labeling for samples with spectra but missing labels without stating whether any real-fuel spectra, or synthetic spectra constructed from real-fuel compositions, enter the training set. If any real-fuel-derived sample or label is used in training, the out-of-distribution interpretation fails categorically. The manuscript must explicitly state that all imputed and pseudo-labeled training samples are derived exclusively from pure components and surrogate blends, and ideally provide a training-set audit confirming that no real-fuel spectrum or real-fuel-based label appears in the training or validation data.
  3. [Section 2.4.1, Eq. (6), Figures 8-9] The non-linear blending model of Eq. (6), which is central to synthetic spectra imputation, is validated against measured spectra for only two fuels (F ACE J gasoline and its blend with 10% methanol, Figures 8-9). The weighted-error panels are also hard to interpret because Eq. (7) is not clearly defined. Since imputed spectra are used as training inputs, additional validation across more chemically diverse fuels, or a quantitative error summary over the imputed set, is needed to support the claim that imputed spectra are chemically valid training data rather than a source of systematic bias.
minor comments (5)
  1. [Eq. (7)] Equation (7) is not well defined: the denominator 'P measured' is not introduced, and the expression as printed is dimensionally inconsistent. Please define the weighted error explicitly, including any sums or normalizations.
  2. [Figure 7 and Section 3.1.1] There are inconsistent counts for real fuels: Section 3.1.1 says 135 real fuels, the Figure 7 text says '20 real fuels and 115 real fuel blends,' while the Figure 7 caption says '20 real fuels and their 117 blends.' Please reconcile these numbers.
  3. [Figures 14-27] The axis labels and tick labels in several figures (e.g., Figures 14, 16, 18, 20, 22, 24) are poorly formatted or overlapping, with values such as '5.28' appearing as axis labels. Please regenerate these figures with readable labels and clarify which quantity is plotted on each axis.
  4. [Section 2.5.3] The description of hyperparameter optimization via Optuna gives no ranges, number of trials, or the criterion used for tuning; please provide these details so that the reader can assess whether hyperparameters other than the augmentation strengths were chosen on the test set.
  5. [Data and code availability] The abstract and conclusion state that the dataset is provided, but the main text does not give a repository link or availability statement. Please add an explicit data/code availability section.

Circularity Check

1 steps flagged · score 4.0 of 10

Augmentation strengths and synthetic-data count were selected on the OOD real-fuel test set; the reported 23.9% Fuelprop gain is therefore a selected optimum rather than an independent prediction, while no derivation-level circularity appears elsewhere.

  1. fitted input called prediction [Section 3.3.2-3.3.4, Figures 12-29 (esp. Figs. 14-25 and 29)]
    "Figure 14 shows the effect of varying the range of the multiplicative scaling factor s, as defined in Algorithm 1. Moderate perturbations improved predictive accuracy, with the lowest MAE achieved at s = ±50. The best-performing setting (Figure 15) resulted in an 8.0% reduction in MAE. ... The final evaluation assesses the composite model, referred to as Fuelprop ... with Fuelprop achieving the lowest MAE showing a 23.9% reduction relative to the baseline as shown in Figure 29."

    All augmentation strengths (s, m, c, n, h, v) and the synthetic-data count were selected as minima of OOD MAE on the 135 real fuels (Figures 12-25), and the same OOD MAE is then reported as Fuelprop's 23.9% improvement. The reported OOD prediction is therefore the selected optimum of the very criterion it is meant to validate, not an independent evaluation of a fixed model. The real-fuel test set served both as hyperparameter-selection set and as benchmark, so the headline gain is optimistically biased and cannot be interpreted as an unbiased estimate of out-of-distribution generalization.

full rationale

This is an empirical benchmark, not a derivation, so the extreme circularity patterns (self-definitional equations, imported uniqueness theorems, renaming known results) do not apply. The final RON/MON/DCN predictions are compared with ASTM-measured values, so the outcome is not defined by the input spectra. Pseudo-labeling is a standard self-training loop and is self-referential by design, but it does not by itself make the test result circular. The non-linear blending equation (Eq. 6) is imported from the authors' prior work [72], a self-citation, but it is an externally published experimental model and is checked against measured spectra for two fuels (Figures 8-9), so it is not load-bearing circularity. The one demonstrated circular element is the hyperparameter selection: Section 3.3.2 selects the synthetic-data count by OOD MAE and Section 3.3.3 selects every augmentation strength by OOD MAE, and the same OOD MAE is then reported as the Fuelprop gain. This makes the 23.9% figure a selected optimum, not an independent prediction. A protocol ambiguity remains: Section 2.4.1 describes generating synthetic spectra 'for real fuels,' and the paper never states that imputed or pseudo-labeled real-fuel entries were excluded from training, although Section 2.5.2 says real fuels are reserved for testing; this is an unclarified risk, not a demonstrated reduction. Score 4: one fitted/selected quantity is presented as an OOD prediction, but the central empirical content (new ATR-FTIR dataset, measured spectra, trained CNN) retains independent value.

Assumptions & free parameters 8 free parameters · 6 assumptions · 0 invented entities

The paper introduces no new physical entities. Its quantitative claims rest on tunable parameters: six augmentation strengths and the synthetic data count, all selected by sweeping the held-out real-fuel test set, plus CNN hyperparameters tuned by Optuna that are not reported. The domain assumptions are: literature-compiled labels are accurate ground truth; the Eqn. 6 blending model and class-averaged PIONA-O spectra produce chemically faithful imputed spectra; pseudo-labels from a surrogate-trained model are informative; the augmentations are label-preserving; and UDA consistency training transfers to spectral regression. The independently measured experimental spectra and ASTM-based labels are the strong pillars; the imputed and pseudo-labeled training samples are the weak ones.

free parameters (8)
  • Multiplicative scaling range s = 50
    Selected as the best setting by sweeping OOD test MAE (Figure 15); not fixed by physics or prior work.
  • Mask size m = 250
    Selected by OOD test sweep (Figure 17).
  • Spectral interference scale c = 0.001
    Selected by OOD test sweep (Figure 19).
  • Additive noise level n = 0.001
    Selected by OOD test sweep (Figure 21).
  • Horizontal shift range h = 10
    Selected by OOD test sweep (Figure 23).
  • Vertical shift range v = 0.05
    Selected by OOD test sweep (Figure 25).
  • Number of synthetic data points = best setting not legible in text; reported 10.2% MAE reduction
    Chosen by sweeping the dataset size (Figures 12-13); the exact best value is not legible in the manuscript text.
  • CNN hyperparameters (layer widths, learning rate, weight decay) = not reported
    Tuned with Optuna (Section 2.5.3), but the selected values are not reported in the manuscript.
assumptions (6)
  • domain assumption Property labels compiled from literature and averaged across sources are accurate ground truth for RON, MON, and DCN
    Section 2.1: labels come from different laboratories and ASTM standards (D2699, D2700, D6890, D7170, D7668); averaging assumes no systematic inter-lab bias.
  • domain assumption The non-linear blending model of Eqn. 6 (reference [72]) reproduces condensed-phase absorbance of fuel mixtures
    Eqn. 6 imputes missing spectra used in training; this paper validates the approach on only two real fuels (Figures 8-9).
  • domain assumption Class-averaged PIONA-O spectra represent the spectral diversity of all real fuels in the test set
    Eqn. 4 builds real-fuel spectra from 102 class-averaged component spectra weighted by DHA or NMR composition analyses.
  • domain assumption Pseudo-labels generated by a surrogate-trained model are informative for samples with partially missing target values
    Section 2.4.1: the self-training loop assumes base-model predictions add signal rather than reinforce error.
  • domain assumption The augmentations (masking, shifts, noise, scaling, interference) are label-preserving for RON, MON, and DCN
    Section 2.4.3: consistency training minimizes disagreement between original and augmented spectra, which is only valid if augmentations leave the target properties unchanged.
  • domain assumption UDA consistency training, developed for classification, transfers to spectral regression with a small training set
    Section 2.4.3: the framework is adapted from Xie et al. [79] without evidence that the same gains occur for regression on n=757 spectra.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fuelprop: Fuel property prediction from ATR-FTIR spectroscopic data." pith.science (2026). https://pith.science/paper/43SA54FJ

@misc{pith2026250601601,
  author       = {Pith},
  title        = {Pith review of: Fuelprop: Fuel property prediction from ATR-FTIR spectroscopic data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/43SA54FJ}},
  note         = {Machine review of arXiv:2506.01601}
}
read the original abstract

Synthetic fuels are crucial for decarbonizing the transportation sector. A significant challenge lies in the rapid and efficient characterization of these fuels. Chemometric methods using ATR-FTIR data offer a potential alternative to conventional techniques. This study expands the applicability and performance of chemometric models by providing an extensive ATR-FTIR spectral dataset and exploring various data enhancement strategies. Data enhancement was achieved by semi-supervised data generation, consistency enforcement through unsupervised data augmentation, and data imputation using synthetic spectra blending and pseudo-labeling. Models were trained on surrogate fuels and rigorously tested on real fuels, representing out-of-distribution testing conditions. We believe that this work will enhance the adoption of chemometric models for fuel characterization.

Figures

Figures reproduced from arXiv: 2506.01601 by the authors.

Figure 1
Figure 1. Schematic explaining semi-supervision using unsupervised data augmentation. [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Examples of data augmentation techniques applied to spectra. [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Frequency distribution of real fuel types in the dataset. [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗
Figures from the paper (16 more)
Figure 4
Figure 4. Figure 4: Frequency distribution of target values in the training dataset. [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]
Figure 5
Figure 5. Figure 5: ATR-FTIR spectra of the pure components, categorized by chemical class. [PITH_FULL_IMAGE:figures/full_fig_p020_5.png]
Figure 6
Figure 6. Figure 6: ATR-FTIR spectra of fuel surrogate blends. The top panel shows spectra of [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]
Figure 7
Figure 7. Figure 7: ATR-FTIR spectra of 20 real fuels and their 117 blends. [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]
Figure 8
Figure 8. Figure 8: Comparison of measured and synthetically generated mid-infrared spec￾tra for FACE J gasoline. The weighted er￾ror, calculated as in Eqn. 7, is shown be￾low. The synthetic spectrum is generated by blending PIONA-O class-averaged spec￾tra. 0.0 0.2 0.4 0.6 0.8 1.0 Absorba…
Figure 10
Figure 10. Figure 10: Out-of-distribution performance of a model incorporating synthetic spectra blending compared to the baseline across 20 runs, each with different random seeds. basel i n e pseu do-l abel i n g 3 4 5 6 7 8 M e a n A b s ol u t e E r r o r ( M A E ) 25%~ 75% Ran ge wi th…
Figure 12
Figure 12. Figure 12: Effect of increasing the number of synthetic data points on model perfor￾mance. The boxplot compares the out-of￾distribution MAE across 20 runs, each with different random seeds. basel i n e 50 3 4 5 6 7 8 N umber of syn th eti c data poi n ts M e a n A b s ol u t e E…
Figure 15
Figure 15. Figure 15: Out-of-distribution performance of the best-performing multiplicative scal￾ing augmentation setting (factor range = ±50) compared to the baseline across 20 runs, each with different random seeds. 5. 28 4. 81 4. 82 4. 84 4. 9 4. 97 5.1 4 basel i n e 250 50 1 0 250 50 1…
Figure 16
Figure 16. Figure 16: Effect of varying the mask size m for masking augmentation on out￾of-distribution performance. The box￾plot compares the out-of-distribution MAE across 20 runs, each with different random seeds. basel i n e 250 3 4 5 6 7 8 Mask si ze M e a n A b s ol u t e E r r o r (…
Figure 18
Figure 18. Figure 18: Effect of varying the inter￾ference scale c for spectral interference augmentation on out-of-distribution perfor￾mance. The boxplot compares the out-of￾distribution MAE across 20 runs, each with different random seeds. basel i n e 0. 001 3 4 5 6 7 8 I n terferen ce sc…
Figure 20
Figure 20. Figure 20: Effect of varying the noise level n for additive noise augmentation on out-of-distribution performance. The box￾plot compares the out-of-distribution MAE across 20 runs, each with different random seeds. basel i n e 0. 001 3 4 5 6 7 8 N oi se l evel M e a n A b s ol u…
Figure 22
Figure 22. Figure 22: Effect of varying the hori￾zontal shift range h for horizontal shift augmentation on out-of-distribution perfor￾mance. The boxplot compares the out-of￾distribution MAE across 20 runs, each with different random seeds. basel i n e 1 0 3 4 5 6 7 8 H ori zon tal sh i ft …
Figure 24
Figure 24. Figure 24: Effect of varying the vertical shift range v for vertical shift augmentation on out-of-distribution performance. The boxplot compares the out-of-distribution MAE across 20 runs, each with different random seeds. basel i n e 0. 05 3 4 5 6 7 8 Verti cal sh i ft ran ge M…
Figure 26
Figure 26. Figure 26: Comparison of baseline perfor￾mance, models trained with individual aug￾mentation types, and a model trained with all augmentations combined across 20 runs, each with different random seeds. basel i n e al l 3 4 5 6 7 8 M e a n A b s ol u t e E r r o r ( M A E ) 25%~ …
Figure 28
Figure 28. Figure 28: presents a comparison of the baseline model and models trained with each individual enhancement strategy. The results indicate an improve￾ment in performance as each strategy is incorporated, with Fuelprop achiev￾ing the lowest MAE showing a 23.9% reduction relative t…
Figure 30
Figure 30. Figure 30: Actual vs. predicted values of RON, MON, and DCN for the Fuelprop model, [PITH_FULL_IMAGE:figures/full_fig_p031_30.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

84 extracted references · 72 canonical work pages

  1. [1]

    URL: https: //www.iea.org/reports/world-energy-outlook-2023 , licence: CC BY 4.0 (report); CC BY NC SA 4.0 (Annex A)

    IEA (2023) World Energy Outlook 2023, IEA, Paris, 2023. URL: https: //www.iea.org/reports/world-energy-outlook-2023 , licence: CC BY 4.0 (report); CC BY NC SA 4.0 (Annex A)

    2023

  2. [2]

    Commission, J

    E. Commission, J. R. Centre, M. Crippa, D. Guizzardi, E. Schaaf, F. Monforti-Ferrario, R. Quadrelli, A. Risquez Martin, S. Rossi, E. Vig- nati, M. Muntean, J. Brandao De Melo, D. Oom, F. Pagani, M. Banja, P. Taghavi-Moharamli, J. K¨ oykk¨ a, G. Grassi, A. Branco, J. San-Miguel, GHG emissions of all world countries – 2023, Publications Office of the Europe...

    work page doi:10.2760/953322 2023

  3. [3]

    URL: https://www.eia.gov/outlooks/ieo/

    International Energy Outlook 2023, EIA (Energy Information Adminis- tration), 2023. URL: https://www.eia.gov/outlooks/ieo/

    2023

  4. [4]

    Raimi, Y

    D. Raimi, Y. Zhu, R. G. Newell, B. C. Prest, Global energy outlook 2024: Peaks or plateaus, 2024

    2024

  5. [5]

    Energy Outlook 2024, BP, British Petroleum, 2024

    bp. Energy Outlook 2024, BP, British Petroleum, 2024. URL: https://www.bp.com/content/dam/bp/business-sites/en/ 30 global/corporate/pdfs/energy-economics/energy-outlook/ bp-energy-outlook-2024.pdf

    2024

  6. [6]

    N. O. Kapustin, D. A. Grushevenko, Long-term electric vehicles outlook and their potential impact on electric grid, Energy Policy 137 (2020) 111103. URL: https://www.sciencedirect.com/science/ article/pii/S0301421519306901. doi:https://doi.org/10.1016/j. enpol.2019.111103

    arXiv 2020

  7. [7]

    URL: https: //corporate.exxonmobil.com/energy-and-innovation/ outlook-for-energy

    2023 Outlook for Energy, ExxonMobil, 2023. URL: https: //corporate.exxonmobil.com/energy-and-innovation/ outlook-for-energy

    work page 2023

  8. [8]

    V. Ram, S. R. Salkuti, An overview of major synthetic fuels, Energies 16 (2023) 2834

    work page 2023

Show all 84 references
  1. [10]

    doi:10.1520/D2699-19

    Standard Test Method for Research Octane Number of Spark-Ignition Engine Fuel, Technical Report ASTM D2699 - 15a, ASTM int., 2019. doi:10.1520/D2699-19

    2019 doi

  2. [11]

    doi:10.1520/D2700-19

    Standard Test Method for Motor Octane Number of Spark-Ignition Engine Fuel, Technical Report ASTM D2700-19, ASTM int., 2019. doi:10.1520/D2700-19

    2019 doi

  3. [12]

    Standard Test Method for Determination of Ignition Delay and Derived Cetane Number (DCN) of Diesel Fuel Oils by Combustion in a Constant Volume Chamber, Technical Report ASTM D6890 - 16, ASTM int.,

  4. [13]

    doi: 10.1520/D7170-16

    Standard Test Method for Determination of Derived Cetane Number (DCN) of Diesel Fuel Oils-Fixed Range Injection Period, Constant Vol- ume Combustion Chamber Method, Technical Report ASTM D7170, ASTM int., 2016. doi: 10.1520/D7170-16. 31

    2016 doi

  5. [14]

    doi: 10.1520/D7668-17

    Standard Test Method for Determination of Derived Cetane Number (DCN) of Diesel Fuel Oils—Ignition Delay and Combustion Delay Using a Constant Volume Combustion Chamber Method, Technical Report ASTM D7668, ASTM int., 2021. doi: 10.1520/D7668-17

    2021 doi

  6. [15]

    Coury, A

    C. Coury, A. M. Dillner, Atr-ftir characterization of organic functional groups and inorganic ions in ambient aerosols at a rural site, Atmo- spheric Environment 43 (2009) 940–948

    2009

  7. [16]

    Xia, L.-m

    Q. Xia, L.-m. Yuan, X. Chen, L. Meng, G. Huang, Analysis of methanol gasoline by atr-ft-ir spectroscopy, Applied Sciences 9 (2019) 5336

    2019

  8. [17]

    A. G. Pe˜ na, F. A. Franseschi, M. C. Estrada, V. M. Ramos, R. G. Zarracino, J. C. Z. Lor ´ ıa, A. V. C. Quiroz, Fourier transform infrared- attenuated total reflectance (ftir-atr) spectroscopy and chemometric techniques for the determination of adulteration in petrodiesel/bio...

    2014

  9. [18]

    B. P. Sutliff, S. Goyal, T. B. Martin, P. A. Beaucage, D. J. Audus, S. V. Orski, Correlating near-infrared spectra to bulk properties in polyolefins, Macromolecules 57 (2024) 2329–2338

    2024

  10. [19]

    Tuschel, Practical group theory and raman spectroscopy, part ii: Application of polarization (2014)

    D. Tuschel, Practical group theory and raman spectroscopy, part ii: Application of polarization (2014)

    2014

  11. [20]

    Bolanca, S

    T. Bolanca, S. Marinovi, S. Ukic, A. Jukic, V. Rukavina, Develop- ment of artificial neural network model for diesel fuel properties pre- diction using vibrational spectroscopy, Acta Chim. Slov. 59 (2012) 249–257. PMID: 24061237, retrieved from: https://pubmed.ncbi.nlm. nih.go...

    2012

  12. [21]

    Mevik, R

    B.-H. Mevik, R. Wehrens, The pls package: Principal component and partial least squares regression in r, Journal of Statistical Software 18 (2007) 517–523

    2007

  13. [22]

    Kelly, C

    J. Kelly, C. Barlow, T. Jinguji, J. Callis, Prediction of gasoline octane numbers from near-infrared spectral features in the range 660-1215 nm, Analytical chemistry 4 (1989) 313–20. doi: 10.1021/ac00179a007. 32

    1989 doi

  14. [23]

    A. A. Kardamakis, N. Pasadakis, Autoregressive modeling of near-ir spectra and mlr to predict ron values of gasolines, Fuel 89 (2010) 158–

    2010

  15. [24]

    Swarin, C

    S. Swarin, C. Drumm, Prediction of gasoline properties with near- infrared spectroscopy and chemometrics, SAE Technical Paper (1991) 313–20. doi:10.4271/912390

    1991 doi

  16. [25]

    V. N. Korolev, A. V. Marugin, V. B. Tsaregradskii, Estimation of the petroleum product knock rating by regression analysis of near-infrared absorption spectra, SAE Technical Paper 45 (2000) 1177–1181. doi: 10. 1134/1.1318105

    2000

  17. [26]

    G. E. Fodor, K. B. Kohl, R. L. Mason, Analysis of gasolines by ft-ir spectroscopy, Anal. Chem. 68 (1996) 23–30. doi: 10.1021/ac9507294

    1996 doi

  18. [27]

    J. M. Andrade, S. Muniategui, D. Prada, Prediction of clean octane numbers of catalytic reformed naphthas using ft-m.i.r. and pls, Fuel 76 (1997) 1035–1042. doi: 10.1016/S0016-2361(97)00095-1

    1997 doi

  19. [28]

    S. R. Daly, K. E. Niemeyer, W. J. Cannella, C. L. Hagen, Predicting fuel research octane number using fourier-transform infrared absorption spectra of neat hydrocarbons, Fuel 183 (2016) 359–365. doi: 10.1016/ j.fuel.2016.06.097

    2016

  20. [29]

    J. C. L. Alves, C. B. Henriques, R. J. Poppi, Determination of diesel quality parameters using support vector regression and near infrared spectroscopy for an in-line blending optimizer system, Fuel. 97 (2012) 710–717. doi:10.1016/j.fuel.2012.03.016

    2012 doi

  21. [30]

    Al Ibrahim, A

    E. Al Ibrahim, A. Farooq, Octane prediction from infrared spectroscopic data, Energy & Fuels 34 (2020) 817–826. URL: https://doi.org/10.1021/acs. energyfuels.9b02816. doi: 10.1021/acs.energyfuels.9b02816. arXiv:https://doi.org/10.1021/acs.energyfuels.9b02816

    2020 doi

  22. [31]

    Al Ibrahim, A

    E. Al Ibrahim, A. Farooq, Prediction of the derived cetane num- ber and carbon/hydrogen ratio from infrared spectroscopic data, En- ergy & Fuels 35 (2021) 8141–8152. URL: https://doi.org/10.1021/ acs.energyfuels.0c03899. doi: 10.1021/acs.energyfuels.0c03899. arXiv:https://doi....

    2021 doi

  23. [32]

    Okada, S

    H. Okada, S. T. Sanders, First-order prediction of the relative perfor- mance of infrared (ir) absorption, raman, and combined (ir + raman) spectroscopy for estimating composition and bulk properties of fuel mix- tures, IEEE Sensors Journal 22 (2022) 16046–16054. doi:10.1109/J...

    2022

  24. [33]

    Y. Wang, Y. Ding, W. Wei, Y. Cao, D. F. Davidson, R. K. Hanson, On estimating physical and chemical properties of hydrocarbon fuels using mid-infrared ftir spectra and regularized linear models, Fuel 255 (2019) 115715. doi:10.1016/j.fuel.2019.115715

    2019

  25. [35]

    Dalmiya, M

    A. Dalmiya, M. Sheyyab, J. M. Mehta, K. Brezinsky, P. Lynch, Derived cetane number prediction of jet fuels and their functional group surrogates using liquid phase infrared absorption, Pro- ceedings of the Combustion Institute (2022). URL: https://www. sciencedirect.com/scienc...

    2022 doi

  26. [36]

    J. M. Derfer, C. E. Boord, F. C. Burk, R. E. Hess, W. G. Lovell, R. A. Randall, J. R. Sabina, Knocking characteristics of pure hydro- carbons, Technical Report 225, America Society for Testing Materials,

  27. [37]

    Yanowitz, M

    J. Yanowitz, M. A. Ratcliff, R. L. McCormick, J. D. Taylor, M. J. Mur- phy, Compendium of experimental cetane numbers, Technical Report, National Renewable Energy Lab.(NREL), Golden, CO (United States), 2017

    2017

  28. [38]

    Yanowitz, E

    J. Yanowitz, E. Christensen, R. L. McCormick, Utilization of renewable oxygenates as gasoline blending components (2011)

    2011

  29. [39]

    D. W. Naegeli, D. M. Yost, D. S. Moulton, E. C. Owens, G. K. Chui, The measurement of octane numbers for methanol and reference fuels blends, SAE transactions (1989) 712–722. 34

    1989

  30. [40]

    R. L. McCormick, G. Fioroni, L. Fouts, E. Christensen, J. Yanowitz, E. Polikarpov, K. Albrecht, D. J. Gaspar, J. Gladden, A. George, Se- lection criteria and screening of potential biomass-derived streams as fuel blendstocks for advanced spark-ignition engines, SAE Internation...

    2017

  31. [41]

    W. L. Kubic, J. R. W. Jenkins, C. M. Moore, T. A. Semelsberger, A. D. Sutton, Artificial neural network based group contribution method for estimating cetane and octane numbers of hydrocarbons and oxygenated organic compounds, Ind. Eng. Chem. Res. 56 (2017) 12236–12245. doi:10...

    2017 doi

  32. [42]

    A. G. A. Jameel, V. V. Oudenhoven, A.-H. Emwas, S. M. Sarathy, Pre- dicting octane number using nuclear magnetic resonance spectroscopy and artificial neural networks, Energy & Fuels 32 (2018) 6309–6329. doi:10.1021/acs.energyfuels.8b00556

    2018 doi

  33. [43]

    URL: https://roempp.thieme

    R ¨OMPP-Redaktion, Octan-zahl, 2002. URL: https://roempp.thieme. de/lexicon/RD-15-00161

    2002

  34. [44]

    Egloff, P

    G. Egloff, P. Van Arsdell, Octane rating relationships of aliphatic, al- icyclic, mononuclear aromatic hydrocarbons, alcohols, ethers, and ke- tones, Journal of the Institute of Petroleum (London) 27 (1941) 121– 138

    1941

  35. [45]

    Szybist, B

    J. Szybist, B. West, Update on cooptima light-duty spark-ignition re- search. 2017, 2020

    2017

  36. [46]

    A. M. Schweidtmann, J. G. Rittig, A. Konig, M. Grohe, A. Mitsos, M. Dahmen, Graph neural networks for prediction of fuel ignition qual- ity, Energy & fuels 34 (2020) 11395–11407

    2020

  37. [47]

    Z. Guo, K. H. Lim, M. Chen, B. J. R. Thio, B. L. W. Loo, Predicting cetane numbers of hydrocarbons and oxygenates from highly accessible descriptors by using artificial neural networks, Fuel 207 (2017) 344–351

    2017

  38. [48]

    Hunwartzen, Modification of CFR test engine unit to determine octane numbers of pure alcohols and gasoline-alcohol blends, Technical Report, SAE technical paper, 1982

    I. Hunwartzen, Modification of CFR test engine unit to determine octane numbers of pure alcohols and gasoline-alcohol blends, Technical Report, SAE technical paper, 1982. 35

    1982

  39. [49]

    Neumann, J

    M. Neumann, J. G. Rittig, A. B. Letaief, C. Honecker, P. Ackermann, A. Mitsos, M. Dahmen, S. Pischinger, Fuel ignition delay maps for molecularly controlled combustion, Energy & Fuels 38 (2024) 13264– 13277

    2024

  40. [50]

    Badraa, A

    J. Badraa, A. AlRamadan, M. Sarathy, Optimization of the octane response of gasoline/ethanol blends, Applied Energy 203 (2017) 778–

    2017

  41. [51]

    Solaka, M

    H. Solaka, M. Tuner, B. Johansson, W. Cannella, Gasoline surrogate fuels for partially premixed combustion, of toluene ethanol reference fuels, SAE (2013). doi: 10.4271/2013-01-2540

    2013 doi

  42. [52]

    A. G. A. Jameel, N. Naser, A.-H. Emwas, A.-H. Emwas, M. Sarathy, Predicting fuel ignition quality using 1h nmr spectroscopy and multiple linear regression, Energy & Fuels 30 (2016) 9819–9835. doi: 10.1021/ acs.energyfuels.6b01690

    2016

  43. [54]

    S. H. Won, S. Dooley, P. S. Veloo, H. Wang, M. A. Oehlschlaeger, F. L. Dryer, Y. Ju, The combustion properties of 2,6,10-trimethyl dodecane and a chemical functional group analysis, Combustion and Flame 161 (2014) 826 – 834. URL: http://www.sciencedirect.com/science/ article/p...

    2014 doi

  44. [55]

    Singh, J

    E. Singh, J. Badra, M. Mehl, M. Sarathy, Chemical kinetic insights into the octane number and octane sensitivity of gasoline surrogate mixtures, Energy & Fuels 31 (2017) 1945–1960

    2017

  45. [56]

    Waqas, N

    M. Waqas, N. Naser, M. Sarathy, J. Feijs, K. Morganti, G. Nyrenstedt, B. Johansson, Auto-ignition of iso-stoichiometric blends of gasoline- ethanol-methanol (gem) in si, hcci and ci combustion modes (2017). 36

    2017

  46. [57]

    Waqas, N

    M. Waqas, N. Naser, M. Sarathy, K. Morganti, K. Al-Qurashi, B. Jo- hansson, Blending octane number of ethanol in hcci, si and ci combus- tion modes, SAE International Journal of Fuels and Lubricants 9 (2016) 659–682

    2016

  47. [58]

    Angikath, N

    F. Angikath, N. Naser, S. M. Sarathy, Investigating the effects of c3 and c4 alcohol blending on ignition quality of gasoline fuels, Energy & Fuels 34 (2020) 8777–8787

    2020

  48. [59]

    A. G. A. Jameel, V. C. van Oudenhoven, N. Naser, A.-H. Emwas, X. Gao, S. M. Sarathy, Predicting ignition quality of oxygenated fu- els using artificial neural networks, SAE International Journal of Fuels and Lubricants 14 (2021) 57–86

    2021

  49. [60]

    G. M. Fioroni, M. J. Rahimi, C. K. Westbrook, S. W. Wagnon, W. J. Pitz, S. Kim, R. L. McCormick, Chemical kinetic basis of synergistic blending for research octane number, Fuel 307 (2022) 121865

    2022

  50. [61]

    T. M. Foong, K. J. Morganti, M. J. Brear, G. da Silva, Y. Yang, F. L. Dryer, The octane numbers of ethanol blended with gasoline and its surrogates, Fuel 115 (2014) 727–739. doi: 10.1016/j.fuel.2013.07. 105

    2014 doi

  51. [62]

    D. Kim, C. K. Westbrook, A. Violi, Two-stage ignition behavior and octane sensitivity of toluene reference fuels as gasoline surrogate, Com- bustion and Flame 210 (2019) 100–113

    2019

  52. [63]

    P. L. Perez, A. L. Boehman, Experimental investigation of the autoigni- tion behavior of surrogate gasoline fuels in a constant-volume combus- tion bomb apparatus and its relevance to hcci combustion, Energy & Fuels 26 (2012) 6106–6117. doi: 10.1021/ef300503b

    2012 doi

  53. [64]

    A. S. AlRamadan, S. M. Sarathy, M. Khurshid, J. Badra, A blending rule for octane numbers of prfs and tprfs with ethanol, Fuel 180 (2016) 175–186. doi:10.1016/j.fuel.2016.04.032

    2016 doi

  54. [65]

    Naser, S

    N. Naser, S. M. Sarathy, S. H. Chung, Ignition delay time sensitivity in ignition quality tester (iqt) and its relation to octane sensitivity, Fuel 233 (2018) 412–419. 37

    2018

  55. [66]

    Nicolle, N

    A. Nicolle, N. Naser, T. Javed, N. Rankovic, S. M. Sarathy, Autoignition characteristics of ethers blended with low cetane distillates, Energy & Fuels 33 (2019) 6775–6787

    2019

  56. [67]

    Cannella, M

    W. Cannella, M. Foster, G. Gunter, W. Leppard, F ACE gasolines and blends with ethanol: deteiled charecterization of physical and chemical properties, Technical Report A VFL-24, Coordinating Research Council, 2014

    2014

  57. [68]

    Cannella, C

    W. Cannella, C. Fairbridge, R. Gieleciak, P. Arboleda, T. Bays, H. Dettman, M. Foster, G. Gunter, D. Hager, D. King, C. Lay, S. Lewis, J. Luecke, S. Sluder, B. Zigler, M. Natarajan, Ad- vanced alternative and renewable diesel fuels: characterization of physical and chemical pr...

    2013

  58. [69]

    A. G. Abdul Jameel, N. Naser, A.-H. Emwas, S. M. Sarathy, Surrogate formulation for diesel and jet fuels using the minimalist functional group (mfg) approach, Proceedings of the Combustion Institute 37 (2019) 4663 – 4671. URL: http://www.sciencedirect.com/science/article/ pii/...

    2019 doi

  59. [70]

    C. Lee, A. Ahmed, E. F. Nasir, J. Badra, G. Kalghatgi, S. M. Sarathy, H. Curran, A. Farooq, Autoignition characteristics of oxygenated gaso- lines, Combustion and Flame 186 (2017) 114–128

    2017

  60. [71]

    Aljohani, A

    K. Aljohani, A. A. E.-S. Mohamed, H. Lu, H. J. Curran, S. M. Sarathy, A. Farooq, Impact of exhaust gas recirculation and nitric oxide on the autoignition of an oxygenated gasoline: Experiments and kinetic modelling, Combustion and Flame 259 (2024) 113174

    2024

  61. [72]

    Al Ibrahim, H

    E. Al Ibrahim, H. E. Rekik, A. Farooq, Characterization of non- ideal blending in infrared spectra of gasoline surrogates, Fuel 344 (2023) 128134. URL: https://www.sciencedirect.com/science/ article/pii/S0016236123007470. doi:https://doi.org/10.1016/j. fuel.2023.128134. 38

    2023

  62. [73]

    S. M. Sarathy, G. Kukkadapu, M. Mehl, W. Wang, T. Javed, S. Parka, M. A. Oehlschlaeger, A. Farooq, W. J. Pitz, C.-J. Sung, Ignition of alkane-rich face gasoline fuels and their surrogate mixtures, Proceedings of the Combustion Institute 35 (2015) 249–257. doi: 10.1016/j.proci....

    2015 doi

  63. [74]

    S. M. Sarathy, G. Kukkadapu, M. Mehl, T. Javed, A. Ahmed, N. Nasera, A. Tekawade, G. Kosiba, M. AlAbbad, E. Singh, S. Parka, M. AlRashidi, C. S. Hoa, W. L.Roberts, M. A.Oehlschlaegerd, S. Chih-Jenb, A. Farooq, Compositional effects on the ignition of face gasolines, Combustion...

    2016 doi

  64. [75]

    Javed, A

    T. Javed, A. Ahmed, L. Lovisotto, G. Issayeva, J. Badra, S. M. Sarathy, A. Farooq, Ignition studies of two low-octane gasolines, Combustion and Flame 185 (2017) 152–159. doi: 10.1016/j.combustflame.2017. 07.006

    2017 doi

  65. [76]

    doi:10.1520/D6730-19

    Standard Test Method for Determination of Individual Components in Spark Ignition Engine Fuels by 100-Metre Capillary (with Precolumn) HighResolution Gas Chromatography, Technical Report D6730, ASTM int., 2019. doi:10.1520/D6730-19

    2019 doi

  66. [77]

    Standard Test Method for Determination of Individual Components in Spark Ignition Engine Fuels by 50-Metre Capillary High Resolution Gas Chromatography, Technical Report D6733, ASTM int., 2020. doi: 10. 1520/D6733-01R20

    2020

  67. [78]

    Alnajjar, B

    S. Alnajjar, B. Cannella, H. Dettman, C. Fairbridge, J. Franz, T. Galant, R. Gieleciak, D. Hager, C. Lay, S. Lewis, M. Rat- cliff, S. Sluder, J. Storey, H. Yin, B. Zigler, Chemical and physical properties of the fuels for advanced combustion engine (F ACE) research diesel fuel...

    2010

  68. [79]

    Q. Xie, Z. Dai, E. Hovy, T. Luong, Q. Le, Unsupervised data augmenta- 39 tion for consistency training, Advances in neural information processing systems 33 (2020) 6256–6268

    2020

  69. [80]

    LeCun, Y

    Y. LeCun, Y. Bengio, G. Hinton, Deep learning, nature 521 (2015) 436–444

    2015

  70. [81]

    K. He, X. Zhang, S. Ren, J. Sun, Delving deep into rectifiers: Surpassing human-level performance on imagenet classification, in: Proceedings of the IEEE international conference on computer vision, 2015, pp. 1026– 1034

    2015

  71. [82]

    D. P. Kingma, Adam: A method for stochastic optimization, arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  72. [83]

    Akiba, S

    T. Akiba, S. Sano, T. Yanase, T. Ohta, M. Koyama, Optuna: A next- generation hyperparameter optimization framework, in: Proceedings of the 25th ACM SIGKDD international conference on knowledge discov- ery & data mining, 2019, pp. 2623–2631. 40

    2019

  73. [161]

    doi:10.1016/j.fuel.2009.08.029

    2009 doi

  74. [793]

    doi:10.1016/j.apenergy.2017.06.084

    2017 doi

  75. [1958]

    doi:10.1520/STP225-EB

    doi

  76. [2018]

    doi:10.1520/D6890-18

    doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.