Measurement and Testing Infrared Spectroscopy:
37
Firstly, a research group from the King Abdullah University of Science and Technology in Saudi Arabia produced a model for predicting RON and MON using infrared spectroscopy (IR) of pure components13
. The area of fuel property predictions is
extensive, as direct testing using proper equipment would be far too costly and impractical to conduct13
. An area of study in
this fi eld relates to chemometric-based fi elds areas of research, applying the partial least-squares regression (PLSR) algorithm to predict octane numbers. However, new approaches such as artifi cial neural networks (ANNs) have become increasingly more prominent as they have proven to accurately predict molecule behaviors and other qualities within the compound. To acknowledge the prevalence of ANNs, the group explored several methods of gas-phase IR spectra of hydrocarbons and ethanol to predict the molecular properties of their compound. Then, the team applied its fi ndings to an ANN to demonstrate IR spectroscopy’s viability in octane number prediction. According to the study, the IR spectra of 61 pure hydrocarbon species were employed to create 148 blends of hydrocarbons13
. Then,
the data for each species was collected and run through various algorithms. Either principle component analysis (PCA) or singular value decomposition (SVD) was used to simplify the data, and PLSR was used to derive features from the IR spectra.
Figure 3: A model of the road map for the experiment 13. Importance and Relevance
of Octane Testing: With names and defi nitions properly defi ned, we must now ask why it is so important to continue octane testing development. Fundamentally, octane numbers represent the stability of gasoline and its tendency against knocking. A high octane number indicates a more stable fuel and thus a lower chance of auto ignition during use6
. By accurately rating the octane quality
of fuels in such a way, researchers and consumers alike can gauge the effi cacy of many different fuels. As oil companies distill the number to simply imply the quality of gasoline, drivers benefi t passively from the association of higher octane numbers to higher quality fuels6
.
From a researcher’s standpoint, octane numbers also hold relevance. Measuring an accurate octane rating serves as a valid marker for retracing the creation of that fuel. With exact samples, recreating a successful octane fuel process can be much clearer given a numerical value that resulting compounds must also match. In addition to boosting the consistency of samples, future research building on potential compounds will have a more solid foundation with a traceable pathway towards producing that compound.
Finally, the simplest benefi t of all would be the overall effi ciency of the engine. While the act of reducing knocking is itself a boon to effi ciency, further research has determined a link with the
sensitivity of fuel, the difference between the RON and MON values, and the engine’s effi ciency11
. In a remarkable study
conducted in 2005 by a team from Shell Global Solutions on the concept, the group discovered a positive association between increasing the RON and subsequent sensitivity of the fuel and the engine effi ciency and fuel consumption, with sensitivity more strongly infl uencing energy effi ciency at lower RON values (92) than higher values (98)11
. Therefore, by upholding a higher
sensitivity (i.e. a lower MON for a given RON), fuels can have increased anti-knock performance. Octane number testing continues to lead the industry, and its importance cannot be understated. Without it, fuel research would be decades behind, and fuel effi ciency would not be as high as it is today.
Recent Innovations:
Like all scientifi c innovations, octane testing has experienced a myriad of different innovations to improve the process. As octane numbers are a rather signifi cant factor in fuel and can only be determined experimentally6
, many researchers have
devised methods of predicting octane numbers and other qualities by extrapolating data from previous research to make the rating process easier. This paper will cover several recent advancements in the area of octane research, making note of their signifi cance and positive implications on future octane testing.
Artifi cial Neural Networks:
Advancing a method mentioned in the previous study, a 2023 study conducted by a research group from the University of Massachusetts illustrates an approach to developing a model that predicts octane number and octane sensitivity using ANNs14
.
As previously mentioned, physical octane testing for large sets of potential fuels would be a waste of time, money, and samples. Therefore, this team employed ANNs, which are effective at predicting the properties of molecules without the need for physical samples14
. These networks are trained with quantitative
structure-property relationship (QSPR) descriptors, which inform properties of single-component fuels such as octane and have the unique potential to capture fundamental properties and interactions that would be lost with other datasets. For modeling combustion-related properties of hydrocarbons, numerous methods exist using many other approaches for modeling. Still, research has proven that ANNs have outperformed other models when concerning prediction accuracy of the fi nal product15
.
Furthermore, ANNs have previously been used to predict RON and MON values of gasoline and gasoline blends with great success16
. The study then proposes two different methods for predicting the octane sensitivity (OS) of a given compound: either by predicting RON and MON individually to compute the OS or by cutting out the middleman and directly computing the OS. Afterwards, in both cases, ANNs were trained on the relevant descriptors, and the fuel properties of 278 unique compounds were predicted for both methods.
. Given such percentages, the study concluded that the models performed more effectively when directly predicting OS rather than predicting the individual RON and MON values, as the general error value when computing OS is smaller in comparison to computations performed with two values possessing errors. In addition to their fairly accurate predictions, the models were able to identify the exact origins of reactivity within the molecule’s structure14
, as the QSPR descriptors allow Figure 4: Plots of individual ANNs showing predicted vs experimental values for ROR (RON), MOR (MON), dOS, and OS14 .
for fundamental insights into the properties responsible for molecule reactivity. Finally, the ANNs in the study successfully identifi ed hidden relationships between molecular structure and reactivity, highlighting several high-potential molecular fuels that are currently being studied as viable sources. Needless to say, the artifi cial neural network models were a complete success, and the information contained within the QSPR descriptors provides an incredibly strong foundation for the future chemical analyses of octane fuels.
WWW.PETRO-ONLINE.COM
These results from the experiment were astounding and impactful. From a direct analysis of the data, the ANNs predicted the RON, MON, and OS values effectively, with set percentage errors falling within acceptable measures, as shown in Figure 414
From the results, the study concluded that the model was a success. The group was able to extract the necessary data to predict the effi cacy of the compound from IR spectra and PLSR. Furthermore, the team also demonstrates the use of ANN to better capture octane behavior, capturing errors well within the margin of error for RON and MON values13
. Thus, the effi cacy of
IR spectroscopy as a method of octane number prediction is a rousing success, allowing for future models to be produced using the IR spectra of pure hydrocarbons and other components.
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88
