Abstract:Fuel ignition quality, measured in the form of Derived Cetane Number (DCN), is an important part of integrating fuels, including sustainable aviation fuels, in compression ignition engines. DCN has been correlated with simulated and/or real spectroscopic measurements as well as other physical and chemical properties, but rarely have these correlations developed into a pathway to application. One application of the correlations is the use of miniaturized onboard fuel sensors that could assist, by using predicted DCN, in real-time feedforward engine control. To aid in the application of developing such DCN fuel sensors, Raman spectra coupled with chemometrics and a selection of influential spectral features were investigated. In this study, the Raman spectra were obtained from a database that included jet fuels, jet fuel mixtures, pure hydrocarbon components, and their weighted mixtures. The resulting Raman spectral database from the experimental measurements included spectra of components that span a wide range of DCNs and covered all the expected chemical functional groups present in a standard jet fuel. Chemometric models were developed to associate Raman spectra with DCN in subsets of the spectral range to aid in sensor miniaturization. The models were tested on jet fuels such as National Jet Fuel Combustion Program fuels designated A-1, A-2, and A-3 along with mixtures of jet fuels that spanned a wide range of DCN, simulating fuels that could represent real-world scenarios. An Artificial Neural Network (ANN) model trained on the fingerprint region (500 cm-1 - 1800 cm-1) of the Raman spectra was able to capture the non-linearity of the association between the Raman spectra and DCN with a test R2 score of 0.926, a test MSE of 3.61, and a test MPE of 3.41. Around 97 % of the unseen test samples were predicted within 10 % of the DCN measured with an Ignition Quality Tester. One hundred features of the fingerprint region influencing DCN predictions in the optimal ANN model were extracted using a Global Surrogate (GS) model. A reduced ANN model trained on only these one hundred features performed slightly better with a test R2 score of 0.935, test MSE of 3.19, test MPE of 3.20 and with the entire set of unseen test samples predicted within 10 % of the measured DCN. For assessing applicability of real-time and online DCN sensing, the Raman spectrometer was integrated with a flow cell capable of allowing measurements of DCN in flowing fuel samples and included the optimal ANN model of the fingerprint region and the 100-feature GS-ANN model on a Raspberry Pi computer. A number of unseen F-24/alcohol-to-jet fuel mixtures composed of unknown volumes were tested using the flow cell for DCN, and all of these samples were predicted within 10 % of the measured DCN.

Predicting Octane Number from Species Profiles: A Deep Learning Model

On the Use of Artificial Neural Networks to Model the Performance and Emissions of a Heavy-Duty Natural Gas Spark Ignition Engine

Prediction of Efficient Operating Conditions Inside a Heavy-Duty Natural Gas Spark Ignition Engine Using Artificial Neural Networks

Direct Prediction of Steam Cracking Products from Naphtha Bulk Properties: Application of the Two Sub-Networks ANN

A Least Squares SVM Algorithm for NIR Gasoline Octane Number Prediction

An Artificial Neural Network Model to Predict Efficiency and Emissions of a Gasoline Engine

Application of Neural Network in the Study of Combustion Rate of Natural Gas/diesel Dual Fuel Engine.

Prediction of emission characteristics of diesel/n-hexanol/graphene oxide blended fuels based on fast outlier detection-sparrow search algorithm-bidirectional recurrent neural network

A multi-scale sampling method for accurate and robust deep neural network to predict combustion chemical kinetics

Predicting fuel research octane number using Fourier-transform infrared absorption spectra of neat hydrocarbons

Prediction of Ignition Delay Times of Jet A-1/hydrogen Fuel Mixture Using Machine Learning

Study on Hydrocarbon Fuel Ignition Characterization Based on Optimized BP Neural Network

Experimental study on the spray characteristics of octanol diesel and prediction of spray tip penetration by ANN model

Artificial neural network models for forecasting the combustion and emission characteristics of ethanol/gasoline DFSI engines with combined injection strategy

Synergistic effect analysis on sooting tendency based on soot-specialized artificial neural network algorithm with experimental and numerical validation

Augmented flame image soft sensor for combustion oxygen content prediction

A comprehensive study on the accuracy and generalization of deep learning-generated chemical ODE integrators

Artificial neural network based chemical mechanisms for computationally efficient modeling of kerosene combustion

A Raman spectroscopy based chemometric approach to predict the derived cetane number of hydrocarbon jet fuels and their mixtures

Prediction of auto-ignition temperatures of hydrocarbons by neural network based on atom-type electrotopological-state indices.

Application of Elman and Cascade neural network (ENN and CNN) in comparison with adaptive neuro fuzzy inference system (ANFIS) to predict key fuel properties of ABE-diesel blends