Due to the high cost and time required to synthesize alternative fuel candidates for comprehensive testing, an Artificial Neural Network (ANN) can be used to predict fuel properties, allowing researchers to preemptively screen desirable fuel candidates. However, the accuracy of an ANN is limited by its error, measured by the root mean square error (RMSE), standard deviation, and r-squared values derived from a given input database. The present work improves upon an existing model for predicting the Cetane Number (CN) by changing the neuron activation function of the ANN from sigmoid to rectified linear unit (ReLU). This change to the ANN's architecture provides an increase in accuracy by reducing the RMSE by 21.4% (1.35 CN units), the average standard deviation across models by 28%, and increasing the r-squared value by 0.0492 across a wide range of molecular structures. Additionally, by using the ReLU activation function, input data is not required to be normalized, which reduces the likelihood of an inaccurate prediction on future fuel candidates which may have input parameters outside the range of normalization. Increasing the accuracy of the predictive ANN in this way will allow researchers to obtain more accurate fuel property predictions for promising fuel candidates.
Due to the high cost and time required to synthesize alternative fuel candidates for comprehensive testing, an Artificial Neural Network (ANN) can be used to predict fuel properties, allowing researchers to preemptively screen desirable fuel candidates. However, the accuracy of an ANN is limited by its error, measured by the root mean square error (RMSE), standard deviation, and r-squared values derived from a given input database. The present work improves upon an existing model for predicting the Cetane Number (CN) by changing the neuron activation function of the ANN from sigmoid to rectified linear unit (ReLU). This change to the ANN’s architecture provides an increase in accuracy by reducing the RMSE by 21.4% (1.35 CN units), the average standard deviation across models by 28%, and increasing the r-squared value by 0.0492 across a wide range of molecular structures. Additionally, by using the ReLU activation function, input data is not required to be normalized, which reduces the likelihood of an inaccurate prediction on future fuel candidates which may have input parameters outside the range of normalization. Increasing the accuracy of the predictive ANN in this way will allow researchers to obtain more accurate fuel property predictions for promising fuel candidates.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.