Predictive model and near infrared spectroscopy in predicting the diesel fuel properties

Monitoring the diesel fuel properties play an important role in the performance of vehicle engines. Near-infrared (NIR) technology has been investigated as an alternative to monitor the diesel fuel properties. NIR spectroscopy shows an enormous potential for quantitative analysis of complex samples by coupling with artificial neural networks (ANNs). Although a single layer ANN shows promising in the establishing better relationship between a component of interest and NIR spectrum, a different algorithm for updating weight that has been proved to improve the performance of the multilayer could further reveal the potential of single linear layer ANN in NIR spectroscopic analysis. Therefore, this study investigates the performance of a single layer ANN that trained with LevenbergMarquardt (SLM) and that trained with Scaled Conjugate Gradient (SSCG) and compares the proposed methods with multilayer ANN that trained with same learning algori thms. Results were evaluated and discussed with previous studies that used the same data sets to establish the relationship between the NIR spectral data and diesel fuel properties. Finding depicts that the proposed SLM and SSCG were capable of predicting the diesel fuel properties using NIR spectrum without data reduction, and achieving better accuracy in predicting the diesel fuel properties compared with other recent methods. In addition, using a proposed genetic algorithm for data reduction to improve the predictive model of the proposed method.

Name dc:type.qualificationname

mphil

Level dc:type.qualificationlevel

masters

Grantor dc:publisher.institution

Universiti Tun Hussein Onn Malaysia

Year dc:date.issued

2018

Author dc:creator

• Gamal Al-kaf, Hasan Ali

• QA75-76.95 Calculating machines
• TP315-360 Fuel

Language dc:language

en