Prediction of Cetane Number of a Biodiesel Based on Physical Properties and a Study of Their Influence on Cetane Number Department of Mechanical Engineering, Sri Venkateswara College of Engineering

Format:

Conference/Event

Author/Creator:

Suryanarayanan, Saikishan, author.

Conference Name:

2007 Fuels and Emissions Conference (2007-01-23 : Cape Town, South Africa)

Language:

English

Physical Description:

1 online resource

Place of Publication:

Warrendale, PA SAE International 2007

Summary:

Cetane number is one of the indispensable parameters in the study and selection of fuels for CI engines. Hence it is an important criterion for selection of bio-diesels, which exhibit a wide variety of characteristics based upon their source, method of preparation et cetera Since the conventional techniques for evaluating cetane number are tedious, alternate methods are being developed. This paper attempts to find cetane number based on the properties of the bio-diesel so that cetane number can be found without operating an engine. If a correlation between fuel properties and cetane number is established, the influence of each of the fuel properties on cetane number can be analyzed. This paper uses artificial neural networks (ANNs), which are a recently developed computational technique used to correlate non-linear data, to predict cetane number and analyze the influence of the various fuel properties namely density, viscosity, flash and fire points on the cetane number of a bio-diesel and its various blends.Oils produced from seeds of Sunflower, Palm, Pungam plant (Honge) and their Methyl Esters are used for this analysis. Standard diesels with known cetane numbers are used. The fuel properties are found by standard techniques. Different combinations of ANNs are formed by varying the number of hidden layer neurons, activation functions and other network parameters and are trained to predict cetane number from fuel properties using a series of train data. The networks so formed are validated using a series of test data. The best network is chosen based on the least value of mean squared error. By analyzing the weight matrix of the trained network, the relative impacts of the properties on the cetane number are analyzed. It is hoped that this work would aid in the selection of the suitable source and method of preparation to obtain the bio-diesel with the desired physical properties so that optimal cetane number is achieved

Notes:

Vendor supplied data

Publisher Number:

2007-01-0077

Access Restriction:

Restricted for use by site license