Umudike Journal of Engineering and Technology

This study is aimed at prediction of the physical properties of three leaved yam starch using Artificial Neural Network (ANN) soft computing techniques. The drying experiments were conducted at Drying Temperature: 40-75⁰C, Drying Time: 0-240 minutes and air velocity: 1.5- .30 to estimate bulk density, true density, surface area and bulk volume for the dried three leaf yam starch. ANN programming codes were developed in Matlab R2014b. 11 backpropagation ANN algorithms and neurons were trained for the prediction. The result obtained showed that Levenberg–Marquardt backpropagation was able to provide smallest MSE. The optimization of the ANN simulation using LMA algorithm showed that 14 hidden neurons (MSE 0.0000000000000696572) was chosen for bulk density, 11 hidden neurons (MSE 0.000000000000234103) was chosen for true density, 6 hidden neurons (MSE 0.0000000000921779) was chosen as the best bulk volume and 12 hidden neurons (MSE 0.000000000502683) was chosen for surface area. It was concluded obtaining an ANN model that could make reliable prediction on the effect of temperature, time and air velocity on the physical properties of the three leaf yam starch such as bulk density, true density, bulk volume and surface area which are very essential in the design and construction of processing and handling equipment for the starch production