Neural Network and Neuro -Taguchi’s Method

Important parameters on apple drying process are investigated experimentally and modeled employing artificial neural network and neuro-Taguchiʹs method. Experimental results show that the apple drying curve stands in the falling rate period of drying. Temperature is the most important parameter that has a more pronounced effect on drying rate than the other two parameters i.e. air velocity and the thickness of apple slices. In order to model the drying process, a software was developed which uses the error back propagation algorithm for training. At first, the software was used to simulate the time-dependent variations of moisture content using neural network. Then in order to model the time derivation of moisture ratio in break point, the software was utilized in two ways. First, it was used with no use of any optimization method for modeling the process. In the other approach, the software in a hybrid fashion with Taguchiʹs method as an optimization method is utilized to correct weight matrix entries. The results demonstrate that the use of neuro-Taguchiʹs method can give some improvements over neural network accuracy as compared with conventional neural networks approach. By using neuro- Taguchiʹs method, error is reduced by about 46.4%.