Modeling the Drying of a High-Moisture Solid with an Artificial Neural Network

Abstract

The byproduct so-called alperujo (a solid−liquid byproduct generated from olive oil extraction by two-phase centrifugation) is difficult to dry due to its high moisture content. An artificial neural network (ANN) was developed to model the drying of alperujo in fluidized bed dryers. The ANN, a three layer perceptron (3 inputs, 4 hidden nodes, and 1 output) with back-propagation updating, serves to predict the moisture of the output solid at a time t + T from known input data at time t. The input data are the actual values of the input air temperature, fluidized bed temperature, and output solid moisture. T is the sampling time. The ANN learning, topology optimization, and verification are described in this paper. The essential data to design the ANN were taken from runs in a bench-scale dryer, and the ANN validation was carried out by applying basic statistical criteria and the standard tests of Mann−Whitney, Kruscal−Wallis, and Kolmogorov−Smirnov. The results of these tests, which compare the real with predicted moisture data, demonstrate that the fluidized bed dryer is well-modeled by the ANN (prediction error of 4.5%).