Prediction of Accurate Values for Outliers in Coal Drying Experiments

Abstract

Coal drying is a quite important process from both burning efficiency and granulation perspective. Therefore, coal drying experimentation processes always attract researchers from various fields. Those experiments are quite costly since they require expensive laboratory equipment and considerable labor hour. Even if the costs of experiments are tolerable, often long experiment periods and large number of experimentation will cause serious problems for prompt academic results. During the analysis of experiments, researchers convert the results into graphical form. However, when creating charts, it is observed that some of the results diverge from the others abnormally marking some measurement as outliers. In such cases, experiments should be repeated to eliminate the effects of these abnormalities. Due to high costs and time constraints, repetition of an experiment is not preferable in general. To predict the accurate values for outliers and overcome issues generated by these abnormalities, artificial neural network (ANN) is employed in this study and tolerable deviations and acceptable experimental costs are reached by using ANN.