STUDI FERMODELAN DAN SIMULASI FLOTASI BATUBARA DENGAN KOMPUTASI JARINGAN SYARAF TIRUAN (ARTIFICIAL NEURAL NETWORK)

• Main Author: Sanwani, Edy
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/1439
• Institution: Institut Teknologi Bandung

Froth flotation is the well established technique for treating coal fines below 0.5 mm. Evaluation of coal flotation performance by conventional models (empirical and analytical models) needs detail laboratory experiments, good and detail understanding kinetics, etc. Anew modeling technique was introduced to overcome the limitation of the conventional models. The new model was performed based on the artificial neural network computation. Feedforward backpropagation method was applied to calculate the derivation of the model. Achievement of the modeling was significantly influenced by the topology of input and output connection. The accuracy of the feedforward backpropagation neural network (tbnn) computation results in predicting the performance of a batch coal flotation process was evaluated with respect to experimental results. It was found that it was characterized by the number of training patterns. The best result of the study was obtained with the error of 2.20%, 4,61%, 3.98%, and 2. 11%, respectively for combustible recovery, ash content and inherent moisture of clean coal produced, and yield of the process. On the other hand the iteration convergence and the rate of computation were significantly determined by the network parameter adjustment which included the connection weight initiation, the number of training patterns, the hidden layers of neuron, the rate of learning process and the value of its momentum.