Pemodelan Jaringan Syaraf Tiruan Untuk Memprediksi Kinerja Pengering Pati Sagu Tipe Pneumatic Conveying Ring Dryer

Pneumatic conveying ring dryers (PCRD) have been designed for drying sago starch. To facilitate the development of further designs, an accurate prediction model is required. The objective of this research is to develop a model of artificial neural network (ANN) to predict the efficiency of drying on the sago starch dryer of PCRD type. The ANN model consists of 12 input neurons, hidden layers, and 1 output neuron. Hidden layer variations are performed on the ANN model with network structure 12-5-5-1-1, 12-15-15-1-1, and 12-25-25-1-1. The learning algorithm uses a trainln type backpropagation with logsig activation function. The ANN model was trained and tested using 81 data sets (54 sets of training data and 27 sets of test data). The test results show that the comparison between the results of the prediction model of ANN with observation obtained r2 train value of 0.998, and r2 test in 0.916. The results of optimization of the ANN model obtained by the mean value of Mean Square Error (MSE) train and test (0.063 and 0.232), Root Mean Square Error (RMSE) train and test (0.251 and 0.482), Mean Absolute Deviation (MAD) train and test (0.063 and 0.232), Mean Absolute Error (MAE) training and test (0.209 and 0.393), and Mean Relative Error (MRE) train and test (2.414 and 4.609). The best network structure is 12-25-25-1-1. This shows that the ANN model is able to predict the efficiency of drying, so it is feasible to be used for the design development on the sago starch dryer of PCRD type.