Rancang Bangun Spektroskopi Cahaya Tampak Untuk Penentuan Kualitas Susu Dengan Menggunakan Jaringan Syaraf Tiruan

Abstract

Measurement to determine the natural cow’s milk quality using the neural network method has been carried out. The milk quality can be determined from its composition. The main components of milk are protein, fat, sugar (lactose), mineral, and water. The visible light spectroscopy is one of several equipments that can used to determine milk composition. The light source with various wavelength through on sample, build a spectrum graph of intensity versus wave length. The spectrum informs the milk’s chemical content. Neural network can identify object by learning and recall process. Each contents of milk has different spectrum. The graph is converted to numerical number. The neural network use backpropagation method. The learning process in this method determine synaptic weight and the direction is opposite with signal function as neuron response of given input. To determine synaptic weight need the activated function those is derivable and continue to all function boundary. This characteristic is completed by sigmoid activated function. The neural network program is created by Delphi 5 language. At this program, input layer receive spectrum data, and output layer as components of milk content. Number of layer and node on the hidden layer, and parameters of neural can be chosen in this program. The choice can be found until get least error. After doing the learning process the experimental result shows a optimum result., by network with architecture: 6-5-5, learning rate 1.0, and momentum 1.0.