PENENTUAN KADAR AIR DAN ASAM LEMAK BEBAS BUAH SAWIT BERDASARKAN WARNA MENGGUNAKAN JARINGAN SYARAF TIRUAN

RONGGO KUSWIDIANTO, Dr.Rudiati Evi Masithoh, STP.,M.Dev.Tech ; Dr. Joko Nugroho WK, STP.,M.Eng

2014 | Skripsi | TEKNIK PERTANIAN

Penentuan kadar air dan asam lemak bebas sawit (Elaeis gueneesis sp.) biasanya dilakukan secara konvensional di laboratorium yang membutuhkan waktu dan proses yang lama. Penelitian ini bertujuan untuk menentukan kadar air dan asam lemak bebas sawit berdasarkan parameter warna. Sawit pada berbagai tingkat kemasakan diukur parameter warna R, G, B dan L, a, b, serta kadar air dan kandungan asam lemak bebas. Kadar air dan kandungan asam lemak bebas diprediksi menggunakan metode jaringan syaraf tiruan (JST) backpropagation. Warna R, G, B dan L, a, b, digunakan sebagai input, sedangkan kadar air dan kandungan asam lemak bebas digunakan output. Pelatihan JST menggunakan tingkat pembelajaran (learning rate) 0,3 dengan nilai MSE sebesar 0,0151, hidden layer 1 dengan 40 node dan hidden layer 2 dengan 60 node, serta epoch 1756. Hasil penelitian ini menunjukkan bahwa program yang dirancang dapat memprediksi nilai kadar air dan asam lemak bebas berdasarkan komponen warna buah sawit. Koefisien korelasi grafik regresi antara kadar air dan asam lemak bebas aktual dan prediksi adalah 0,92 dan 0,665. Nilai regresi yang tingi tersebut menunjukkan bahwa model yang diujikan sangat baik.

Determination of water content and free fatty acid of kernel palm (Elaeis gueneesis sp.)is usually done by conventional method in laboratory which takes time and requires long process. This research aimed at determining water and free fatty acid content of palm kernel based on color parameters. R, G, B and L, a, b color parameters, as well as water and free fatty acid content were measured from palm kernel at various ripening. Water and free fatty acid content were predicted using artificial neural network(ANN) method. R, G, B and L, a, b were used as inputs, while water and free fatty acid contentwere used as outputs. The result of backpropagation neural network was trained with 0,3 learning-rate level, with MSE value is 0,0151. Hidden layer 1 value is 40, and hidden layer 2 value is 60 was happened on 1756 of epoch. Results showed that ANN could predict water and fattty acid content of kernel palm based on color. Correlation coefficients between actual and predicted values of water dan free fatty acid content were 0,92 and 0,665, respectively. It show that the model was good for tested.

Kata Kunci : Sawit, Kadar air, Asam lemak Bebas, RGB, Lab, Jaringan Syaraf Tiruan, Backpropagation