Widyaningtyas, Shinta and Dr. Sucipto,, STP. MP. and Yusuf Hendrawan,, STP. M. App. Life Sc. PhD. (2018) Deteksi Campuran Kopi Reguler Dalam Kopi Luwak Menggunakan Metode Dielektrik dan Pengolahan Citra Digital. Magister thesis, Universitas Brawijaya.
Abstract
Kopi adalah bahan baku minuman yang banyak dikonsumsi dan penting bagi perekonomian negara produsen kopi. Salah satu jenis kopi yang dikenal mahal dan langka adalah kopi luwak (palm civet coffee) sehingga rawan dicampur dengan kopi reguler. Deteksi campuran kopi reguler dalam kopi luwak secara konvensional menggunakan analisis kimia bersifat destruktif, mahal, membutuhkan preparasi sampel dan waktu lama. Saat ini, konsumsi ekstrak green beans menjadi trend baru karena dianggap rendah kalori. Pengukuran total fenol green beans membantu mengukur aktivitas antioksidan. Selain itu, kopi memiliki rasa asam yang identik dengan pH. Trend mengonsumsi ekstrak green beans kopi membuat pengukuran pH dilakukan karena banyak konsumen kopi sensitif terhadap keasaman kopi terutama kopi arabika. Hal ini memberi peluang perancangan peralatan sederhana, cepat, akurat, dan non destruktif untuk menduga persentase campuran kopi reguler dalam kopi luwak, total fenol, dan pH. Perancangan alat penduga persentase campuran kopi reguler dalam kopi luwak, total fenol, dan pH dapat menggunakan metode dielektrik dan metode pengolahan citra digital. Metode dielektrik untuk mengakuisisi data biolistrik biji kopi dan metode pengolahan citra untuk mengakuisisi data citra. Pendugaan persentase campuran kopi reguler dalam kopi luwak, total fenol, dan pH perlu integrasi dengan metode untuk mempelajari pola data seperti Jaringan Saraf Tiruan (JST). Penelitian ini dilakukan dengan mencampur green beans kopi luwak dan kopi reguler pada presentase 0% kopi luwak, 10% kopi luwak, 30% kopi luwak, 40% kopi luwak, 50% kopi luwak, 70% kopi luwak, 90% kopi luwak, dan 100% kopi luwak. Hasil penelitian menunjukkan 5 fitur terpilih sebagai input JST data biolistrik yaitu impedansi (Z), induktansi seri (Ls), induktansi paralel (Lp), resistansi seri (Rs), dan resistansi paralel (Rp) menggunakan metode correlation attribute evaluator. Topologi JST terpilih untuk data biolistrik yaitu 5 – 40 – 40 – 3 (5 input, 40 node hidden layer 1, 40 node hidden layer 2, 3 output) dengan learning rate 0,1 dan momentum 0,9 fungsi pembelajaran trainlm, fungsi aktivasi tansig pada hidden layer dan output layer. Topologi terpilih menghasilkan R pelatihan 0,98902; R validasi 0,98204 dan MSE pelatihan 0,0099; MSE validasi 0,0479. Hasil penelitian deteksi pemalsuan kopi luwak menunjukkan 5 fitur terpilih sebagai input JST data citra yaitu lain red sum mean, value sum mean, S_HSL sum mean, blue variance, dan hue variance menggunakan metode ReliefF. Data citra menggunakan JST menghasilkan topologi terpilih yaitu 5 – 30 – 40 – 3 (5 input, 30 node hidden layer 1, 40 node hidden layer 2, 3 output) dengan learning rate 0,1 dan momentum 0,5 fungsi pembelajaran trainlm, fungsi aktivasi tansig pada hidden layer dan purelin pada output layer. Topologi terpilih menghasilkan R pelatihan 0,99502; R validasi 0,97933 dan MSE pelatihan 00085; MSE validasi 0,0442. Hasil penelitian menunjukkan MSE validasi metode dielektrik sedikit lebih rendah dibandingkan metode pengolahan citra digital. Hal ini membuat kedua metode non destruktif ini berpotensi sebagai sensor dalam menduga persentase campuran kopi reguler dalam kopi luwak, total fenol, dan pH.
English Abstract
Coffee is a raw material for beverages that are consumed and important for the economy of coffee producing countries. One type of coffee that is known to be expensive and rare is palm civet coffee so it is sensitive to be mixed with regular coffee. Detection mixtures of regular coffee in palm civet coffee conventionally uses chemical analysis that is destructive, expensive, requires sample preparation and need more times. At present, consumption of green beans extract is a new trend because it is considered low calories. Total phenol measurements of green beans to measure antioxidant activity. In addition, coffee has an acidic taste that is identical to pH. The trend of consuming green coffee extract makes measurements of pH carried out because many coffee consumers are sensitive to the acidity of coffee, especially arabica coffee. This gives the opportunity to design simple, fast, accurate and non-destructive equipment to predict percentage mixtures of regular coffee in palm civet coffee, total phenol, and pH. Design of percentage mixtures of regular coffee in palm civet coffee, total phenol, and pH can use dielectric method and image processing. Dielectric method for acquiring bioelectric data of coffee bean and image processing to acquire image data. Prediction of percentage mixtures of regular coffee in palm civet coffee, total phenol, and pH need integration with methods to study data patterns such as Artificial Neural Networks (ANN). This research was conducted by mixing green beans, palm civet coffee and regular coffee at 0% civet coffee, 10% civet coffee, 30% civet coffee, 40% civet coffee, 50% civet coffee, 70% civet coffee, 90% civet coffee, and 100% civet coffee. The results showed 5 selected features as ANN input for bioelectric data namely impedance (Z), series inductance (Ls), parallel inductance (Lp), series resistance (Rs), and parallel resistance (Rp) using correlation attribute of evaluator method. Selected ANN topology for bioelectric data 5 - 40 - 40 - 3 (5 inputs, 40 hidden layer 1 nodes, 40 hidden layer 2 nodes, 3 outputs) with 0.1 learning rate and 0.9 momentum, trainlm learning function, tansig activation function in the hidden layer and output layer. Selected topologies gave R training at 0.98902; R validation at 0.98204 and MSE training at 0.0099; MSE validation at 0.0479. The results showed 5 selected features as ANN input for image data namely red sum mean, value sum mean, S_HSL sum mean, blue variance, and hue variance using ReliefF method. Selected ANN topology for image data 5 - 30 - 40 - 3 (5 inputs, 30 hidden layer 1 nodes, 40 hidden layer 2 nodes, 3 outputs) , 0.1 learning rate and 0.5 momentum, trainlm learning function, tansig activation in hidden layer and purelin in output layer. Selected topologies gave R training at 0.99502; R validation at 0.97933 and MSE training at 00085; MSE validation at 0.0442. The results showed that MSE validation of dielectric method was slightly lower than the image processing method. This makes the two non-destructive methods potentially as sensors to predict percentage mixtures of regular coffee in palm civet coffee, total phenol, and pH.
| Item Type: | Thesis (Magister) |
|---|---|
| Identification Number: | TES/572.437/WID/d/2018/041810816 |
| Uncontrolled Keywords: | Biolistrik, Citra Digital, Kopi Luwak, Jaringan Saraf Tiruan, Pemalsuan,Artificial Neural Network, Bioelectric, Image Processing, Palm Civet Coffee, Percentage of Mixtures |
| Subjects: | 500 Natural sciences and mathematics > 572 Biochemistry > 572.4 Metabolism > 572.43 Energy metabolism > 572.437 Bioelectrochemistry |
| Divisions: | S2/S3 > Magister Teknik Industri Pertanian, Fakultas Teknologi Pertanian |
| Depositing User: | soegeng sugeng |
| Date Deposited: | 09 Aug 2022 07:50 |
| Last Modified: | 09 Aug 2022 07:50 |
| URI: | http://repository.ub.ac.id/id/eprint/193088 |
![[thumbnail of SHINTA WIDYANINGTYAS.pdf]](http://repository.ub.ac.id/style/images/fileicons/text.png) |
Text
SHINTA WIDYANINGTYAS.pdf Restricted to Registered users only Download (8MB) |
Actions (login required)
 |
View Item |