Katı margarin bitmiş ürün yağsız kuru madde analizinin sonucunun yapay nöral ağ kullanılarak tahmin edilmesi
Prediction of the total salt and milk percentage at the finished product quality control stage of margarine using artificial neural network model
- Tez No: 197145
- Danışmanlar: DOÇ.DR. ERDOĞAN ALPER
- Tez Türü: Yüksek Lisans
- Konular: Bilim ve Teknoloji, Kimya Mühendisliği, Science and Technology, Chemical Engineering
- Anahtar Kelimeler: Artificial neural networks, backpropogation algorithm, Bayesianregularization, margarine
- Yıl: 2005
- Dil: Türkçe
- Üniversite: Hacettepe Üniversitesi
- Enstitü: Fen Bilimleri Enstitüsü
- Ana Bilim Dalı: Kimya Mühendisliği Ana Bilim Dalı
- Bilim Dalı: Belirtilmemiş.
- Sayfa Sayısı: 127
Özet
.$7,0$5*$5ø1%ø70øùh5h1
Özet (Çeviri)
PREDICTION OF THE TOTAL SALT AND MILK PERCENTAGE AT THEFINISHED PRODUCT QUALITY CONTROL STAGE OF MARGARINE USINGARTIFICIAL NEURAL NETWORK MODELMetin TantalkayaABSTRACTIn this study, it is aimed to design an appropriate artificial neural network model topredict the total salt and milk percentage at the finished product quality controlstage of Yayla margarine manufactured by Turk Henkel at the affiliated facility sitein Izmir. Results of the free fatty acid, color (red), melting point, moisture content,pH and salt analyses were decided to be used as input variables in the neuralnetwork to predict the output that is the total milk and salt percentage. Datarequired for the network training and test sessions were collected from productquality control sheet which was generated for each finished product batchseparately. A three layer feed forward back-propogation network was chosen forprediction. After several training and test sessions, optimum performance wasobtained with the use of model which had 6 neurons in input layer and 30 neuronsin its hidden layer. However, generalization ability of this model seemedinsufficient, therefore Bayesian regularization was used to improve it. Aftercompletion of design, neural network model was tested with 385 different data set,each containing input data which have never been introduced to the model before,and the model was able to respond 90% of those inputs by an actual error lessthan 20% which is considered to be acceptable. The neural network model is alsoable to predict the total salt and milk percentage by an absolute mean error 9.98%.Since both input and target data are noisy, performance of that model isconsidered to be successful but should be further developed.
Benzer Tezler
- Mide kanseri ve beslenme
Gastric cancer and nutrition
ZEKİ KEMEÇ
Tıpta Uzmanlık
Türkçe
2005
OnkolojiDicle Üniversitesiİç Hastalıkları Ana Bilim Dalı
DOÇ.DR. ABDURRAHMAN IŞIKDOĞAN
- Katı yağlarda bazı eser elementlerin alevli atomik absorpsiyon spektrometresi ve/veya grafit fırın atomik absorpsiyon spektrometresi ile tayini
Determination of some trace elements in solid oil samples by using flame atomic absorption spectrometry or/and graphite furnace atomic absorption spectrometry
ALPER CAN BABAÇ
Yüksek Lisans
Türkçe
2019
KimyaTrakya ÜniversitesiKimya Ana Bilim Dalı
PROF. DR. YASEMİN BAKIRCIOĞLU KURTULUŞ
- Katı ve sıvı yağ tüketiminin plazma lipoprotein fraksiyonları üzerine olan etkilerinin araştırılması
Investigation of effect of saturated and unsaturated fats on lipoprotein fractions
RECEP GÖKÇE
Doktora
Türkçe
1997
BiyolojiDumlupınar ÜniversitesiBiyoloji Ana Bilim Dalı
PROF. DR. MUSTAFA YÖNTEM
PROF.DR. İDRİS AKKUŞ
- Katı ve sıvı yağların lipid peroksidasyonu ve antioksidan sistemler üzerine olan etkilerinin araştırılması
Peroxidation of solid and liquid lipids and the investigation of their effects on antioxidant systems
SERDAR ERGÜN
Doktora
Türkçe
2004
BiyolojiDumlupınar ÜniversitesiBiyoloji Ana Bilim Dalı
PROF.DR. AHMET ÖZATA
Y.DOÇ.DR. MUSTAFA YÖNTEM
- Storage stability, quality and open-dating of fats and oils
Katı ve sıvı yağların saklama dayanırlılıkları, kalitesi ve raf ömrü
AHMET KAYA
