PREDICTION OF NICKEL (II) IONS REMOVAL EFFICIENCY FROM LEACHATE USING ARTIFICIAL NEURAL NETWORK

Year 2011, Volume: 6 Issue: 1, 398 - 405, 01.02.2011

Beytullah Eren

Abstract

This paper presents the development of an artificial neural network (ANN) model for the prediction of the removal efficiency (Re %) of Nickel (II) ions from leachate based on 90 experimental data sets obtained in a bench scale experiments. The ANN models developed in this study used three input variables including initial concentration of Ni (II) ions, adsorbent dosage, and contact time for predicting corresponding Re %. The performance of the ANN models were assessed through mean square error (MSE), mean absolute error (MAE), and coefficient of determination (R2). The ANN model was able to predict Re % of Ni (II) ions with a tangent sigmoid transfer function (tansig) in hidden layer with 10 neurons and a linear transfer function (purelin) in output layer.The Levenberg–Marquardt algorithm (trainlm) was found as the best training algorithm with a minimum MSE of 0,00049. The modeling results indicated that there was an excellent agreement between the experimental data and predicted values.

Keywords

Adsorption, Artificial Neural Networks, Modeling, Nickel (II) Ions, Zeolite,

SIZINTI SUYUNDAN NYKEL (II) YYONLARI GYDERYM VERYMYNYN YAPAY SYNYR A?LARI YLE TAHMYN EDYLMESY

Abstract

Bu çaly?mada syzynty suyundan Nikel (II) iyonlarynyn giderim verimini (% Re) tahmin edebilen bir Yapay Sinir A?y (YSA) modeli geli?tirilmi?tir. Modelin geli?tirilmesinde kullanylan 90 adet deneysel veri laboratuar ölçekli deneylerden elde edilmi?tir. YSA modelinin geli?tirilmesinde girdi parametresi olarak syzynty suyu ba?langyç Nikel (II) iyonu konsantrasyonu (mg/L), adsorbent miktary (gr) ve temas süresi (dk) parametreleri, çyky? parametresi olarak Nikel (II) iyonu giderim verimi (% Re) kullanylmy?tyr. Geli?tirilen modelin etkinli?i Ortalama Karesel Hata (OKH), Ortalama Mutlak Hata (OMH) ve determinasyon katsayysy (R2) gibi istatistiksel parametreler kullanylarak belirlenmi?tir. Geli?tirilen tüm modeller arasynda % Re için en iyi tahmin kabiliyetine sahip olan YSA modelinin tek gizli katmanly, 10 i?lem elemanly (3-10-1) ve ö?renme algoritmasy olarak Levenberg–Marquardt geri yayylym algoritmasyny (trainlm) kullanan a? mimarisine sahip oldu?u belirlenmi?tir. Geli?tirilen 3-10-1 a? mimarili YSA modelinden elde edilen tahminlerin ölçüm sonuçlary ile istatistiksel açydan kar?yla?tyrylmasy ile modelinin çok iyi bir tahmin yetene?ine sahip oldu?u ve bu amaçla kullanylabilece?i görülmü?tür.

Keywords

Adsorpsiyon, Modelleme, Nikel (II) Yyonlary, Yapay Sinir A?lary, Zeolit,