Journal of Applied Sciences1812-56541812-5662Asian Network for Scientific Information10.3923/jas.2010.2678.2682RashidR.RadzaliS.R.Abdul RahmanB.Mohamed EsivanS.M.1220101021Measurement of biological variables in a process is a key
to efficient control and supervision of the bioprocess. In a process of protein
production such as erythropoietin (EPO), it is crucial but difficult to measure
EPO concentration using direct or on-line measurements. EPO concentration is
usually measured through laboratory analysis where expensive costs of test kit,
tedious and long time analysis are the biggest obstacles. Artificial neural
network software sensor was developed to estimate EPO concentration based on
other measured variables such as biomass, substrate or by-product in EPO production.
Radial Basis Function was utilized to map nonlinear mapping between the input
and output parameters. This study deals with effect of input numbers and spread
constant on radial basis performance. It is found that different number of inputs
and spread constant significantly affect the performance of the predictive model.
The high values of coefficient of determination, R from regression analysis
also proved that this model successfully mapping the nonlinear relationship
between the input and output variables.]]>Bernard, O., Z. Hadj-Sadok and D. Dochain,2000Chen, S., C.F.N. Cowan and P.M. Grant,1991Cheruy, A.,1997Choi, D.J. and H. Park,2001Dacosta, P., C. Kordich, D. Williams and J.B. Gomm,1997Eckardt, K.U.,1996Golobic, I., H. Gjerkes, I. Bajsic and J. Malensek,2000Lanouetta, R., J. Thibault and J.L. Valade,1999Moghadas, R.K and K.K. Choong,2008Rashid, R., H. Jamaluddin and N.A.S. Amin,2006Vanek, M., P. Hrnc, J. Vovsik and J. Nahlik,2004Solihin, M., I. Wahyudi and A. Albagul,2006Jianxu, L. and S. Huihe,2002