Conductivity as a measured parameter has been used to evaluate the model of a non-linear process (spherical tank) at three different levels. The model was generated using a neural network backed propagation technique. These model parameters were used to design a neural- model predictive controller. For closed loop control of the process the algorithm were simulated in MATLAB. The performance of NMPC controller was compared with Smith Predictor controller and IMC controller based on rise time, settling time, overshoot and ISE and it was found that the NMPC controller is better suited for this process. The experimental and the model value were subjected to error analysis and the error was found to vary from 4.8 to 5.1 percent as flow rates varied from 2LPM to 5LPM. © 2012 IFSA.