The weakest link in the a control system is the control valve as it is the only moving part and due to the presence of friction it is subjected to varying dynamics quite often. The major causes of control valve problems are nonlinearities such as stiction, backlash, and deadband. Due to these, the valve movement does not follow the control signal accurately but deviates from it. In other words, the relationship between controller output (OP) and process variable (PV) often changes. These nonlinear and discrete dynamics of valves introduces artificial oscillations in the control systems which severely affects the performance. Detection and diagnosis of the problems associated with control valves is essential reduce the variability in the product quality in process industries. In this work, a statistical method is proposed to distinctly identify and quantify the valve hysteresis, stiction and process disturbances based on the previous plant data. Adaptive transforms are built to model OP-PV relationship which is then used in predictive control framework to reject the disturbances otherwise cause by oscillating valve output.