This paper presents a study of the parameters estimation for pressure's model in steam heated cylinder based on the experimental input-output data. From the practice, there exists different between the model designed for pressure control loop and the actual pressure process, that difference is revealed after a long time in performance. By using the new model in which the integral part in the old model is replaced by the inertia part, the paper uses the practical data to identity parameters of new model based on the gray box identification method. To estimate the parameters of this new model based on practical input-output data, the new model in the time domain is firstly converted into the Laplace domain to avoid formulating the identification algorithm with the differential operator, which is known to be numerically challenging in the presence of noise. Second, to get the linear model in s-domain with pressure and valve position signals filtered the authors use m stable and causal filter in series, this model next is converted to the time domain to obtain the model which is filtered describing the relations between the pressure and valve position of process not affected by the noise. From the final linear model, using the algorithm N4SID the parameters of the new model finally were estimated. With many model transformations but the proposed method is subsequently successfully validated in using experimental data, the estimated model exhibit good fit to the input-output data. This study results allow the redesign of the controller in pressure control loop to obtain better quality control.