To address the challenge of establishing a stable removal function for double-sided polishing to predict the finished surface profile, we use the coordinate transformation method to derive the relative velocity distribution equations for the upper and lower surfaces of the component. Subsequently, static pressure distributions on both surfaces are simulated using ANSYS software. The simulation data is then imported into Matlab and fitted with a polynomial method to determine the time-varying pressure distribution formulas for the component's surfaces. Based on the Preston equation, an expression for the correction coefficient K is derived. The value of the correction coefficient K is calculated to be 2.588×10^-15 from four sets of polishing experimental data, enabling the construction of a predictive model for the surface pattern in double-sided polishing processes. The predictive model is ultimately validated through machining experiments. The experimental results indicate that the error in predicting the PV (Peak-to-Valley) value accounts for 1.07% to 7.4% of the actual PV value after processing, demonstrating good agreement between the predicted model and the actual post-processing surface pattern.