Magnetorheological finishing （MRF） is an important process for reducing surface damage and efficiently suppressing the surface error in conical mirrors. However， because of the serious distortion and complex distortion law of the removal function for MRF of conical surfaces， existing methods have failed to establish the removal function model directly and effectively， resulting in a low surface convergence efficiency. In this study， the influence mechanism of the conical curvature effect on the distortion of the MRF removal function was analyzed. In addition， the analytical rule of the normalized characteristic parameters of the removal function for the mean curvature was studied. Furthermore， the removal function deduction scheme from plane to cone was established. This method integrates the evolution laws of the length and width of the conical MRF removal function， volume removal rate， and peak removal rate into the deduction to more comprehensively reflect the distortion characteristics of the MRF removal function under the conical curvature effect. It avoids the measurement of physical and chemical characteristic parameters and the solving of complex equations and nonlinear problems， providing an effective， low-cost method for the removal function deduction of conical MRF under actual conditions. The results of repetitive spot sampling experiments show that the deduction errors of the characteristic parameters of the conical removal function are between 3.20% and 12.02%， demonstrating the strong applicability of the proposed deduction method.