An efficient method to predict the convection heat transfer coefficients on the top surface of the engine piston is proposed. The method is based on the inverse method of the thermal conduction problem and uses a numerical optimization technique. In the method, the heat transfer coefficients are numerically obtained so that the difference between analyzed temperatures from the finite element method and measured temperatures is minimized. The method can be effectively used to analyze the temperature distribution of engine pistons in case when application of prescribed-temperature boundary condition is not reasonable because of insufficient number of measured temperatures. A hollow sphere problem with an analytic solution is taken as a simple example and accuracy and efficiency is demonstrated. The method is applied to a practical large liquid petroleum gas(LPG) engine piston and the heat transfer coefficients on the top surface of the piston is successfully calculated. Resulting analyzed temperature favorably coincides with measured temperature.