Various industries widely use the internal combustion engine as a type of power machine. The cylinder liners and piston rings are the key components of internal combustion engines. Their wear state directly affects the performance and life of the internal combustion engine. Therefore, numerical simulation and analysis of the wear state of cylinder liners and pistons in internal combustion engines is of great significance for optimizing the engine to improve its reliability and durability. This article models the dynamics of the cylinder liner-piston ring system of an internal combustion engine, concentrating on the piston’s second-order motion and the cylinder liner’s vibration response. Then, we integrate the multi-scale feature extraction technique and the combined recurrent neural network (RNN) architecture to design a prediction model of cylinder liner surface wear, testing the feasibility of the optimization scheme for the cylinder liner and piston ring performance of an internal combustion engine. The optimized scheme using this paper’s method increased the minimum oil film thickness by 45.67% compared to the pre-optimization, while also reducing the friction loss work by 3.2%.