Neural Network Prediction of Reciprocating Friction for X-Type Seals

Abstract X-type sealing rings are widely used in sealing systems. They operate in complex environments and play a crucial role in maintaining the safe operation of the entire mechanical system. Nevertheless, the prediction for the friction force in X-type sealing rings relied on methods that were computationally complex and cumbersome. This made it difficult to quickly and accurately estimate the working state of the sealing rings in practical engineering applications, thereby affecting the safe operation of mechanical structures. To address this challenge, this article developed a neural network-based predictive model coupled with finite element analysis to accurately estimate the friction force in X-type seals under reciprocating motion. In this work, ansys finite element analysis software was used to obtain the reciprocating friction force data of X-type sealing rings with different sizes under different medium pressures. Then, a neural network prediction model was applied to analyze and predict the data. Finally, experiments were conducted to verify the predicted friction force data again. The results showed that the finite element simulation method could obtain friction force data relatively accurately and could be used as a data source for neural network training. The neural network-based predictive model coupled with finite element analysis could predict the friction force of the sealing rings during operation quite accurately. The prediction error was roughly maintained within 5%. Compared with the pure numerical calculation method, this approach greatly reduced the complexity of computation and had higher practical value in engineering applications.