In order to investigate the friction coefficient variation law of the cylinder liner-piston ring friction pair in the internal combustion engine under complex working conditions,the reciprocating friction motion between the cylinder liner-piston ring of the internal combustion engine system was simulated using the SRV high-temperature tribotester according to the temperature and pressure of actual working conditions.The friction coefficient of the cylinder liner-piston ring friction pair under different temperature and pressure was obtained,and the viscosity of lubricating oil at different temperatures was tested.A BP neural network prediction model was constructed with friction contact surface temperature,load and lubricant viscosity as input and friction coefficient as output.The results demonstrate that the BP neural network has a high prediction accuracy,and the error value of the test set is stable within 1%,which meets the engineering accuracy requirements.