Objective  This paper puts forward a method for monitoring and evaluating the lubrication performance of a marine stern tube bearing which combines a lubrication performance decay model and support vector machine (SVM) algorithm.

  Methods  Aiming at difficulties in the monitoring and recognition of the lubrication regimes of stern bearings, a bearing lubrication decay numerical model is established and validated with experimental data. The effects of load, roughness and radius clearance on the lubrication decay mechanism are then investigated. Based on the SVM algorithm, a lubrication regime classifier is constructed; the hyperparameters are optimized through a grid search algorithm; the datasets of different lubrication regimes are used for training; and lubrication regimes for stern bearings are evaluated.

  Results  The results show that with the increase of external load, roughness and radius clearance, the critical speed of the deterioration of the bearing lubrication regime increases, the working range of hydrodynamic lubrication (HL) decreases and the working range of mixed lubrication（ML） increases. The lubrication regime recognition model is then verified by the simulation dataset, and the proposed lubrication regime recognition method has an accuracy rate of 96.88%.

  Conclusion  The method proposed herein can monitor the lubrication performance characteristics of marine stern tube bearings and effectively identify the optimal bearing lubrication regime.