Marine sliding bearing residual life prediction method based on transfer learning

Abstract

The invention provides a marine sliding bearing residual life prediction method based on transfer learning, and the method comprises the steps: 1, carrying out the anti-seizure performance test of marine sliding bearings made of different materials, and collecting the multi-sensor signal data of the sliding bearings made of different materials in the whole degradation process; 2, time domain and frequency domain multi-sensor feature extraction is carried out, a multi-dimensional multi-sensor feature vector is formed, and normalization processing is carried out; and 3, building a transfer learning life prediction model, reducing the difference between two domains by introducing domain classification loss, MMD loss and CORAL loss, carrying out iterative training on the residual life prediction model by adopting an RMSprop optimization algorithm, and finally outputting a residual life prediction result of the target domain material sliding bearing. According to the method, the first-order and second-order characteristic differences between failure data distributions of sliding bearings made of different materials are reduced, and meanwhile, the universality and prediction precision of the life prediction model are improved.

Description

technical field [0001] The invention relates to the technical field of bearing residual life prediction, in particular to a method for predicting the residual life of marine sliding bearings based on migration learning. Background technique [0002] In marine diesel engines, sliding bearings play an important role in supporting dynamic loads and converting the reciprocating motion of the piston into the rotational motion of the crankshaft. Fatigue failure is a common damage form of plain bearings under long-term high cyclic dynamic load conditions. In addition, the harsh lubrication state caused by excessive cyclic dynamic load will induce a sudden increase in the temperature of the back of the shoe, aggravate the degradation of the sliding bearing, and even cause major failures such as shaft holding and burning of the sliding bearing in severe cases. If the remaining service life of the bearing can be accurately predicted and the degradation status of the bearing can be tr...

AI Technical Summary

Problems solved by technology

Due to the differences in fatigue resistance, seizure resistance, wear particle embedding, compliance, etc. among sliding bearing materials, the degradation and failure data of sliding bearings of different materials may be located in different marginal probability distributions, and for each It is difficult to collect sufficient failure samples for sliding bearings made of different materials

However, most data-driven lifespan prediction methods require a large number of training samples for learning, and assume that the training samples and test samples have the same domain, ignoring the impact of the distribution difference between the two on the prediction accuracy

Therefore, it is difficult to obtain a robust life prediction model for the diversity of sliding bearing materials and the difficulty in accumulating samples

Images