The temperature rise prediction of the friction components in multi-disc clutches is of great significance for ensuring the stable operation of the transmission system. A hybrid prediction method for clutch temperature rise based on trend decomposition is proposed. The period term and trend term in the data are decomposed by using the exponential moving average decomposition and the RevIN method to reconstruct the distribution of the input data. Feature extraction is carried out with the aid of the CNN to enhance the feature learning ability of the model. The Multi-Layer Perception (MLP) model and the Crossformer model are adopted to predict the trend and period terms respectively. The final prediction result is obtained by synthesizing the two parts of the output. By comparing with three representative models, it is verified that this method has a relatively high prediction accuracy, and the average performance is improved by at least 23.5%.

multi-disc clutches, radial temperature rise, failure forecast, deep learning