Abstract: As a core loading component of the flap hinge mechanism in aircraft,the remaining useful life (RUL) of aviation self-lubricating bearings directly affects operational safety during aircraft takeoff and landing phases.To address the issues of low accuracy and weak generalization in traditional data-driven methods,this paper proposes a remaining useful life prediction method for aviation flap hinge mechanisms that integrates modern temporal convolutional networks with Transformer.By establishing an experimental platform to collect acoustic emission signals from self-lubricating bearings under actual flight conditions,the method leverages the dilated causal convolution characteristics of modern temporal convolutional networks to achieve global feature extraction from long-term time series data.Combined with Transformer’s self-attention mechanism,it performs refined mining of local micro-features and deep-level characteristics,constructing an end-to-end life prediction model.Experimental results demonstrate that on datasets collected under actual operating conditions,the model’s root mean square error (RMSE) remains consistently below 2%,showing 30%~50% improvement in prediction accuracy compared with deep learning baseline models.The research outcomes confirm the reliability and stability of the proposed method.