Abstract: In view of the harsh operating conditions of rolling bearings, the difficulty in extracting effective features, as well as the fact that the first prediction time (FPT) in bearing life prediction is difficult to determine and the accuracy of a single model is low. A bearing life prediction method combining bearing degradation state estimation and 1DCNN-BiGRU-SA is proposed. Firstly, the variational mode decomposition (VMD) optimized by the triangulation topology aggregation optimizer (TTAO) is utilized, combined with the Pearson correlation coefficients to perform noise reduction and reconstruction on the signal. Secondly, the time-domain degradation features of the reconstructed signal are extracted. The precision determination of bearing FPT is realized by constructing a degradation feature fusion mechanism, and then the evolution law of the degradation state of the bearing throughout its entire life cycle is characterized more accurately. Finally, a 1DCNN-BiGRU model embedded in the self-attention (SA) mechanism is constructed, and a time series composed of multiple time-domain features is used as input to predict bearing life. The XJTU-SY rolling bearing accelerated life test data set is used to verify the effectiveness of the proposed method. The research results show that the proposed life prediction method has higher prediction accuracy.