现有轴承健康指标构建方法大多仅依托振动监测数据获取性能退化变量，难以全面反映轴承的实际健康状态，进而影响剩余寿命预测精度。针对此问题，提出一种基于多源数据融合的轴承剩余寿命预测方法。首先，采用飞蛾扑火优化算法确定变分模态分解的最优超参数，将数据分解为预定数量、特定区间内的子序列，充分挖掘多维数据中隐藏的时序特征，并融合轴承振动与温度信号特征，构建复合健康指标。其次，通过随机模型刻画轴承健康指标的退化演变过程，采用对数似然函数最大化法完成模型初始参数的估计求解。最后，利用贝叶斯公式实现模型参数的实时更新，得到轴承剩余寿命的概率分布结果。基于轴承全寿命试验数据集开展仿真验证，结果表明所提健康指标能够准确反映轴承健康状态，有效提高轴承剩余寿命预测的准确性。

Most of the existing methods for constructing bearing health indicators rely solely on vibration monitoring data to obtain performance degradation variables, which makes it difficult to comprehensively reflect the actual health status of the bearing and thus affects the accuracy of remaining useful life prediction. To address this issue, a bearing remaining useful life prediction method based on multi-source data fusion is proposed.Firstly, the Moth-Flame Optimization algorithm is used to determine the optimal hyperparameters of the variational mode decomposition, decomposing the data into a predetermined number of subsequences within a specific interval, fully exploring the hidden time series features in multi-dimensional data, and fusing the vibration and temperature signal features of the bearing to construct a composite health indicator. Secondly, a random model is used to describe the degradation evolution process of the bearing health indicator, and the initial parameters of the model are estimated and solved by maximizing the log-likelihood function. Finally, the Bayesian formula is used to update the model parameters in real time, obtaining the probability distribution result of the bearing's remaining useful life. Simulation verification is carried out based on the bearing full life test data set. The results show that the proposed health indicator can accurately reflect the health status of the bearing and effectively improve the accuracy of bearing remaining useful life prediction.