Abstract: In the field of structure health monitoring, the data asynchrony problem caused by heterogeneous sampling frequencies of monitoring equipment remains a critical challenge. The direct fusion of unregistered data can significantly compromise the reliability of evaluation results, leading to distortions in characterizing the actual in-service state of structures. To address this issue, this study proposes a Bayesian inference driven method for the temporal registration of multi-source asynchronous data and the real-time structural damage detection. The method proposed introduces a sequential asynchronous data fusion framework. By establishing a multi-timescale discretization model based on a fusion period, the state space and observation equations for each sensor node are formulated. An unscented Kalman filter is then employed for real-time structural parameter identification using multi-source asynchronous data. To verify the effectiveness of the method proposed, a comprehensive validation system is constructed, including a three degree of freedom shear structure model, a three-span bridge model, and a shaking table test on a scaled 12 storey reinforced concrete frame model. The research results demonstrate that the method proposed can achieve an accurate identification using multi-source asynchronously sampled data. Moreover, the method proposed maintains a satisfactory estimation performance even when low frequency sampling measurements are used.