Abstract:To improve the accuracy and stability of noise and vibration transfer path identification in complex ship structures, a reverse multi-stage transfer path analysis method is proposed. The method establishes input–output relationships using operational transfer path analysis and introduces a truncated total least squares algorithm based on singular value decomposition to correct ill-conditioned matrices and enhance computational robustness. By applying hierarchical decomposition, the method identifies energy contributions between subsystems and constructs a multi-level noise transfer model. Experimental verification in an anechoic water tank shows that the proposed approach accurately reconstructs forward and reverse transfer characteristics, validates reciprocity of transfer paths, and effectively identifies abnormal energy transmission and faults under isolator failure conditions. The results demonstrate that this method significantly improves the reliability and precision of transfer path identification and provides strong technical support for ship vibration reduction and condition monitoring.