AR model-based crosstalk cancellation method for operational transfer path analysis

To reduce the crosstalk effects between the signals at reference points in operational transfer path analysis (OTPA) and improve the reliability of solutions, a novel OTPA method based on auto-regressive (AR) model is proposed in this paper. The key to this method is to obtain the power spectrum by AR model for estimating the prior transmissibility function matrix. Firstly, in view of the lack of noise reduction process in conventional crosstalk cancellation method, singular value decomposition (SVD) is applied to reduce the noise of original signals. Secondly, Burg algorithm is used to calculate AR model parameters. Thirdly, auto-power spectrum and cross-power spectrum are obtained by AR model and the periodogram method respectively. Fourthly, the transmissibility function matrix is obtained by Landweber iterative method based on corrected signals at reference points. Finally, the proposed method can significantly reduce the crosstalk effects, resulting in more accurate evaluation of transfer path contributions. Generally, this study can provide accurate and reliable evidences for vibration & noise monitoring and reduction for mechanical systems, and thus benefit for the active control of vibration & noise.