A Support Vector Regression-based Operational Transfer Path Analysis Method for Detecting Vibro-acoustic Sources with Uncertain Path Contributions

摘要

Operational Transfer Path Analysis (OTPA) is a widely-used approach in engineering to detect the dominant source of vibration or noise by comparing the contributions of several pre-defined vibrationoise transfer paths. Because of its dependency on the quality of signals, this approach may identify incorrect path contributions, particularly when random noise is mixed into the collected data. A number of deterministic techniques, e.g. singular value truncation, have been proposed to mitigate this problem. However, the performances of these techniques rely heavily on careful parameter-tuning. To address this issue, a modified OTPA method is presented in this paper which uses support vector regression (SVR) to evaluate the inherent uncertainties associated with the path contributions. Apart from its data-based, automatic way for parameter definition, the main advantage of the proposed method is that it transforms each contribution from a single value to an adjustable interval, so that the reliability of the predictions could be measured and further examined. The effectiveness of the proposed method is verified against traditional OTPA by a simple acoustic emitter-and-receiver numerical example.