Statistically Data-Driven Operational Transfer Path Analysis

Abstract

Conventional model-driven operational transfer path analysis (OTPA) cannot update and optimize itself based on data characteristics, which weakens its analysis accuracy and reliability. Inspired by data-driven idea of learning from data, this paper develops statistically data-driven OTPA. First, by considering the statistical distribution characteristics of the potential data errors according to the central limit theorem, the factors affecting the transmissibility error are analyzed and summarized. Then, by constructing the objective function and performing iterative optimization, more and more suitable data is gradually found, and the OTPA model is updated and optimized synchronously. Finally, the iteration is terminated when the data is not updated, and a statistically driven OTPA model is obtained. The validation is carried out on simulation, test bed and gas turbine vibration datasets, and robust accuracy improvements are observed. The proposed method helps to improve the analysis accuracy and reliability of OTPA without adding additional practical work, and is expected to enhance the generalization performance of OTPA in more complex mechanical systems and application scenarios.

Keywords: Operational transfer path analysis, Statistically data-driven method, Central limit theorem, Total least squares, Multi-rotor gas turbines.