This study aims to improve simulation efficiency for status and fault monitoring in gas seal systems.
A rapid calculation method for seal flow fields based on non-invasive model order-reduction is proposed. First, training data were obtained by calculating sample working conditions in the full-order model (FOM) with a typical spiral groove gas seal as the research object. Then, singular value decomposition and genetic aggregation algorithms were used to perform dimensionality reduction and interpolation fitting on the training data. Finally, select an appropriate sample size and the number of modes for constructing a reduced-order model (ROM) based on error prediction. A flow field ROM consisting of working condition-modal coefficients and modal bases is established.
The simulation performance of the ROM is evaluated by the error of results between the FOMs and ROMs. The results of this study indicate that the absolute error of the seal face pressure prediction does not exceed 0.01 MPa, with a maximum relative error below 0.3%. The maximum pressure in the slot area increased by 59.57% compared to the inlet pressure. In addition, the calculation time for the ROM has been reduced to less than 1 s.
The data-driven flow field ROM described in this paper provides method support for model simplification and the construction of gas seal digital twins, effectively solving the problem of real-time monitoring and fault detection in gas seal systems.
The peer review history for this article is available at: https://publons.com/publon/10.1108/ILT-06-2025-0288/
