A winding deformation detection method based on identification of the nonlinear vibration system of transformer

This paper aims to propose an efficient and reliable method to detect the mechanical faults in transformers from vibration signatures and relative electric inputs. In the previous work, a single-input and single-output (SISO) Hammerstein-type model can be developed for identifying the nonlinear transformer vibration system, with individual vibration sources activated. According to the previous work, it can be found that the mechanical faults have a significant influence on the property of the linear part in that Hammerstein-type model. And changes in the model order can indicate the variance of characteristics of the linear part. The order determination method based on the Lipschitz criterion is thus presented and utilized to estimate the model orders, which are considered as a mechanical faults indicator. The proposed method is tested on a transformer, giving promising results for fault detection of the transformer windings.