A New Gearbox Fault Diagnosis Method Based on Gramian Angular Field and ConvNeXt

Abstract

The gearbox plays a critical role in the wind turbine, and its failure may lead to the shutdown of the entire power generation system. Therefore, it is crucial to diagnose gearbox faults accurately. In this paper, we convert the time series fault diagnosis problem into an image classification problem. First, the series are transformed to images by polar coordinates using Gramian Angular Summation Field (GASF), which preserves the original temporal characteristics of the data. The images are subsequently processed by the ConvNeXt model to perform image classification. The effectiveness and flexibility of the proposed method have been validated through several comparative experiments on the HUST gearbox dataset.