Optimization Method for Evaluating Generator Wedge Tightness using a Learning Model

The stator wedges of large-scale generators must remain tight during operation to protect the windings from vibrations. This study aims to improve the traditional qualitative wedge evaluation method that relies on manual hitting and auditory assessment, by employing a portable device to acquire wedge tightness condition data. The Gaussian mixture model (GMM) algorithm was utilized to perform clustering based on wedge tightness conditions, followed by an evaluation of classification accuracy using the optimal clustering model. This research provides a comparative evaluation of classification performance by optimizing GMM clustering for data-driven quantitative assessment of wedge conditions.