State classification for worm reduction gears based on vibration signal

Abstract

With the development of artificial intelligence technologies and the increment of structural complexity of modern mechanical systems, introducing machine learning technology into state classification of rotation machinery has become a hot and vital realm. This research elaborates on the acquisition of source data, steps of processing the data including sample division and transformation of data structure, feature extraction, machine learning models training, examines availability of source data and data features, and undertakes research on applying three machine learning algorithms and different model training methodologies to identify the working conditions and the fault patterns of Worm Reduction Gears and conduct rigorous validating on trained models, comparison and analysis of verification results of different scenarios and come to certain conclusions, eventually revealing performance disparities in different algorithms performance and training methods.