This paper presents a study on fault identification in self-aligning roller bearings. For this purpose, a test bench was used for experiments and data acquisition, utilizing a system composed of a microcontroller board and a MEMS-type accelerometer sensor. The main objective is to determine which Machine Learning techniques are most efficient in fault identification. The kNN (K-Nearest Neighbors) and Decision Tree algorithms were tested, with statistical descriptors applied to the acquired data. Additionally, a feature reduction study was conducted using the PCA (Principal Component Analysis) technique and the ANOVA (Analysis of Variance) statistical method for time-domain data analysis.The data were also transformed into the frequency domain before being applied to the algorithms. The kNN algorithm with frequency-domain data achieved the best result, with an accuracy of 98%.
Fault Analysis in Self-Aligning Roller Bearings Using Machine Learning Techniques with kNN and Decision Tree Algorithms
73 Views
6 Downloads