Manutenção preditiva por análise de vibração em motores industriais utilizando floresta randômica

Abstract

This study explores the application of machine learning techniques in diagnosing faults in rota- ting machinery, which are responsible for increasing costs and reducing productivity in industrial environments. The main objective is to develop machine learning models with accuracy above 90% and compare them with each other. The data used were obtained from the Mafaulda rotating machinery fault simulator, and the predictions were made using supervised data. The data collected from experimental accelerometers were transformed from the time domain to the frequency domain using the Fast Fourier Transform (FFT). Additionally, data simplifications were performed to ensure that the available computational resources were sufficient for data processing and modeling. During the modeling process, the Principal Component Analysis (PCA) technique was applied, which increased the model's accuracy by 10.8%. Subsequently, the models Random Forest, Logistic Regression, Gradient Boosting, and Random Classification were compared, with Gradient Boosting achieving the best performance, reaching an accuracy of 95%.