Enhance fault identification in rotary equipment using Machine Learning algorithms

• Published in: 2024 5th International Conference on Innovative Trends in Information Technology (ICITIIT) pp. 1 - 6
• Main Authors: Sangeetha, V, Chaudhari, Shilpa Shashikant, Tanupriya, R, Theertha, K, Varsha, S D, Vishnupriya, C
• Format: Conference Proceeding
• Language: English
• Published: IEEE 15.03.2024
• Subjects: Costs; fast Fourier transform; Fast Fourier transforms; Fault diagnosis; Industries; machine learning; Machine learning algorithms; Predictive models; rotary equipment; vibration analysis; Vibrations
• DOI: 10.1109/ICITIIT61487.2024.10580376

Rotating machinery is essential to production and is utilized across many sectors. Compressor, pump, and motor productivity all impact an industry's profitability. A single industry can lose {\}2.3 million annually due to unplanned downtime. In some industries, these costs are even higher, with unplanned downtime losses reaching tens of thousands of dollars per minute. To address these types of issues, the proposed system uses machine learning algorithms to identify the faults in rotary equipment. In this paper, the proposed model applies machine-learning algorithms to classify the faults in the model. Fast Fourier transform (FFT) analysis is carried out on vibration signals to separate them into constituent spectral components. A Decision tree and Random Forest classifier are used to predict anomalies and classify them into different fault types. The proposed system predicts specific fault types based on vibration diagnosis which compared to the general features of temperature, speed, and acceleration gives a much more accurate analysis. The proposed models provide users with previously unattainable levels of insight that accurately predict specific potential failures and replace the failing components just before they are due to fail. The results indicate that the Multi-fault classification using random forest classifier achieves higher accuracy of 86 percent when compared to decision tree classifier of 84 percent.