Development of a systematic failure analysis in Bearings using Neural Networks

• Grant number: 24/16835-9
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: February 01, 2025
• End date: January 31, 2026
• Field of knowledge: Engineering - Mechanical Engineering - Mechanical Engineering Design
• Principal Investigator: Wilson Carlos da Silva Junior
• Grantee: Carolina Yumi Siroma
• Host Institution: Instituto Federal de Educação, Ciência e Tecnologia de São Paulo (IFSP). Campus Guarulhos. Guarulhos , SP, Brazil

Abstract

The detection of bearing failure signals plays a crucial role in the industrial field, directly impacting the performance and reliability of mechanical equipment. In most cases, the acquisition of failure signals is conducted using high-cost commercial accelerometers, making it impractical for applications that could be easily implemented in mass-produced factory equipment. For failure detection, the application of neural networks is an emerging method. Thus, this study proposes to evaluate and compare the use of a low-cost MEMS accelerometer, the MMA7361, for obtaining temporal vibration signals. An experimental test bench is being developed for fault analysis using artificial intelligence, where the individually encapsulated sensors will be attached to a bearing housing that will simulate the following conditions: normal condition, inner race defect, and rolling element defect. To verify the failure signatures in the bearings, an algorithm composed of a Convolutional Neural Network (CNN) will be used, with an expected accuracy of 95%.