Doctoral Student Studies Green Approach to Gear Fault Diagnosis

Gear failures are a common problem in heavy machinery, often leading to costly downtime, excessive noise, and energy losses. Conventional detection systems depend on battery-powered vibration sensors, which have limited lifespans and create environmental waste. To tackle this challenge, Ahmedabad University doctoral student Jogin Dhebar developed an eco-friendly, AI-powered diagnostic system that requires no external power.

Jogin introduced a TENG-based Time-Frequency Mode Embedded Convolutional Neural Network (T-TFCNN). The Triboelectric Nanogenerator (TENG) is a self-powered sensor capable of generating electrical signals directly from the vibrations of rotating machinery. These raw signals are then processed using Variational Mode Decomposition (VMD) to extract meaningful patterns. Finally, a one-dimensional Convolutional Neural Network (CNN) classifies the gear faults with remarkable precision.

Jogin tested the solution in the Mechanical Design Lab to validate the theory by building a complete experimental setup that included a 3HP induction motor and gearbox. He simulated real gear conditions such as healthy gears, missing teeth, chipped teeth, and root cracks. A custom-made TENG sensor was mounted directly on the gearbox to collect data, which was used to train and test the AI model. This hands-on process allowed Jogin to integrate mechanics, data science, and artificial intelligence knowledge, applying it to real-world challenges.

The system achieved 100 per cent accuracy in diagnosing gear faults at 20 Hz and 98.33 per cent accuracy at 30 Hz. It also offered a solution that was sustainable, scalable, and ready for industrial application.   

Jogin's work, presented at the International Mechanical Engineering Congress and Exposition (IMECE-2025) by the American Society of Mechanical Engineers (ASME), offers a sustainable and highly reliable method for maintaining smooth machine operation. Developed under the guidance of Professor Akhand Rai​​​​​​, the solution explored in the study "Gear Fault Diagnosis Based on Triboelectric Nanogenerators and Convolutional Neural Network" proves eco-friendly and effective in diagnosing gear faults.