Volume 45, No 3, 2023, Pages 487-502

Download full text in PDF

Prediction of Gearbox Oil Degradation Based on Online Sensor Data and Machine Learning Algorithms


Kunal Kumar Gupta , S.M. Muzakkir

DOI: 10.24874/ti.1491.06.23.08

Received: 6 June 2023
Revised: 16 July 2023
Accepted: 22 August 2023
Published: 15 September 2023


In most of the gearboxes, mixed lubrication conditions prevail, and to avoid the wear of gear surfaces, oil additives like extreme pressure, anti-wear, anti-rust and antioxidant additives are used. The lubricant additives form a lubricant film on gear surfaces, minimize metal-to-metal contact and protect the surfaces. But in this process, the lubricant additives are consumed, and oil quality deteriorates causing degradation of wear-prevention lubricant functionality. The degradation of lubricant with time, even without its usage is alarming and it has been reported in the present manuscript. To observe the consequence of degraded gear-oil on gear surface, an experimental setup has been developed. The results of experiments, conducted on commercially available two-gear oils have been detailed. Three cases of single stage spur-gear pair were considered: (1) Loaded with 40 Nm torque value and operated at 1200 rpm for 198 hours duration. (2) Loaded with 50 Nm torque value and operated at 500 rpm and. (3) Accelerated conditions generated by adding 0.0025 %v of mild (36% concentration) Hydrochloric acid in the lubricant in addition to accelerated conditions specified in case 2. For the cases 2 & 3, the setup was run for 90 minutes duration. The dataset of this study includes five parameters namely time, humidity, temperature, oil quality and generated Fe debris. Machine learning techniques have been used to reduce the actual number of experiments by applying LR, DTR, KNNR, RFR, ANN and SVM in predicting the Oil degradation rate.


Lubricating ageing, Additive depletion, Wear, LR, DTR, KNNR, RFR, ANN

Latest issue


Volume 45
Number 3
September 2023

Crossref logo


TiI news RSS 2.0

Table of contents