Deep Learning-Based Prediction and Analysis of the Influence of Lubrication on Gear Efficiency

Abstract

This paper involves an analysis and prediction of the effects of spur gears on efficiency based on the use of deep learning. The efficiency of the gear transmission is highly influenced by friction of the two meshing tooth faces, the formation of films of lubricant, power loss, and temperature of operation. Five lubrication conditions were studied in this work: unlubricated reference condition, grease, low-viscosity oil, medium-viscosity oil, and high-viscosity oil. Calculations of torque and speed were used to calculate input power, output power, power loss, and gear efficiency using the experimental data. Alongside, three deep learning algorithms, such as Deep Neural Network, 1D-Convolutional Neural Network, and Long Short-Term Memory network, were implemented to forecast gear efficiency and power loss based on various lubrication circumstances. The experimental data revealed that gear performance was greatly enhanced by the presence of lubrication as opposed to the absence of lubrication. The lowest efficiency of 87.9, highest power loss of 75.9 W and highest operating temperature of 58o C was observed with the unlubricated gear pair, whereas the lowest power loss of 37.2 W and the highest efficiency of 95.4 and 95.7 was observed with the low and high viscosity oil respectively. Medium-viscosity oil produced the best performance with the highest efficiency of 97.7 percent, the lowest power loss of 14.7 W, and the lowest operating temperature of 27 C compared to the unlubricated condition which had higher efficiency, lower power loss, and operating temperature. Deep learning outcome indicated that Deep Neural Network model had the most optimal prediction performance with an efficiency prediction RMSE of 0.29, power loss RMSE of 0.78 W and an R. 2 value of 0.98. The 1D-CNN and LSTM models made acceptable predictions, although, their errors were larger compared to the DNN model. In general, the findings support the view that adequate lubrication enhances the efficiency of gears by preventing the occurrence of frictional losses and heat. The article also shows that deep learning can become an efficient method to predict the gear performance and find the most appropriate lubrication condition.