|
Volume 10, Number 1, January 2025 , pages 35-50 DOI: https://doi.org/10.12989/acd.2025.10.1.035 |
||
|
Artificial neural network modelling for predicting tribological properties of Al8090/TiB2/C composites using optimized hyperparameters |
||
Mohamed Zakaulla
|
||
| Abstract | ||
| This study introduces a new framework that utilizes artificial neural networks (ANN) to analyze data and forecast the tribological properties of Al8090/TiB2/C composites. For training a multi-layered artificial neural network (ANN), a total of 1920 input datasets are used. These datasets are created by combining six input parameters, including the volume of Al8090 matrix, Titanium diboride, and Graphene, as well as the load, sliding speed, and sliding distance. The corresponding output consists of specific wear rate and coefficient of friction. A surrogate model for predicting the tribological properties has been developed by optimizing the hyperparameters to enhance the accuracy of the model's predictions. The results of the ANN-based approach validate that the proposed model has a mean absolute percentage error of 3.42% for the predictions of specific wear rate in dry sliding wear test scenarios, and a MAPE of 0.28% for the predictions of coefficient of friction. | ||
| Key Words | ||
| aluminium; artificial neural network; hyperparameters; graphene; tribology | ||
| Address | ||
| Mohamed Zakaulla: Department of Mechanical Engineering, H.K.B.K College of Engineering, Bangalore 560045, India/ Visvesvaraya Technological University, Belagavi, Karnataka, India | ||
