<p>This study focuses on the application of machine learning to predict the wear rate of rubber compounds under different operating conditions. Estimation of rubber wear rate has practical importance in various applications such as tires, shoes, wiper blades among others. Developing a mathematical model to predict the loss rate of rubber materials due to wear will provide a useful tool for engineers to better estimate the lifespan of rubber products, improve their wear resistance, and reduce economic losses. In this study, the experimental results are obtained from a sliding friction and wear test machine that has been designed and developed with the purpose of testing rubber samples on different substrates. In this test machine, the sliding velocity and normal load on the rubber sample can be accurately adjusted and controlled. The friction force is continuously measured using a load cell during the experiments. The wear loss of the rubber compound is measured using a precise weighting scale more than nine times for each set of controlled parameters (speed, normal load, and substrate surface) to ensure the accurate measurement of steady state wear rate. The wear tests are performed on three types of surfaces with different roughness parameters. The surface roughness parameters of each substrate have been obtained from height profile measurement using an optical profilometer with sub-micron precision. The steady state wear loss is measured at a total of eight different sliding speeds. After collecting the experimental data, a statistical analysis is conducted to determine the correlation between the inputs (speed and substrate surface roughness) and the output (the wear loss per unit distance). The experimental data set is randomly divided into three subsets for training, validation, testing of the Machine Learning (ML) models. In all the ML algorithms,70% of experimental data is used for training, 15% for validation, 15% for testing​. Three different ML algorithms in MATLAB are utilized to analyze the collected data. The first ML method used in this study is Artificial Neural Network (ANN) with Levenberg-Marquardt backpropagation algorithm. This backpropagation algorithm is selected due to its superior performance for wear rate prediction after comparing its results with other backpropagation algorithms such as Bayesian regularization and scale conjugate gradient. The second ML algorithm is Support Vector Machine with Gaussian kernel, and the third algorithm is Boosted Tree with eight leaves (terminal nodes). The most efficient algorithm is determined by comparing the root mean squared error of all three ML models. The results conclude that Artificial Neural Network outperforms the other two machine learning algorithms for rubber wear prediction.</p>

Estimation of Rubber Wear Rate Using Three Different Machine Learning Algorithms

Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2021/volume/85543

ASME 2021 International Mechanical Engineering Congress and Exposition (IMECE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.