Multi-response characterization of ultra-thin strip rolling process-machine learning approach

The prediction of machining parameters attracted more researchers and industrial quality experts due to the accurate optimizations with the developing techniques every second. The present paper involves multi-response characterization of the Roll Force, von Mises Stress and PEEQ (Equivalent Plastic Strain) i.e., output responses with the aid of input parameters viz., percentage reduction, Coefficient of friction (COF) and Roll Speed in Ultra-Thin Strip rolling using Machine Learning (ML) models. The aim of this research is to train the machine learning framework to predict the output responses through fivefold cross validation and hyper parameter optimization. The research focuses on a hybrid approach i.e., experimental results are fed to ABAQUS and to the Machine learning framework. The simulations generally take 20—40 h based on the input parameters. As the ML framework is integrated, the prediction time is reduced to 17 s, which is superior to any mathematical approach.