Quantitative prediction of residual stress and hardness in case-hardened steel based on the Barkhausen noise measurement

The aim of this study is to predict residual stress and hardness of a case-hardened steel samples based on the Barkhausen noise measurements. A data-based approach for building a prediction model proposed in the paper consists of feature generation, feature selection and model identification and validation steps. Features are selected with a simple forward-selection algorithm. A multivariable linear regression models are used in predictions. Throughout the selection and identification procedures a cross-validation is used to guarantee that the results are realistic and hold also for future predictions. The obtained prediction models are validated with an external validation data set. Prediction accuracy of the prediction models is good showing that the proposed modelling scheme can be applied to prediction of material properties.

:► We study the quantitative prediction of material properties. ► Predictions are based on the Barkhausen noise measurements. ► A prediction scheme is proposed and used successfully. ► Features that are selected and used in prediction models are meaningful. ► Residual stress and hardness are predicted with good results.