Just another WordPress site - Ruhr-Universität Bochum

Abstract

Statistical approach for the automated regression analysis of creep experiments is presented. The proposed method is based on combination of statistical algorithms for the change point analysis in nonstationary time series and regression analysis and it consists from three main steps. First, several statistical methods for the change point analysis have been tested for the automated selection of representative data sample. Then, a regression analysis has been performed based on the selected measurement points. During this step, three different functions have been considered for the modelling of creep behaviour. Finally, the most appropriate method for the automated data selection and the most-appropriate modelling function for the creep data have been identified based on several statistical criteria. All calculations have been performed in open-source software R. Moreover, to store the original experimental data files and selected representative data samples, a relational database using a combination of MySQL, Apache and PHP has been developed. The proposed approach of treating creep experiments provide large possibilities to speed up the data preparation for their further analysis and reduce the human factor and thus avoid possible errors. Moreover, the obtained results are consistent, objective and reproducible.