Optimization of Pipe Support Position by Using Machine Learning

Abstract

Pipe supporting devices are very significant on maintaining not only pipe structure but also whole structure on thermal power plant or nuclear power plant. Selection and proper placement of supports is the one of the most difficult tasks on power plant design. Recently, many of skilled engineers have been retired in these areas and human resources shortage has become a social problem. Therefore, using accumulated know-how, automated pipe-support arrangement system is needed and optimization of design is also needed condensing past technics. In this study, we have developed optimization system of pipe-support combination to properly decide their position and type, using machine learning and deep learning. As a result, it is proven that machine learning and deep learning are sufficiently useful in design and their arrangement optimization of pipe-support on thermal power plant or nuclear power plant. In this study, PyCaret as a tool of automated machine learning has been used and it has been proved that LightGBM is the best algorithm.