Intelligent identification system for boiling dynamics from acoustic emission signal using machine learning technique

Abstract

Many heat transport and heat removal systems in various engineering fields adopt the boiling heat transfer to maximize their functions. Thus, many researchers have been devoted to investigating the fundamental phenomena beneath boiling heat transfer, and the identification of boiling regime is one of interesting subjects to be studied. To achieve clear identification of the boiling regimes and understand overall boiling behaviors over a wide region of the heated surface of interest, the present study employs an acoustic emission (AE) measurement technique. In this study, the feasibility of AE measurement to the identification of boiling regimes is tested and the machine learning algorithm is developed to establish the intelligent identification system for further application to 3D printing-based glass test facility in UNIST. The AE feature of each boiling regime characterized by statistical and spectral parameters. Different acoustic features generated from different boiling regimes including nucleate boiling, transition boiling, boiling crisis, and film boiling are characterized by its measured AE signal with respect to various parameters. The spectral AE signal from each boiling regime provides input and target parameters for the training system using neural network method. Finally, the trained identification system effectively classifies the boiling regimes using AE signals.