Using Artificial Neural Networks to Develop a Predictive Method From Complex Experimental Heat Transfer Data

Abstract Artificial neural networks are employed to develop a predictive algorithm using experimental heat transfer data for a complex situation. The data of Marto and Anderson has used to illustrate the process. This data is from a series of experiments investigating the boiling heat transfer from a vertical bank of tubes in refrigerant 114 with variable amounts of oil present. Both finned and unfinned tubes were investigated. The network was trained with a partial set of the available data. The prediction obtained using the trained network was then compared to the remaining experimental data. The artificial neural network provided an excellent predictive method.