Neural Network Control of Temperature Profile by Impinging Jet Arrays

Abstract

The wall temperature profile in the flow field of an impinging two-jet array has been controlled using a neural network. The jet excitation is achieved by injection and suction through fine slits co-located at the nozzle exits to obtain a desired wall temperature profile. The wall temperature profile is determined "uniquely" by the excitation pattern so that the flow field is essentially considered as the "function" with the excitation pattern as the input and the wall temperature profile as an output. A neural network learns the inverse function of the flow field via offline learning and online learning, and is then serves as the controller. As a result, the wall temperature distribution is controlled with high accuracy and this demonstrates the applicability of control on the convective heat transfer process.