Neural network based prediction of heat flux profiles on STRIKE

Abstract

The instrumented calorimeter STRIKE (Short-Time Retractable Instrumented Kalorimeter Experiment) has been designed with the main purpose of characterizing the SPIDER (Source for Production of Ion of Deuterium Extracted from Radio Frequency plasma) negative ion beam in terms of beam uniformity and divergence during short pulse operations. STRIKE is made of 16 1D Carbon Fiber Composite (CFC) tiles, intercepting the whole beam and observed on the rear side by infrared (IR) cameras. The front observation presents some drawbacks due to optically emitting layer caused by the excited gas between the beam source and the calorimeter, and the material sublimated from the calorimeter surfaces due to the heating itself. This paper proposes a Neural Network-based approach to solve the inverse non-linear problem of determining the energy flux profile impinging on the calorimeter, considering the 2D temperature pattern measured on the rear side of the tiles. Most of the conventional methods used to evaluate the inverse heat flux are unbearably time consuming; since the objective is having a tool for heat flux evaluation for STRIKE real time operation, the need to have a ready-to-go instrument to understand the beam condition becomes stringent. For this reason, in this paper, a Multi-Layer Perceptron has been used to solve the problem. Once properly trained, the neural networks provide a fast evaluation of the impinging flux. Furthermore, there is no need to optimize any parameter since this operation is already included in the self-adjustment of the network weights during the training. The achieved results show the reliability of the proposed method both with stationary and non-stationary heat fluxes.