Investigating the influence of combustion flow field position on intelligent enhancement in moiré fringe analysis using deep learning

Abstract

Moiré tomography, holds a pivotal role in visualizing structures and inverting key parameters for flow fields, with its robust stability and extensive measuring range. And, extracting information from moiré fringes with the highest possible accuracy and efficiency remains one of the most challenging problems. In this study, an intelligent moiré fringe analysis method based on U-Net is proposed, which can accurately and efficiently extract wrapped phase information for moiré fringes. To verify the feasibility of the proposed method, three sets of experiments were conducted, resulting in the acquisition of more than 3600 frames moiré fringes. By training the fringes, three prediction models are obtained, and these models are subsequently utilized for prediction. The results manifest that the average RMSE of the predictions for the corresponding three groups is less than 0.220, when the training set and the test set have matching flow fields in the optical path. However, the prediction effect is not ideal if there is no match, which means the predictive performance is influenced by the position of the flow field in the optical path in the training set. Therefore, it is essential to consider the influence of the flow field position in the process of achieving intelligence in moiré tomography. In a word, the pertinent conclusions drawn in this paper provide valuable insights for enhancing the intelligence level of moiré tomography.