Work

Deep learning-based quantification of turbulence and equivalence ratio effects on cellular morphology in premixed hydrogen/air flames

Year: 2026

Type: article

Abstract: Experiments were conducted in a fan-stirred constant-volume combustion vessel to investigate the effects of equivalence ratio and turbulence intensity on cellular structures and instabilities of spherically expanding turbulent premixed hydrogen/air flames. A few-shot transfer-learning segmentation method based on Cellpose 2.2.2 was adopted to process high-speed shadowgraph images, with dedicated datasets for model fine-tuning and generalization assessment. The trained model extracted cell number, total cell-boundary length and cell area to characterize cellular refinement during flame propagation. Correlations of wrinkling coefficient and normalized average cell area with effective Lewis number, normalized turbulence intensity and Karlovitz number were examined. Multicollinearity and regularized multivariable analyses indicate these parameters act as coupled indicators rather than independent control variables. Two dimensionless flame-structure parameters were proposed, and Peclet-number-based scaling showed regime-dependent exponent trends consistent with flame propagation and front structural evolution.

Source: International Journal of Hydrogen Energy

DOI: 10.1016/j.ijhydene.2026.157261

Authors: Weilin Zheng, Bangying Xue, Wen Zeng, Fan Xie, Haoran Zhao

Institutions: Shenyang Aerospace University, Northeastern University

Language: English

Topic: Combustion and flame dynamics

Subfield: Computational Mechanics

Field: Engineering

Domain: Physical Sciences

SDG: Affordable and clean energy

Open Access status: hybrid

Estimated APC paid: $3,730