Deep learning based techniques for flame identification in optical engines

Rufino, Caio Henrique; Coraça, Eduardo Moraes; Lacava, Pedro Teixeira; Ferreira, Janito Vaqueiro

doi:10.1177/14680874221106976

Cited by 2 publications

(1 citation statement)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sun et al (2022) did flame edge detection for premixed, diffusion and energetic material flames at atmospheric pressures in thermography, single-channel and RGB images with fairly good contrast to the background. In the context of combustion diagnostics research and for more demanding image conditions, there exist some applications of ML-based flame front detection and segmentation in optical SI-engines by ; and Rufino et al (2023) for pressures up to 0.38 MPa. However, to the knowledge of the authors, no ML-based application considering planar laser induced fluorescence images of the OH radical (OH-PLIF) and data recorded at high pressure conditions exists in the literature.…”

Section: Introductionmentioning

confidence: 99%

Deep learning-based image segmentation for instantaneous flame front extraction

Strässle,

Faldella,

Doll

2024

Exp Fluids

View full text Add to dashboard Cite

This paper delves into the methodology employed in examining lean premixed turbulent flame fronts extracted from Planar Laser Induced Fluorescence (PLIF) images at elevated pressures. In such flow regimes, the PLIF signal suffers from significant collisional quenching, typically resulting in image data with low signal-to-noise ratio (SNR). This poses severe difficulties for conventional flame front extraction algorithms based on intensity gradients and requires intense user intervention to yield acceptable results. In this work, we propose Convolutional Neural Network (CNN)-based Deep Learning (DL) models as an alternative to problem specific conventional methods. The pretrained DL models were fine-tuned, employing data augmentation, on a small annotated dataset including a variety of conditions between SNR $$\approx$$ ≈ 1.6 to 2.6 and subsequently evaluated. All DL models significantly outperformed the best-scoring conventional implementation both quantitatively and visually, while having similar inference times. IoU-scores and Recall values were found to be up to a factor $$\approx$$ ≈ 1.2 and $$\approx$$ ≈ 2.5 higher, respectively, with $$\approx$$ ≈ 1.15 times improved Precision. Small-scale structures were captured much better with fewer erroneous predictions, becoming particularly pronounced for the lower SNR data investigated. Moreover, by applying artificially modeled noise, it was shown that the range of image conditions in terms of SNR that can be reliably processed extends well beyond the images included in the training data, and satisfactory segmentation performances were found for SNR as low as $$\approx$$ ≈ 1.1. The presented DL-based flame front detection algorithm marks a methodology with significantly increased detection performance, while a similar computational effort for inference is achieved and the need for user-based parameter tuning is eliminated. It enables a very accurate extraction of instantaneous flame fronts in large image datasets where supervised processing is infeasible, unlocking unprecedented possibilities for the study of flame dynamics and instability mechanisms at industry-relevant conditions.

show abstract

Section: Introductionmentioning

confidence: 99%

Deep learning-based image segmentation for instantaneous flame front extraction

Strässle,

Faldella,

Doll

2024

Exp Fluids

View full text Add to dashboard Cite

show abstract

Combustion of Hydrated Ethanol and Gasoline RON95 Ultra-High Pressure Direct Injection in Optical Access SI Engine

Malheiro de Oliveira,

Mendoza,

Martelli

et al. 2024

SAE Technical Paper Series

View full text Add to dashboard Cite

<div class="section abstract"><div class="htmlview paragraph">High and ultra-high pressure direct injection (UHPDI) can enhance efficiency gains with flex-fuel engines operating on ethanol, gasoline, or their mixtures. This application aims to increase the engine’s compression ratio (CR), which uses low CR for gasoline due to the knocking phenomenon. This type of technology, involving injection pressures above 1000 bar, permits late fuel injection during the compression phase, preventing auto-ignition and allowing for higher compression ratios. UHPDI generates a highly turbulent spray with significant momentum, improving air-fuel mix preparation, and combustion, resulting in even greater benefits while minimizing particulate matter emissions. This study aims to develop ultra-high-pressure injection systems using gasoline RON95 and hydrated ethanol in a single-cylinder engine with optical access. Experimental tests will be conducted in an optically accessible spark ignition research engine, employing thermodynamic, optical, and emission results. In the present work, the spark plug was placed in the lateral, so the ignition and part of the flame propagate close to the cylinder wall, and it will exchange with greater heat to the wall than the flame portions that propagate towards the central region of the chamber. Therefore, the flame front propagates at different speeds; causing stretching and wrinkling that can lead to instabilities and cyclic variability. To address this issue, this work presents experimental results that, through the images post-processing of flames under a SOI (start of injection) sweep strategy in the compression phase to closer of the spark ignition, associating the non-uniform propagation velocity of the flame with the cyclic variability. The fuel impingement on the wall was critical in this scenario, which led to higher soot concentrations and diffusive flames for gasoline. It was found that the injection close to the spark plug enhances the heat release, and combustion stability, decreasing soot emissions. Total unburned hydrocarbons (THC), Nitrous oxides (NOx), aldehydes, and soot emissions decreased for end of injection events closer to the spark ignition. This trend opposes the increase observed in CO emissions.</div></div>

show abstract

Deep learning based techniques for flame identification in optical engines

Cited by 2 publications

References 30 publications

Deep learning-based image segmentation for instantaneous flame front extraction

Deep learning-based image segmentation for instantaneous flame front extraction

Combustion of Hydrated Ethanol and Gasoline RON95 Ultra-High Pressure Direct Injection in Optical Access SI Engine

Contact Info

Product

Resources

About