TL;DR
This paper investigates how the "edge effect" – the region where flame signals become weak – impacts the accuracy of reconstructing flame temperature and soot volume fraction using CCD camera data. The authors found that ignoring this edge region significantly degrades accuracy, particularly for soot volume fraction, and propose using the flame image boundary as a unified reconstruction boundary for better results.
What Problem Does This Paper Address?
Accurately understanding combustion processes is crucial for improving efficiency and reducing harmful emissions. Flame temperature and soot volume fraction are key parameters that reveal combustion state and pollutant formation. While optical diagnostic methods like CCD cameras offer non-invasive measurement, reconstructing these parameters from the captured radiation signals is challenging. A significant hurdle is the "edge effect," where the faint radiation signals at the flame's periphery can lead to substantial errors in reconstruction, especially for soot. This paper aims to quantify this edge effect and propose a more robust reconstruction method.
Methods
The study employs a numerical reconstruction model based on an axisymmetric flame assumption. The flame cross-section is divided into concentric rings, and the radiation intensity captured by a CCD camera is modeled. The core of the method involves solving an inverse problem to derive the radiation source term, from which temperature and soot volume fraction are calculated using colorimetry. The paper specifically investigates the impact of excluding outer rings (representing the flame edge) from the reconstruction domain. They also introduce stochastic errors to simulate measurement noise in both radiation intensity and the coefficient matrix to assess the system's robustness. The LSQR algorithm is used to solve the ill-posed inverse problem.
Datasets
The study uses a simulated dataset based on the line-of-sight measurement of a typical coaxial diffusion flame, as reported in reference [45]. This dataset provides the "exact" (ground truth) distributions of flame temperature and soot volume fraction for comparison. No new experimental datasets were generated for this specific study; rather, the existing data was used to validate the numerical model and its findings regarding edge effects.
Key Insights and Ideas
- Edge Effect is Critical: The flame's edge, where radiation signals are weak, is not a negligible region. Ignoring it leads to a significant increase in reconstruction error, especially for soot volume fraction, making the results unreliable at the flame boundary.
- Soot is More Sensitive: The reconstruction of soot volume fraction is considerably more sensitive to the edge effect and measurement errors than flame temperature.
- Unified Boundary is Better: Using the overall flame image boundary as the reconstruction boundary, rather than arbitrarily excluding outer rings, yields more satisfactory and accurate 2D distributions of temperature and soot.
- SNR Matters, But Edge Effect Dominates: While lower Signal-to-Noise Ratios (SNR) degrade accuracy, the impact of neglecting the flame edge is more pronounced and detrimental to reconstruction quality.
- Coefficient Matrix Noise is Less Influential: Errors in the coefficient matrix (which is related to the reconstruction boundary choice) have a less significant impact on accuracy compared to errors in the radiation intensity itself.
Key Results
- When the outermost ring (17th) was ignored, the radiation intensity loss was ~1%. When the 16th and 17th rings were ignored, it was ~3%.
- Ignoring the flame edge caused relative reconstruction errors to increase from the flame center to the edge.
- The relative reconstruction error for soot volume fraction at the boundary reached up to 45.21% (ignoring 17th ring) and 80.97% (ignoring 16th and 17th rings).
- Even with a low SNR of 46 dB, the maximum average relative reconstruction error for temperature was 0.11% and for soot volume fraction was 2.06%.
- Using the flame image boundary as the unified reconstruction boundary provided satisfactory accuracy, even at low SNRs.
- The study found no obvious regularity between reconstruction accuracy and the extension of the reconstruction boundary when using the image boundary.
Limitations and Future Work
- Axisymmetric Assumption: The model assumes axisymmetry, which might not hold for all real-world flames. Future work could extend this to asymmetric flames.
- Simplified Radiation Model: The model relies on certain assumptions about soot scattering and absorption. More complex radiative transfer models could be explored.
- Numerical Study: The findings are based on numerical simulations. Experimental validation with real flame data would strengthen the conclusions.
- Unified Boundary Optimization: While the image boundary was proposed as a unified boundary, further investigation into optimizing this boundary choice for different flame types and conditions could be beneficial.
- Low Soot Regions: The paper notes that low soot volume fractions in non-sooting regions can influence accuracy. Further research could explore specific strategies to mitigate this.
Summary updated