Simultaneous soot multi-parameter fields predictions in laminar sooting flames from neural network-based flame...

Area of Science:

• Combustion Science
• Optical Diagnostics
• Machine Learning Applications

Background:

• Soot formation in flames is a critical phenomenon impacting combustion efficiency and emissions.
• Accurate measurement of soot parameters like volume fraction, temperature, and particle size is essential for understanding and controlling combustion processes.
• Traditional diagnostic methods for these parameters can be complex, costly, and time-consuming.

Purpose of the Study:

• To develop and validate a novel neural network-based method for simultaneously diagnosing multiple key parameters in laminar sooting flames.
• To utilize flame luminosity as an input for predicting soot volume fraction, temperature, and primary particle diameter.
• To establish a more efficient, lower-cost, and high-fidelity diagnostic approach for combustion and reacting flows.

Main Methods:

• Development of a Bayesian optimized back propagation neural network (BPNN).
• Application of the BPNN to flame luminosity data for prediction.
• Assessment of the method's feasibility and robustness using numerical modeling.
• Validation of the approach with experimental data from laminar diffusion sooting flames.

Main Results:

• The BPNN model accurately predicts the planar distribution of soot volume fraction, temperature, and primary particle diameter.
• High prediction accuracies were achieved: up to 114 K for temperature, 0.25 ppm for volume fraction, and 2.56 nm for particle diameter.
• The method demonstrated robustness through both numerical and experimental validation.

Conclusions:

• The proposed machine learning-assisted optical diagnostics provide an efficient and cost-effective solution for multi-parameter diagnosis in sooting flames.
• This approach paves the way for high-fidelity, simultaneous measurement of critical combustion parameters.
• The study highlights the potential of AI in advancing combustion research and diagnostics.