Lean blowout detection using topological data analysis.

Area of Science:

• Combustion engineering
• Data analysis
• Dynamical systems

Background:

• Modern lean premixed combustors operate under ultra-lean conditions to meet emission standards.
• This operation increases susceptibility to lean blowout (LBO), a critical failure mode.
• Existing LBO prediction techniques are primarily developed for single-burner systems and are less effective for multi-burner configurations.

Purpose of the Study:

• To address the challenge of early lean blowout (LBO) detection in multi-burner combustors.
• To introduce and evaluate Topological Data Analysis (TDA) as a novel tool for real-time LBO prediction.
• To demonstrate the effectiveness of TDA across various combustor configurations.

Main Methods:

• Application of Topological Data Analysis (TDA) to combustion dynamics data.
• Analysis of TDA metrics during the transition to lean blowout.
• Investigation of sublevel set TDA metrics with low sampling-rate signals.

Main Results:

• Established LBO detection techniques are less effective for multi-burner combustors.
• TDA metrics are computationally inexpensive and exhibit monotonic trends approaching LBO.
• Sublevel set TDA metrics show robust monotonic changes even with low sampling rates.

Conclusions:

• Topological Data Analysis (TDA) provides a computationally efficient and reliable method for real-time LBO prediction in multi-burner combustors.
• TDA enables fine-tuning of the LBO safety margin, enhancing operational safety.
• TDA facilitates the use of simple, low-cost sensors for effective combustion monitoring.