Data Analytics Method For Detecting Extinction Precursors To Lean Blowout In Spray Flames

Aircraft engines must always maintain a margin between the operating equivalence ratio and the lean blowout boundary because flame-out presents a significant risk to the safety of the aircraft. It is believed that flames undergo a series of extinction/re-ignition phenomena before blowout. Previous attempts to characterize these phenomena have not been universally accepted. The approach presented here is from data analytics and consists of three parts: data curation, fault detection, and an adaptive alarm reliability assessment. The data curation filters the nonstationary behavior from photomultiplier tube signals recorded from a combustor test-rig, thereby reducing the number of false alarms. The filtered data is used to develop a fault detection algorithm that detects changes in the statistical properties of the signal. This results in alarms that serve as precursors of impending blowout. By leveraging information from previous blowout occurrences and the currently observed signal, the reliability of these alarms is updated in an adaptive manner. Through this methodology, combustion system operators are provided a means for assessing the proximity of blowout in a probabilistic manner.

Automated summary

This paper introduces a method for monitoring aircraft engine flame-out through data analytics, including data curation, fault detection, and adaptive alarm reliability assessment. Alarms are used as precursors of impending flame-out, updated in an adaptive manner based on information from previous occurrences and currently observed signals, providing a probabilistic means to assess the proximity of flame-out.