Active Control of Lean Blowout Using an ANN-Based On-Line Direct Controller

Combustor Lean Blowout (LBO) is a significant challenge in modern military gas turbine combustors. LBO can occur either during rapid engine deceleration or during operation at idle conditions with water/steam ingestion. Current practice is to design adequate safety margins to avoid the onset of blowout. However, significant engine performance improvements can be realized though the development of a system that identifies and prevents blowout. Such a closed-loop, active control system is proposed for development in this SBIR. In Phase I, we will focus on sensing LBO and its precursors. Detailed experiments will be conducted to identify chemiluminescence and pressure signal characteristics that precede blowout. A prototype gas turbine injector being developed for JSF application under Navy sponsorship will be used to provide a realistic and representative test environment. In Phase II, the complete system will be developed and demonstrated. A novel Artificial Neural Network (ANN) on-line direct controller will be developed to identify LBO precursors in real time and control the fuel injector response. An ANN-based direct controller does not require a system model and the controller automatically adjusts to changing dynamic system conditions. For LBO control, an innovative plasma-based flame stabilization system will be incorporated into the fuel injector and used to stabilize the flame beyond the lean blowout limits by introducing a corona discharge at the injector exit.