Ravi Rajamani

Estimating Gas Turbine Internal Cycle Parameters Using a Neural Network

We show that a neural network can be successfully used in place of an actual model to estimate key unmeasured parameters in a gas turbine. As an example we study the combustion reference temperature, a parameter that is currently estimated via a nonlinear model inside the controller and is used in a number of critical mode-setting functions within the controller such as calculating the fuel-split between various manifolds. We show that a feedforward neural network using simple back propagation learning can be built for estimating combustion reference temperature. The neural network matches the accuracy of the current estimate; and it is more robust to errors in its internal parameters. This is advantageous from the point of view of implementation since a number of errors creep in when running the algorithm on a digital controller, and an estimator that is not robust with respect to such errors can degrade the performance of the whole system.Copyright

Closed Loop Minimization of NOx Emissions in a Gas Turbine

This paper describes a system for minimizing NOx emissions in gas turbines. The system uses an optimization technique implemented on a portable computer. No hardware changes are necessary in the combustion system. The system was developed for an LM6000 Dry Low Emissions (DLE) engine. However, it can be applied across the board on any LM series DLE engine manufactured by the Industrial Aero-derivatives (IAD) division of GE Power Systems (GEPS).Copyright

Model-Based Detection of Leaks and Blockages in Pipes

This paper deals with a model-based strategy for detecting leaks and blockages in a network of pipes. The fault detection and isolation (FDI) system uses a Kalman filter and a model of the piping system to decide whether the system is operating in a normal or failed state, and to distinguish the type of fault. While the idea of using Kalman filters is quite old, its application in the present case is novel, as is the formulation of the FDI system that uses only one model. The theory is backed by experimental validation on a test rig.Copyright

Multivariable Control of Industrial Gas Turbines: Field Test Results

Rejecting, or reducing, the effect of external disturbances on process parameters is an important problem in control design. In this paper we apply multivariable control techniques to reduce the effect of input disturbances, such as variations in the line frequency, on key internal parameters of an industrial gas turbine. The parameter we are most interested in is the combustion reference temperature, an estimated variable that is used by the controller to schedule division of fuel to various fuel nozzles and determine switching points between combustion modes. The dynamic response of this parameter correlates well with the dynamic response of fuel air ratio inside the combustor. Therefore, an important step in improving combustor performance is better regulation of the combustion reference temperature. We show that the use of a multivariable controller in place of the existing decentralized controller makes the disturbance rejection problem much easier to solve. As the gas turbine is inherently a multivariable system — i.e. the inputs, fuel and air, are coupled to the outputs, power and exhaust temperature — this result is not entirely surprising. We use a frequency domain, control design technique known as the Edmund’s method. The linear models are obtained using system identification techniques. We present results from a field test of the controller (implemented on a GE Frame 7E turbine) in the form of data comparing the response of the multivariable controller with that of the existing (decentralized) controller. These results clearly show that by using a multivariable controller the effects of the external disturbances can be reduced by a factor of 3 when compared with the existing design.© 1996 ASME