Maximum Likelihood Virtual Sensor Based on Thermo-Mechanical Internal Model of a Gas Turbine

This paper proposes an iterative double-step method for the modeling of a gas turbine (GT). The approach is presented to address a specific application with a great industrial impact, however, it can be extended to different physical systems. As regards GT modeling, the first step is based on generalized maps and thermodynamic laws that allow an algebraic static estimation of flows, temperatures, and pressures of each GT section. The second step is based on a Kalman filter (KF) that corrects these static estimations exploiting all available measurements and introducing thermodynamic and mechanical equilibrium. The model has been trained and validated on a massive data set created using a numerical propulsion system simulation (NPSS)-based design tool, which contains the turbine geometrical and mechanical data. Finally, the quality of the model has also been evaluated by exploiting field data taken from existing plants.