A Multivariate State Estimation Technique and Multilayer Forward Neural Network Framework on Steam Turbine Valve Flow Curve Simulation

Abstract

The Multivariate State Estimation Technique (MSET) is firstly implemented as the fault detection technique for monitoring parameters of the steam turbine valve. Second, healthy operation states are remained to form a preliminary dataset for the valve curve simulation. Then, the Forward neural network (FNN) based piecewise linear fitting method is used to simulate the steam turbine valve flow curve. With a MSET-FNN framework proposed, the simulation accuracy will be enhanced by removing the unhealthy operational states. In addition, the simulation process will gain a speed up with the utilization of GPU.