Estimating the rigid component of the turbine-generator shafts torsional speed using the multisensor data fusion technique

A new approach to estimate the rigid component of the turbine-generator shafts torsional speed was developed to obtain a good filtering of the rotation speed by the weighted average method in the multisensor data fusion, based on the redundancy and mutually compensatory property of the sensing data of the multiple speed sensors distributed along the shafts. A rigid speed fusion strategy was derived from the modal decomposition of the rotation speed, which determined the optimal weighting factors by a minimax problem. The sensor fault detection and isolation technique was introduced to improve accuracy and tolerance of the data fusion strategy. The proposed approach has been used in the filtering process based on the on-line monitoring data of the torsional vibration testing for a 300 MW turbine-generator unit. It is shown that the multisensor data fusion of the rigid speed is an available counter measurement against the interaction of the shaft torsional vibration with the turbine-generator control system for its high performance and facile implementation.