A multiscale data reconciliation approach for sensor fault detection

When used separately on the sensor data of processes like nuclear reactors, reconciliation with fault detection and isolation strategy gives noise-corrupted estimates, wavelet transformation erroneous inferences about operating point process under conditions. Aiming to solve these challenging problems, a hybrid multi-scale scheme that combines transform is proposed in this work. The method uses steady-state framework assumption consistent algebraic relationships among coefficient data. role multivariate techniques obtaining relationships, online faults, orthogonal decomposition, coefficients demonstrated. It shown reconciled estimates obtained from very closely represent true behavior as problems respect random noise, high-frequency components due influence faults signal are alleviated. effectiveness quantitatively established when applied ex-core neutron detector advanced heavy water reactor various simulations.