Data Validation and Validated Uncertainty Estimation of Multifunctional Self-Validating Sensors

Abstract

A novel strategy of using polynomial predictive filters coupled with validated random fuzzy variable (VRFV) is proposed for the online measurements validation and validated uncertainty (VU) estimation of multifunctional self-validating sensors. The polynomial predictive filters-based data validation approach is applied to measured data records for multiple potential faults detection, isolation, and recovery (FDIR). The corresponding raw measurements are then validated online to avoid a disaster caused by the incorrect or poor quality ones. Further, the normal signals with sudden changes can also be distinguished from the true faults. As a good measurement practice in the self-validating sensor, VU will be associated with the validated measurements values (VMV) to improve reliability. A novel framework by means of VRFV is proposed for online VU expression, in which negative effects of different faults are fully considered. This VRFV-based uncertainty estimation method provides more confidence levels together with confidence intervals, and is also more convenient than the traditional guide to expression of uncertainty in measurements (GUM). As a more general theory, the VRFV has taken both the nonrandom and random contributions to VU into account. A real experimental system of multifunctional self-validating sensors is designed to verify the performance of the proposed strategy. From the real-time capacity and data validation accuracy, a performance comparison among different methods is conducted. Results demonstrate that the proposed scheme provides a better solution to the data validation and online VU estimation of multifunctional self-validating sensors under both normal and abnormal fault situations.