A multifunctional self-validating sensor

Multiple different quantities are often measured by the multifunctional sensor in industrial applications. The sensor may not be truly healthy or fault-free, which could reduce the reliability of measurement results. In this paper, a new prototype of multifunctional self-validating sensor was put forward, and detailed intelligent functions were illustrated and generalized. Specifically, we 1) employ the multivariable relevant vector machine (MVRVM) coupled with faults detection, isolation and recovery (FDIR) technology for the final validated measurement value (VMV), in which the polynomial predictive filters with low computation complexity is used for the data validation and then the incorrect measurements are validated or corrected on line. This MVRVM is very suitable for multiple measured components; 2) propose a novel random fuzzy variable (VRFV) based uncertainty evaluation strategy for the on-line validated uncertainty (VU) estimation, in which negative effects of different faults are fully considered. As a more general theory, VRFV has taken both the nonrandom and random effects into account; 3) define some measurement value status (MVS) to imply how the VMV are obtained, which enhance the security in application. The methodology presented is illustrated on an experimental system for hydrogen concentration, temperature and humidity, and results demonstrate that the proposed scheme provides a good solution to the status self-validation of the multifunctional self-validating sensor under both normal and abnormal faults situations.