A novel signal reconstruction strategy of multifunctional self-validating sensor

Abstract

Aiming at the desired status self-validation of traditional multifunctional sensor, a novel multifunctional self-validating sensor functional model is employed to improve the measurement reliability. Detailed self-validating functions which consist of faults detection, isolation and recovery, validated uncertainty estimation and health levels evaluation of sensors are presented, especially the proposed multivariable relevance vector machine (MVRVM)-based signal reconstruction emphasized in this paper. Being different from traditional single measured physical signal, MVRVM has expanded into simultaneous reconstruction of multiple physical variables with one sparser model. Compared with previous one output with single model, the computational burden of this paper is much lower, which benefits the on-line status validation of sensors. The working principle of MVRVM is emphasized for multiple measured signals reconstruction, which is very suitable for the final validated measurement values of multiple measured components. A real experimental system of multifunctional self-validating sensor was designed to produce the actual samples, and further verify the proposed methodology. Experimental results demonstrate that the proposed strategy could provide a good solution to the signal reconstruction of multifunctional self-validating sensors under both normal and off-normal situations.