Grey bootstrap method for data validation and dynamic uncertainty estimation of self-validating multifunctional sensors

The accuracy and reliability of multifunctional sensor outputs directly influence the running state and performance of measurement and control systems in chemical processes. Given their importance, self-validating multifunctional sensors are presented to improve the reliability of measurements in operation. A novel strategy based on the grey bootstrap method (GBM) is proposed for the online data validation and dynamic uncertainty estimation of self-validating multifunctional sensors. The data validation algorithm and the working principle based on GBM are applied for multiple faults detection, isolation and recovery (FDIR). The proposed FDIR scheme can simultaneously isolate multiple faults of multifunctional sensors and accomplish failure recovery with high accuracy and good timeliness. Moreover, it has a good performance of discriminating between fault-free signals with sudden changes and undoubted faults. On account of the unknown probability distribution and small sample size, the traditional expression of uncertainty has limitation in dynamic measurements. As a data-driven method, the GBM can evaluate the measurement uncertainty from poor information without prior information about the probability distribution of measurand in real-time. The performance of the proposed strategy is verified by computer simulations and a real experimental system of chemical gas concentration monitoring. Through the comparison of different methods, the results show that the GBM has superiority for the data validation and dynamic uncertainty estimation of self-validating multifunctional sensors.