Prediction of Sensor Faults and Outliers in IoT Devices

Abstract

Internet of Things (IoT) is tremendously growing and interacting with the physical world in the era of Industry 4.0. In near future, billions of companies will have advanced communication technology and it will increase the growth of critical systems. The accuracy measurement of the functionality of a critical system is a challenging job. Fault Tolerance (FT) is a major concern to ensure the dependability, availability and reliability of critical systems. Faults should be predicted and controlled proactively to lessen failure impact on the critical systems. To predict these failures and use the relevant procedure to avoid it before it actually occurs, FT techniques are used. These techniques are implemented in critical systems to avoid failures as the security of systems is more important than the reliability of systems. It minimizes the effect of faults that are being investigated. FT techniques work on a concept that if the system is built differently then it should fail differently. If a redundant variant fails then atleast the other one should give a satisfactory result. This study exhibits an analysis of existing FT techniques like N-version programming, Recovery blocks and N-self-checking programming. A critical study of sensor faults and outliers prediction models in IoT is presented. A bibliometric analysis is also carried out on 716 Scopus indexed publications to analyze the current research trends in this domain.