Research On The Reliability Guarantee Mechanism Of The Internet Of Things Monitoring System

With the development and popularization of Internet of Things technology,the reliability of Internet of Things system has become a concern.In the Internet of Things system,wireless sensor network is an important part of it,which is mainly used to collect various environmental parameters.The failure of sensor nodes will lead to unreliable Internet of Things monitoring system.Data are the basis for making scientific decisions and providing services.If the quality of data is poor,the decision may be unreasonable.Ensuring the data quality is the basis for ensuring the reliability of the Io T monitoring system.This dissertation mainly focuses on the reliability guarantee mechanism of its monitoring system.Considering the Io T-based haze monitoring system as the background,the reliability of sensors under different fault-tolerant models is analyzed,and an outlier detection algorithm based on MARS and probabilistic programming for Io T-based monitoring data is proposed.The dissertation is summarized as follows.1.The fault tolerance mechanisms of different backup structures and different backup modes are proposed.In the Internet of Things monitoring system,sensors are easily to fail when deployed in harsh environments.In order to ensure the reliability of the system,different fault tolerant models of sensor structure are proposed,and fault tree and Markov chain are selected for reliability analysis.The simulation results show that compared with the hot backup redundant system,the cold backup system is more reliable.From the perspective of system reliability,heterogeneous substitution scheme is superior to homogeneous scheme.2.To improve the data quality of Internet of Things,a multi-variable outlier detection method based on Multivariate Adaptive Regression Spline(MARS)model and probability programming is proposed.This method can combine multiple variables and detect outliers through a model.Firstly,a multivariate adaptive regression spline model is built to generate the research residuals.Then,a general univariate outlier detection model based on full Bayesian inference is established by using probabilistic programming method with residual as input.The results show that the probabilistic programming model can detect more accurate outlier points,and the model provides a probability distribution with confidence interval.