Research On System Reliability Based On Reachability Analysis And Distance Algorithm

The computer systems bring us convenience,but also are threats to our lives.Therefore,it is necessary to research on the reliability of computer systems.In this thesis,we mainly study the reliability of two types of computer systems: the reliability of the hybrid systems and the telemedicine systems.The hybrid systems are widely used in the military equipment and industrial automation field.It will be a threat to our lives if a system malfunction occurs in these systems.The purpose of the reachability analysis of hybrid systems is to verify if the hybrid systems can operate safely in any environment.Wearable devices combined with the telemedicine systems are used to monitor patients in real time.We detect the data transmitted by the wearable devices to ensure the accuracy of diagnosis.Therefore,it is important to ensure the accuracy of data transferred by wearable devices.For the reliability of hybrid systems,the concept and application of hybrid systems are introduced firstly.Then,several hybrid system models such as hybrid automata are introduced.Afterwards,we summarize two types of formal validation techniques: theorem proving and model checking.Formal verification is an important method to ensure the reliability of hybrid system,and reachability analysis is the basis of formal verification.We formalize hybrid zones for rectangular hybrid systems reachability analysis.The hybrid zones are constrained by a series of linear inequalities which satisfying the strict restriction condition.The hybrid zones are proved to be closed over the state reachability operations for hybrid system.Finally,the reachability analysis algorithm is proposed.For the fault detection of wearable devices,the reasons of data exceptions and several traditional anomaly detection methods are introduced firstly.Then,the method of fault detection based on distance which uses time series,kernel density estimation and Hidden Markov Model is proposed.Finally,the simulation results show that the proposed method is an effective algorithm and it possesses better detection accuracy with a low false alarm rate by comparing with the method based on Mahalanobis distance.