Research On Fault Fuzzy Clustering Method For Reducing False Alarm

The BIT technique is widely used to improve the test and diagnostic capability of the system. However, the high false alarm rate (FAR) existing during the test process becomes one of the major factors to limit the application of BIT. Consequently, the priority research area of BIT technique focuses on the reducing the FAR which is also the bottleneck breakthrough of BIT technique.In the study, all the work is performed to reduce the FAR of BIT. The detailed work is as followed.1. The research foundation of this paper is Markov triple-state model which can increase total fault diagnosis rate (TFDR) and reduce the false alarm rate (FAR). Based on the theory of this model, the states of a system are divided into health state, quasi-fault state and fault state.2. According to the Markov triple-state model, the key work concentrates on the determination of the fault threshold. A new objective functions is firstly created considering the main factors of fault diagnosis, including TFDR, FAR, running cost, running risk, maintenance cost. The most optimum fault threshold is obtained by solving the objective function and some examples are used to prove the accuracy of the solution.3. The fuzzy equivalent matrix based on transitive closure and F statistics are applied to sort the standard fault samples and standby detection samples. In the meanwhile, the fault type and the reason of standby detection samples are obtained and the fault causes of quasi-fault state and fault state are also realized using this method.4. The theoretical method of variable average n is presented to solve the isolation of transient fault, intermittent fault and false alarm in the case of discrete time and discrete state by solving some examples. It is also available to decrease the false alarm caused by transient and intermittent fault.5. The Spool Fault Diagnosis Platform is realized with MATLAB. It can be concluded that the Spool Fault Diagnosis Platform and the methods used in this paper are available and reliable through the experimental results of spool.