Research On Fault Diagnosis Of Electronic System Based On Entropy And Graph Model

The research for large-scale systems has experienced a spike in demand due to the rapid increase of electronic systems which have more functions and complex structure. The system fault diagnosis becomes a hot topic including fault propagation, location, and test points selection. In addition, security, as an extensiton of fault diagnosis proplem, has gained more and more attention all over the world. This paper is organized as four aspects, the research of fault propagation path, test points selection and optimization, hierarchical fault diagnosis, and hardware security metrics.1. Research on fault propagation method of electronic system based on small world method. Combining the entropy and small world network algorithm, we proposed a single soure multipath algorithm which is suitable for the large-scale complex system fault path search algorithm. The entropy of the indegree, outdegree was used to sort each node. The second step was to initialize the failure node as the target node and use the chararctistic of small world network to analyze the weak node. Thus, we will get all possible fault nodes and the corresponding fault propagation path. The iterations of the algorithm is a few and also there is no additional memory space required, so time cost and space complexity is few. Entropy first search (EFS) is introduced to speed up the search process in finding the possible shortest path and then the algorithm can converges as soon as possible. In addition, the update algorithm for dynamic shortest path search is proposed for practical use. The experiments in large random sparse graphs show the efficiency.2. Research on entropy-based test points selection and optimaziton. Test points selection method based on the fault dictionary for analog circuit is widely studied. We proposed the concept of simplified entropy, which uses statistics method instead of the complicated formula of entropy. Simplifying the computation of entropy is equivalent to only 1/6 of the traditional entropy algorithm. We also prove the feasibility and accuracy of this method. In addition, multidimensional search method is adopted to find potential test points sets that can isolate the potential faults. The proposed method is also adaptive to the complexity of fault dictionary. In each iteration of the proposed algorithm, the dimension of multidimensional search could be changed according to the complexity of fault dictionary. Second, we proposed an algorithm based on improved information flow model and path searching to slove the system level test points selection. We first remove the untestable faults of the feedback loop, then use optimizing path searching algorithm to select the test points. With the iteration of the algorithm, the scale of system is getting smaller. Thus, the method can be applied to large-scale complicated system applications.3.Rearch on hierarchical graph-based fault diagnosis for electronic systems. As an important branch of fault diagnosis method, the fault diagnosis method based on graph theory is widely used and studied by many scholors. Most complex electronic systems are usually a sparse matrix after modeling, so we proposed a new type of graph representation including indegree and outdegree representation, and used the concept of the unreachable and unstarable groups to greatly reduce algorithm space. By extending the small-world network and the application of entropy, the entropy first search algorithm is introduced in finding the possible shortest path, followed by the algorithm convergence as soon as possible. In terms of the efficiency and benefits of the proposed method, they can be shown and tested in the experiments by compring with the classic and intelligent algorithm. Finally, the proposed method was applied to XX radar system, and the fault diagnosis results prove the proposed method.4. The analysis on the extension of fault diagnosis:hardware security. Hardware security is correlated with fault diagonis. Differ with other problem in fault diagnosis, hardware security is hard to indentify. In fact, security belongs to systems fault problems and become new and important areas. In this paper, we proposed a metric of system security susceptibility comparison. These evaluation results could be used to analysis the potential security risks of equipment, and comparing the effect of different attack methods. We present a theoretical comparison of power-attack susceptibility of two common DAC architectures using metrics derived from information theory. The capacitor-divider DAC also demonstrates a lower susceptibility compared to the charge-sharing DAC. Simulation results demonstrate the computation method. This method will be meaningful for the future studies.