Fault Analysis Of Circuits Based On Data Mining

With the development of science and continuous improvement of automation technology, small to people’s daily lives, large to aviation, aerospace, military defense, electronic devices become increasingly widespread, its safety and reliability are gotten more and more attention. And the reliability of electronic equipment is directly related to its own circuit system, it is necessary to do some fault analysis of the circuit system for the purpose of improving itself reliability and security by detecting whether there is fault in the system, if there is, we need to locate the failed components position which could guide us to the fixes and improvements of the circuit system. Besides, systematic and objective circuit failure analysis can guide the design of the circuit system to a certain extent, has very important significance.The main idea of this paper is to process normal and fault simulation data information obtainning through circuit simulation using Binary k-means clustering algorithm, and then classify the potential failure modes of the test circuit system through an improved nearest neighbor classification algorithm and classification algorithm based on rules, and finally propose a method of locating failure component through multiple probe points, which can solve the problem of positioning of fault components.The main contents are as follows:Firstly, this paper introduces the basic idea of data mining techniques, including the basic process of data mining, construction and preprocessing of the data set, the existing common classification methods and clustering methods. In addition, some of the concepts and analytical methods of the circuit fault analysis are described from the perspective of digital circuits, analog circuits and digital-analog hybrid circuits.Secondly, this paper introduces ways of normal and fault simulation of circuit system, including the methods of motivation design, the modeling of normal and fault components,pretreatment of normal circuit simulation, fault injection technology and storage means of simulation results.Thirdly, this paper do some analysis to the simulation results using the ideas of data mining, including data extraction from the CSDF data files and parameterization of the extraction data, on this basis, this paper does some unsupervised classification to the potential fault simulation results using Binary k-means clustering algorithm, then combining the results of classification with the observation of the actual output waveform, builds a training set and applies this data set in an improved nearest neighbor classification algorithm and classification algorithm based on rules, which can construct potential fault classification models for the test circuit system. Based on the previous work, applying a locating method of failure component through multi-probe points proposed in this paper to the sample circuit, verifying this method has a good performance in the terms of accurate positioning of fault components.At the end, this thesis introduces the software implementation of circuit fault analysis based on data mining, to some extent, this software can improve the effiency of fault analysis personnel.