Research On Fault Diagnosis For Analog Circuits Using Uniform Node Voltage Increment Ratio Vectors

Although many remarkable achievements have been established in the field of analog circuits fault diagnosis, however, it is lagged behind the fault diagnosis of digital circuits. In last decades, many diagnosis methods have been successfully developed for analog circuits. It is well known that fault dictionary is the most valuable approach among many diagnosis methodologies and will be widely used in the future. For fault dictionaries, the evident character is their simplicities. But, this is no yet the case for the constructions of these dictionaries. In general, there are three major defects of those classical fault dictionaries, (1) the compromise between fault coverage and dictionary length, wider fault coverages needs bigger dictionaries, (2) disability of parametric fault diagnosis, many dictionaries can diagnose hard fault only, and (3) inaccuracy with tolerances. On the other hand, the artificial intelligence techniques using has become more and more important in the field of automating fault diagnosis in these last decades. Based on circuit theory, geometric theory and fuzzy techniques, this paper address itself to linear analog circuit diagnosis by means of using Uniform Node Voltage Increment Ratio vectors as fault signatures to overcome the defects mentioned above, and the research results are as follows.1. The classical DC and AC dictionaries are further studied and their major limitations are pointed out.2. The concept of Fault Couple is proposed and used for test nodes selection. Three algorithms are designed for test nodes selection and faults isolation respectively. Compare to the nodes selection with ambiguity sets, less test nodes numbers are required by using the concept of Fault Couple. Consequently, the dictionary length can be reduced. With the reduction of the dictionary size, both the storage space and the diagnosis time are saved.3. The concept of Uniform Node Voltage Increment vector is proposed. It is proved that the Uniform Node Voltage Increment vectors are only determined by the locations of the faulty components, and are independent of the parameter deviations of the faulty components. From the Uniform Node Voltage Increment vectors, the Uniform Node Voltage Increment Ratio vectors are deduced. Except all characters of an Uniform Node Voltage Increment vector, an Uniform Node Voltage Increment Ratio vector has the benefit of the independent of parameter deviation direction. Consequently, an uniform fault dictionary can be constructed for both catastrophic and parametric faults diagnosis with those Uniform Node Voltage Increment Ratio vectors. The Uniform Node Voltage Increment Ratio dictionary has the prominent advantages of wide fault coverage and capability of diagnosis for those un-predefined faults.4. During these last decades, automating fault diagnosis using artificial intelligence techniques has become an important research field. This paper develops some of these techniques in order to fill in the tolerance gap. In case of considering tolerances, the precision will be decreased dramatically. First, some points of view in geometry are extended to analog fault diagnosis, and four geometric models are designed for faults isolation. Second, fuzzy techniques are used and two fuzzy diagnosis models are designed. Either the geometric models or the fuzzy models are based on the Uniform Node Voltage Increment Ratio vectors. The promising experimental results manifest the effectiveness and the correctness of these models.