Researche On Methods Of Fault Diagnosis For Analog Circuits Based On Phasor Analysis

Although technology in the analog circuit area has been greatly developed in the last few decades, there is still a lack of design methods in fault diagnosis and design-for-testability(DFT) for analogy circuits, due to multiple factors such as its relatively more sensitive performance with respect to circuit variables and stimulations, the lack of suitable fault models, the continuously variable component parameters as well as component fault tolerance and measurement errors. This leads to a relative great portion of the cost for system development and testing in the areas of analog-digital mixed signal system design and manufacturing. Therefore, the testing and fault diagnosis of analog circuits is one of the most challenging research areas.The main objective of this thesis is to improve the testing technology of analogy circuit, with a primary focus on the fault diagnosis of analog circuits. By using the advance technologies, such as phasor analysis, multi-frequency voltage sensitivity analysis, wavelets analysis and support vector machine(SVM), this research mainly investigated the methods of fault modeling, extraction of fault characteristic and optimal diagnosis of multi-fault for analog circuits, with some innovative achievements for practical applications, which further provides theoretical supports to practical engineering problems such as the online testing and fault diagnosis for analog circuits.1. The phasor circle fault model for analog circuit is studied and improved based on the slope fault model and phasor analysis method, and the mathematic depiction of the parameters are given for phasor circle model to construct fault dictionary. The whole phasor circle model can be constructed by stimulating only a few fault sampling data, which helps to reduce the construction complexity of fault modeling significantly and also optimizes the Fault Detection Ratio(FDR) of both parametric faults and catastrophic faults for analog circuits.2. With the considerations on the fault isolation and ambiguity groups of the phasor circle fault model, this thesis conducted a research on the formation cause of ambiguity groups, and the classical method of DFT by bypassing. Meanwhile, based on the idea of redundant design, the novel method of DFT by bypassing redundant components is proposed, which effectively solves the problems of ambiguity groups in fault diagnosis of analog circuits.3. Owning to the parameter tolerance and testing error, the internal relation between phasor circle model and slope fault model is studied. And then a fault modeling method based on the multi-dimensional complex space is proposed. This method combines the phasor circle fault model and the slope fault model. It not only increases the distances between different fault trajectories and improves the Fault Isolation Ratio(FIR), but also provides new ideas for multiple-fault modeling and diagnosis in analog circuits.4. The computational method of Mahalonobis distance for general normal distribution is investigated, based on the theory of mathematical statistics and principal component analysis(PCA) method. Then, combining with the phasor analysis method, using the real part and imaginary part of response signal as individual sample data, two weighted algorithms of Mahalonobis distance based on information entropy and the contribution of principal component variance are proposed. The simulating results on benchmark circuit show that both methods are helpful to solve the challenging problems of fault characteristic extraction in analog circuits.5. This thesis also investigated the multi-fault diagnosis technology of analog circuits, and then an improved fault dictionary diagnosis method is proposed based on multi-frequency sensitivity. In this method, fault dictionary is constructed based on the sensitivity relationship between the response voltage on a single test point and the parameter of limited potential fault components. Furthremore, the Hybrid Invasive Weed Optimization(HIWO) algorithm for SVM is studied and improved, on this basis, by using wavelet package technique to extract fault characteristics, this thesis proposes a new multi-fault diagnosis method of analog circuits, which improves the effects of fault diagnosis.