Analog fault diagnosis of actual circuits using neural networks

Fault diagnosis of analog circuits based on wavelet packets

“…In this implementation, the difference of the LP and HP offers a coarse resolution illustrating the basic structure of the FS and fine resolution showing the finer detail of the FS, respectively. At each level, the approximations and details are regrouped into a pyramid structure known as the Laplacian pyramid [10,11]. , then…”

“…Unlike the wavelet transform [10,11] which gives octave band decompositions of the frequency axis, the wavelet packet analysis can apply the refiner tree whose structure depends on the signal properties, and among these, the full tree divides the frequency axis into equal-bandwidth components yielding a linear division of the spectrum similar to the short-time Fourier Transform (STFT).…”

“…However, little literature has been available in the fault analysis and diagnosis of analog circuits using wavelets. References [10,11] proposed a neural network based fault diagnosis method using 0-7803-8560-8/04/$20.00©2004IEEE wavelets as a preprocessor to reduce the number of input features to a manageable size. Unfortunately, the decomposition based on wavelet transforms [10,11] of a signal into approximation signals and detail signals is limited to the approximate signals.…”

Diagnosis of Incipient Faults in Nonlinear Analog Circuits

“…It is more difficult to diagnose soft faults than hard faults because the features of the soft fault cases of the CUT are not significant. In recent years, many methods such as the fault dictionary method [5,6], the neural network [7][8][9][10], fuzzy analysis [11,12], the FNLP method [13], the wavelet preprocessing [14], the test-point node selection [15] , PCA method [16] and the support vector machine algorithm [17][18][19] have been presented for fault diagnosis of analog circuits.…”

Probabilistic Graphical Models for the Diagnosis of Analog Electrical Circuits

“…Examples are: the fault dictionary approach, which collects a set of common or relevant faults and associates them with (sets of) measurements by which they can be identified [2], the model-based diagnosis of digital circuits based on constraint propagation and an assumption-based truth maintenance system (ATMS) [8], and the simulation of a circuit for different predefined faults to generate training data for a classifier, for example, an artificial neural network [1,13]. In particular the diagnosis of digital electrical circuits is well-developed.…”