The -equivalence class approach to analog fault diagnosis problems

A new linear algorithm is given for branch fault diagnosis problems of analog circuits. It is proved rigorously that the algorithm is as powerful as any nonlinear algorithm with probability one. The algorithm is applied to several nontrivial examples and it works very well. A new numerically reliable method is also given of checking if a vector belongs to a linear subspace, which is an important part of the algorithm as well as the existing linear algorithms.