A general method for analog test point selection using multi-frequency analysis

An important problem that arises in analog testing for fault dictionary approach is the test point selection. It consists of selecting a minimum set of test points to achieve the highest percentage of isolation. By using the concepts of ambiguity set and integer-coded dictionary, it can be formulated as a combinatorial optimization problem. On the basis of the formulation, this paper develops a general and accurate method for test point selection. This is achieved by presenting a new ambiguity set partition method based on hierarchical clustering and an improved entropy index method. Besides, multi-frequency analysis is also incorporated in the described method. The proposed method is tested and compared with exhaustive search through its application to two benchmark circuits. The results indicate that, compared with the extant methods, our method search for a test point set that not only includes a minimum number of test points but also has a good performance in terms of diagnostic accuracy and fault isolation rate. Meanwhile, the proposed method can be extended to deal with the problem of test point selection for some specific fault model.