Design of Controlled Random Tests with the Given Hamming Distance

The article considers the issues of testing computing systems and their components. It shows the low efficiency of probabilistic testing due to the failure to use the available information about both the test object and previously generated test sets. It is noted that it is promising to use information about previous test sets when constructing controlled probabilistic tests. A class of controlled probabilistic tests with a small number of test sets is identified and studied. Known standard controlled probabilistic tests with a small number of test sets are analyzed. A method for generating controlled probabilistic tests with a given Hamming distance is considered, the basis of which is the representation of binary test sets as a set of symbols of other number systems. Providing a given value of the Hamming distance for a controlled probabilistic test in a non-binary number system guarantees the same value for the binary interpretation of the generated test sets. The efficiency of the proposed test construction method is experimentally analyzed and confirmed for the case of testing memory devices for the presence of complex faults of mutual influence.