Stochastic and statistical analyses of the distribution of fault coverage in random-pattern testing

In this paper, we analyze the distribution of fault coverage in random-pattern testing. While our discussion here overlaps in part with several previous works, some new aspects for the evaluation of fault coverage distributions will be shown by introducing a nonparametric approach of the new statistical technique. We introduce a stochastic variable that maps the events of detection and non-detection of each fault in a random-pattern testing into the integers 1 and 0 respectively. Based on this variable, we establish a stochastic evaluation model and investigate the estimation of the distribution of fault coverage. We show that the expected number of detected faults and/or fault coverage can be estimated precisely from the test results for individual test patterns in one test sequence. We propose a computational approach for nonparametric estimation of the distribution of fault coverage. Re-sampling from the results for individual test patterns with about 2 to 10 test sequences enables us to effectively generate the histogram for the distribution of fault coverage.