中文摘要：

失效数据常被用于评估软件的质量、监测和预测软件的运行情况,不同的测试方法对失效数据的影响是研究的重点。采用随机测试、分支覆盖测试和分块覆盖测试这3种不同的测试方法选取测试用例集,运用Daikon动态地获取程序不变量,再从这些不变量中提取失效数据,比较分析哪种方法可以获取更多的失效数据。通过实验得出结论：在3种测试方法中,随机测试方法可以获得更多的失效数据。

英文摘要：

Failure data are often used to evaluate the quality of software,monitor and predict the operation situation of software.The effect of different testing methods on failure data is the focus of the current studies.Firstly,this paper adopts three different testing methods：random testing,branch-coverage testing and block-coverage testing to select test case set,respectively.Secondly,program invariants for each method are obtained dynamically by Daikon tool,and the failure data are then extracted from program invariants.Finally,the number of failure data gotten from each method is compared.It comes to the conclusion that random testing can get the most failure data among the three methods.