Generating effective test data to show that a faulty structure could produce a failure poses many complexity problems. Code
reviews are known as an effective method for detecting potential critical program structures. But detecting this structures
in a code review takes a lot of effort and is error prone.
This paper presents a method to detect potential faulty structures by program modeling. Therefore we build a database containing
informa- tions about classified faults. For analyzing the program we generate a first order predicate logic model implemented
in horn clauses. We show how to detect the fault structures defined in the database and how to reduce the effort for the code
review by slicing the relevant parts of the program.