12.226. Many code packages are available for the development and analysis of system logic models. Many are specialized depending on the application desired. Although a discussion of such applications is beyond our scope, it is useful to mention several characteristics of fault trees that affect modeling.

12.227. The concept of cut set is important in fault tree analysis. In a collection of basic events, called a cut set, if all of these basic events occur, the top event is guaranteed to occur. In words, a cut set is defined as a set of system events that, if they all occur, will cause system failure. For example, in the simplified fault tree shown in Fig. 12.16, if we have the rupture of both headers A and В, the top event will occur and we have a cut set. A path set is complementary to a cut set in that it consists of a group of events (or failures) which must not occur in order to ensure that the top event will not happen. The path set usually includes events in addition to those in the cut set.

12.228. Since a large system may have thousands of cut sets, it is nec­essary to simplify the analysis by identifying and eliminating subsets which tend to duplicate simpler logic paths and therefore are not essential. There­fore, one class of computer program is used to generate minimal cut sets, usually by Boolean manipulation techniques. A classic code for this pur­pose, known as qualitative evaluation, is MOCUS [29]. Although various codes are available to assist in the actual construction of fault trees by using automated procedures, some level of manual involvement is usually required.

12.229. The next requirement is to carry out a quantitative evaluation by introducing probabilities to the logic gates. For example, an extensive data base exists for the reliability of mechanical and electrical components which can be used in a suitable computer model to calculate the failure probability of the top event. As pointed out in §12.217, the probabilities of each event in a cut set are multiplied together which could result in some very small values to be added at OR logic gates. Thus, some sim­plification is possible. A typical code for quantitative evaluation is WAM — BAM developed by the Electric Power Research Institute [30].

12.230. Various other codes are available for uncertainty analysis and consideration of common cause failures. Since often no data base is avail­able to establish probabilities for many logic gates, expert system proce­dures have been used in recent studies as described in the following section. At any rate, the establishment of confidence levels for the results continues to receive major modeling development attention.