Abstract

Fault Trees (FT) are widely used for reliability analysis. Static FT who have no temporal and/or functional failure dependencies are solved by using Binary Decision Diagrams. A FT extended with gates capable of modelling temporal and/or functional failure dependencies is called a Dynamic FT. A Dynamic FT is usually solved by a conversion to a Markov Chain (MC). Unfortunately a MC will have an effect called state space explosion. This means that the number of states of the MC can become extremely large. This is an undesirable effect. Bayesian Networks (BNs) is a different method to calculate unreliability. This method requires less states than a MC and is an attractive alternative to compute unreliability. But since most systems are drawn as (D)FT, it would be nice if the user can still use (D)FTs but calculate the unreliability via BNs. The constructed tool (called IDIC) provides this functionality. The user can draw his FT in the GUI of the tool and by pressing a button he lets the program calculate the unreliability of his system by translating the FT to a BN and analysing it.

IDIC consists of two main parts, the conversion and the GUI. In the GUI the user can draw several FT constructs (eg. several types of gates and basic events) with the help of the package Jgraph. These constructs can be edited, deleted, moved and added to a central drawing area. Saving and loading functionality is present, therefore drawn FTs can be stored and loaded for future use. The conversion is a step-by-step process. The first step is reading a specific formatted file with the help of the ANTLR package. ANTLR uses the information from the file to construct a standard formatted BN. This BN will be analysed by the package SMILE. The BN is translated from standard format to SMILE format and analysed. IDIC will read the drawn FT, convert it to a BN and analyse it. The results, which is the unreliability per state, is displayed in the GUI.

IDIC has been tested via two case studies. The results of these case studies are good. Compared with the tool Galileo, which uses MCs to analyse the FT, the results are similar. The differences are smaller than a thousandth in one case and a millionth in the other. But the time needed is the key factor here. The first case required more than 8 minutes for Galileo to solve using MCs, IDIC only uses 10 seconds to solve this case. The other case study was similar in computation time.

IDIC is a prototype, the algorithms, data structures and packages used are not the fastest, least memory/space consuming. They are however easy to use, intuitive and well documented. Therefore although IDIC is functional, there are several ways to improve several components. This thesis lists a few of those possible improvements.

Additional Resources

1. The paper