Method of researching the most probable reason of a leading sign by the propagation of knowledge in a causal network

摘要

The method is used for detecting the causes of a predetermined incident, particularly for detecting the causes of a particular, especially emphasised leading sign. For this purpose, a causal network is built up from simple components, for the individual network nodes of which the probabilities of occurrence are calculated taking into consideration all relevant influences. The knowledge on the presence or non-presence of a node is propagated through the causal network by taking into consideration all events which can influence the probability of a network node. To calculate the probabilities of occurrence of the network nodes, numerical probability values are needed which are obtained as approximations from samples, which in each case are collected more favourably under the condition of the presence of a real system status and for a limited network environment. Finally, a so-called independence analysis makes it possible to determine the actual causal agent of the leading sign and, if necessary, also a causal concatenation by repeating the process.