| This thesis is based on the project to bulid a hall for workshop of metasyntheticengineering for correlational study of disasters. The HWME consists of expert system,knowledge management system and machine system. The primary issue to build KMS ishow to represent various forms of disaster knowledge which should be oragnized andmanaged properly by certain methods. Two kinds of representative knowledge have beenselected as research objects, one is relevance knowledge about disasters, the other isuncertain knowledge of disaster-chains. Knowledge have been represented based onsemantic network combined with appropriate theorys which are chosen judging bydifferent knowledge backgrounds.A specific semantic network has been built for research on relevance knowledgeabout disasters. To manage and mine it, a special calculating model is established bydrawing on concept of semantic similarity, which is used to quantitate the strength ofrelationships between different knowledge nodes. After making a test, the model isproved to be available and efficient.To research uncertain knowledge representation of disaster-chain, a prototype ofcause-effect disaster-chain based on sematic network has been built. By using bayestheory, a probability propagation model has been established. After making a modelingexample, we have analysed the advantages and limitations of the bayes model. Toimprove the ability of representation a special theory called soaking reasoning isintroduced. After describing soaking reasoning mechanism of problem solving, we applyit to solve the problem about representation of dynamic disaster-chain uncertainknowledge. It has been proved that soaking reasoning model is more efficient inrepresenting dynamic uncertain knowledge according to the result of programming andsimulating. At last, comparisons are made between the two kinds of models. The conclusion is that each has its own significances under different situations indisaster-chain knowledge representation. |
PDF Full Text Request