Research On Knowledge Modeling And Evolution Reasoning In EDSS

Emergency decision system could be regarded as an open complex giant system with characteristics of multi-autonomy, multi-factor, multi-scale and multi-variability. It includes rich and deep complicated scientific problems. Five key scientific urgent problems with Chinese emergency management research in 5 to 10 years had been proposed by academician Fan Wei-Cheng. Theoretical research and application in real circumstance given by this paper are centering on and spreading out from one of the five, that is, theory and method of multi-target emergency decision-making in the scientific problems in a complex environment.The paper is including four parts:(1) Through the analysis of typical emergency cases, the self-rethinking of great challenges in emergency decision support system (EDSS) would be done. Based on consulting a large number of literatures, trends and current issues in EDSS are summarized. Then, following the decision principle"bounded rationality"given by master Simon, this paper puts forward a three-layer emergency decision model which includes information platform layer, business layer, and user interface layer. The information platform layer is the shared data represented by multi-dimension knowledge integrated temporal logic and ontology technique; the business layer is multi-target emergency decision based on the hybrid evolutionary strategy with double-gene mutation; the user interface layer, its different designs would be given to satisfy users with different needs.(2) At the information platform layer, two mainstream knowledge representations, temporal knowledge and ontology knowledge have been modeled to increase the accuracy of decision knowledge representation.①Centering on the temporal logic, the modeling fundamental theory and representations of temporal knowledge in the emergency decision system have been deeply discussed. From the logical foundation of temporal representation, related theories such as temporal elements and temporal logic, definition and layers of knowledge, representation methodology and common temporal knowledge representations have been briefly introduced. Then, temporal knowledge in the emergency decision system is divided into two parts: one is to describe the fact; the other is to describe the event relationship. A quintuple knowledge representation has been designed to show temporal information correlated with time in this system. It also affords the formalized representation of different time information to represent sequential relationship knowledge and make decision more accurately.②Surrounding the ontology technique, fundamental theories including definition, description logic, modeling primitives, description language and editing tool have been introduced, and a modeling method, domain knowledge of emergency decision, which is straightforward and convenient, has been proposed in this paper. Then, the emergency plans and the ontology knowledge model of fire emergency have been studied and built particularly, and the applicability of this modeling method has been validated by the creation of significant ontology knowledge. This method gives a better interaction between developers and analysts, and greatly increases the normativity and accuracy of domain ontology establishment.(3) At the business layer, the evolutionary reasoning mechanism of emergency decision-making and the implement of parallel string matching algorithm have been studied in this paper to conquer the difficulty in the rapidity of automatic decision-making.①First, a novel parallel string matching algorithm based on r- continuous bits matching rule has been presented. Second, the matching probability between r-continuous bits pattern string and text string has been analyzed and computed in theory. This algorithm is implemented in cluster environment. The experiment result shows, the new algorithm can accelerate ontology matching with the increase of data size. The application of parallel string matching algorithm in ontology matching can take advantage of parallelism, and new bounding points between parallel field and ontology field would be found.②Focusing on reasoning mechanism of evolutionary strategy, a hybrid evolutionary algorithm with double-gene mutation based on (μ+λ+κ)?ES and (μ+λ) ? ES has been proposed by translating the automatic emergency decision-making to an optimize problem of finding the best decision rule parameters. There are two operators in this algorithm: Gauss and Cauchy. Gauss mutation operator is used in the parent population with the best objective function value, and Cauchy mutation operator is used in the one with the worst. A careful search is preceded. Both theories and experiments could validate that the hybrid mutation operators can increase convergence speed and accuracy on the premise of keeping population variety.(4) A multi-target fire emergency decision optimization mathematics model, which aims at the shortest response time and the minority damage, has been built. It has been proved to be valid. A multi-target fire emergency decision rapid evolutionary reasoning process has been implemented by the combination of the improved hybrid evolutionary algorithm and the ontology data. Meanwhile, technical difficulties in practice such as the case preprocessing, the ontology reasoning interface and the data operation in the algorithm module etc. have been solved.This paper has got some research findings already, but some aspects of this subject are still worth deeply studying and exploring:(1) In this paper, the establishment of knowledge model is based on the static knowledge elements, the study of dynamic description logic can be used to perfect the formalization of regularity knowledge and procedure knowledge. Then, the satisfiability of multi-dimension knowledge model has been validated in theoretic and emergency decision dynamic knowledge model is given. In this case, the knowledge completeness could be achieved by emergency decision knowledge.(2) The research and application of ontology matching become a hot topic with many challenging tasks. Further work lies in finding richer semantic matching problems by computing semantic distance, adding ontology knowledge attributes or weight factors etc. Moreover, the matching between ontology and other kinds of data models during the process of the transformation from plain text, database or XML to ontology will come under review.