Research And Application Of Fast Hybrid Intelligent Model Based MMDB

For the problems of difficulties with knowledge acquisition and management,reasoning not quickly and lacking of self-learning in expert system, in-depth analysisand researches for key technologies of Main Memory database (MMDB), rule-basedknowledge representation and retrieval, rule-matching algorithm, learning algorithmbased on artificial neural network. Moreover, the corresponding algorithms andstructures are optimized. A fast hybrid intelligent model based on MMDB, expertsystem and artificial neural networks is and proposed to solve the problem ofintelligence applications timeliness required in the fields of engineering controls, quickdecision-making telecommunications and financial.Because traditional database can not meet the requirement of effectiveness of dataprocessing in time-critical applications and the memory have increased capacity andlower prices, MMDBs are more and more widely used as foundation of a real-timedatabase. An optimization T-Trees and concurrency control techniques of T-Trees areproposed. Data organization and management, index technologies and concurrencycontrol are analyzed. Some experiments are made to show the higher performance ofoptimized algorithms.The architecture of a rule-based expert system, the knowledge representationmethods of expansion production, as well as rule matching algorithm and reasoningstrategies are studied. Three reasoning mechanisms and comprehensive function-basedconflict resolution mechanisms is analyzed. Uncertainty reasoning technology, as wellas the rules of consistency maintenance algorithm is studied. The most widely used rulematching algorithm-RETE algorithm is analyzed. The core idea and process of the fastmatching algorithm is introduced. An optimized RETE algorithm is proposed to reducethe space and time complexity of the matching process, enhanced the efficient ofreasoning. Combined with the rule-based expert system, MMDB and artificial neuralnetworks, a fast hybrid intelligent model is proposed. The structure and principles of themodel and a hybrid knowledge representation are analyzed. Using the hash index andimproved T-tree index in MMDB to improve knowledge retrieval capabilities in processof rule matching. The BP network uses data to training and can be pruned. And then therule can be extracted from the network to make up for the lack of self-learning ability ofthe expert system.Because there is much kind of data that have complex relationships, a two-stagemission assessment program based, the hybrid intelligent model is proposed to obtainthe status of the aircraft during the aircraft mission. In the evidence acquisition phase,the technology of distributed access to data is used, and then uses different models toobtain evidence. In reasoning stage, it brings a variety of evidence (telemetry, orbit,images, etc.) together and obtains conclusions of the assessment by reasoningcomprehensively with fast hybrid intelligent model. The result of the simulationexperimentation shows that: the system can fulfill the real-time assessment task.