Research Of Key Problems Of Knowledge Based Engineering On Product Design

Knowledge is the basic of artificial intelligence, so artificial intelligence problem solving is based on knowledge. They are the key problems to study in area of artificial intelligence, how to acquire knowledge from real world, how to reasonably represent captured knowledge as computer code for convenient storage, and how to reason with the knowledge to solve real problems, i.e., knowledge acquisition, knowledge representation and reasoning using knowledge. In 1990’s, computer technique, artificial intelligence and CAD developed and integrated to form the research of Knowledge Based Engineering (KBE). And soon, KBE has become one of the research highlights both in science circles and enterprise circles.The thesis analyzes the key problems of knowledge engineering, and studies deeply about knowledge modeling, knowledge representation, knowledge reasoning and knowledge reproduction. The main innovations include:(1) QFD-based knowledge modelingTake QDF as the main thread, use scenario analysis, KJ (Affinity Diagram) method and AHP method, construct QFD-based knowledge model for product design. Translate customers’requirements into technical factors, and map the value ordering into technical method choosing, so, satisfactory design plan can be determined.(2) Object-oriented and frame-based knowledge representationThe complexity of product design needs not only numerical calculation but also non-parameter knowledge such as heuristic experience. Normal single knowledge representation cannot represent all the needed knowledge comprehensively and effectively. The thesis proposes object-oriented and frame-slot-based product design knowledge representation, to integrate the advantages of object-oriented and frame paradigm.(3) Case-based and rule-based reasoningThe thesis proposes intelligent knowledge reasoning mechanism for product design to solve the problem of technical difficulties and long design period. First, based on the product parameters determined by QFD knowledge modeling and expertise, use Case-Based Reasoning to search the case base for product design plan which is most similar to the technical parameters by QDF. At last, use rule-based reasoning to modify and determine product design plan, based on the rules in rule base by experts and design engineers.(4) Knowledge reproduction using genetic algorithm based on chaotic switchOptimal design of complex product is a hybrid, discrete, multi-variable, non-linear and multi-target and planning problem. Generic algorithm is a typical intelligent calculation method, suitable for non-linear problem. But due to its limitation, it is very difficult to optimize product design. The thesis introduces chaos to generic algorithm, to use chaotic searching effective gene. Use chaotic iteration, search individuals gene base dynamically, vary the found alleles, and reproduce a brand new individual.Modern product design is a creative knowledge-driven process, including knowledge inheritance, integration, creation and management. On the application side, the thesis takes transformer intelligent design as a case, designs the object models of transformer main part and component parts, data and attributes models represented as slots, designs case representation of product and component parts models, indexing mechanism, similarity model and case searching algorithm, designs rule-based reasoning network, rule representation and rule-based transformer design case modifying algorithm. At last, the thesis gives transformer optimal design model, proposes a solution model of genetic algorithm based on chaotic switch, tested with real case.