Research On Intelligent Semantic Feature Modeling Based On Semantic Representation

In the Semantic Feature Modeling system, the feature not only encapsulates a great deal of geometric and topological information, but also maintains all engineering and functional information in different stages including design, analysis, manufacture and other processes of products. Thus old feature representations are hard to meet the need of high efficient management and organization of model data, and increase the difficulty of intelligent design.Based on the study of semantic feature modeling system, cellular model and other related research achievements, this paper presents a data representation of models and features for semantic feature modeling system. And it studies the use of data representation in management of semantic features, representations of model knowledge, optimization of feature parameters and conceptual design. The main contents are as follows:Firstly, to improve the ability of maintaining and managing the feature semanteme in semantic feature modeling system, a new model representation is presented, which is based on the cellular model and feature semanteme. It represents the feature semantemes in the model with parameter set, the surface creation function, attribute set and constraint set, simplifies the computation model of the feature interactive operations and validity maintenance via using semantic surface to divide all geometric faces in features into several sets in which faces have the same or similar constraint relations, simplifies the Boolean and features operations including insertion, deletion and modification by cellular model and semantic owner list, subdivides the feature surfaces accurately by rebuilding the scope of the feature semanteme and all geometrical faces in it, increases the efficiency of testing the model’s validity by judging whether the semantic surfaces are intact with cellular model. Secondly, to improve the efficiency of accumulation, update and reuse of model knowledge in intelligent semantic feature modeling system, an expression method of model knowledge based on semantic representation is presented. It creates the semantic dependence graph with the semantic representation and the constraint relations among features in the model, generates the model’s semantic knowledge via extracting the independent semantic node set and the strong relation set from model’s semantic dependence graph, adjusts the component ratio of all kinds of knowledge in knowledge base automatically via regulating the threshold of the strong relation semantic node set dynamically. And it determines the behavior type by the constraint relations between the feature modified currently and the one corresponding to behaviors, generates the behavior knowledge with the behavior type and the structures of the semantic dependence graph and then evaluates the used knowledge with results of behavior transition.Thirdly, to improve the designing efficiency of the semantic feature modeling system, a method of feature parameter optimization based on semantic representation is presented. It represents the feature model with semantic representation, reasons the value scope of feature parameters with the features’semanteme and the model’s validity, evaluates and selects models by the sub-semanteme of features, generates new model entities via swapping and mutating the features’sub-semanteme and filters the invalid models with features’semanteme.Fourthly, to improve the efficiency and level of conceptual design in semantic feature modeling system, a semantic representation based method is presented. It represents the feature as "black box" with n input and m output, describes the plan of conceptual design with the input and output terminal of the model to be solved and generates the plan of conceptual design via connecting "black boxes" with the theory of power transfer.At last, all algorithms have been verified on HUST-CAID system. And experiments show that these algorithms can satisfy the need of modeling with semantic feature, detecting of feature interaction, maintaining of model’s validity, optimizing of feature parameters and conceptual designing very well.