Research On The Method For Transforming Manufacturing Data Of Aircraft Parts Into Knowledge

The type,amount,and quality of knowledge in the manufacturing knowledge base are the key to achieve the intelligence of process planning.How to effectively and quantitatively evaluate the capability of knowledge in the knowledge base and continuously obtain the requisite knowledge to promote the continuous improvement of knowledge capability are two important issues that need to be solved urgently.In order to solve these two key technologies,this paper carries out the research on knowledge transformation from manufacturing data of aircraft parts based on the knowledge capability measurement.The main work includes the following aspects:(1)The framework for transforming manufacturing data into multi-granularity knowledge is proposed.The composition of manufacturing data and the knowledge contained in the manufacturing data are analyzed respectively from the organizational form and the relationship between the various elements.After that,the multi-granularity knowledge model is established to clarify the internal relations and enhance the generic reusability of knowledge.Then,the generic process is presented to transform single granularity manufacturing data into multi-granularity knowledge,i.e.,analysis of manufacturing data,identification of target knowledge and construction of new knowledge.At the same time,the knowledge coverage is defined by the knowledge capability to conduct quantitative evaluation,which provides reference for further requirement of knowledge transformation and improves the efficiency of knowledge acquisition.(2)The method for transforming fabrication order data into multi-granularity knowledge is developed.A multi-granularity fabrication order design knowledge model is established using the method of tuples.Based on the generic process of knowledge transformation,specific implementation steps are established: data disassembly,state determination and knowledge construction.With regard to state determination,similarity measure methods for different granularity knowledge are proposed to reduce the redundancy in the transformation process,including similarity calculation for attribute value of the different data types,similarity calculation for the feature vector,and similarity calculation for the process flow.As a novel approach,combing sequence alignment with edit distance is proposed to calculate similarity exactly between two process flows.Due to the consideration of more comprehensive influencing factors than other existed methods,the algorithm has the higher accuracy.For the construction of new knowledge,according to states of part and process flow,the corresponding knowledge construction process is established to realize the preservation of different granularity knowledge through the operations of creation,replacement,and reference.(3)The method of knowledge transformation from unstructured detection data is established.Based on the analysis of the compositional structure of test data and target knowledge,the generic process of knowledge transformation is further refined,including four steps: structural transformation,relation mapping,state determination,and knowledge fusion.In view of the structural transformation,three steps are constructed through comparison analysis of point cloud model and design model,point cloud data fitting,and XML(Extensible Markup Language)file transfer.For state determination,a method of calculating similarity based on feature sensitivity is developed,in which the weight of each feature is obtained by the Sobol method.Compared with other similarity calculation methods,the advantages of this method with high matching accuracy are illustrated.With respect to knowledge fusion,the mechanisms of redundancy elimination and conflict resolution are established to ensure the uniqueness and consistency of knowledge.(4)The quantitative measurement model of manufacturing knowledge capability is constructed.The evaluation of knowledge capability is mainly characterized by knowledge coverage.On the definition analysis of knowledge coverage,the calculation method is provided with two mainly steps.First,the total number of expected knowledge unit is determined through the establishment of classification relationship among discrete-valued features and the discretization of continuous-valued features.Second,the distance between knowledge units is analyzed for effective coverage.Based upon the quantitative measure result of knowledge capability,the strategy of improving knowledge capability can be effectively developed according to the expected goal.