| Abundant process knowledge is accumulated in discrete manufacturing system,which has the characteristics of multi-source,heterogeneous and high redundancy,which makes the process of process knowledge discovery and reasoning in discrete manufacturing system difficult.Most of the existing researches focus on knowledge discovery and the optimization of reasoning objective function,do not fully consider the feature reuse in the process of reasoning optimization,and the utilization of massive process data is not perfect.Starting from the representation of process knowledge,this paper proposes a typical processing route clustering method based on process coding,excavates the internal implicit relationship between process routes,and puts forward a knowledge reasoning model based on knowledge map on the basis of clustering results,extracts the features of process knowledge from text input information,and matches the features of the constructed process knowledge map,Based on the results of knowledge reasoning,a knowledge driven process parameter optimization method is proposed.The specific research contents are as follows:(1)The process knowledge composition of discrete manufacturing system is analyzed,the definition of typical process route is proposed,and the process route information model is defined according to the information contained in the process route.Aiming at the problem of multi-source heterogeneous process knowledge discovery,the process route data are preprocessed such as specification and elimination.On the basis of process coding,the similarity between different process routes is studied,and the similarity matrix is used to describe the similarity between different process routes.Based on the proposed similarity decision factors,a typical processing route discovery method considering clustering distance is proposed.Based on the clustering results,a clustering cost evaluation strategy is proposed to improve the clustering effect of typical routes and ensure the effectiveness of process route clustering.(2)Considering that the knowledge map has the characteristics of efficient retrieval and intelligent reasoning,according to the above typical process route mining results,a process route oriented knowledge map construction method is proposed.A three tuple process knowledge representation model is established to describe the mapping relationship between process knowledge ontology and knowledge map.On this basis,part knowledge map,feature knowledge map and machining process knowledge map are constructed.Further,a knowledge reasoning model of extraction fusion matching is proposed,which comprehensively considers the content of input text information,and generates reasoning results about machining features and machining processes under the feature matching of part features and manufacturing requirements.The experimental results show that the proposed knowledge reasoning model has better reasoning accuracy and realizes the knowledge query reasoning of text information.(3)Considering the internal relationship between process parameters and machining process of discrete manufacturing system,a self-organizing mapping optimization algorithm integrating collaborative training is proposed.In the process of process parameter optimization,the randomly generated sample particles are often used to optimize the objective function,without considering the correlation between the existing process parameters.To solve the above problems,under the framework of self-organizing mapping algorithm,the superior information of process parameters in evolution is regarded as a knowledge unit,the reuse effect of existing process parameter knowledge units is calculated through hyperbolic space distance,and the process parameter category center with differential information is formed by collaborative training to realize the reuse of process parameter history information.Based on the proposed algorithm,the energy consumption problem of discrete manufacturing system is solved.The effectiveness of the method is verified by comparison and simulation experiments. |
