Research On Digital Twin Modeling Of Casting Process Driven By Data And Knowledge

The foundry industry is an important industry in the national economy.However,while the foundry industry creates huge economic wealth,it also consumes a lot of resources and has a serious impact on the environment.Therefore,the implementation of green manufacturing projects and the green transformation of the foundry process have become a key strategy for the sustainable development of the foundry industry.However,the current overall situation is that there is still a lack of effective means that can effectively identify the key links and key influencing factors of energy conservation and emission reduction in the process.At the same time,the phenomenon of information flow faults in the production system is widespread in enterprises,which has become a bottleneck restricting the green transformation of processes and the in-depth implementation of green manufacturing projects,and has hindered the sustainable development of the foundry industry.Building a digital twin of the foundry process is an effective way to break through this bottleneck.However,in view of the complexity of the casting process,it is very difficult to establish a logical model of the process by means of mechanism analysis.The description model does not have the ability to deal with process data.Although the datadriven method avoids the difficulty of mechanism analysis,the model also has the ability to process data.But this kind of learning model has black box characteristics,and they are not interpretable.Although it can output effective results,it is unable to analyze the influencing factors and interactions of the results,and cannot effectively guide the green planning of the process and the green improvement of the process.Therefore,how to realize the digital twin modeling of the casting process is the core problem of solving the interaction and integration of the physical space and information space of the process,promoting the green transformation of the casting process and the in-depth implementation of green manufacturing engineering.Aiming at the digital twin modeling problem of the foundry process,this paper uses machine learning,knowledge fusion and other technologies to study the digital twin modeling of the foundry process driven by data and knowledge.The main research work is as follows:(1)Combined with the functional requirements of the digital twin of the foundry process,the concepts of the digital twin of the foundry process and the process scene are clarified,and the framework of the digital twin model of the foundry process composed of the process digital model,the process scene data model and the process scene load calculation model is designed.Based on the analysis of the structural characteristics of the process,a modeling method based on multi-level directed acyclic graphs is used to construct a digital model of the casting process;on the basis of the analysis of process data attributes,a scene-based process data model is constructed.(2)Based on the production rules,the casting process knowledge is characterized,and the ruled Petri net is further used for element mapping and knowledge reasoning,which is used to construct the feature engineering of the process.Furthermore,the training data set is constructed according to the feature engineering,the XGBoost algorithm and the SHAP model feature analysis method are used to perform the iterative modeling based on feature resolution,and the calculation model of the casting process scene is established,which supports the calculation of resource and environmental load data for process scenarios and the interpretation of model features.(3)Designed and developed a digital twin verification system for the foundry process to verify the functional realization of the digital model,scene data model and calculation model.