Research On Condition Monitoring Method Of Marine Crane Boom Manufacturing Process Based On Digital Twin

Marine crane is a special crane for carrying out transportation operations in the Marine environment,mainly used for ship-to-shore,ship-to-ship cargo transport transfer,replenishment at sea and other important tasks.The boom is an important part of the cargo transport and transfer of marine crane.In the manufacturing process,there are problems such as lack of transparency in condition monitoring,difficulty in operation and maintenance control,low value of industrial interconnection data utilization,and the accuracy and efficiency of welding defect detection need to be improved.Digital twin technology can realize the high-fidelity modeling of information space to physical space,and solve the problems of perfect matching and tracking of information space to physical space,dynamic real-time response,and collaborative optimization.Therefore,it is an urgent problem to study the condition monitoring method of marine crane boom manufacturing process based on digital twin.The dissertation focuses on the condition monitoring method of marine crane boom manufacturing process based on digital twin.The condition monitoring system architecture of marine crane boom manufacturing process based on a digital twin is constructed.The method of establishing the digital twin high-fidelity model of the manufacturing process and realizing the physical-information multi-level mapping mechanism is proposed.The adaptive updating and optimization mechanism of digital-model interaction driven by streaming data is studied.A modified deep-learning welding defect identification method based on hyperparameter optimization is proposed.The development of a digital twin system for enterprise production needs is researched.Theoretical and technical application support for digital twin in marine crane boom manufacturing process of virtual-real dynamic mapping,process data visualization,state monitoring,anomaly detection,and defect identification is provided.The main research contents are as follows:(1)Given the characteristics and intelligent requirements of the condition monitoring of marine crane boom manufacturing process,the condition monitoring system architecture of marine crane boom manufacturing process based on a digital twin is built with data-driven and business packaging and integration as the core,including the physical layer,communication layer,data layer,business layer,and service layer.The representation meaning of each layer in the system architecture and the data coupling relationship between layers are expounded in detail.According to the manufacturing requirements and the characteristics of the system architecture,three key technologies with specific characteristics of marine crane boom manufacturing process are identified as the research objects of condition monitoring service application.(2)Aiming at the difficulties in implementing high-fidelity mapping caused by complex processes and site complexity in the manufacturing process of marine crane boom,a digital twin high-fidelity model construction and incidence relation mapping method is proposed to achieve a comprehensive description and complete expression of manufacturing resources in the manufacturing process of marine crane boom.Define product manager and resource manager components,build the PPR(Product Process Resource)structure tree of the manufacturing process,introduce sense manager components to endow the PPR structure tree with real-time performance,and build a high-fidelity model of the production and manufacturing resources involving man-machine-product-material-rule-environment-measurement,all-factor,allbusiness,and all-asset.The incidence relation mapping mechanism of the three managers is studied to realize the multi-level representation mapping relationship of static characterization,dynamic characterization,real-time characterization,and prescient characterization of the manufacturing process in information space.Focusing on the real-time mapping of the digital twin,the dissertation expounds on the real-time data-driven mechanism of the fully automated manufacturing process and the critical event-driven mechanism of the non-fully automated manufacturing process and improves the correlation mapping mechanism of three managers.(3)Aiming at the low anomaly detection accuracy caused by concept drift when the degradation of marine crane boom manufacturing equipment,an anomaly detection method of welding equipment based on digital mode interaction adaptive updating and optimization is proposed to realize the model construction of equipment behavior transformation and the improvement of anomaly detection accuracy.The logical model of a single device working independently and multiple devices working together is optimized.The mechanism of device behavior state change is researched,and the behavior rules of the twin model resource manager are enriched.Based on the real-time stream data obtained by the perception manager,an adaptive update and optimization framework based on the integrated machine learning detector is constructed.The improved Isolation Forest algorithm and hierarchical finite state machine are used to realize the adaptive interaction between the flow data and the anomaly detection model,and the effectiveness of this method in improving the anomaly detection accuracy of concept drift is verified.(4)Aiming at the low accuracy and efficiency of detection and identification of welding defects in marine crane boom,a multi-strategy Modified Grey Wolf Optimization(MGWO)algorithm is proposed to optimize the combination of convolutional neural network hyperparameter configuration to realize intelligent recognition and classification of image sets of typical weld defect types.GWO is modified by using a nonlinear search strategy,dynamic weight factor updating strategy,and dynamic memory factor updating strategy.The optimal configuration combination of convolution kernel size,convolution kernel number,pooling type,neuron number,and other hyperparameters of typical convolutional neural network architecture is solved to realize intelligent detection and recognition of typical weld defect types and verify the effectiveness of this method in improving the detection accuracy and recognition efficiency of a welding defect.(5)To meet the needs of an equipment manufacturing enterprise,develop a digital twin status monitoring system for the marine crane boom manufacturing process.Analyze the functional positioning of the system,design the specific development technical framework,and conduct business integration and encapsulation of key technologies such as the proposed highfidelity model construction and correlation mapping method,the anomaly detection method of digital mode interaction adaptive updating and optimization,and the convolutional neural network method of hyperparameter configuration combination optimization,to realize the data perception and database design of the manufacturing process.Develop multi-dimensional displays of the manufacturing process,automatic equipment monitoring,welding manufacturing process visualization,and other system function modules.Finally,through the application of the digital twin system in enterprises,the theoretical system of the digital twin method for condition monitoring of marine crane boom manufacturing process is proved to be feasible and effective.