Research On Digital Twin Construction And Rul Prediction Method Of Continuous Casting Equipment

Continuous casting equipment,as a large complete set of equipment in industrial production,has a complex structure and harsh operating environment.Equipment failures and damages often occur,leading to unstable product quality and increased production costs.Using digital twin technology to maintenance continuous casting equi pment can expand the data dimension of equipment and improve the level of intelligent operation and maintenance.This thesis combines digital twin technology to study the digital operation and maintenance method of continuous casting equipment.By constructing digital twin of continuous casting equipment and predicting the remaining useful life(RUL)of continuous casting roller,it provides auxiliary decisionmaking for intelligent operation and maintenance of continuous casting equipment,which has practical engineering application value and theoretical research significance.(1)Aiming at the problems of digital operation and maintenance management and RUL prediction of continuous casting equipment,the composition of digital twin of industrial equipment and the construction method of virtual and real synchronous operation model are analyzed.The research framework and technical route of digital twin construction and RUL prediction under continuous casting machine condition are given.The key technical difficulties such as the construction of virtual and real synchronous operation model,the generation of twin data and the RUL prediction of continuous casting roller are analyzed,and the research scheme of corresponding problems is clarified.(2)Aiming at the problems of complex construction process and inconsistent construction methods of digital twin virtual and real synchronous operation model of industrial equipment,a data interaction protocol specification based on NC-Link is designed for synchronous operation of virtual model and physical entity.The expression paradigm of motion structure relationship of virtual model under the protocol specification is given,which is used to quickly establish the pose and mechanical motion relationship between virtual model elements.Through standardized protocol model expression,the state parameter synchronization of virtual and real moving bodies is realized.Taking the arc section of continuous casting machine as an example,the effectiveness of the method of driving virtual and real synchronous operation based on protocol data is verified,which lays a digital model foundation for the study of the operation mechanism of digital twin.(3)In order to grasp the working conditions of key components of continuous casting equipment in real time,a digital twin containing sensor acquisition data,theoretical operation data and finite element simulation data is constructed.The calculation method of theoretical operation data and the acquisition method of simulation data under simulated real working conditions are analyzed.In order to solve the problems of poor timeliness and complex calculation of finite element simulation results,a simulation dataset is constructed.Support vector regression combined with k-fold cross validation is used to train the network model to enrich the real-time data of twins and provide multi-source data basis for the RUL prediction of key components of continuous casting equipment.(4)In order to improve the operation and maintenance management efficiency of continuous casting equipment and realize the real-time RUL prediction of continuous casting machine rollers,a multi-layer combined network learning model based on Convolutional Neural Network(CNN),t Bi-directional Long Short-Term Memory(BiLSTM)network and Multi-Head Self-Attention(MHSA)mechanism was constructed.The CNN is used to initially extract short-term time series information,feature information and expand data dimensions,and then the BiLSTM is used to extract long-term time series information.Finally,the MHSA mechanism is used to mine the hidden information in the extracted feature information.The network model was trained using the integrated multi-source caster operation data,and the effectiveness of the continuous casting roller RUL prediction model was proved through data testing.Compared with the CNN-BiLSTM model,the mean absolute error(MAE)of the RUL prediction of the model was reduced from 4.14 to 1.97.Compared with the original data,combined with multi-source fusion data,the MAE of the RUL prediction of the model could be further reduced to 1.58.Finally,the validity and feasibility of the proposed method are verified by constructing the digital twin prototype system of the segment of the continuous caster,which provides strong support for the operation and maintenance management of the segment of the continuous caster,and lays a theoretical foundation for the construction and application of the digital twin of industrial equipment.