Digital twin driven production progress prediction for discrete manufacturing workshop

Abstract

• This article focuses on a novel synchronous update method for digital twin workshop model with integrating learning approach. • The representation model is established to describe the performance degradation rules of discrete manufacturing workshop. • An Adaboost-DNN-LSTM based synchronous update method is proposed to guarantee the consistency between DTM and physical workshop. • A competitive election mechanism is designed to select base learner among DNN and LSTM between historical and dynamic sample. In make-to-order manufacturing enterprises, accurate production progress (PP) prediction is an important basis for dynamic production process optimization and on-time delivery of orders. Digital twin technology offers an enabling tool for PP analysis. Although the production process can be observed, analyzed, and controlled in real-time by digital twin model (DTM), there exist some uncertain events, degradation of manufacturing elements, and abnormal disturbance in physical workshop (PW), which would cause the deviation between DTM and PW performance and affect the prediction accuracy of PP. Synchronous evolution of DTM for precision holding to ensure the consistency between DTM and the performance of PW, and guarantee the accuracy of DTM is still a challenging issue, especially when dealing with new dynamic samples for complex production environment of discrete manufacturing workshop (DMW). This article focuses on how to effectively construct DTM synchronous update methods based on dynamic sample data for DMW. This study proposes a representation model of performance degradation and an Adaboost-DNN-LSTM based synchronous update model with competitive election mechanism to enhance the accuracy of PP prediction with time in industrial environment. The experiment is conducted in the realistic production dataset, which demonstrates that the proposed synchronous evolution model has good performance for realizing the synchronization of the performance of physical workshop in industrial environment, and can greatly improve the prediction ability for PP.