Digital Twin Process Manufacturing
Aims & Scope
The process industry is the pillar of national economies and includes the chemical, iron and steel, and non-ferrous industries. Digital Twin creates the core of smart manufacturing by integrating advanced sensing, communication, and data mining technologies. It facilitates complicated decision making in all aspects of the process industry, including supply chains, product quality, energy scheduling, and equipment diagnosis. Digital Twin has greatly facilitated the modeling and optimization of manufacturing processes, but it also brings a number of challenges, e.g., how to integrate mechanism knowledge with industrial big data in the modeling of industrial process and how to deal with multiple and coupled objectives in the optimization of the production process. This session aims to showcase innovative research from both theoretical and practical perspectives, exploring the challenges and opportunities brought by the integration of Digital Twin technology and process manufacturing.
The session will focus on the following points
• Digital Twin key technologies in Process Manufacturing
• Applications of Digital Twin in Process Manufacturing
• Application of artificial intelligence in Process Manufacturing
• Design and application of Digital Twin platforms in Process Manufacturing
Session Chairs
Presentations
-
Assistant ProfessorUniversity at BuffaloTitle: Networked data analytics in cyber manufacturing systemsAbstract To be confirmed
-
Assistant Research FellowThe Hong Kong Polytechnic UniversityMachine Learning assisted Design of High Entropy AlloysAbstract To be confirmed
-
Associate ProfessorChongqing UniversityTitle: Digital Twin-Driven Scheduling Optimization for Steelmaking and Continuous Casting WorkshopAbstract To be confirmed
-
ProfessorNortheastern UniversityTitle: Digital Twin and Scheduling of Parallel Walking Beam Reheating FurnacesAbstract Digital twin is a novel technology with a promising prospect in smart manufacturing. It can realize high-fidelity mapping of physical entities, data fusion of physical and virtual space, and real-time monitoring of physical entities. This makes it easy to make optimal decisions and precise control. The steel industry has long production processes, complex equipment operation mechanisms, and dynamic production environment. Due to the lack of digital representation of its production process, many optimization and control schemes obtained by existing methods have large deviations in their implementation processes. How to provide accurate digital representation and make optimal decisions based on it is a great challenge. The development of digital twin (DT) technology in industrial scenarios provides a solution to this issue. In the steel production process, reheating furnaces are highly energy-consuming facilities connecting the continuous casting and hot rolling process. This work proposes a five-dimension DT faced to parallel walking beam reheating furnaces. In it, we construct a three-stage model for virtual entities. It can accurately obtain the charging/discharging times and real-time positions of slabs along with simulating the slabs’ logistics processes. An adaptive large neighborhood tabu search method based on DT, named as ALTS-DT, is further proposed for parallel reheating furnace scheduling (PRFS) problems.
-
Associate ProfessorUniversity of Science and Technology BeijingTitle: Industrial time-series data generationAbstract In recent years, big data technology has driven the digital and intelligent transformation in various fields. However, problems such as poor data quality and high labeling costs are still prevalent in most industrial scenarios. The insufficiency of data severely affects the performance of various machine learning algorithms, making it difficult for them to meet the requirements of high precision and high reliability in industrial environments. In addition, industrial data generally exhibits complex characteristics such as temporality and strong coupling, which further limits the applicability of traditional data augmentation methods. Addressing three key issues in industrial data—complex spatiotemporal correlations, insufficient samples, and difficulty in embedding domain knowledge—this report proposes a solution to the problem of time-series data generation in industrial processes based on generative adversarial networks.