Digital twin simulation for quality and efficiency evaluation on industrial manufacturing: A Theoretical Analysis

Abstract

This paper presents an integrated intelligent manufacturing quality and efficiency assessment system that leverages real-time data collection, knowledge graphs, and machine learning to optimize assessments and intelligently analyze production processes. A digital twin simulation platform is utilized for precise mapping of physical environments to virtual models, enhancing manufacturing quality and efficiency. The system's objectives include developing a data collection and analysis framework for interoperability of heterogeneous data sources, employing advanced machine learning models for event extraction and causal relationship identification, and constructing a comprehensive causal graph for in-depth quality analysis. A novel method integrating graph convolutional neural networks with knowledge graphs is proposed for precise quality and efficiency assessment, featuring a knowledge graph embedding-based GCN model and a regularized loss function for improved accuracy. The system also incorporates a digital twin approach for accurate simulation, with a focus on online data collection, network resource allocation, and the creation of virtual models through digital modeling techniques. A communication network and data management system are designed to enhance M2M communication, using MILP and cloud computing for network adaptability and data real-timeness. The project aims to automate the design process with STEP standards and build an IMT digital twin system for comprehensive machine tool monitoring, ultimately achieving the intelligent manufacturing goal of "controlling the real with the virtual."