In recent years, China's steel industry has formed a three-dimensional technology integration which includes information technology integration, process technology integration, and life cycle technology integration. These technology integrations are the basis of informationization. Therefore, the steel industry is an industry that is closest to intelligence, has the best conditions for achieving intelligentization, and has the most urgent demand for intelligentization.” On September 27, Academician of the Chinese Academy of Engineering and Professor of Northeastern University Wang Guodong made the remark above in the 2019 (5th) China Iron and Steel Industry Internet + Promotion Conference, which was co-organized by China Metallurgical News Agency and Tangshan Industrial and Information Technology Bureau, and sponsored by Qian’an Jiujiang Wire Co., Ltd.
Platform architecture of intelligent manufacturing in steel industry
Wang Guodong introduced the platform architecture of intelligent manufacturing in the steel industry. In his view, intelligence has the following characteristics:
First, it has the ability to perceive, that is, the ability to perceive the outside world and acquire external information. This is a prerequisite and necessary condition for generating intelligent activities.
Second, it has the ability to remember and think, that is, to store the perceived external information and the knowledge generated by the thinking, and to use existing knowledge to analyze, calculate, compare, judge, associate, and make decisions.
Third, it has the ability to learn and adapt, that is, through the interaction with the environment, continuously learn and accumulate knowledge, so that they can adapt to environmental changes. Fourth, it has the ability to make decisions, that is, to respond to external stimuli, form decisions and convey corresponding information.
The core of intelligent control is the Information Physics System (CPS). The overall architecture of the system is: perception (physical) layer, network layer, computing layer. By integrating advanced information technology and automatic control technology such as sensing, computing, communication and control, it constructs a complex system in which people, machines, materials, environment, and information elements are mapped, timely interacted, and efficiently coordinated in physical space and information space, to achieve on-demand response, fast iteration, and dynamic optimization of resource allocation and operation within the system. At present, the embedded industrial control system of metallurgical enterprises is composed of physical layer, network layer, computing control layer and application layer.
Wang Guodong pointed out two principles for intelligent manufacturing platform construction. The first one is to make full use of existing system and existing data of enterprises, adopt advanced intelligent manufacturing technology, enhance network interconnection and data collection and processing, and develop digital twin of CPS system, namely digital perception. Intelligent manufacturing platform construction could upgrade the existing system of enterprises into an intelligent system with self-learning, self-adaptation and self-organization, and realize the whole process intelligent manufacturing and promote the high-quality development of enterprises.
The second principle is to implement true intelligence. In the process of implementing the full process and integrated intelligent manufacturing, enterprises should gradually and steadily carry out the software transformation to existing enterprise resource planning (ERP), manufacturing execution system (MES), process control, basic automation, etc, under the premise of meeting the functional requirements of these systems. In this way, these systems will be capable of sensing, memorizing, thinking, learning, adapting, and behavioral decision-making, so as to become truly intelligent systems.
“The steel industry is a process industry. Therefore, the information system of the steel industry must be full-process and integrated, so as to realize the comprehensive analysis, control and management of material flow, information flow and energy flow.” Wang Guodong emphasized.
He further pointed out that with the enrichment of technical means and the improvement of computing power, the way to solve problems will change from traditional ways of dividing a complicated problem into relatively simple ones at different levels, to ways of the flat development featuring “cross-layer and cross-domain”.
The flat two-tier structure is composed of “cloud intelligence layer” and “local decision-making layer.” The cloud intelligence layer is characterized by the functions of low real-time MES, ERP and BI (Business Intelligence), able to achieve comprehensive coordination of multiple objectives, including quality, equipment capability, cost, resources and human resources. The “local decision-making layer”, on the other hand, is characterized by the functions of high real-time PCS (process control system), BA (building equipment automatic control system), and sensors, supplemented by new technologies such as artificial intelligence and big data, computing power improvement and real-time performance. An online real-time dynamic adjustment between the two layers ensures the improvement of quality and efficiency in a flexible and coordinated way.
In order to better the flat development, enterprises need edge computing. The application of edge computing can realize intelligent control on near-field devices, equipments and systems, solve the uncertainty caused by network delay, reduce the burden on the cloud, and at the same time take into account the coordination of cloud computing and terminals.
The core of steel intelligent manufacturing platform is digital sensing system.
Iron and steel enterprises are faced with some problems in the process of intelligent development. Wang Guodong pointed out that the steel industry is a multi-process, long-flow process industry, and the steel production process is a full process “black box”, because chemical reactions, physical changes within the continuous casting and rolling work pieces in the smelting reactor and related important parameters are unseen, intangible, and undetectable, and there is an extreme lacking of real-time perception of internal information. Most of the models used in current system are theoretical models or empirical models, which are subject to uncertainty, poor prediction accuracy, and difficulty in achieving high-precision, dynamic coordination control. Rough, even blind, control decision-making is also a problem. Furthermore, changes in production conditions and equipment operating status lead to complex changes in production conditions, and a more complicated relationship among process input conditions, state variables and control objectives.
“Digital Perception provides us with new ideas and feasible methods to overcome the insurmountable obstacles of modern information technology such as big data, Internet, cloud computing, AI (artificial intelligence), and provides an ideal solution to accurate sensing and control for metallurgical and other industries.” Wang Guodong further stated that “under the control of information flow, material flow and chemical flow react chemically or physically, therefore, solving the mathematical model based on physical-chemical reaction with sufficient accuracy could obtain a unified solution of material flow, energy flow and information flow.
Due to the characteristics of uncertainty, strong nonlinearity and extremely complex nature, we must adopt the combination of mathematical model and AI/big data, take the mathematical model as the basic line, and apply AI and empirical big data optimized mathematical model to achieve high-efficiency digital sensing and crack the black box system of steel industry.
The basic components of a CPS system include sensors, actuators, and decision control units. The basic components combined with the feedback loop control mechanism constitute the basic functional logic unit of CPS, and perform the most basic detection and control functions. The digital sensing system: based on big data, artificial intelligence, Internet of materials and physical-chemical model, the giant digital sensing system realizes the digital twin of all physical quantities in the real physical world.
The decision control unit can make intelligent decisions according to the digital sensing system, the sensing information and the user input rules, and issue a process control parameter adjustment instruction to ensure that the control target value is obtained under the established conditions.
Wang Guodong pointed out:
The digital twin is the core of the physical system. The digital sensing system of Industry 4.0 can fully sense and control the whole process of information flow, material flow and energy flow. The whole process integrated CPS system can solve problems such as breakpoints and data offline automatic collection, and realize whole process integrated control and coordination.
The main functions of intelligent manufacturing system:
The first is in-depth perception of information, that is, to achieve accurate prediction of process quality and online state perception;
The second is data processing and storage, that is, to achieve whole process quality data collection and processing;
The third is network interconnection, including 5G Internet of materials, Internet connection to all materials and people;
The fourth is equipment diagnostic maintenance, that is, to achieve equipment health status diagnosis and intelligent maintenance;
The fifth is quality control optimization, that is, to achieve quality abnormality traceability and production process optimization;
The sixth is precise coordinated control, that is, to achieve quality precision control and multi-process coordination and intelligent optimization;
The seventh is intelligent optimization decision-making, that is, to achieve the intelligent optimization decision-making that realizes the integration of the horizontal production, supply and marketing, and the vertical and horizontal levels.
To build a steel industry information system, in accordance with the process of steel production, an intelligent agent capable of intelligent sensing, decision-making, and control should be established in major production links to achieve intelligent control of various links. These intelligent agents include: raw material processing agent, Blast furnace smelting agent, steelmaking-continuous casting agent, hot-rolling agent, cold-rolling agent, and heat-treating agent. Steel enterprises should integrate various agents into a multi-agent system that cover the entire process, to achieve cohesion, coordination, balance, and optimization among these agents, Wang Guodong suggested.
