Katja Windt, SMS group, Germany

Abstract:
Today, remote operation of entire plant complexes is no longer a vision. Fundamental technological developments from the field of digitization and automation, such as, 5G or hybrid process modelling or self-learning capabilities, make it possible to fundamentally re-design the operation of highly complex, industrial plants. They offer the possibility to bundle highly qualified experts in centralized control centers. For manufacturers operating at multiple locations, this can bring major benefits. In addition to business benefits such as increased productivity and product quality, centralized training centers can improve the quality of employee education, while the physical distance between the plant and centralized training or control centers increases employee safety. The use of advanced predictive technologies in the fields of plant maintenance, energy, emissions and product quality also makes it possible to further increase production sustainability by increasing plant availability, lowering overall energy consumption, reducing emissions and improving product quality. To get there it means to change the way production plants are operated and operator teams work. First, it needs transparency. Data and insights into the current planned operation, equipment and process condition and production plans are a crucial prerequisite. All information must be gathered, suitably stored and presented in centralized control rooms in which one control pulpit operates several production steps centrally. Additional AR/VR tools support both operators and on-site crews in the way they interact and communicate. Ultimately, it needs operator teams with a diverse skill set. Data science like AI, metallurgical process knowhow and operator experience need to come together in order to understand the gathered data, derive matching actions and execute those successfully.

Alexander Thekale, Primetals Technologies Germany, Germany

Co-Author:
Christian Horn, Primetals Technologies Germany
Martin Kerschensteiner, Primetals Technologies Germany
Dominik Wassermann, Primetals Technologies Germany
Andreas Bauer, Primetals Technologies Germany

Abstract:
In steel manufacturing a core task of operators today is plant supervision. Cameras are a cost-effective way of displaying crucial parts of the plant to give operators an overview of ongoing production processes. However, the sheer overload of information by dozens of simultaneous video streams can be a challenge for workers and could lead to delayed interventions upon problems. Machine learning as a part of artificial intelligence has proven to be an effective solution for addressing this problem. By applying computer vision models and techniques, machines can gain a high-level understanding from input images and videos. Enriched with domain specific knowledge and additional plant information, it becomes a digital assistant or digital expert. It enables detecting and notifying operators on critical conditions of involved processes and components to increase quality and reduce downtime; digital experts can even actively interact with the process. In order to be able to use machine learning based digital assistants in an industrial environment, a robust framework must be created that is suitable for continuous operation. With this paper we present an approach on how digital assistants can be deployed for industrial applications. We describe the various challenges and outline an intuitive process to build a production-ready solution which is integrated into the existing plant software infrastructure. In addition, we illustrate how several aspects like monitoring and versioning can be realized. Using a real-world example of a digital assistant, we demonstrate the successful realization of our solution.

Roman Markus Holler, PSI Metals Austria GmbH, Germany

Co-Author:
Joachim Gnauk, PSI Metals GmbH
Hannes Sperl, PSI Metals GmbH

Abstract:
Blockchain technology is a potential game changer for the entire chain of the manufacturing industry. Over the past years, reliable tracking of goods using blockchain technology has become a major topic as it enhances transparency and traceability of products while revealing information like the history and origin of the products. While it is important to use blockchain technology to create a reliable and transparent traceability, such transparency partly reveals confidential information about production processes. This kind of scenario is not peculiar to special steel producers supplying sensitive industries, but also to conventional production chains where confidential information might be derived from the published information. In our paper, we will reveal how zero-knowledge proofs (ZK proofs) create increased transparent tracking while keeping confidential data. The standard EN 10373 describes certification of metallic products complying with EN 10204 using computational models based on sensor data captured during the production processes. By combining EN 10373 with blockchain technology and ZK proofs, producers can successfully compute the product’s quality without revealing anything about the production details. Keywords: Blockchain, Zero-Knowledge Proofs, Product Tracking, Transparency, Computational Models

Marcus Neuer, VDEh-Betriebsforschungsinstitut GmbH, Germany

Co-Author:
Andreas Wolff, VDEh-Betriebsforschungsinstitut GmbH

Abstract:
Modern process industries is driven by decision making, decisions which need solid factual understanding of a situation. Through the recent successes of artificial intelligence, mainly due to its subfield of machine learning, prediction and knowledge models are available to support those decisions. Explainable AI (XAI) developed techniques for fusing expert knowledge, existing physical equations and easily to interpret relations to provide more insight into the result generation of machine learning models. Counterfactual AI provides a way to work with the available data and to introduce straightforward means of mathematical logic and conditional probability into real. Such systems can answer hypothetical question like „if the tension at the rolling would not have been too high, would the strip would still have been damaged“. This requires the AI system to first adopt a perspective that is not present in the current data set evaluated, second, to utilise all previously learned information and third to combine it in form of Bayesian conditional probabilities to actually predict the scenario in which the questioned parameters would have occurred. Counterfactual AI is therefore able to scan its own machine learning input variables for „what-if“ scenarios, which can be straightforwardly applied to selected use cases in steel processing and process industry in general. In more global context, Counterfactual thinking in autonomous systems is one step towards Causal AI. The latter combines common machine learning approach on data, causal inferences on the found relationships including counterfactual questions and lastly physics-informed approaches with analytical process models. The talk will give details on the described techniques and presents real world applications of the discussed methods. It will cover different processes of the steel production route like continuous casting and rolling, to display the advantages brought by Causal and Counterfactual AI.

Niklas Reinisch, RWTH Aachen University, Germany

Co-Author:
Tarik Viehmann, RWTH Aachen University
David Bailly, RWTH Aachen University
Gerhard Lakemeyer, RWTH Aachen University
Gerhard Hirt, RWTH Aachen University

Abstract:
In recent years, many production processes, including open-die forging, have been digitalized, resulting in process data and process knowledge being available in many places. However, this knowledge is often decentralized, e.g., in individual companies, and cannot be consolidated in one database easily due to stakeholder interests (secrecy, ownership, etc.) or the volume of data. In order to nevertheless draw added value from all available process data, a multi-agent system (MAS) is presented using the example of an open-die forging. Based on an ontology, the individual agents can communicate semantically with each other to exchange information. A top-level service broker manages the available services and forms the link between agents and user interface. In this manuscript, a MAS-demonstrator is presented that provides two example services, the material choice of a forging as well as the design of an optimal pass schedule. For this, the MAS uses process data along the production chain of an open-die forging as well as an ontology designed to represent this production related data. In the MAS multiple agents are included, representing different stake holders along the production chain like the steel retailer, the steel producer and the open-die forge.

Joaquín B. Ordieres Meré, Universidad Politécnica de Madrid, Spain

Co-Author:
Andreas Wolff, VDEh-Betriebsforschungsinstitut GmbH
Antonio Bello-García, Universidad de Oviedo
Stefano Dettori, Scuola superiore di studi universitari e di perfezionamento Sant'Anna

Abstract:
Productivity in modern metals plants and processes depends on sophisticated computer-controlled automation systems that have become powerful, and ubiquitous. It is part of the Internet of Things (IoT). In the case of automation, the computers that make production smarter also make it more vulnerable to external interference. Manufacturers have become more vulnerable to cyber-attacks after shifting to Cloud infrastructure and services, since from 2017, there have been approximately 382 new vulnerabilities, and additionally, the crackers have had the tendency of exploring targeting vulnerabilities before the security research team and responsible software vendors realize their presence. After reviewing several RFCS research projects in the context of the RFCS EU funded dissemination project named ControlInSteel, it becomes clear that many of the created models for forecasting can be useful for estimating bias between expectations and measured values. Unlike earlier analytical attempts to find more effective model representations, the cloud oriented Operational Technologies provide scalable solutions enabling different applications. Such applications not only are useful for low level monitoring activities, but also to create higher level of representation of data, mainly product oriented with high traceability on the low-level data. This paper wants to present the different capabilities that cloud OT solutions enable, both in process monitoring, OT oriented cybersecurity, and high-level data representation. All the analysis will be carried out by using a case study from the hot strip mill for long products in the context of the EU funded Autosurveillance project.

Sheetal Birla, Falkonry, United States

Abstract:
Steel manufacturing processes are heavily instrumented, generating large volumes of automation data in the form of time series to the tune of over 5 million data points per second per plant. The time series data contain adequate information to represent the state of a physical system and production process at any given point in time. However, prevalent data-driven fault detection methods require significant setup efforts and expert inputs for modeling every known state of the system. This paper presents a novel self-supervised AI approach that does not require any setup effort and is capable of monitoring every existing process parameter and asset metric at high speed. Our approach utilizes a deep learning architecture based on Convolutional Variational Autoencoders (CVAE) that can start learning from small amounts of data to identify excursions, can automatically and incrementally learn as the underlying behavior of the asset changes and can process millions of measurements a second across thousands of time series. The automated time series AI informs plant operations of conditions that require human attention and provides diagnostics of underlying issues - leading to informed production and maintenance decision-making. Self-supervised AI overcomes the challenge the conventional machine learning method faces scaling to the needs of steel manufacturing by accommodating the challenges of constant equipment, environment, and product changes that hinder classical supervised learning methods. This paper will show this new AI in commercial steel manufacturing operations today.

Tarun Mathur, ABB AB, India

Abstract:
Artificial Intelligence (AI) is a key enabler for amplifying return on investment for digital technologies. There are many examples of big data, cloud platforms and data integration investments in the steel industry though much fewer for intelligent applications that convert this data into value. The complex nature of operations in a steel plant along with unavailability of large historical data pose a significant challenge to deploying artificial intelligence projects. This paper details several applications of AI in steel manufacturing for operational excellence, process performance and reliability improvement. The first application concerns improvement of manual and complex operation of the steelmaking shop using data from all meltshop processes, ladles, cranes etc. to build a machine learning model that predicts temperature losses for heats. It then prescribes the lifting temperature to the ladle furnace operator resulting in better superheat compliance, thus improving average caster speed and reducing energy input at ladle furnace. The second application uses data-based models to predict and control process performance by utilizing the well-known Model Predictive Control framework. This kind of application works in a closed loop with the automation system, correcting PID setpoints towards a more profitable operating range. These applications work well for slower and continuous processes such as those in pellet or sinter plant, grinding, reheat furnaces etc. The third application relates to how reliability can be improved by combining asset and process information. It’s applied in downstream rolling processes to analyze coil and process data in different phases, make correlations and label the operations data, and then use this labeled data to make predictive models. Some of these models are also prescriptive, providing the operator with exact action needed to avoid an anomalous situation.

Nikolaos Matskanis, CETIC ASBL, Belgium

Co-Author:
Guillaume Ginis, CETIC ASBL
Sebastien Dupont, CETIC ASBL
Rami Sellami, CETIC ASBL

Abstract:
Steel production involves separate manufacturing sites for each production phase and various stakeholders in the supply chain. All these sites are consuming and producing heterogeneous data that is often not efficiently exploited (no data aggregation, coalition), which impede the added value of a global overview of processes, quality and security assessment. In addition the growing use of automation and interconnection of Cyber-Physical Systems causes an increase of the attack surface and new challenges and security risks for industrial control systems.A platform with global overview of production quality and security over multiple sites that is multi-tenant and multi-actor and supports the aggregation/integration of data, would facilitate the anomaly detection in all stages of production and plants involved. Additionally allow application of remediation strategies to all involved sites. We propose a solution based on the FADI framework for deploying and orchestrating a Big Data management and analysis platform for each plant in the production chain. Such platform instance is called a federated FADI instance. Then, a global, federal FADI instance will be configured in order to communicate with the federated instances and combine received data for ensuring an overall overview on the production chain and anomalies detection. By interconnecting these instances, we will obtain a FADI Federation. FADI leverages the Collaboration, Automation, Lean, Measurement and Sharing principles of the DevOps approach to ensure that software is produced with a high level of quality and speed by integrating development (Dev) and operations activities (Ops). In order to address the new security threats, the FADI framework will be augmented with security services to secure the software development life cycle, improving the immunity of the system. This DevSecOps approach automates security activities in the software continuous integration and deployment pipeline including security response processes for assisting operators and security remediation processes using the FADI Vacsine tool.

Joachim Gnauk, PSI Metals GmbH, Germany

Co-Author:
Jonas Meinke, PSI Metals GmbH
Jan Guhl, Gestalt Robotics GmbH
Robert Piontek, GEFERTEC GmbH

Abstract:
The KIKA-IPK project (funded by the German Federal Ministry of Education and Research) aims to provide a cloud platform as a standard for industrial AI applications. This platform will interconnect industry partners and AI third party service providers in such a way that specific problems in production can be analyzed via this platform in a standardized way. The resulting AI application either will be available to the industrial customer remotely or deployed to the on premise system for offline use. The research project aims at developing an AI cognition support system for in-process control, which will enable a more resource efficient process and material configuration through self-learning correlations of signal characteristics with process properties. The machine operator's empirical knowledge of the connection between tangible quality features on the one hand and process characteristics on the other is modelled by machine learning methods. This is demonstrated by the example of additive manufacturing of steel parts.

Dieter Bettinger, Primetals Technologies Austria, Austria

Co-Author:
Martin Schaler, Primetals Technologies Austria
Petra Krahwinkler, Primetals Technologies Austria
Christian Tauber, Primetals Technologies Austria
Angelika Klinger, voestalpine AG
Harald Fritschek, Primetals Technologies Germany
Christoph Feilmayr, voestalpine AG
Magdalena Schatzl, K1-MET GmbH
Clemens Staudinger, voestalpine AG
Ross Goldberg, Midrex Technologies, Inc.

Abstract:
When using AI-based results for decision making or decision support, the reliability and a basic understanding of the underlying reasoning process is essential. Due to the involved complex metallurgy, thermodynamics and material parameters, modelling ironmaking processes with first principles methods is demanding. Artificial Intelligence (AI) and other data driven methods offer solutions for such complex systems. While AI- Algorithms are powerful, they tend to be black boxes. This can be overcome with Explainable AI and by considering Meta-Information. In this paper we present, how successful AI based applications for sinter plants, blast furnaces and direct reduction plants are built using such transparency techniques and how they are integrated into state-of-the art decision-support systems.

Alessandro Stenico, PSI Metals GmbH, Germany

Co-Author:
Akriti Malla, PSI Metals GmbH
Robert Jäger, PSI Metals GmbH

Abstract:
At the age of Industry 4.0 and complex supply chain ecosystems, production management systems face a double challenge: first, they need to be re-designed to address increasing requirements for supply chain flexibility and resilience: business process flows and manufacturing strategies need to be more adaptive than ever, and this translates to the production management system. Second, they must incorporate technological developments that came along with the fourth industrial revolution, including dedicated industrial artificial intelligence (AI) and data science services, in order to leverage the added value lying in Big Data and support digitally driven manufacturing processes. In Quality Control, for instance, the aim of zero defects and no downgrades can be approached by enhanced process and product insight paired with the capability of fast, smart, and sustainable reactions, including online prediction and prescription, transparently executed. In Production Planning, work-in-progress (WIP) material residence times in stockyards and manufacturing lead times could be predicted by dedicated machine learning services, while smart autonomous agents could automatically adjust planned production schedules based on such results. Yet, to support such capabilities, production management systems must allow metals producers to model their business processes in workflows and integrate plug-and-play services. They demand collaboration between independent solutions, real-time visibility through embedded analytics, business logic configurability and the possibility of scope extensions via low-code approach. All of this needs to be supported by automated deployments, regardless if on-premise or in the cloud, and seamless upgrades ensuring minimal downtime during upgrade processes. Last but not least, they must provide state-of-art encryption mechanisms and resilience to cybersecurity threats. This paper explores the IT infrastructure and business process management functions that a metals production management system 4.0 must provide, to plainly support and leverage the promises of the industrial internet revolution.

Falk-Florian Henrich, Smart Steel Technologies GmbH, Germany

Co-Author:
Lucas Corts, Smart Steel Technologies GmbH

Abstract:
Conventional scheduling methods utilized in the steel production process, primarily semi-manual in their nature, tend to result in operational challenges. They are characterized by a lack of accurate and reliable quality prediction capabilities and cannot cope with the complexity of dynamic synchronization of upstream and downstream operations. An insufficient level of automation and a limited planning horizon present challenges in effectively harmonizing the planning of melt shop, casting and rolling operations. As a result, the direct and hot charging rates of an integrated steel plant are often low, leading to suboptimal energy utilization and high inventory and slab handling costs. To overcome these challenges, advanced scheduling systems that employ a combination of automation and reliable quality prediction have been developed. These systems can adjust dynamically in real-time to mitigate upstream deviations, which leads to increased hot charging rates and improved energy efficiency. By using automated scheduling, high utilization and coordination of equipment across all process stages can be accomplished. Smart Steel Technologies will elaborate on how an advanced scheduler utilizes mixed integer solvers and artificial intelligence models to optimize the scheduling process. Intelligent cost functions take into account quality predictions and all influencing restrictions from relevant aggregates. As a result, precise scheduling that boosts productivity and lowers energy consumption in difficult market conditions can be achieved.

Heinz-Josef Ponten, PSI Metals GmbH, Germany

Co-Author:
Rudolf Felix, PSI Fuzzy Logik & Neuro Systeme GmbH
Sebastian Grob, PSI Metals GmbH

Abstract:
There is no doubt that high-quality steel is indispensable in today's industrialized society, and steel producers have always been confronted with the dilemma of meeting the market requirements for new and thus more complex steel grades in an economically attractive way. A further dimension, the obligation of the steel industry to transform to carbon-neutral and more energy-efficient production, is added, which is becoming a tightrope walk due to the current situation on the energy markets. So how can steel producers optimize production processes, manage their heat schedules, save energy, comply with CO2 regulations and stay competitive with changing market demands? This paper discusses how collaborative smart software helps to optimize energy usage and hence influences CO2 reduction targets while minimizing production and material costs. An Online Heat Scheduler creates a detailed work schedule for all planned heats, including all required treatment and transport steps, their durations, the assignment of required production facilities and operating equipment as well as forecasts for required energy and material demands. Furthermore, it secures the required throughput and eliminates unpredicted discontinuity or sequence interruption at the caster line in case of unplanned standstills or delays. This process ensures efficient process quality by means of intelligent data collection, analysis and balance between target and criteria conflicts. Our approach is a proven industry standard, which helps plant managers flexibly manage planned downtime/maintenance work, plant problems, Hot Metal, DRI or Oxygen availability, Electrical Energy demand forecast and usage which leads to significant reduction in energy consumption and costs. Keywords: Decarbonization; Energy; CO2-Emission; Forecast; Qualicision; Heat Scheduling; Shopfloor

Jens Brandenburger, VDEh-Betriebsforschungsinstitut GmbH, Germany

Co-Author:
Miguel Gutierrez, Universidad Politécnica de Madrid
Joaquin Ordieres, Universidad Politécnica de Madrid
Alessandro Maddaloni, Institut Polytechnique de Paris
Valentina Colla, Scuola superiore di studi universitari e di perfezionamento Sant'Anna
Vincenzo Iannino, Scuola superiore di studi universitari e di perfezionamento Sant'Anna
Christoph Schirm, Thyssenkrupp Rasselstein GmbH
Dirk Müller, Thyssenkrupp Rasselstein GmbH
Erwin Sirovnik, Thyssenkrupp Rasselstein GmbH
Andreas Wolff, VDEh-Betriebsforschungsinstitut GmbH
Ahmad Rajabi, VDEh-Betriebsforschungsinstitut GmbH

Abstract:
Within the RFCS project DynReAct new concepts to improve the flexibility of production scheduling in flat steel production were developed and demonstrated at the tin-plate production site of thyssenkrupp Rasselstein in Andernach providing multiple production steps with free choice of multiple plants for production of a certain product. The selected concept follows a hybrid scheduling approach combining three planning levels with different planning horizons as well as planning accuracies to provide robust production plans on the one hand and flexible reaction strategies in case of unexpected events or decreasing plant performances on the other hand.

Marc Schwarzer, PSI Metals GmbH, Germany

Abstract:
Metals producers with complex routings (e.g. various finishing options, alternative “sister” shops) require reactive Planning and flexible Material Allocation. Dynamic Order Dressing (OD) provides several advanced features, which combined into the proper overall system architecture, can provide the vital inputs for an optimized Planning decision. Initially a Sales Order Item Position is elaborated into multiple alternative Production Order (PO) variants, each of which is representing an individual unique routing. Within each PO variant OD is calculating multiple Material Demand Variants (modelling alternative dimensions and cutting factors) that can lead to additional PO variants. Traditionally, PO Variants had to be pre-determined on Sales Order pre-processing level, e.g. using high level rules that suggest using three PO Variants to model different plant routings. With Dynamic Order Dressing PO Variants can now be determined dynamically during the Material Demand calculation. For instance, the routing determination returns multiple alternatives at certain decision points and after you determined the exact intermediate and final materials, you can dynamically generate additional auxiliary PO’s (first calculate the exact Pipe dimensions and pieces and afterwards generate an according Coupling PO). Finally Due Date Quoting (DDQ) will evaluate all PO Variants and according to configuration determine their potential due date and select the best PO Variant according to optimization criteria. At later time Material Allocation is executed, preferably using the primary Material Demand Variant of the PO Variant selected by DDQ. The additional Variants, however, extend the room for solutions significantly, reducing unassigned stocks for the plant. This comes especially handy when a plant purchases many intermediate materials of different dimensions externally. Once such alternative intermediate material design is selected for allocation, this will automatically trigger a Material PO generated by Forward Dressing to ensure optimal routing, production and quality instructions for this non-standard material.

Fuqiang Wei, Shougang Group Company, China

Co-Author:
Jian Zhang, Shougang Group
Jiangtao Zheng, Shougang Group
Xiangjun Meng, Shougang Group
Chao Wang, Shougang Group
Jiangli Yang, Shougang Group

Abstract:
Abstract：This paper summarizes the change of steel coil transportation technology in strip mill plant in recent 20 years, analyzes the traditional steel coil transportation modes and their technical characteristics, introduces the new intelligent vehicle powered by super capacitor, compares this new type vehicle with the traditional technology, and briefly describes the engineering case and practical application of the new technology, It is concluded that rail transportation and intelligent are the development trends of steel coil transportation in steel plant.

Rolf Lamm, Minteq International GmbH, Germany

Abstract:
Continuous changes in the economic environment and the increasing number of EAF plants accompanied by competitive pressure require steel producers to introduce innovative measures to reduce costs, CO2 emissions and improve safety. The paper introduces the latest development of an automatic and continuous refractory maintenance system for the electric arc furnace (EAF), tailored of the requirements of modern steel production at an American steelplant. The automatic system has eliminated the disadvantages and inherent in intermittent refractory maintenance and follows the “No Person on the floor”- Safety-Philosophy. The functionality of the SCANTROL™ 4.0 system (laserscanner driven measurement of refractory thickness, visual representation of scanned results and intelligent material application) has significant enhanced productivity, working conditions and decision-making capabilities of steel operators. The 5th generation Laserscanner technology with more than 10 million measuring points per Scan in a furnace enables wear to be determined with a very high degree of accuracy and, thanks to the high measuring point density, joints and cracks can also be detected. With the help of the measurement data, an intelligent program calculates exactly the critical areas that need to be repaired. Furthermore, the quality and the required quantity of the repair gunning material is proposed. The information thus obtained is used to automatically control a gunning robot for repair of the refractory lining. This machine applies the repair compounds exactly where they are needed in the furnace. The overall effect at the steelplant has included: • Reduction in total refractory consumption • Increased furnace availability by reducing “Power Off” delays. • The ability to effectively maintain all areas of the furnace • Extending the available production period between brick relines • Improved operational safety • Integration of the determined data into an “Industry 4.0” environment

Delphine Rèche, VDEh-Betriebsforschungsinstitut GmbH, Germany

Co-Author:
Christina Minari, Dimasimma Srl
Matteo Chini, Pittini S.p.A.
Denis Azzano, Pittini S.p.A.
Loris Bianco, Pittini S.p.A.
Stephan Lindemann, Salzgitter Flachstahl GmbH
Christian Clees, Multikopter.de
Peter Lundin, Swerim AB
Roberto Piancaldini, Rina Consulting – Centro Sviluppo Materiali S.p.A
Alexander Dunayvitser, VDEh-Betriebsforschungsinstitut GmbH
Kevin Sze, Swerim AB
Antonella Angrisani, Rina Consulting – Centro Sviluppo Materiali S.p.A

Abstract:
Although the use of unmanned vehicles for industrial applications is discussed a lot lately, systems in operation can hardly be found in the steel industry and common knowledge in the plants about the specific capabilities of UVs is poor. The objective of ROBOINSPECT project is to introduce novel robotic inspection systems in the European steel industry. For this purpose, existing technologies based on unmanned vehicles (UVs) are assigned to inspection tasks and technological gaps are closed by further developments. The aims are to reduce downtimes and increase occupational safety. Unmanned aerial (UAV) and ground (UGV) vehicles are developed for semi-autonomous operation during running production in confined and hazardous areas, using custom indoor navigation concepts. Software is provided to accelerate damage analysis of facilities and processes as well as to detect online plant misalignments. In this project, different UV types are used and specifically designed for different types of inspection tasks to cover a wide range of inspection work in the European steel industry. A drone has been developed together with improved software to be able to carry out indoor inspections without using a GPS navigation system. Another use case which is studied in this project is the use of ground vehicle together with a robotic arm equipped with a specific camera system developed for automatic inspection purposes. In addition, for another use case, a video microscope has been developed to analyse dynamic displacements of plants during operation by subpixel amplification (for example during steel rolling application). Intelligent software for image analysis to handle large data streams of visual material generated by the UVs’ sensors are also examined and developed here. To conclude, the (UVs) will help in production to carry out inspection and maintenance tasks in order to ensure highest safety for personal as well as time and money savings.

Arno Haschke, Primetals Technologies Germany, Germany

Co-Author:
Andreas Maierhofer, Primetals Technologies Germany
Sebastian Kündinger, Primetals Technologies Germany

Abstract:
"Over the last few years, several promising products on the market have been offering artificial intelligence or predictive maintenance to overcome the difficulties of monitoring the condition of assets in a plant. The key to creating the methods and analysis, which allow for reliable monitoring, rests in an understanding and knowledge of the process and technology of a production line. Reliable monitoring is accomplished using state-of-the-art methods, e.g., data analytics, for analyzing and creating suitable responses to operating conditions and disturbances. Primetals Technologies has developed a digital assistant that uses the understandings mentioned above to monitor any assets in any plant. Additionally, it can integrate existing monitoring systems, adapt to various data sources, or combine different analysis results to solve more complex monitoring demands in one central system. Besides reliable and automatically monitoring 24/7/365, the new system delivers solution-oriented information, which helps operators make the right decision and act fast in case of an alert. It provides this additional information based on built-in know-how to make anticipating and overcoming potential malfunctions easier."

Florian Hollensteiner, Primetals Technologies Austria, Austria

Co-Author:
Anna Theresa Strasser, Primetals Technologies Austria
Inge Aasheim, Primetals Technologies Austria
Gabriel Lenna Do Nascimento, Primetals Technologies Austria
Klaus Stohl, Primetals Technologies Austria
Michael Weinzinger, Primetals Technologies Austria
Irina Tolkacheva, Primetals Technologies Austria
Moritz Ortner, Primetals Technologies Austria
Georgiy Pervushin, Primetals Technologies Austria
Josef Boehm, Primetals Technologies Austria
Horst Puchner, Primetals Technologies Austria

Abstract:
One of the goals of maintenance management is to maximize the return on investment in assets. This leads to an opportunity that transforms maintenance and repair service from a cost factor, into an activity which influences profitability and leads to a competitive advantage. Metallurgical companies worldwide are using condition monitoring, computerized maintenance management systems solutions or 24/7 remote services. But how to make your maintenance even smarter? Primetals Technologies integrates these processes and methods to add more value to them according to the synergy law: the whole is greater than the sum of the parts. More than that, these methods have been successfully implemented in current steel productions and brought numerous results including unplanned and planned downtime decrease, total cost of ownership reduction and equipment availability growth. An example including achievements of this smarter maintenance management solutions are presented in this paper.

Christian Mengel, SMS group, Germany

Co-Author:
Sebastian Richard, SMS group
Ivars Valdmanis, SMS group
Lars Gillgren, SSAB AB

Abstract:
Digitalization in metallurgical industry is a focus in current development enabling process engineers and maintenance staff to optimize their processes and help solving problems in their daily work. Condition monitoring systems collecting relevant data from the process is aiming at monitoring the health of the main components. SMS groups Genius CM program package is offering several monitoring functions in the field of vibration analysis and mechanical and hydraulic machine parts making use of field units for local data collection and analyzing process data to derive information on the condition of the equipment. Within the rolling equipment of hot strip mills the conditions of the mill stands is of major importance. Bad mechanical conditions can lead to rolling problems e.g. roll positioning issues, thickness errors, cambered strips, tailend cobbles, strip surface defects or unplanned roll changes. New monitoring functions have been developed and applied focusing on the integrity of the mill stand. These are analyzing - the parallelism of the rolls indicating unacceptable wear conditions in the stands - the hysteresis of the roll positioning equipment monitoring the conditions in the force transmission components - and the roll force measurement equipment. Model based calculations and helping functions are targeting at optimal user support to determine the most critical equipment conditions and focus on the relevant steps for maintenance and required service actions. The paper gives a short introduction on the modular Genius CM software package and describes the new developed monitoring functions. Application cases of different hot rolling mills will give an insight of the possibilities for trouble shooting and measures to detect and solve acute problems in daily mill operation.

Christian Dengler, Paul Wurth S.A., Luxembourg

Co-Author:
Olivier Mersch , Paul Wurth S.A.
Volodymyr Kuskov, Paul Wurth S.A.
Dirk Malcharek, Paul Wurth S.A.
Fabrice Hansen, Paul Wurth S.A.

Abstract:
Following the old proverb that you cannot manage what you cannot measure, tracking of important production metrics is a requirement for staying competitive in the global market for iron and steel. The digital transformation promises new insights for production improvements through, among others, an increase in advanced data analytics, newer and smarter sensors and an improved exchange and display of information. Leveraging such potential through the realization of value adding digital solutions poses its own challenges for many plant operators in the iron and steel sector, foremost the lack of available work force with the required qualifications and the development time and costs for such software. For technology suppliers, this implies developing and delivering tools that provide immediate benefit and sufficient flexibility to integrate new diagnostics for specific problems or plants. Paul Wurth is continuously improving and developing such tools. Recent developments of such tools and their added value for ironmaking are topics of this article. In the article, we present how we reduce the time to product for condition monitoring and process optimization solutions using our low-code development platform DataXpert and show the general approach for the development of such systems using this platform. We then provide tangible examples of existing template solutions in ironmaking that are developed using this platform. Those template solutions deliver a set of existing functionalities and can be adapted or extended to the needs and requirements of specific plants. Added value is illustrated on several examples: monitoring of high level metrics for slag granulation, diagnostic analytics for coke machines or predictive maintenance for the bell less top.

Boris Marcukaitis, DANGO & DIENENTHAL, Germany

Abstract:
Forging manipulators are large machines used in open and closed-die forging and ring rolling. They can easily weigh as much as 1,000 tons and are used to position workpieces of up to 350 tons, moving them safely in all six degrees of freedom. They can only fulfill their function if the axes are reliable and precise, and the hydraulic control system works perfectly. Such highly dynamic machines are typically the bottleneck in manufacturing. If the manipulator stops working, so do the press and furnaces. This usually also impacts downstream processes like heat treatment and machining. The growing need to save both on maintenance and spare parts calls for new methods to keep systems operational. To increase the availability of these machines, DANGO & DIENENTHAL is developing a system for the early detection of failures of individual machine components. The system involves the installation of additional measuring systems to the standard sensors already installed. Various test programs employ a predefined movement profile to check the individual functions of the machine and record the registered times, vibration spectra, and pressure curves. The measured values thereby obtained are stored and compared with the results of the machine in new condition. The results are evaluated and a new remaining service life is calculated for the individual components. DANGO & DIENENTHAL will be able to offer this optional new functionality in the near future. The option increases machine availability and prevents unplanned system failures. It alerts customers to an impending breakdown, allowing them to obtain spare parts in time. Once these parts are available, the faulty parts can be replaced during the next scheduled shutdown.

Mark Haverkamp, SMS group, Germany

Co-Author:
Sebastian Kemper, SMS group
Thomas Nerzak, SMS group
Andre Peschen, SMS group
Karsten Weiss, SMS group

Abstract:
Digitalization is currently the focus of the steel industry in order to optimize plant planning, production, maintenance, service and training. Due to the large system dimensions and according complexity, intelligent solutions are in demand today more than ever. The "XR Plant Inspector" (XRPI) is such a solution. It was developed by SMS group to visualize complex plant models in 3D. In contrast to conventional CAD programs, this software can display entire steelworks in full detail by techniques deduced from the game world. XRPI creates a digital image of the real system, which e.g. helps to analyze maintenance tasks quickly and effectively. During training, XRPI supports participants in surveying the complete system virtually and collaboratively, thereby significantly enhancing qualification. Through this immersive training, the customer gains an immediate deep insight and can identify directly with his system, even in an early planning stage. Furthermore, the XRPI facilitates the commissioning process. Here, open questions, changed boundary conditions and solution approaches can be documented and discussed through the direct connection to the 3D data, preventing miscommunication. But that's not all, the software has recently been expanded to include important additional features, in particular the link to other SMS plant apps, providing an coherent overall concept to the customer. The XRPI new developments include e.g. the automated 3D model creation, the integration of 3D animations, an interface to eDoc (direct link to technical plant documentation) as well as an open MQTT interface, which enables the connection to other SMS apps such as Genius CM (predictive maintenance app). The publication describes these new developments. The use cases listed here give an insight into the wide range of potential applications and advantages for the plant operator.

Dieter Bettinger, Primetals Technologies Austria, Austria

Co-Author:
Klaus Stohl, Primetals Technologies Austria
Bernhard Voglmayr, Primetals Technologies Austria
Harald Fritschek, Primetals Technologies Germany
Martin Schaler, Primetals Technologies Austria
Thomas Kronberger, Primetals Technologies Germany

Abstract:
"Ironmaking facilities are very capital-intensive plants. There is a variety of smart solutions that provide guidance and support to avoid critical operational situations, warn in advance of system failures or equipment problems and support efficient maintenance. For new plants such advanced systems are frequently considered as essential, as they support cost efficiency and reduce the carbon footprint. However, also for plants in operation and in particular for plants at the end of the life cycle these systems provide important guidance to extend the lifetime avoiding disturbances, failures and critical situations. Various solutions from smart sensors to relevant aspects of process control and process optimization systems as well as predictive maintenance systems including digital assistants are presented. The possibilities to integrate legacy systems with advanced technologies and paths to an efficient interplay are illustrated and discussed. "

Matt Anderson, Primetals Technologies US, United States

Co-Author:
Paras Patel, Primetals Technologies US
Ruth Kirkwood-Azmat, Primetals Technologies US

Abstract:
"Primetals Technologies Long Rolling is at the forefront of applying the latest smart digital solutions in the industry. Their vision systems, developed over several years, provide an extremely strong, smart digital portfolio with a focus on improving safety, reliability and maintenance, while increasing production, yield and quality. In addition, these systems provide previously unattainable insights into the rolling process. Prevention is far more economical than problem fixing. Primetals Technologies promotes a proactive, predictive, and preventive maintenance approach for all our customers. Vibration analysis allows accurate prediction for mill maintenance; enabling improved downtime scheduling. Data is collected both periodically and continuously by our proprietary condition monitoring and analysis systems. Long-term partnerships have been developed with our customers, using our vast predictive service knowledge and over a century of long-rolling experience. Sub-optimal roll cooling can lead to premature roll wear, reduced yield, and catastrophic failure. Primetals Technologies has developed a new roll cooling system with a quick-change, additive manufactured cartridge, which along with sensors, helps to ensure rolls are receiving consistent, optimized cooling. With increasing focus on safety, operators are restricted from approaching the mill to perform manual checks on section size, equipment status or adjust settings. A new guide series provides continuous feedback on section size, equipment condition and allows remote adjustments. In addition, an alternative section measurement system has recently been brought to market. The main focus of this portable device is again to keep operators away from the mill during production, while providing accurate, objective measurements. Other recent developments include crop length optimization, Laying Head/Stelmor vision control system, and a roll-change robot. The paper gives practical examples how the implementation of smart digital solutions in long rolling leads to an improved plant performance, optimized maintenance strategy and safer operation."

Sebastian Koldorf, MAGMA Gießereitechnologie GmbH, Germany

Co-Author:
Erik Hepp, MAGMA Gießereitechnologie GmbH
Evgenii Shvydkii, MAGMA Gießereitechnologie GmbH

Abstract:
Today, simulations for improving process conditions and the quality of continuous cast products are a generally accepted tool. State of the art simulation tools provide quantitative insights into the flow, solidification and stress formation in continuous casting processes. This includes the entire casting process, starting from the tundish and the flow into the mold, up to the solidifying strand, that is withdrawn through various cooling zones. This provides important information about quality and productivity when evaluating process alternatives. The classical use of simulation solutions has evolved to a higher degree of automation and virtual process optimization. Usage of digital twin technology in combination with integrated statistical tools enables a systematic virtual testing to identify an ideal process window. This enables the expert to identify the significant process parameters and to investigate the level of their influencing effect, without performing expensive and time-consuming trials on the shop floor. Based on this knowledge, it is possible to optimize continuous casting processes that are both, cost-effective and robust with respect to process variations. This paper will discuss the modelling of modelling of electromagnetic stirring (EMS) and its impact on the flow behavior in the strand. Another focus will be the evaluation integrated stress calculation to avoid the formation of cracks. Stress simulation is also used to described the gap formation in the mold and the corresponding change of primary cooling conditions. The results are shown using industrial examples for a bloom caster. With the availability of the integrated process knowledge of a digital twin, it is possible to identify optimal operating points for quality improvements and productivity increases. The objective is to derive a comprehensive model for online monitoring and optimized dynamic online control of the cooling and solidification process. KEYWORDS: VIRTUAL EXPERIMENTATION –CONTINUOUS CASTING –DIGITAL TWIN –ONLINE CONTROL –ELECTROMAGNETIC STIRRING

Volker Diegelmann, VDEh-Betriebsforschungsinstitut GmbH, Germany

Co-Author:
Roman Kuziak, Łukasiewicz Research Network – Upper Silesian Institute of Technology
Andrij Milenin, AGH University of Science and Technology
Szczepan Witek, AGH University of Science and Technology
Maciej Pietrzyk, AGH University of Science and Technology
Krzysztof Bzowski, AGH University of Science and Technology
Łukasz Rauch, AGH University of Science and Technology
Christian Trappmann , MANNSTAEDT GMBH
Andreas Falck, MANNSTAEDT GMBH
Nora Egido Perez, Sidenor Investigación y Desarrollo
Victor Manuel Santisteban Mendive, Sidenor Investigación y Desarrollo
Hagen Krambeer, VDEh-Betriebsforschungsinstitut GmbH
Monika Feldges, VDEh-Betriebsforschungsinstitut GmbH
Nils Hallmanns, VDEh-Betriebsforschungsinstitut GmbH

Abstract:
During rolling, straightening and thermal processes internal stresses arise impairing the products material properties and causing material distortion due to stress relief mechanisms. The characteristics of those effects are still associated with a high degree of uncertainty. To overcome these uncertainties, an improved and highly advanced process technology is required combining hybrid process models (physical and statistical) with a virtual plant model (digital twin) that enables online simulation of material conditions and properties, e.g. residual stresses responsible for deformations, by use of soft sensors. The locally varying degrees of deformation and temperature profiles during production lead to problems regarding shape stability, straightness and level of residual stresses. To predict the effect of the different processing steps on the quality of the products more accurately, online capable residual stress determining techniques are mandatory to control processes from a perspective of internal stresses as the main influencing factor for shape and straightness. A hybrid approach is chosen combining material, physical and data-based process models to a soft-sensor for residual stresses. It comprises experimental material, analytical and statistical process data to generate knowledge for the realization of a “residual-stress-based” process strategy. A product related container gathers all relevant data and serves as a repository for relevant models to calculate non-measurable data. The data and model container is realized today by the digital twin technology. As a direct online measurement of residual stresses is not possible, the development of a soft-sensor for residual stresses will solve this technological gap and can be used for further process control.

Fu-Yuan Hsu, National United University, Taiwan, Province of China

Co-Author:
Chi-Ming Hung, Metal Industry Research and Development Center
Yu-Lin Feng, Metal Industry Research and Development Center

Abstract:
In the solidification simulation of shape casting, many sophisticated commercial modeling packages are accurate enough to predict the shrinkage porosity in the modeling results. However, it is difficult to produce autonomously a precise shape and geometry of an optimized feeder to avoid shrinkage porosity from the results solely. In this study, a so-called “casting shape analyzing technique (CSAT)” was developed and applied for the prediction of the solidification time for castings with complicated shapes. Using this technique, an optimized feeder for a complicated shape of casting could be calculated and generated. This feeder design is part of the casting process control for producing a good quality casting without shrinkage defects during solidification. For casting with complicated shapes, the relationship between the shape factor and the mold constant was established. Using this equation, the relationship between the solidification time and the geometry of a shape casting was correlated. As a result, solidification time and a required feeder volume could be predicted for an arbitrary casting shape. In this technique, an exact volume of a feeder could also be derived from the magnification equation. The precise shape and volume of the three-dimensional feeder geometry were autonomously produced from the equation. Four different sizes of ductile caliper castings with the same shape were successfully validated using this technique. It accurately predicted the required feeder size for these four calipers. The method and procedure to predict the solidification time and required feeder size for castings with complicated shapes were proposed in this study eventually. Keywords: process quality control, ductile iron casting, shape factor, mold constant, feeder design, shrinkage porosity.

Joonas Ilmola, University of Oulu, Finland

Co-Author:
Jari Larkiola, University of Oulu
Oskari Seppälä, University of Oulu
Joni Paananen, University of Oulu

Abstract:
Digitalization is taking a bigger role in the steel industry. Models for predicting mechanical properties, metallurgical phenomena, roll forces and microstructure have been commonly used in development of novel steel grades. These individual models may predict certain phenomena thoroughly, but input values are usually based on an assumption or on a “good guess”. To produce reliable boundary conditions for these models of individual phenomenon a virtual rolling model is developed. This model computes the whole process of the hot strip mill from roughing to accelerated water cooling on a run-out table. Strip location and temperature evolution is calculated continuously. Thermal and thermomechanical (rolling stands) boundary conditions are placed according to process layout. Input data for the model is automatically read from raw process data. Rolling parameters are calculated using a coupled ARCPRESS model which calculates normal and frictional shear stress distributions in the roll gap predicting roll forces and displacements of the work roll surface. Calculations for static recrystallization are considered in evolution of flow stress. Phase fractions during water cooling are calculated as well. The virtual rolling model minimizes the need for parameter speculation for each simulated process. All the input values are read from actual process data and the metallurgical and mechanical state of the strip are computed through-out the whole process.

Johan Björkvall, Swerim AB, Sweden

Co-Author:
Patrik Strandberg, Outokumpu Stainless AB
Magnus Heintz, Swerim AB
Erik Sandberg Sandberg, Swerim AB

Abstract:
Process models for the simulation and control of metallurgical processes have evolved rapidly in recent decades. Today’s state-of-the-art models are very advanced and can describe real-time chemical and physical phenomena in a straightforward way. However, uncertainties in raw-material properties (chemical composition, specific energy consumption, etc.) limit the process models’ ability to correctly describe the outcome of a particular heat or production sequence given a raw material mix and an operational procedure. Furthermore, the absence of measurements and control of parameters influencing the efficiency of the operational procedure (heat status, lining status, heel status, etc.) also contributes to uncertainties regarding the effect of using specific raw materials. This paper presents a method for supervising raw-material properties based on statistical evaluation of process model calculation errors concerning using different raw materials. The method is applied to detect and correct errors in the estimated chemical composition of charge materials in an electric arc furnace at a stainless-steel plant. A web-based tool for presenting alarms and alternative calculated chemical composition has been developed. Results show that during tests of this tool in industrial trials, the model calculation errors are reduced by 11-16 % by following the tool’s recommendations.