Marcus Neuer

Verified

Marcus verified their affiliation via an institutional email.

Learn more

Verified

Marcus verified their affiliation via an institutional email.

Learn more

• Dr. rer. nat. Dipl.-Phys.
• Lecturer at RWTH Aachen University

Technical processes are described by stochastic variables. They reflect the uncertainty that every measurement process is subject to. This chapter introduces both the stochastic fundamentals for working with uncertainties and probabilities, as well as the necessary statistical tools to draw initial conclusions from them. For machine learning, these...

Starting with a review of the principal component analysis (PCA), the chapter explores canonical algorithms of unsupervised learning. It presents cluster approaches like K-Means, Mini-Batch-K-Means and the t-Student distributed neighbour embedding (t-SNE). For all techniques, Python examples are described that help to understand the application of...

Data preprocessing deals with those works necessary to put data in machine learning algorithms. Keeping a physics-informed perspective, we also introduce important transformations and filters to improve the performance of learning techniques. Normalization and triggering are discussed, together with convolution filters for moving average and deriva...

The chapter focuses on supervised machine learning, where algorithms learn to project input data on dedicated desired values. Starting with a simplistic evolutionary algorithm, the abstract idea of learning and its mathematical and programmatic concepts are presented. This approach is then applied to the Least Mean Squares adaptive filter, which is...

While machine learning is a subfield of AI, this chapter discusses several semantic methods to support the learning models and to enable explainability. Synonyms, taxonomies and ontologies are introduced. They digitize the true meaning of something as knowledge models, allowing us to store this meaning and make it accessible for computer programs l...

Data is fundamental for modelling technical processes and production steps. Following a short historical review, the chapter introduces ways to describe data and to characterize it. Structured, semi-structured and unstructured data are discussed and their relevance in common applications is explained. All necessary concepts are reviewed from the pe...

While machine learning is one subfield of AI, this chapter discusses several methods from another subfield, semantic modelling. It supports the learning models and is an enabling technique for explainability. Synonyms, taxonomies and ontologies are introduced. They digitize the true meaning of something as knowledge models, allowing us to store thi...

Lanthanum bromide, typically doped with Cerium or Strontium (LaBr3:Ce,Sr) has become a versatile detector material in the high resolution market of scintillators. With an FWHM often better than 3%, the material is a notable attractor for homeland security applications. Due to its intrinsic activity of La-138, this scintillator shows a complex backg...

The paper presents an anomaly detection approach for cyber attacks on radiation detectors. It is based on specifically trained Restricted Boltzmann Machines (RBM) and Autoencoders (AE). Radiation detectors are relevant for various security missions. Some of them are integrated in networks and therefore prone to cyberattacks. The latter is a threat...

The term cybernetic spectroscopy detector (CSD) refers to a sensorial unit that seamlessly integrates into large-scale detection solutions. It interacts, communicates and even controls other detectors or components, while providing models and data for the interplay with these units.. Such a smart detector is elementary for realising portal solution...

Maschinelles Lernen und künstliche Intelligenz sind omnipräsente Begriffe zur Verbesserung von technischen Prozessen. Die praktische Umsetzung an realen Problemen gestaltet sich aber oft schwierig und komplex. Dieses Lehrbuch erklärt Lernverfahren anhand von analytischen Konzepten im Zusammenspiel mit vollständigen Programmierbeispielen in Python...

Anomaly detection plays a central role in monitoring industrial processes. The common understanding is that the time series of a variable contains visible traces of anomalies - which can be selected and validated by suited algorithms

Nuclide identification codes have been extensively tested for their performance regarding special nuclear material (SNM) in the past. A main focus was put on the identification quality for masking scenarios. Normally strong medical masks are mixed with weak traces of the SNM, to find the limiting ratio, at which the identification code can no longe...

A blockchain algorithm approach is shown that hashes the original data and stores the hash information as part of the spectrum file. The procedure is based on the spectroscopic information, mathematical transformations and even additional information from the detector itself. The resulting N42.42 compatible file remains in its standardised form and...

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, but often the results of existing AI models requir...

Unsupervised learning strategies have proven successful for detecting anomalies in various applications. UMAP, the uniform manifold approximation and projection technique, is a transformation that maps complex input data to a reduced dimensional map. It is applied to real-world spectral data of Plutonium, measured with Sodium Iodide (NaI:Tl), Lanth...

A gamma enrichment analysis is presented that utilises physics-informed neural networks (PINN). Such networks exploit existing mathematical formulas and analytic transformations, to provide the best possible starting for a machine learning training - they hybridise the machine learning perspective with established analytic relationships. Autoregres...

Mobile radiation detection requires large volume detector equipment and high sensitivity. The paper presents a shapelet extraction method that is applied to the time series spectral data of such a mobile search system. The detection capability of a system is often limited by varying background effects. With the technique, time pattern are analysed...

Der Vortrag blickt auf 10 Jahre angewandter Forschung auf den Gebieten der Digitalisierung, des maschinellen Lernens und der Optimierung von Regelungen zurück. Der Digitale Zwilling wird als Kernelement einer autonomen, modell-basierten Metaregelung identifiziert, da er in der Lage ist seine simulativen Modelle einzusetzen, um direkt in Produktions...

Semantic tools are applied to evaluate the problem-solution-impact dependencies in EU research projects. A set of 46 projects was selected, which focused on automation and control research for steel production. The talk reports on assorted results and indicates how steel production could improve by adopting new technologies.

Evaluation of the most relevant methods used in RFCS projects regarding automation downstream.

All projects disseminated by ControlInSteel were carefully assessed with respect to their taxonomical dimensions.

Dimensional analysis yields a clear disadvantage of worker safety and worker performance related impact. Compared to quality improvement, which was by far impacted by the highest number of projects, enabling technologies, power consumption and emission reduction are considered to be underrepresented as well.

With over 45 high profile research projects, funded by the Research Fund for Coal and Steel (RFCS), the project team revisited the problems, solutions, barriers and impact of these works. Using semantic tools like taxonomies and ontologies, it was possible to establish a database with project results at www.controlinsteel.com to allow people, stake...

Measuring soil moisture has been in focus of research works in the field of radiation physics for a long time. Some previous works use neutron detectors stationed on soil fields. They analyse the ratio of thermal and epithermal neutrons, as cosmic neutrons are attenuated by the water contained in the soil. Contrarily to those concepts, the present...

Automation plays a key role for the production processes in steel industry. With a complex and especially interdependent chain of transformation, heating, cooling and refinement stages, each process features a diversity of individual challenges with respect to automation. Especially the Industry 4.0 efforts and the groundbreaking changes induced by...

Industry 4.0 introduced methods for the efficient integration of data across steel plants. An automatic exploitation of this data for process automation is therefore now of central importance. Driven by the latest research and an increasing amount of available industrial process data, diverse techniques of machine learning, a subfield of artificial...

In December 2014, the BBC [1] and the Heise Security Service [2] unanimously reported that a blast furnace in a German steelworks had suffered "massive damage" following sabotage of the plant's computer network. This information was based on a report by the German Federal Office for Information Security (BSI) [3]. Today, in the age of Industry 4.0,...

German presentation to an expert audience in automation at the iba Tag 2021. This talk is part of the ControlInSteel dissemination activities from 2020 - 2022.

The implementation of artificial intelligence faces different challenges of infrastructural, data related, security related and social scope. These aspects are discussed, reflecting on the requirements of introducing such a technology in a broader way. Machine learning, as subfield of artificial intelligence, benefits from advances in Big Data scie...

An example of anomaly detection for industrial production processes is shown, based on a variational autoencoder. Starting with a simple probability corridor, the concept of autoencoding is introduced and variational encoding is reviewed. With an input enrichment that utilizes difference optimization of a two class test set, the varational encoder...

Steel production relies on profound understanding of the involved processes as well, as on a consequent adoption of groundbreaking technologies. One such technology, igniting an impressive manifold of solutions for data analysis, optimization and process control, is machine learning. Following in the wake of Digitalisation and Big Data, it appears...

One main challenge in the application of modern digital concepts is to bring together two very different worlds: the real production world and the cyberspace. The latter is a virtual world which helps to acquire and manage data, calculates scheduling, allocates products to customers and sets up the processes. Their combination forms a cyber-physica...

The invention relates to a measuring roller (1) for determining a property of a strip-type product, in particular a metal strip, guided over the measuring roller.

Artificial intelligence is impacting various areas of research and industry. In spectroscopic applications, the results and measurements are heavily influenced by means of algorithms. These help to interpret the results of a measurement by introducing prior knowledge, a different perspective or even fuse information gained from secondary sensors. T...

A method is shown to improve the accuracy of dose rate measurements for handheld instrumentation over the range of 50keV to 3MeV. Dose rate estimation in handheld devices is usually done by summing up the channel data of the radiation spectrum. In this sum, weights are applied to correct for the sensitivity of the used detector and allowing for a d...

A machine learning method is shown that deconvolves spectra in a single step procedure. This method is based on a special neural network, Bernoulli type restricted Boltzmann machine (RBM). Using sophisticated Geant4 simulations to produce mono energetic line response vectors of a Sodium Iodide detector, this RBM is trained to learn the detector res...

A technique is shown to estimate the location of a source with a single detector. The method exploits an angular histogram of counts calculated by the fusion of the digital compass data with spectroscopic quantities like the count rate. The perimeter around the user is divided into N discrete segments. Storing the count rate in dependency of the an...

Cyber-attacks on the European steel producers have become a real threat, where massive production breakdowns are rare, but very expensive events. Companies prepare for these problems, often only focussing on the IT level, where firewalls, internet restrictions or staff compliance rules are enforced to secure the on-premise networks. But in the era...

Auxialiary aggregates like the cooling of rolls or ventilation systems show reasonable potential for economic and ecological impacts. But those units have only been scarcely considered in optimisation attempts before and there was an urgent need to explore how auxialiary aggregates can be efficiently automatized by smart control strategies. The tal...

Elaborating the difference between digital shadow and digital twins, the talk introduces concepts and technologies for putting digital twins into practice in the process industries. While many people think that digital twins are more or less simulations, the talk also points out the importance of interaction with the processing, which is one defin...

The broad progress of digitalisation provides an increasing amount of online available data, often large data streams, from all relevant production processes. In many cases, this data is stored and used for posterior data analytics to explore the reasons for malfunctions or product degradation. The talk presents new search and monitoring techniques...

In the challenging international steel market, European steel producers must face strong competition. Technological leadership, especially regarding production process optimization helps companies to be ahead of the market game. One key enabler is modern process automation, which aims for yield increase, better energy efficiency, constant or improv...

Cyber-attacks on the European steel producers have become a real threat, where massive production breakdowns are rare, but very expensive events. Companies prepare for these problems, often on the IT level, where firewalls, internet restrictions or staff compliance rules are enforced to secure the on premise networks. For the staff this topic often...

The concept of digital twins is presented. Twins of products, machines and sensors can talk to each other, not only exchanging data but also results of prediction models. In this sense, the twin utilises not only existing information, it autonomously evaluates mathematical or statistical models, that help to anticipate potential future states. Th...

An introductory lecture on digital twins. The difference between digital shadow and digital twin is discussed. The fundamental ability of a twin to interact with the process chain of its real counterpart is presented, following a rigorous development of the communication stability using graph theory and its relation to a Dirac-notation based state-...

Various methodological aspects and decision modules developed by the VDEh Institute for Industrial Research which show the benefits of digitization of the steel industry have been presented. Understanding the structure and framework conditions of interdepartmental production data, smart management of a big data array, semantic modeling of the techn...

• Neue Technologien in der Stahlindustrie erfolgreich erproben und zum Einsatz bringen • Digitale Zwillinge von Produkten und Maschinen werden zu aktiven Komponenten und ermöglichen neue Lösungsstrategien • Wohin führen uns z.B. Technologien wie Blockchain oder Tangles? Werden durch sie sogar neue Produkte möglich?

The description of physical processes and perhaps nature as such was always a main desire and motivation that propelled research in computer science. Simulation complexity grew, providing the ability to compute an increasing number of difficult equations. The mathematical discipline of numerics could unfold its full potential and it´s fundament was...

Digital twins can become players in cross-process optimization tasks. To do that, a passive data twin must become more than a static data collection. This aspect of “activity” is realized when twins are implemented as agents, small autonomous computer programs. Such agents can essentially interact with their environment, either by communication or...

The paper presents a distributed intelligence-based system concept for modeling and self-optimizing the production of flst steel products. The concept is implemented through a multi-agent system approach. In particular, three types of autonomous agents are modelled, each playing a different role in the steel production chain. The agents are designe...

Steel is produced via a series of different, interdependent processes. In the past, most of these processes have been individually optimized, but rarely considered constraints of follow-up aggregates. A huge improvement potential can therefore be achieved by optimizing the process chain as such, introducing smart processes and smart products that a...

Werkstoff 4.0 ist schon keine Vision mehr: Welche Möglichkeiten Werkstoffdaten im Prozess und bei der Verarbeitung bieten, hat sich in Forschungsprojekten gezeigt. Davon können Medizinprodukte profitieren. https://medizin-und-technik.industrie.de/digitalisierung/digitaler-zwilling-werkstoff-4-0/

The rise of renewable energy sources increases power fluctuations in the power network due to daily and seasonal effects which have sensible impacts on pricing and tariffs. According to the Smart Grid Initiative the closer cooperation between electrical energy suppliers, retailers, network operators and authorities on one side and consumers on the...

Deontic logic allows the formal description of obligation and permissions. A reasoning algorithm is presented, that applies these formal concepts for enhancing the performance of an existing nuclide identification algorithm. Radio-isotope identification devices rely on a robust deduction of those nuclides, contributing to the radiation field. Tradi...

Most spectroscopic applications, especially homeland security relevant equipment relies on well-calibrated spectral data. A Kalman filter is applied to the known problem of stabilization, which means online calibration of a scintillation detector to a predefined reference peak. The method uses stochastic assumptions on the measurement uncertainties...

Kommunikation ist ein zentrales Element um die Integrationsbestrebungen der Industrie 4.0 Vision zu realisieren. Agenten sind ein Werkzeug um Prozess- und Werksübergreifende Kommunikation zu ermöglichen, gleichsam erlauben sie eine verteilte, dezentral Optimierungsstrategie anzusetzen. Der Weg hin zum digitalen Zwilling wird exemplarisch für die St...

Digitalisierung ist Antrieb einer sehr umfassenden Wandlung. Um die Vielzahl an gesammelten Daten einer Produktion gewinnbringend zu nutzen, benötigen wir den digitalen Zwilling des Produkts. Er wird zum Werkstoff 4.0, und verleiht dem Produkt Intelligenz und die Fähigkeit selbst in seinen eigenen Produktionsablauf einzugreifen. Kommunikation, das...

Es gibt weltweit einen erhöhten Bedarf an Stahl, aber die Stahlherstellung ist ein enorm anspruchsvoller und kostenintensiver Prozess, bei dem gute Qualität schwer zu erreichen ist. Die Verbesserung der Qualität ist noch immer die größte Herausforderung, der sich die Stahlbranche gegenüber sieht. Das EUProjekt PRESED (Predictive Sensor Data Mining...

Digitalization induces far reaching consequences, leading to a novel industrial era envisioned as Industry 4.0. Complete computational coverage of the production chain and the products fosters vertical and horizontal integration, allowing a cross-process, holistic optimization for the production as such. Products and processes become intelligent by...

Digitalization induces far reaching consequences, leading to a novel industrial era envisioned as Industry 4.0. Complete computational coverage of the production chain and the products fosters vertical and horizontal integration, allowing a cross-process, holistic optimization for the production as such. Products and processes become intelligent by...

Pickling lines are often considered as a bottleneck of steel production. Increasing the throughput has been the declared goal of many research projects. Yet, high quality pickling results depend on a complicated mixture of constraints like careful choice of pickling speed, suited entry temperature and a good estimate about how many scale resides on...

The European steel industry identified digitization as a key enabling technology not only for maintaining its world-wide leading role with respect to product quality, but also for pushing the limits and raising efficiency, reliability and quality to a next level. Recently, many steel companies updated and overhauled existing infrastructures and pre...

European steel is decisively driven by product quality and customer demand. Industry 4.0 is yet a blurry term, nowadays being praised for bringing disruptive advantages for modern manufacturing industries in the next decades. Many people are overwhelmed by the diversity invoked by the word Industry 4.0 and although many applications are emerging, p...

Final report for the EU RFCS project I2MSteel conducted from July 2012 to December 2015.

A technique is presented to optimize the throughput of a pickling line by combining recent Industry 4.0 architecture with process models. So-called autonomous agents act as virtual representatives of the coils and the processes. These smart objects have access to the surface inspection data after hot rolling and after the pickling line, as well as...

The steel industry is an important consumer of electrical energy having a significant impact on the electricity network and accounting to a significant part of production costs. Thus, there is the opportunity of closer cooperation between grid operators and steel industry to improve the power consumption prediction and actively contribute to a secu...

A technique is presented to algorithmically evaluate prompt gamma neutron activation spectra, which were produced through excitation of specific material samples. The excitation is done with a neutron generator that provides a switchable, artificial form of neutron radiation. To evaluate these spectra, a mathematical transformation based on derivat...

A method is presented that uses spectroscopic data and the sensorial output of a gyroscope to create a soft sensor for the incident direction of the radiation. The method can be applied by any handheld radiation detection equipment that is continuously moved by the operator in some regular waving or panning motion. The gyroscope output consists of...

Scattering and attenuation have significant impact on the spectrum. For submersible nuclide identification devices, it is mandatory to still provide high quality identification results, even if the instrument is in the water. A technique is shown, to treat the effects of water scattering. This is done by combining the so-called dynamic derivative c...

A multi-agent system for the steel production is presented. It re-allocates off-spec products to existing alternative orders. The system consists of three main Industry 4.0 components: agents, service-oriented architecture (SOA) and a semantic modelling of the process chain. Using economic principles like virtual marketplaces, the solution of the r...

Daily production of hot-rolled steel is analyzed from the point of view of its correspondence to aimed strip thickness. New distribution of steel products using 12M Steel is suggested. Possible technological roots for hot rolled strip are displayed with explanation of operating algorithm by ArcelorMittal for efficiency and suitability of taken tech...

Stahl wird in einer Reihe verschiedenster Prozesse hergestellt. Die vorliegende Arbeit beschäftigt sich mit der Umplanung von Stahlprodukten im Falle von Prozessfehlern oder Störungen. Eine alternative Prozessroute wird vorgeschlagen, in Abhängigkeit der noch sinnvoll erreichbaren Produktqualität und den synchron gerade vorliegenden Aufträgen durch...