Keywords: Large Scale Complex Systems, Networked Systems, Non Linear Control Systems
Abstract: In this paper, we consider the swarm-control problem of spatially separating a specified target agent within the swarm from all the other agents, while maintaining the connectivity among the other agents. We specifically aim to achieve the separation by designing the movement algorithm of an external agent, called a shepherd, which exerts repulsive forces to the agents in the swarm. This problem has potential applications in the context of the manipulation of the swarm of micro- and nano-particles. We first formulate the separation problem, where the swarm agents (called sheep) are modeled by the Boid model. We then analytically study the special case of two-sheep swarms. By leveraging the analysis, we then propose a potential function-based movement algorithm of the shepherd to achieve separation while maintaining the connectivity within the remaining swarm. We demonstrate the effectiveness of the proposed algorithm with numerical simulations.

Keywords: Networked Systems, Large Scale Complex Systems, Linear Control Systems
Abstract: Distributed cooperative full-state observers estimate the state of a system based on partial estimation at each local sensor node and information exchange among local estimation. An asynchronously distributed cooperative full-state observer is proposed based on a gossip protocol, randomly selecting communication time and a communication node pair. The designed distributed observer achieves the mean square convergence of estimation error.

Keywords: Networked Systems, Distributed Parameter Systems, Optimal Control
Abstract: Coordination of mobile multi-agent systems has been studied intensively for many different purposes. These include swarm robotics, transportation, control of human crowds, and even understanding of collective behavior of animal groups. Mainly, the flocking behavior of multiple agents has been investigated under usually idealized conditions regarding information availability. This paper proposes a model predictive control framework to control flocking agents with limited information. The flocking system is composed of a leader agent and agents that have no information about the group objective and cannot perceive the unknown operation area. The leader agent is a part of the group. However, it has prior information about the group objective and can sense the environment. With the proposed control scheme, the leader steers the flocking agents like a shepherd dog to accomplish the group goal. The presented control scheme is validated in simulation examples.

Keywords: Linear Control Systems, Large Scale Complex Systems, Power and Energy Systems
Abstract: In this paper, we completely describe flatness of the simplest class of interconnected systems: we consider an interconnection of (controllable) single-input linear subsystems in dimension two with a star topology and a linear interconnection dynamics {(with no inputs acting directly on the interconnection variable)}. First, we observe that even if each subsystem is flat, flatness of the global interconnected system is not necessarily preserved. Then, we give necessary and sufficient verifiable conditions for flatness of the interconnected system. When the interconnected system is flat, we analyze how its flat output depends on the interconnection variable and how it can be expressed in function of the flat outputs of each subsystem. Finally, we show how our results can be applied to electrical power systems.

Keywords: Power and Energy Systems, Large Scale Complex Systems
Abstract: Analysis of the small-signal stability of power system is commonly based on the swing equation model. Due to the special structure of the power grid swing equations, an equilibrium set corresponding to a so-called frequency equilibrium has to be analyzed. We present two new and short proofs for two tests by Skar concerning both the nonuniformly and the uniformly damped system. Based on the latter we derive a simple stability test for uniformly damped systems, which does not rely on eigenvalue computations. The nonhyperbolicity of the equilibria in original coordinates is tackled by the concept of normally hyperbolic invariant manifolds. The derivations are completely based on the theory of quadratic pencils.

Keywords: Large Scale Complex Systems, Stochastic Systems, Networked Systems
Abstract: In this paper, we set up a mathematical framework for the modelling and control of complex cyber-physical ecosystems. In our setting, cyber-physical ecosystems (CPES) are cyber-physical systems of systems that are highly connected. CPES are understood as open and adaptive cyber-physical infrastructures. These networked systems combine cyber-physical systems with an interaction mechanism with other systems and the environment (ecosystem capability). The main focus will be on modelling cyber and physical interfaces that play an important role on the control of the emergent properties like safety and security.

Keywords: Networked Systems
Abstract: This paper aims at modeling the optimal behavior for an attacker that has the objective to maliciously manipulate the output of an industrial power plant, which is fed via a public network to a digital twin in order to reconstruct the state. In particular, we consider a scenario where the attacker seeks a tradeoff between two conflicting objectives: dealing the maximum damage in terms of the norm of the estimation error for the observer and keeping the magnitude of the variation of the systems’ output to a minimum. In doing so, we assume the observer is equipped with a bad data detector and the attacker must choose the injected signals in a way that guarantees that the bad data detection condition is not triggered.

Keywords: Networked Systems, Large Scale Complex Systems, Control Design
Abstract: Cyber-Physical Systems are complex systems that integrate physical processes and communication networks like critical infrastructures and industrial plants. Unfortunately, the integration of physical and cyber layers causes also possible issues such as the increased surface of cyber-attacks. A possible example of CPS is the industrial world, which is one of the most important targets of cyber-attacks. Applying protecting architecture as the ones developed for the IT world is not possible due to the specific features of the industrial environment. However, the paper exploits the features of the industrial communication networks to develop an industrial intrusion detection system named Smart Security Probe. This solution has been designed to detect possible anomalies in the network traffic and to help to infer possible anomalies in the data related to the physical processes. S2P has been tested and validated in an environment made of two Programmable Logic Controllers and two Supervisory Control and Data Acquisition systems that are controlling four simulated tanks. The anomaly detection is based on a couple of Interlaced Extended Kalman Filters that are distributed among the controllers and exchange data securely through the Smart Security Probe. The results demonstrate the feasibility of the proposed solution.

Keywords: Power and Energy Systems, Large Scale Complex Systems, Control Design
Abstract: Recently, the concept of resilience of electrical infrastructures has been introduced to quantify the ability of the grid to resist, adapt to, and rapidly recover after the occurrence of high-impact and low-probability (HILP) events. Various surveys discuss the state of the art on the resilience of distribution networks (DNs), which are a subsystem of the electrical infrastructure particularly susceptible to HILP events. It emerges that automation has a central role in guaranteeing and enhancing DNs resilience, although a classification of the existing contributions is missing. To fill this gap, in this paper we review the literature contributions regarding decision and control methods to enhance the resilience of DNs. We classify the reviewed approaches into tactical/strategic and operational level ones, we group them by the time of application and type, and finally we provide a detailed discussion and comparison of the available methods, highlighting open issues and future trends in this field.

Keywords: Safeprocess, Human Machine Systems
Abstract: Due to their essential role, critical infrastructures (e.g., water, gas, and power distribution systems) are subject to persistent monitoring in order to ensure their operational continuity. Because of this, they constitute appealing targets for malicious attackers who carry out physical or cyber attacks with the aim of compromising such critical systems. In this work, we provide a novel framework for an enhanced risk assessment process for critical infrastructures, which is based on the Analytic Hierarchy Process. Specifically, the proposed solution consists of a quantitative framework for site-specific risk assessment, and follows an approach designed to consider the presence of heterogeneous subsystem characterized by different degree of relevance in the infrastructures. A simulation campaign is carried out in a test-range environment, which emulates the behaviour of a water treatment system, in order to prove the effectiveness of the approach

Keywords: Modelling and Control of Biomedical Systems, Manufacturing Plant Control, Computers for Control
Abstract: The goal of this paper is to explore proposals coming from different EU-RFCS research funded projects, in such a way that cybersecurity inside the steel industry can be increased from the Operational Technology area, with the current level of adopted Information Technology solutions. The dissemination project ControlinSteel has reviewed different projects with different strategies, including ideas to be developed inside the AutoSurveillance project. Advanced control process strategy is considered and cloud based solutions are main analysed alternatives. The different steps in the model lifecyle are considered where different cloud configurations provide different solutions. Advanced techniques such as UMAP projection are proposed to be used as detectors for anomalous behaviour in the continuous development / continuous implementation strategy, suitable for integration in processing workflows

Keywords: Manufacturing Modelling for Management and Control, Enterprise Integration and Networking, Modelling, Identification and Signal Processing
Abstract: Manufacturing systems in steel industries are characterized by high complexity and high temperature and pressure conditions during production. Industries have to speed up their production to meet the market’s demand for products in a fast changing economy. To prevent breakdowns in the manufacturing lines and further economic loss, steel industries utilize preventive maintenance approach and early replacement of equipment, which is expensive and not optimal. Preventive maintenance can be beneficial in the steel industry and reduce costs, if it is supported by information gathered from previous breakdowns in the production line, such as condition of equipment, environment and further data that can be collected. In this work, historical data and data collected from a digital twin representation of the manufacturing line from Pittini, a steel making industry in Italy, were utilized to gain information on the conditions before a breakdown in the production line. Furthermore, we present a cloud based framework created by utilizing the information and data for optimization and real-time driven preventive maintenance approach and remote control.

Keywords: Control Design, Adaptive and Learning Systems, Networked Systems
Abstract: The ControlInSteel project, a cooperation of four research institutes, revisited research projects of the last 20 years focusing on automation and control solutions applied to the downstream steel production route. During this investigation, several insights were found regarding the solutions strategies, the problems encountered, and which types of solutions were beneficial for specific problems. For this analysis, 46 projects were systematically reviewed. Taxonomies for the problem space, the solution space, the barriers and issues and the impact were developed and each project categorized along these taxonometrical dimensions. As a result, the interdependencies between solutions and impact could be analysed in a quantifiable way, which led to a new way of evaluating project success. It also brought new insights about the most promising techniques already applied and those techniques, that have been apparently not yet been applied to steel production, although being highly successful in other domains. This leads to potential future research chances for the steel production and their complex process chains. The paper will also finally demonstrate how a similar taxonometrical approach can be used to conserve expert knowledge in automation to feed a truly artificially intelligent control solution not only exploiting machine learning methods but essentially using machine reasoning on top of the digitized expert knowledge to achieve improved process automation.

Keywords: Optimal Control, Modelling and Control of Environmental Systems, Control Design
Abstract: Steel production in integrated steelworks involves the simultaneous production of various by-products, including process off-gases that are usually exploited for generating electricity in the internal power plant, heat and steam. Their discontinuous production is managed through complex network, gasholders and torches, which must be managed with stringent operational constraints. In this paper we present a supervision and control system designed to optimize the economic management of the distribution of process off-gases that also allows minimizing the environmental impact. The system implements a digital twin based mainly on machine learning techniques, including Echo State Networks, and a hierarchical optimization system, which first level is based on an economic model predictive approach and the second level is based on the economic hybrid model predictive control. This system allows to effectively maximize the use of off-gases while minimizing the environmental impact of their use up to 97%.

Keywords: Manufacturing Modelling for Management and Control, Modelling, Identification and Signal Processing, Control Design
Abstract: In steel manufacturing, the rolling step defines the major properties of final products of the steel industry which are delivered to a wide range of different industrial sectors. Modern rolling mills, to reach high degree in process supervision and efficiency, installed sensor equipment that delivers masses of data and information about the process, the product and its quality at high sample rate and at high spatial resolution. But there is a misalignment between the applied high-tech equipment and the online exploitation of such measurements to describe their impact into the product properties. To make use of the high amount of online available data, on 2018 started the RFCS project NewTech4Steel aimed to enhance process stability and product quality in steel production by exploitation of break-through technologies for real-time monitoring, control and forecasting inspired by Big Data concepts. Within this project, on the basis of the strict correlation between the cold rolled strip flatness and downstream process productivity (HDG & painting line), it has been investigated at Marcegaglia Ravenna plant the case-study of managing the massive amount of production data and making them operational by leveraging machine learning algorithm in order to optimize the global productivity. The project allows fast and online data processing and in turn fulfills the plant line needs for: prediction of defect related to manifested coil flatness imperfections and real-time prediction of process parameters during the zinc coating to avoid break/sideslip on critical conditions. The solution created in NewTech4Steel is inspired by Lambda Architecture as the unified platform for data and analytics which is composed of three layers: batch processing layer for offline data, serving layer for preparing indexes and views and speed layer for real–time processing. The work compiled the necessary measures to integrate the new technologies into the existing IT systems especially under the aspects of brown-field implementation, the connection of newly developed systems to existing ones, and the requirements for HMIs for user interaction and visualization.

Keywords: Human Machine Systems, Large Scale Complex Systems, Adaptive and Learning Systems
Abstract: Under the push of the European Green Deal, a transformation to “make things” is taking place. The Danieli Intelligent Plant DIP has the ambition to enable the Sustainability of metals manufacturing. DIP is a new fully immersive control and supervision model based on augmented, mixed and virtual reality sensing and processing technologies. Processes, machines and equipment are holistically integrated with operators/decision makers using data-driven approaches and AI and ML. This allows to extensively control, supervise processes and machines characterized by large degree of autonomy to optimize processes, quality and maintenance operations. Firstly, a new concept of HCI (Human Computer Interface) has been realized based on the evolution of 3Q philosophy allowing local and remote monitoring and control of processes through the same approach customized for different technological scenarios. Remote cooperation, assistance and maintenance is also part of the environment for having no-man-on-floor increasing safety and putting the focus on control and human centrality. Not less important, the extensive use of mobile solutions (wireless Local Control Station) have an important role to reduce the human factor whenever a shop floor operation is mandatory.

Keywords: Modelling and Control of Environmental Systems, Manufacturing Modelling for Management and Control
Abstract: Resource consumption is an important topic for steelmaking industry, which is spending significant efforts to reduce its environmental impact and improve its competitiveness. Water is largely exploited in steelworks for indirect and direct cooling, specific surface treatment, and fumes washing and cooling. It is already reused and recycled after restoring its quality through treatments for temperature and/or pollutant reduction. However, sometimes water networks are not optimized due to outdated water treatments, lack of continuous monitoring, and water network management strategies often based on experience without automation. In recent years, new water treatments, simulation, and optimization tools are becoming available, together with a stronger awareness of the importance of online parameters monitoring. Therefore, improvement of water cleaning, reuse, recycling, and consequent reduction of impact related to water exploitation are potentially achievable. The introduction of innovative treatments must be tested before their implementation in steel plants and the exploration of their behavior in different operating conditions is fundamental. The presented work addresses this topic through the application of several models of operational units, developed in OpenModelica environment and aggregated into a plant simulator. The simulator was used in different case studies related to an Italian plant to assess the impact of new filtering technology for reducing suspended solids on the analyzed water networks and test the effects of different operating configurations on the treatment efficiency. The introduction of new filtration technology leads to environmental and economic advantages due to freshwater intake reduction and water management improvement.

Keywords: Adaptive and Learning Systems, Modelling, Identification and Signal Processing, Manufacturing Modelling for Management and Control
Abstract: The highly individualized production processes for long products in the steel industry is subject to a variety of influencing variables with mutual interactions in a complex manner. To handle this complexity, modern data mining methods can be used for a highly efficient analysis of process data, to detect process anomalies in the process data, e.g. from rolling mills by statistical pattern recognition. This paper proposes a data-based strategy for detecting process anomalies within a hot rolling mill for long products. Suitable data is identified and selected from existing sensors and processed within a new database. This central database is used to train classification algorithms. The reliability of two prominent classifiers based on Principal Component Analysis (PCA) and One-Class Support Vector Machines (OC-SVM) has been evaluated. From the comparison in this respective use case, it has been concluded that satisfying results can be obtained, but PCA is highly dependent on the data distribution. The OC-SVM has also been implemented and tested and offers advantages when the data sets have a more complex distribution.

Keywords: Control Design, Non Linear Control Systems, Robust Control
Abstract: The prospect of modifying neural activity in a principled way, could facilitate the understanding of brain functions and the development of medical treatments. To predict the dynamics that underlie the different brain activities, several neurobiological models have been proposed, either focusing on individual cells or whole populations. In this context, control systems are a powerful tool to provide a correct articulation between inputs, i.e. neural stimuli, and observables, i.e. system outcomes. Based on well-established neurobiological hypotheses, this study presents a control framework to regulate a neural-mass activity, with potential uses for pattern tracking, such as, rhythm evoking and phase synchronisation. Being these mechanisms closely connected with real brain computations, this study is carried out using a meaningful perspective in terms of biological interpretation. To this end, the Wilson-Cowan model is used, where the input stimuli is elicited through light signals applied to genetically modified neurons that express light-gated actuators. Thus, this study states a crucial proof of concept towards a future experimental application of the control framework for neurobiological systems.

Keywords: Modelling and Control of Biomedical Systems, Modelling, Identification and Signal Processing
Abstract: Thyroid cancer is a type of disease that affects the thyroid gland. The primary diagnosis of thyroid tumors based on histopathological patterns can be ambiguous in some cases, so the use of machine learning techniques might improve the classification of thyroid diseases. Moreover, the lack of high sampled datasets makes the classification issue even more complex. This paper proposes a comparative evaluation of two classical machine learning techniques and one Bayesian network framework. We use Exploratory Data Analysis techniques and oversampling methods for data preprocessing and overfitting reduction. Results show that the use of Bayesian network frameworks can help in integrating prior expertise knowledge in the classification problem and build new hypotheses about features interaction.

Keywords: Modelling and Control of Biomedical Systems, Non Linear Control Systems
Abstract: In vitro simulators give the chance to mimic specific human physiological/unphysiological conditions to test medical devices, accelerating innovation cycles, and rapidly exploring new effective solutions. Specifically, in this work, we consider a Pulse Duplicator in use at the Healing Research Laboratory, at the University of Padova, Italy. It allows assessing the performance of prosthetic heart valves under simulated cardiac conditions by generating a controlled pulsatile flow obtained by setting specific system characteristics such as the peripheral resistance and the compliance. In order to conduct effective prosthetic heart valve tests, the Pulse Duplicator has to be manually tuned to provide suitable hemodynamics waveforms (i.e. flow and pressure waves). To assist this time-consuming and not trivial task, we present a Phase Plane Analysis of a particular signal (i.e., the left ventricular pressure) in order to capture additional hemodynamic characteristics of clinical interest that can serve system tuning purposes.

Keywords: Modelling and Control of Biomedical Systems, Computers for Control, Human Machine Systems
Abstract: The approach of strategies aimed at the development of an artificial pancreas (AP) to control glucose levels in patients with type I diabetes mellitus are often insufficient and unsatisfactory. High glucose levels after postprandial episodes that aggressively try to be compensated with high insulin rates, not only cannot guarantee blood glucose levels within ranges characteristic of healthy subjects, but also expose the patient to possible hypoglycemic conditions. Many of these approaches including artificial intelligence, control theory - based on techniques could experiment an unexpected poor behavior since most of them do not include neither meal intake announcements nor meal size information as suggested for later studies. In order to overcome that issue some meal count-detection algorithms reported in scientific publications have shown not only a good performance on blood glucose regulation but less hypoglicemia and hypoglucemia events too. In this paper, we cover supervised learning based techniques for meal detection and carbohydrates counting. For meal detection, decision boundaries are shown for each algorithm in order to get better understanding on how the algorithms perform. Otherwise, for meal estimation, regression algorithms are trained with purpose of compare the prediction for several scheduled meals looking for the one which fits better the prediction against real measured glucose curve within a postprandial time horizon. The algorithms are validated on Food and Drug Administration (FDA)-Approved UVA/PADOVA Type 1 Diabetes Simulator.