- Email: [email protected]
A Smart Maintenance tool for a safe Electric Arc Furnace
12th IFAC Workshop on Intelligent Manufacturing Systems December 5-7, 2016. Austin, TX, USA 12th IFAC Workshop on Intelligent Manufacturing Systems December 5-7, 2016. Austin, TX, USA 12th IFAC Workshop on Intelligent Manufacturing Systems December 5-7, 2016. Austin, TX, USA Available online at www.sciencedirect.com December 5-7, 2016. Austin, TX, USA
ScienceDirect IFAC-PapersOnLine 49-31 (2016) 19–24 A A Smart Smart Maintenance Maintenance tool tool for for a a safe safe Electric Electric Arc Arc Furnace Furnace A Smart Maintenance tool for a safe Electric Arc AFumagalli*, Smart Maintenance tool forColace**, a safeMaurizio Electric Arc Furnace Furnace Luca Marco Macchi*, Cristian Rondi**, Alessandro Alfieri*
Luca Fumagalli*, Marco Macchi*, Cristian Colace**, Maurizio Rondi**, Alessandro Alfieri* Luca Luca Fumagalli*, Fumagalli*, Marco Marco Macchi*, Macchi*, Cristian Cristian Colace**, Colace**, Maurizio Maurizio Rondi**, Rondi**, Alessandro Alessandro Alfieri* Alfieri* *Department of Management, Economics and Industrial Engineering, Politecnico di Milano, P.za Leonardo da Vinci 32, *Department of Management, Economics and Industrial Engineering, Politecnico di Milano, P.za Leonardo da Vinci 32, 20133 Economics Milano, Italy - Contact author's email: [email protected] *Department and Industrial Engineering, Politecnico 20133 Economics Milano, Italy author's email: [email protected] *Department of of Management, Management, and- Contact Industrial Engineering, Politecnico di di Milano, Milano, P.za P.za Leonardo Leonardo da da Vinci Vinci 32, 32, 20133 Milano, Italy Contact author's email: [email protected] 20133 Milano, Italy - Contact author's email: [email protected] **Tenaris Dalmine S.p.A., Piazza caduti 6 luglio 1944 n.1 - 24044 Dalmine (BG), Italy **Tenaris Dalmine S.p.A., Piazza caduti 6 luglio 1944 n.1 - 24044 Dalmine (BG), Italy **Tenaris **Tenaris Dalmine Dalmine S.p.A., S.p.A., Piazza Piazza caduti caduti 66 luglio luglio 1944 1944 n.1 n.1 -- 24044 24044 Dalmine Dalmine (BG), (BG), Italy Italy Abstract: The production process in steel making involves plenty of variables to know the actual status Abstract: The production process in steel making involves plenty of variables to know the actual status in plant operations and the deviations from optimal involves and safe operating conditions. The deviations depend in steel of variables to know the status Abstract: The production in plant operations and the process deviations from making optimal involves and safe plenty operating conditions. The depend Abstract: The production process in steel making plenty ofgeneral, variables tocomplexity knowdeviations the actual actual status on diverse uncertainties due to the material variability and, more in the of modeling in plant operations and the deviations from optimal and safe operating conditions. The deviations depend on diverse uncertainties due to the material variability and, more in general, the complexity of modeling in plant operations and the deviations from optimal and safe operating conditions. The deviations depend the diverse physical and chemical transformations occurring in the production assets. The transformations are on due to the material and, in general, the complexity of the physicaluncertainties and chemical occurring in themore production assets. transformations are on diverse uncertainties duetransformations toof themeasurable material variability variability and, more in in general, the The complexity of modeling modeling partially observable by means parameters. The “man theassets. loop” is then essential to cope the physical and chemical transformations occurring in the production The transformations are partially observable by means of measurableoccurring parameters. Theproduction “man in theassets. loop” The is then essential to cope the and chemical transformations in the the transformations are withphysical the tasks of monitoring, controlling and diagnosing asset health status. Today, these tasks can be partially observable by means of measurable parameters. The “man in the loop” is then essential cope with the tasks of monitoring, controlling and diagnosing the asset health status. Today, these tasksto can be partially observable by means of measurable parameters. The “man in the loop” is then essential to cope aidedthe by tasks the plant automationcontrolling as a lever to support supervision andhealth decision by the operators. The project with of monitoring, and diagnosing the asset status. Today, these tasks can be aided by tasks the plant automationcontrolling as a lever to support supervision andhealth decision by the operators. The project with the ofpaper monitoring, and diagnosing the asset status. Today, these tasks can be reported by this was developed in this context, leading to enhance the decision-making capabilities. aided by the plant automation as aa lever to support supervision decision by the operators. The project reported by this paper was developed in this context, leading to and enhance the decision-making capabilities. aided by the plant automation as lever to support supervision and decision by the operators. The project In particular, a condition monitoring tool was deployed by seeking the the opportunities provided by plant reported by paper developed in this context, leading to decision-making capabilities. In particular, a condition monitoring tool deployed by seeking the the opportunities provided by plant reported by this this paper was was developed in thiswas context, leading to enhance enhance decision-making capabilities. automation enhanced according to Industry 4.0 paradigm. The tool is now running on an Electric Arc In particular, aa condition monitoring tool was by seeking opportunities provided by plant automation enhanced according to Industry 4.0deployed paradigm. The tool the is now running on an Electric Arc In particular, condition monitoring tool was deployed by seeking the opportunities provided by plant Furnace of the steel making plant of Tenaris Dalmine, in Italy. The project used the experience and automation according to Industry 4.0 paradigm. tool is now running on an Electric Arc Furnace of enhanced the steel making plant of Tenaris Dalmine, in The Italy. The project used the experience and automation enhanced according to Industry 4.0 paradigm. The tool is now running on an Electric Arc knowledge gained by Tenaris Dalmine process and maintenance operators as foundation for initiating the Furnace of the steel making plant of Tenaris Dalmine, in Italy. The project used the experience and knowledge gained by Tenaris Dalmine process and maintenance operators as foundation for initiating the Furnace of the steel making plant of approach. Tenaris Dalmine, in Italy. The project used theengineering experience and and conceptualization of a novel and Thus, the tool – resulting from knowledge gained by Tenaris Dalmine process and maintenance operators as foundation for initiating the conceptualization of a novel and approach. Thus, the tool – resulting from engineering and knowledge gained Tenaris Dalmine processplant and maintenance operators as foundation for initiating the implementation of by the concept – is integrating automation with intelligent data analytics, as a result conceptualization of aa novel and approach. Thus, the tool resulting from engineering and implementation of the concept – is integrating plant automation with–– intelligent data analytics, as a result conceptualization of novel and approach. Thus, the tool resulting from engineering and of close collaboration between– Tenaris Dalmine and the Manufacturing group ofanalytics, the Department of implementation of is plant automation with data as result of close collaboration between– Tenaris Dalmine the Manufacturing group ofanalytics, the Department of implementation of the the concept concept is integrating integrating plantand automation withdiintelligent intelligent data astoaa detect result Management, Economics and Industrial Engineering of the Politecnico Milano. The tool enables of close collaboration between Tenaris Dalmine and Manufacturing group of the Department of Management, Economics and Industrial Engineering of the Politecnico di Milano. Theoftool enables to detect of close collaboration between Tenaris Dalmine and Manufacturing group the Department of incipient failures through monitoring of the furnace panels and of the hearth/bottom of the furnace; thus, Management, Economics and Engineering of di Milano. The tool enables to detect incipient failures through monitoring of the furnace panels and of the hearth/bottom of the furnace; thus, Management, Economics and Industrial Industrial Engineering of Politecnico Politecnico di Milano. The tool enables topeople detect it allows to enhance the operations by avoiding downtimes that may lead to risky consequences on incipient failures through monitoring the furnace panels and of the hearth/bottom of the furnace; thus, it allows to enhance the operations by of avoiding downtimes that may lead to risky consequences on people incipient failures through monitoring of the furnace panels and of the hearth/bottom of the furnace; thus, safety. The tool has the to be considered an example of smart that maintenance and its implementation reflects a it allows to operations by avoiding downtimes may to consequences people safety. The tool has the to be considered an example of smart that maintenance its implementation reflects it allows to enhance enhance operations bysystems avoiding downtimes may lead leadand to risky risky consequences on on peoplea path towards building cyber-physical in production. safety. The has be an of path towards building in production. safety. The tool tool has to tocyber-physical be considered considered systems an example example of smart smart maintenance maintenance and and its its implementation implementation reflects reflects aa path towards building cyber-physical systems in production. © 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. path towards building Based cyber-physical systems in production. Keywords: Condition Maintenance, Production system safety, Fault tree analysis, Data analytics. Keywords: Condition Based Maintenance, Production system safety, Fault tree analysis, Data analytics. Keywords: Condition Based Maintenance, Production system Keywords: Condition Based Maintenance, Production system safety, safety, Fault Fault tree tree analysis, analysis, Data Data analytics. analytics.
1. INTRODUCTION 1. INTRODUCTION 1. INTRODUCTION INTRODUCTION The deployment of a1.condition monitoring tool on an Electric The deployment of a condition monitoring tool on an Electric Arc deployment Furnace (EAF) of the steel makingtool plant of Tenaris The of a condition monitoring on an Arc deployment Furnace (EAF) of the steel makingtool plant Tenaris The ofin a condition monitoring on of an Electric Electric Dalmine, located Dalmine (Italy), is presented by this Arc Furnace (EAF) of the steel making plant of Tenaris Dalmine, located in of Dalmine (Italy), is presented by this Arc Furnace (EAF) the steel making plant of Tenaris paper. In steel making plants, the furnace is the most relevant Dalmine, located in Dalmine by paper. In steel making plants, the(Italy), furnaceis ispresented the most relevant Dalmine, located in Dalmine (Italy), is presented by this this asset inInterms of safetyplants, issues. Due to the characteristics of paper. steel making the furnace is the most relevant asset inInterms of safetyplants, issues. Due to the characteristics of paper. steel making the furnace is the most relevant the furnace – running continuously at high temperatures and asset in terms of safety issues. Due the characteristics of the furnace – running continuously atto high temperatures and asset in terms of safety issues. Due to the characteristics of in harsh environmental conditionsat –, many components the furnace – running continuously high temperatures and in harsh environmental conditions –, many components the furnace – running continuously at high temperatures and cannot be visually inspected, thus an intelligent maintenance in harsh environmental conditions –, components cannot be visually inspected, thus an intelligent in environmental conditions –, many many maintenance componentsa tool,harsh featuring real-time monitoring capabilities, represents cannot be visually inspected, thus an intelligent maintenance tool, featuring real-time monitoring capabilities, represents a cannot be visually inspected, thus an intelligent maintenance solution to increase visibility on, andcapabilities, keep under represents control, thea tool, featuring real-time monitoring solution to increase visibility on, andcapabilities, keep under represents control, thea tool, featuring real-time monitoring asset health status. The implementation pathunder of thecontrol, intelligent solution to increase visibility on, and keep the asset health status. The implementation pathunder of thecontrol, intelligent solution to increase visibility on, and keep the maintenance tool and the results achieved by its use in the asset health status. The implementation path of the intelligent maintenance tool and the results achieved by its use in the asset health status. The implementation path of the intelligent plant are herein presented. maintenance tool and plant are herein presented. maintenance tool and the the results results achieved achieved by by its its use use in in the the plant are herein presented. plant presented. Safetyare is herein claimed as an important issue in process design and Safety is claimed as an important issue in process design and in the operations in an theimportant chemicalissue process industrydesign (Zhaoand et Safety is as in process in the operations in an theimportant chemicalissue process industrydesign (Zhaoand et Safety is claimed claimed as in must process al.,the 2009), and the risks for workers be carefully in operations in the chemical process industry (Zhao al., 2009), and the risks for workers must be carefully in the operations in etthe process industry (Zhao et et considered (Barreto al.,chemical 1997). To this end, maintenance is al., 2009), and the risks for workers must be carefully considered (Barreto etrisks al., 1997). To this end, maintenance is al., 2009), and the for workers must be carefully charged with a high et level of responsibility andmaintenance organizational considered (Barreto al., 1997). To this end, is charged with a high et level of responsibility andmaintenance organizational considered (Barreto al., 1997). To this end, is involvement, encompassing also issues related to the: i) asset charged with a high level of responsibility and organizational involvement, encompassing also issues related to the: i) asset charged with a high level of responsibility and organizational lifecycle optimization, ii) also safety management and iii) involvement, encompassing issues related i) lifecycle optimization, ii) also safety iii) involvement, encompassing issuesmanagement related to to the: the:and i) asset asset environmental management. lifecycle optimization, ii) environmental management. lifecycle optimization, ii) safety safety management management and and iii) iii) environmental environmental management. management.
Today, these organizational requirements can be supported Today, these organizational requirements can be supported by a mature state of the art favouring the implementation of Today, these organizational requirements can supported by a mature of the art favouring the implementation of Today, thesestate organizational requirements can be be supported CBM (Condition Based Maintenance) programs as well as by by a mature state of the art favouring the implementation of CBM (Condition Based Maintenance) programs as well as by by a mature state of the art favouring the implementation of the novel capabilities promised by the transformation brought CBM (Condition Based Maintenance) programs as well as by the novel capabilities promised by the transformation brought CBM (Condition Based Maintenance) programs as well as by about by capabilities the Industry 4.0 paradigm. Indeed, “twobrought of the the novel promised by the transformation about by capabilities the Industry 4.0 paradigm. Indeed, “twobrought of the the novel promised by 4.0 the are transformation characteristic features of Industry computerization by about by the Industry paradigm. “two of the characteristic features of4.0 Industry 4.0 areIndeed, computerization by about by the Industry 4.0 paradigm. Indeed, “two of(…). the utilising cyber-physical systems and intelligent factories characteristic features of Industry 4.0 are computerization by utilising cyber-physical systems and intelligent factories (…). characteristic features of Industry 4.0 are computerization by Maintenance is one of systems the application areas, referred to as utilising cyber-physical and intelligent factories (…). Maintenance is one of the application areas, referred to as utilising cyber-physical systems and intelligent factories (…). Maintenance is 4.0” (Kans etapplication al., 2015).areas, In other scientific Maintenance one of the referred to Maintenance is 4.0” (Kans et al., 2015). In other scientific one of4.0 the isapplication referred to as as context, Maintenance related to areas, what is defined as Maintenance 4.0” (Kans et al., 2015). In other scientific context, Maintenance 4.0 et is al., related to what is defined as Maintenance 4.0” (Kans 2015). In other scientific Smart Maintenance (Lee etisal.,related 2006).toCall thisisevolution as context, Maintenance 4.0 what defined as Smart Maintenance (Lee etisal.,related 2006).toCall thisisevolution context, Maintenance 4.0 what defined as you want, the new Maintenance 4.0 solutions areevolution worth to be Smart Maintenance (Lee et al., 2006). Call this as you want, the new Maintenance 4.0 solutions areevolution worth to be Smart Maintenance (Lee al., 2006). Call this as considered as new a lever to et foster the achievement of relevant you want, the Maintenance 4.0 solutions are worth to be considered as new a lever to foster the achievement of relevant you want, the Maintenance 4.0 solutions are worth to be results in asset and maintenance management (Deloux et al., considered as lever to foster the achievement of results in asset maintenance management (Deloux et al., considered asetaaand lever to and foster the achievement of relevant relevant 2008; Huynh al., 2012 Djurdjanovic et al., 2003). results in maintenance management (Deloux 2008; al., 2012 and Djurdjanovic et al., 2003). et resultsHuynh in asset assetet and and maintenance management (Deloux et al., al., 2008; Huynh et al., 2012 and Djurdjanovic et al., 2003). 2008; Huynh al., 2012 Djurdjanovic et al., 2003). Various workset have beenand presented, and scientific literature Various works have been presented, and scientific literature is plentyworks of innovative approaches, implicitly or explicitly Various have presented, and scientific literature is plentyworks of innovative implicitly or explicitly Various have been beenapproaches, presented, scientific literature referred toofas innovative Maintenance 4.0, Smart and Manufacturing, or with is plenty approaches, implicitly or explicitly referred toofas innovative Maintenance 4.0, Smart implicitly Manufacturing, or with is plenty approaches, or explicitly the mosttotraditional idea of4.0, e-Maintenance (Lee et al., 2006 referred as Maintenance Smart Manufacturing, or with the mosttotraditional idea of4.0, e-Maintenance (Lee et al., 2006 referred as Maintenance Smart Manufacturing, orMany with and Muller et al., idea 2008, Guillén et al.,(Lee 2016). the most traditional of e-Maintenance et al., 2006 and most Muller et al., idea 2008, Guillén et al.,(Lee 2016). Many the traditional of e-Maintenance et al., 2006 innovative solutions about prognostics (e.g. Abichou et al., and Muller et 2008, Guillén al., 2016). Many innovative solutions about prognostics (e.g. Abichou et al., and Muller et al., al.,(e.g. 2008, Guillénet et et al., 2016). Many 2012), diagnostics Fumagalli al., 2011) and mixed innovative solutions about prognostics (e.g. Abichou et al., 2012), diagnostics (e.g. Fumagalli et al., 2011) and mixed innovative solutions about prognostics (e.g. Abichou et al., reality tools (e.g. Espindola et al., 2011 and2011) Espindola et al., 2012), diagnostics (e.g. Fumagalli et al., and mixed reality tools (e.g. Espindola et al., 2011 and Espindola et al., 2012), diagnostics (e.g. Fumagalli et al., 2011) and mixed reality reality tools tools (e.g. (e.g. Espindola Espindola et et al., al., 2011 2011 and and Espindola Espindola et et al., al.,
Copyright@ 2016 IFAC 27 2405-8963 © 2016 2016, IFAC IFAC (International Federation of Automatic Control) Copyright@ 27 Hosting by Elsevier Ltd. All rights reserved. Peer review under responsibility of International Federation of Automatic Copyright@ 2016 IFAC 27 Control. Copyright@ 2016 IFAC 27 10.1016/j.ifacol.2016.12.155
2017 IFAC IMS 20 December 5-7, 2016. Austin, TX, USA
Luca Fumagalli et al. / IFAC-PapersOnLine 49-31 (2016) 19–24
2013) are indeed available, and such smart maintenance systems are expected to progressively enter in the industry in various sectors.
transformation affecting the asset degradation (i.e. cooling panels of the furnace). From a methodological perspective, we can say that the Smart Water Monitoring leads to envision information about the furnace available in the tool. It is the result of a first step towards building the CPS perspective: it can be asserted that the actual implementation reflects, at least, part of the implementation path required for CPS in manufacturing. Using, as a reference model, the 5C architecture for implementation envisioned by (Lee et al., 2015), we may assert that the implementation path has been initiated through the first two levels, i.e. the Smart Connection level and the Data-to-Information Conversion level. This paper especially concentrates on the Data-to-Information Conversion level, presenting the novel approach resulting in it.
As stated by Warren Bennis, “the factory of the future will have only two employees, a man and a dog. The man will be there to feed the dog. The dog will be there to keep the man from touching the equipment” (Fisher 1991). Being quite far away from this long-term vision, the focus nowadays should be on the practical application of a few but effective technological solutions to pave the way towards the next generation of manufacturing systems. In other words, the ambitious aim of the factory of the future still stands. It can be also agreed in steel making with a special emphasis on the safe operations of the production assets: keeping the human as far as possible from inappropriate interventions on the machines/plants is an imperative, eventually leading to keep the optimal asset use and to avoid introducing risks for the operators.
The approach used to develop and implement the condition monitoring tool is summarized in three main steps: i)
Fault tree analysis, to understand how the system under study can fail and the related critical scenarios to identify the key measurable parameters (i.e. key variables) to monitor and map the asset behaviour; ii) Engineering and deployment of the State Detection and Health Assessment tools (part of the overall Smart Maintenance tool); iii) Implementation and integration of the tools in the existing ICT architecture of the plant. The steps enabled to create the analytics now running the Smart Water Monitoring. They are discussed through the following structure of the paper. Section 2 introduces the furnace and its key elements. Section 3 summarizes the use of fault tree analysis, to identify the deviations in the process parameters and, thus, to select the signals needed in order to represent the asset behaviour in its degradation. The signals are then the relevant input of the tool. Section 4 presents the development of the tools for state detection and for health assessment. Section 5 illustrates the implementation of the Smart Maintenance tool in ICT architecture. Section 6, through three real cases, describes the rich output information for the process supervisor and for the maintenance engineer provided by the tools. Eventually, Section 7 provides conclusions, envisioning some future steps towards the implementation of CPS in EAFs.
To move towards this end, a Smart Maintenance tool should provide as much as possible a complete visibility of the asset health status, thus avoiding a need of maintenance operators’ intervention in the nearby of the asset. This research aims at presenting a proof of usefulness of such a type of tool, by discussing a practical implementation in the steel making industry as well as by providing evidences of how the data coming from the plant floor level can be transformed into relevant information, namely relevant indicators, elaborated and displayed at an upper level for managing decisions. The paper grounds on some first information provided by Barreto et al., 1997; Colace et al., 2013 and Colace et al., 2015. The tool has been developed in an industrial environment, at Tenaris Dalmine plant, with the purpose of its practical use in the day-by-day operations. The measurable parameters are the relevant ones for applying a condition-based operation of the furnace: if operating conditions are approaching risky situations, they can be identified thanks to the tool, revealing the asset health status, eventually leading to the subsequent operational decisions by the operator. More specifically, the tool aims at monitoring the parameters of the furnace on the cooling panels and on the refractory bottom/hearth. To this regard, in this paper we will refer to the tool also as a Smart Water Monitoring according to its primary functionality. This communicates the value of the functionality: the tool helps managing the information gathered through the analysis of data resulting from the monitoring of the cooling water of the EAF; therefore, building on such a monitoring, information extraction leads to a “smart water” that conveys the relevant indicators for the asset health status. This type of analysis enables to make evident how the concept of digital twin, that is typical of Cyber-Physical Systems (CPS), is actually being implemented in the case. More specifically, it is remarkable a reduction of distance between the physical transformation – happening in the cooling water – and even physical-chemical transformation – occurring in the furnace and influent for the physical transformation of the water – and the cyber level implementing the analytics: it could be asserted that the monitored water, used as the conveying entity, is acting as a digital twin indirectly linking to the physical and chemical
2. FURNACE DESCRIPTION The EAF is constituted by a lower sheet metal keel (lowerpart) coated with refractory bricks, which serves to contain the liquid steel, and is cooled by a cage, with a structural function, that supports panels cooled by a water circuit (Ferro et al., 2007). In the heating process, the panels cooled by a water circuit could be stressed out, and therefore fail with a leak of water in the furnace, which is finally leading to risky situation. The panels are particularly relevant for the present research. In Tenaris Dalmine plant, there are two electric furnaces with the following characteristics: capacity 105 t, diameter 6.1 m, weight of the structure (considering also refractory bricks) 230 t. The characteristics of each furnace allow to obtain 98 t 28
2017 IFAC IMS December 5-7, 2016. Austin, TX, USA
Luca Fumagalli et al. / IFAC-PapersOnLine 49-31 (2016) 19–24
of liquid steel, with an average power request of 67 MW per cast, and a maximum power peak of 89 MW. The casting cycle is around 38.5 min and this corresponds to a productivity above 150 t/h. The main contribution of energy needed for melting the scrap is supplied from the electricity for about 2/3 and for 1/3 by chemical energy supplied by the lances and the burners on board. In particular: melting 100 t of steel requires approximately 40 MWh of electricity; the chemical process is instead controlled by a system that monitors gases exiting the furnace and optimizes the injection by lances, providing oxygen; more specifically, the chemical system is a set of devices that provides multipoint injection of oxygen, methane and carbon. It can be easily understood that a lot of energy is provided to melt the steel and this impacts on the stress that the panels can suffer.
21
graphical representation of the logical path that connects the unwanted event – namely the “root” of the tree, also said “top event” – with the events that are its root causes. The technique follows a hierarchical methodology which first connects the top event with the events that are directly intended as its root causes; such events (intermediate events) are in turn related to other events that are the underlying causes. The hierarchical tree construction is completed when it has reached the so-called events / basic causes – that is to say, the "leaves" of the tree. For the proposed case, FTA was built relying both on experience of the industrial experts and theoretical issues analysed by academic experts, involved in the research. The identified top events are the holing of the panels and the presence of the water in the bottom/hearth. In figure 3 and 4 an extracted part of the FTA related to the two top events identified is presented. Some boxes of the FTA are provided with notes, helpful in order to document the experience and knowledge synthesized during the first project meetings.
Fig. 1. Rendering of the EAF
Fig. 3. Extract of fault tree analysis: holing of the panels
The panels are positioned around the furnace circumference on two different rows. To this regard, they are defined by the Tenaris practitioners as “down panels”, when they are in the first row close to the hearth/bottom, and “up panels” when they are in the second row, close to the furnace roof. According to the fault tree analysis, “down panels” reveal to be less stressed since they are most of the time covered by the steel slag. On the other hand, the “up panels” are quite stressed, and they may be broken by different mechanisms, namely: suddenly broken by a splash, progressively degraded to the exposition when not covered by the steel slag, stressed by irradiation by electric arc.
Fig. 4. Extract of fault tree analysis: presence of water in the bottom/hearth For each critical scenario related to a top event, the key measurable parameters – i.e. the key parameters that best describe the behaviour of the assets – are identified. The objective is to define a limited group of variables, in the remainder key variables, highly representative of the asset / scenario or some of its features. The key variables identification is a relevant phase of the State Detection and Health Assessment tools deployment. In fact, the cooling panels’ behaviour is not easy to be modelled. Sensors can be hardly installed due to harsh conditions, thus few information about water temperature, flow and pressure can be collected. Moreover, within the panel, the water goes under a change of physical status, creating also areas in which there is water vapour, making very difficult to deploy a physical model of the panel itself. For these reasons, in this industrial case the identification of key variables related to an anomalous behaviour of the panels is an important phase to approach problem solving. Once identified the key variables and completed the feasibility study, the monitoring tool needs to be engineered with specific algorithms in order to assess in which state the furnace is working. The FTA clearly leads to the decision of developing two different part of the tool,
Fig. 2. Rendering of the furnace with panels 3. FAULT TREE ANALYSIS AND KEY VARIABLES RELATED TO CRITICAL SCENARIOS The Fault Tree Analysis (FTA) has been developed by the U.S. Air Force and then used in various sectors in order to support the reliability analysis of a technical system. The analysis according to the FTA technique is based on the development of a tree, called Fault Tree, which allows a 29
2017 IFAC IMS 22 December 5-7, 2016. Austin, TX, USA
Luca Fumagalli et al. / IFAC-PapersOnLine 49-31 (2016) 19–24
during the heating cycles due to steel melting, the reference feature of the key variable indicator is defined; in this way, the expected values of the key variable indicator at the generic time t are known and it is possible to build a control chart with upper and lower thresholds. Subsequently, for this part only SD model is developed, as the alarm is enough to take a decision, i.e. to stop the EAF because exceeding thresholds means the presence of water in the bottom/hearth (which is clearly a risky situation).
inclusive of their respective algorithms: i) Smart Water Monitoring for panels; ii) Smart Water Monitoring for refractory hearth. 4. ENGINEERING AND DEPLOYMENT OF THE STATE DETECTION AND HEALTH ASSESSMENT TOOLS For the top event “Holing of the panel”, the Smart Water Monitoring for panels tool has been developed. In particular, State Detection (SD) and Health Assessment (HA) models are developed for this first tool: their main characteristics are summarized in the remainder. We start with the SD model. 1. An alarm threshold model has been defined with a SD aim. In particular, a control chart has been deployed to take under control the key measurable parameter for each panel. 2. All the critical events that go out of the control chart are counted in a “critical events counter”. The alarm is then displayed when such a counter overcomes the defined threshold. 3. In order to fine tune the SD model, the alarm is also parametrized according to the number of critical events within a certain timeframe. Starting from the variables of the SD model for panels, it is then possible to engineer the HA model. 1. The model is based, for each panel, on the overlapping area between two distributions of a key feature extracted by the analysis of variables related with cooling water of the panels. It enables matching the most recent features (used as process signatures resulting from smart water monitoring) with those features observed during normal and faulty asset behaviour. 2. Variables have been manipulated with data analysis approaches in order to obtain a Gaussian distribution. Thus, once obtained the distributions, it is possible to sketch on a graph two curves: the first one as reference feature/signature, coming from data of not stressed operating conditions of the panel; the second one as feature/signature representative of the actual asset health status due to the operating conditions. 3. For each panel, the construction of the reference feature and the actual feature are needed in order to compute the overlapping area, finally corresponding to an indicator that is used to measure the actual overlap. The tool also compares the value of the average key variables in the period of analysis with its reference average, being the maximum difference an empirical value. The set of information from the two different algorithms delineates a matrix built upon four quadrants with corresponding scenarios/alert levels. Regarding the top event “presence of water in the bottom/hearth” a SD model is engineered as Smart Water Monitoring for refractory hearth/bottom. In particular, it is worth remarking that, in the feasibility study, it has been identified that the presence of water in the bottom of furnace (i.e. on the hearth/bottom) determines an anomalous trend of a key variable indicator. Therefore, based on historical data,
5. IMPLEMENTATION AND INTEGRATION OF THE TOOLS IN THE EXISTING ICT ARCHITECTURE The state detection and health assessment models engineered are then coded into the Smart Maintenance tool, and the output must be shown to the process operator/maintenance engineer in the control room. Therefore, the Smart Water Monitoring is developed in such a way that is accessible as Level 2 tool in the ICT architecture (see figure 5).
Fig. 5. Architecture of Tenaris Dalmine existing ICT systems The tool is web-based and fully integrated with the suite used by plant operations in Tenaris Dalmine. This fully realizes the paradigm postulated since many years about integration of Condition Based Maintenance solutions within the information system for maintenance management (Fumagalli et al., 2009). Moreover, the Smart Water Monitoring has been developed as highly parametrized tool: it is thus possible to apply it on other similar furnaces, even with differences in configurations. It would be relevant for further developments to other EAFs, beyond Tenaris Dalmine. 6. TESTING PHASE: REAL CASE ANALYSIS For a more complete and clear illustration of the tools and its rich output information three real cases are presented. In a first case, during the testing phase of the developed tools, there were some water tests to identify the presence of water in the hearth/bottom. During the heating cycles of testing days there are abnormal features of key variables. The tests started in the heating cycle number 16 (“campagna 16” in the legenda of Fig. 6), where the values of the key 30
2017 IFAC IMS December 5-7, 2016. Austin, TX, USA
Luca Fumagalli et al. / IFAC-PapersOnLine 49-31 (2016) 19–24
variable of castings during the cycle remained lower than the expected values.
23
variables bigger than the respective thresholds, there is an alert level 4 (i.e. the 4th quadrant of the risk matrix). This is the worst scenario in which the actual process is different from the reference process, and the actual feature is moving to the right, i.e. this represents a deviation leading towards more degradation. In other words, the process seems to stress the panel more than the normal operating conditions. This arises attention: the operator should be aware of the fact that that the actual maintenance policy would be no longer suited; indeed, it is recommended increasing inspections to control/prevent any possible damage.
Fig. 6. State detection tool for hearth/bottom. This can be highlighted by comparison with the reference values and the past heating cycle (number 12 – “campagna 12” in the legenda). At the beginning of the next campaign (number 17 – “campagna 17” in the legenda), the values of key variable have an anomalous feature compared to the reference values. It is due to the water presence in the hearth/bottom, until all the water vaporizes and the key variable reaches the reference value. In a second case, during a planned inspection, two holes of small dimensions were discovered in a panel.
Fig. 9. State detection tool for panels.
Fig. 7. Picture of the panel took in the inspection Looking in the days prior to the inspection, the SD tool for hearth/bottom shows how the key variable is lower than the expected value of the reference feature, over the minimum threshold. After some time, the key variable returns to the expected values because, probably, the holes were covered and clogged by dross, thus there were no more water leakage.
Fig. 10. Health assessment tool for panels. 7. CONCLUSIONS Production process in steel making industry involves many variables and operators cope with the tasks of monitoring, controlling and diagnosing the health status of the production assets. This often results in high difficulties to effectively analyse the current assets’ behaviour, to detect and diagnose process anomalies. This eventually impacts on the capability to quickly take appropriate control actions. The experience and knowledge gained by Tenaris Dalmine process and maintenance staff gave the foundations for initiating a novel approach, finally introducing the herein presented Smart Maintenance tool. The tool enables to detect incipient failures through monitoring of furnace panels and hearth/bottom, before causing a downtime that may bring disastrous consequences on people safety, integrity of assets and non-continuity of the production cycle. The fact that the ICT infrastructure was pre-existent, as well as the importance of safety issues in such an asset operation, represented good reasons for investing in the tool as solution for the company, bringing innovation for maintenance. Today, the tool is
Fig. 8. Zoom of state detection tool for hearth/bottom To better understand this case, it is interesting to evaluate the behaviour of the SD tool for the panels for the same event: figure 9 shows the register of the critical events counter with some key variables bigger than the upper threshold in the days before the inspection, i.e. castings which have stressed out the panel and probably determined the holes. It is evident that the panels appeared to be subject to stress in these days (signal in Fig. 8 out of control band). In the third case, the panel of this real case seems to change its actual feature. Fig. 10 shows the health assessment tool in which the actual Gaussian feature (red line) and the reference feature (black line) are compared. As the rate of overlapping area is low and the difference between the average of the key 31
2017 IFAC IMS 24 December 5-7, 2016. Austin, TX, USA
Luca Fumagalli et al. / IFAC-PapersOnLine 49-31 (2016) 19–24
running and the aim is to correlate the critical events counter and the overlapping areas (i.e. the indicators provided by the tool) to the useful life of the panels. In the Smart Water Monitoring for Panels, the validated HA and SD algorithms are the starting point in order to continue towards this direction. To this end, it is necessary to continue the testing phase in order to collect the data needed to characterize the life of the panels up to their replacement/repair. The tracking of the panels life will then serve to map the information of Overlapping Areas (resulting from Health Assessment model within Smart Water Monitoring for Panels tool) and the critical events counter (resulting from State Detection model within Smart Water Monitoring for Panels tool) with the actual degradation occurred/observed in the panel. As a vision for the future, the final aim will be the building of an acceptable degradation function of the panels. This would be the basis for the definition of a reliability model of the panels themselves – to express the probability not to degrade before a time t –, thus for the computation of the subsequent risks. On the whole, it could be eventually possible to obtain (by a computational model) a prognostic risk prediction. A further extension of the Smart Maintenance tool could be built by reflecting on the vision of the social network of machines suggested by Lee et al., 2015. An extension of the analytics approach would be helpful to provide EAFs with self-comparison ability. This will be useful for performance rating among the fleet owned by the company, and for a similarity identification between the performance and assets operating condition, eventually useful to develop a powerful predictive analytics of the remaining useful life based on the different operating conditions. On the scientific level, it would enable continuing the implementation path towards building cyber-physical systems in production.
Djurdjanovic D, Lee J and Ni J. (2003) Watchdog agent – an infotronics-based prognostics approach for product performance degradation assessment and prediction. Adv Eng Inform; 17: 109–125. Espíndola D., Fumagalli L., Garetti M., Botelho S., Pereira C., (2011). Adaption of OSA-CBM Architecture for Human-Computer Interaction through Mixed Interface, INDIN 11 conference, Caparica, Lisbon, Portugal, 26-29 July 2011, 485 - 490. Espíndola D., Fumagalli L., Garetti M., Pereira C.E., Botelho S., Ventura Henriques R. (2013) A model-based approach for data integration to improve maintenance management by mixed reality. Computers in Industry, Volume 64, Issue 4, May 2013, 376-391. Ferro L., Giugliano P., Galbiati P., Memoli F., Giavani C., Maiolo J. (2007). The electric arc furnace of Tenaris Dalmine: from the application of the new technologies of digital electrode regulation and multipoint injection to the dynamic control of the process. 16th IAS Steelmaking Conference, 6th – 8th November 2007, Rosario, Argentina, 59-72. Fumagalli L., Macchi M., Rapaccini M. (2009). Computerized maintenance management systems in SMEs: A survey in Italy and some remarks for the implementation of Condition Based Maintenance, IFAC Proceedings Volumes (IFAC-PapersOnline), 13 (PART 1), pp. 1615-1619. Fumagalli L., Ierace S., Dovere E., Macchi M., Cavalieri S., Garetti M. (2011). Agile diagnostic tool based on electrical signature analysis, IFAC Proceedings Volumes (IFAC-PapersOnline), 18 (PART 1), pp. 14067-14072. Fisher M. (1991) The millionaire's book of quotations. p. 15 Guillén A.J., Crespo A., Macchi M. & Gómez J. (2016). On the role of Prognostics and Health Management in advanced maintenance systems. Production Planning & Control, Huynh KT, Anne B and Christophe B. (2012). Adaptive condition-based maintenance decision framework for deteriorating systems operating under variable environment and uncertain condition monitoring. Proc IMechE, Part O: J Risk and Reliability. Epub ahead of print 7 November 2012. Kans M, Galar D and Thaduri A. (2015). Maintenance 4.0 in Railway Transportation Industry. Proceedings of the 10th World Congress on Engineering Asset Management (WCEAM 2015) Springer International Publishing, page 317-331, isbn 978-3-319-27064-7. Lee J, Behrad Bagheri, Hung-An Kao. (2105). A CyberPhysical Systems architecture for Industry 4.0-based manufacturing systems, Manufacturing Letters, Volume 3, January 2015, Pages 18-23, ISSN 2213-8463. Lee J, Ni J, Djurdjanovic D, et al. (2006). Intelligent prognostics tools and e-maintenance. Comput Ind 2006; 57(6): 476–489. Muller A, Crespo Marquez A and Iung B. (2008). On the concept of e-maintenance review and current research. Reliab Eng Syst Safe, 93(8): 1165–1187. Zhao J, Cui L, Zhao L, et al. (2009). Learning HAZOP expert system by case-based reasoning and ontology. Comput Chem Eng, 33: 371–378.
REFERENCES Abichou B., Voisin A., Iung B. (2012). Choquet integral parameters inference for health indicators fusion within multi-levels industrial systems: application to components in series. 2nd IFAC Workshop on Advanced Maintenance Engineering, Service and Technology, AMest'12, Sevilla, Spain Barreto SM, Swerdlow AJ, Smith PG, et al. (1997). A nested case-control study of fatal work relate injuries among Brazilian steel workers. Occup Environ Med; 54:599– 604. Colace C., Fumagalli L., Pala S., et al. (2013). An intelligent maintenance system to improve safety of operations of an electric furnace in the steel making industry. Chem Eng, 33: 397–402. Colace C., Fumagalli, L., Pala, S., Macchi, M., Matarazzo, N. R., & Rondi, M. (2015). Implementation of a condition monitoring system on an electric arc furnace through a risk-based methodology. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability, 1748006X15576441. Deloux E, Castanier B and Berenguer C. (2008). Maintenance policy for a deteriorating system evolving in a stressful environment. Proc IMechE, Part O: J Risk and Reliability; 222: 613–622. 32
ScienceDirect IFAC-PapersOnLine 49-31 (2016) 19–24 A A Smart Smart Maintenance Maintenance tool tool for for a a safe safe Electric Electric Arc Arc Furnace Furnace A Smart Maintenance tool for a safe Electric Arc AFumagalli*, Smart Maintenance tool forColace**, a safeMaurizio Electric Arc Furnace Furnace Luca Marco Macchi*, Cristian Rondi**, Alessandro Alfieri*
Luca Fumagalli*, Marco Macchi*, Cristian Colace**, Maurizio Rondi**, Alessandro Alfieri* Luca Luca Fumagalli*, Fumagalli*, Marco Marco Macchi*, Macchi*, Cristian Cristian Colace**, Colace**, Maurizio Maurizio Rondi**, Rondi**, Alessandro Alessandro Alfieri* Alfieri* *Department of Management, Economics and Industrial Engineering, Politecnico di Milano, P.za Leonardo da Vinci 32, *Department of Management, Economics and Industrial Engineering, Politecnico di Milano, P.za Leonardo da Vinci 32, 20133 Economics Milano, Italy - Contact author's email: [email protected] *Department and Industrial Engineering, Politecnico 20133 Economics Milano, Italy author's email: [email protected] *Department of of Management, Management, and- Contact Industrial Engineering, Politecnico di di Milano, Milano, P.za P.za Leonardo Leonardo da da Vinci Vinci 32, 32, 20133 Milano, Italy Contact author's email: [email protected] 20133 Milano, Italy - Contact author's email: [email protected] **Tenaris Dalmine S.p.A., Piazza caduti 6 luglio 1944 n.1 - 24044 Dalmine (BG), Italy **Tenaris Dalmine S.p.A., Piazza caduti 6 luglio 1944 n.1 - 24044 Dalmine (BG), Italy **Tenaris **Tenaris Dalmine Dalmine S.p.A., S.p.A., Piazza Piazza caduti caduti 66 luglio luglio 1944 1944 n.1 n.1 -- 24044 24044 Dalmine Dalmine (BG), (BG), Italy Italy Abstract: The production process in steel making involves plenty of variables to know the actual status Abstract: The production process in steel making involves plenty of variables to know the actual status in plant operations and the deviations from optimal involves and safe operating conditions. The deviations depend in steel of variables to know the status Abstract: The production in plant operations and the process deviations from making optimal involves and safe plenty operating conditions. The depend Abstract: The production process in steel making plenty ofgeneral, variables tocomplexity knowdeviations the actual actual status on diverse uncertainties due to the material variability and, more in the of modeling in plant operations and the deviations from optimal and safe operating conditions. The deviations depend on diverse uncertainties due to the material variability and, more in general, the complexity of modeling in plant operations and the deviations from optimal and safe operating conditions. The deviations depend the diverse physical and chemical transformations occurring in the production assets. The transformations are on due to the material and, in general, the complexity of the physicaluncertainties and chemical occurring in themore production assets. transformations are on diverse uncertainties duetransformations toof themeasurable material variability variability and, more in in general, the The complexity of modeling modeling partially observable by means parameters. The “man theassets. loop” is then essential to cope the physical and chemical transformations occurring in the production The transformations are partially observable by means of measurableoccurring parameters. Theproduction “man in theassets. loop” The is then essential to cope the and chemical transformations in the the transformations are withphysical the tasks of monitoring, controlling and diagnosing asset health status. Today, these tasks can be partially observable by means of measurable parameters. The “man in the loop” is then essential cope with the tasks of monitoring, controlling and diagnosing the asset health status. Today, these tasksto can be partially observable by means of measurable parameters. The “man in the loop” is then essential to cope aidedthe by tasks the plant automationcontrolling as a lever to support supervision andhealth decision by the operators. The project with of monitoring, and diagnosing the asset status. Today, these tasks can be aided by tasks the plant automationcontrolling as a lever to support supervision andhealth decision by the operators. The project with the ofpaper monitoring, and diagnosing the asset status. Today, these tasks can be reported by this was developed in this context, leading to enhance the decision-making capabilities. aided by the plant automation as aa lever to support supervision decision by the operators. The project reported by this paper was developed in this context, leading to and enhance the decision-making capabilities. aided by the plant automation as lever to support supervision and decision by the operators. The project In particular, a condition monitoring tool was deployed by seeking the the opportunities provided by plant reported by paper developed in this context, leading to decision-making capabilities. In particular, a condition monitoring tool deployed by seeking the the opportunities provided by plant reported by this this paper was was developed in thiswas context, leading to enhance enhance decision-making capabilities. automation enhanced according to Industry 4.0 paradigm. The tool is now running on an Electric Arc In particular, aa condition monitoring tool was by seeking opportunities provided by plant automation enhanced according to Industry 4.0deployed paradigm. The tool the is now running on an Electric Arc In particular, condition monitoring tool was deployed by seeking the opportunities provided by plant Furnace of the steel making plant of Tenaris Dalmine, in Italy. The project used the experience and automation according to Industry 4.0 paradigm. tool is now running on an Electric Arc Furnace of enhanced the steel making plant of Tenaris Dalmine, in The Italy. The project used the experience and automation enhanced according to Industry 4.0 paradigm. The tool is now running on an Electric Arc knowledge gained by Tenaris Dalmine process and maintenance operators as foundation for initiating the Furnace of the steel making plant of Tenaris Dalmine, in Italy. The project used the experience and knowledge gained by Tenaris Dalmine process and maintenance operators as foundation for initiating the Furnace of the steel making plant of approach. Tenaris Dalmine, in Italy. The project used theengineering experience and and conceptualization of a novel and Thus, the tool – resulting from knowledge gained by Tenaris Dalmine process and maintenance operators as foundation for initiating the conceptualization of a novel and approach. Thus, the tool – resulting from engineering and knowledge gained Tenaris Dalmine processplant and maintenance operators as foundation for initiating the implementation of by the concept – is integrating automation with intelligent data analytics, as a result conceptualization of aa novel and approach. Thus, the tool resulting from engineering and implementation of the concept – is integrating plant automation with–– intelligent data analytics, as a result conceptualization of novel and approach. Thus, the tool resulting from engineering and of close collaboration between– Tenaris Dalmine and the Manufacturing group ofanalytics, the Department of implementation of is plant automation with data as result of close collaboration between– Tenaris Dalmine the Manufacturing group ofanalytics, the Department of implementation of the the concept concept is integrating integrating plantand automation withdiintelligent intelligent data astoaa detect result Management, Economics and Industrial Engineering of the Politecnico Milano. The tool enables of close collaboration between Tenaris Dalmine and Manufacturing group of the Department of Management, Economics and Industrial Engineering of the Politecnico di Milano. Theoftool enables to detect of close collaboration between Tenaris Dalmine and Manufacturing group the Department of incipient failures through monitoring of the furnace panels and of the hearth/bottom of the furnace; thus, Management, Economics and Engineering of di Milano. The tool enables to detect incipient failures through monitoring of the furnace panels and of the hearth/bottom of the furnace; thus, Management, Economics and Industrial Industrial Engineering of Politecnico Politecnico di Milano. The tool enables topeople detect it allows to enhance the operations by avoiding downtimes that may lead to risky consequences on incipient failures through monitoring the furnace panels and of the hearth/bottom of the furnace; thus, it allows to enhance the operations by of avoiding downtimes that may lead to risky consequences on people incipient failures through monitoring of the furnace panels and of the hearth/bottom of the furnace; thus, safety. The tool has the to be considered an example of smart that maintenance and its implementation reflects a it allows to operations by avoiding downtimes may to consequences people safety. The tool has the to be considered an example of smart that maintenance its implementation reflects it allows to enhance enhance operations bysystems avoiding downtimes may lead leadand to risky risky consequences on on peoplea path towards building cyber-physical in production. safety. The has be an of path towards building in production. safety. The tool tool has to tocyber-physical be considered considered systems an example example of smart smart maintenance maintenance and and its its implementation implementation reflects reflects aa path towards building cyber-physical systems in production. © 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. path towards building Based cyber-physical systems in production. Keywords: Condition Maintenance, Production system safety, Fault tree analysis, Data analytics. Keywords: Condition Based Maintenance, Production system safety, Fault tree analysis, Data analytics. Keywords: Condition Based Maintenance, Production system Keywords: Condition Based Maintenance, Production system safety, safety, Fault Fault tree tree analysis, analysis, Data Data analytics. analytics.
1. INTRODUCTION 1. INTRODUCTION 1. INTRODUCTION INTRODUCTION The deployment of a1.condition monitoring tool on an Electric The deployment of a condition monitoring tool on an Electric Arc deployment Furnace (EAF) of the steel makingtool plant of Tenaris The of a condition monitoring on an Arc deployment Furnace (EAF) of the steel makingtool plant Tenaris The ofin a condition monitoring on of an Electric Electric Dalmine, located Dalmine (Italy), is presented by this Arc Furnace (EAF) of the steel making plant of Tenaris Dalmine, located in of Dalmine (Italy), is presented by this Arc Furnace (EAF) the steel making plant of Tenaris paper. In steel making plants, the furnace is the most relevant Dalmine, located in Dalmine by paper. In steel making plants, the(Italy), furnaceis ispresented the most relevant Dalmine, located in Dalmine (Italy), is presented by this this asset inInterms of safetyplants, issues. Due to the characteristics of paper. steel making the furnace is the most relevant asset inInterms of safetyplants, issues. Due to the characteristics of paper. steel making the furnace is the most relevant the furnace – running continuously at high temperatures and asset in terms of safety issues. Due the characteristics of the furnace – running continuously atto high temperatures and asset in terms of safety issues. Due to the characteristics of in harsh environmental conditionsat –, many components the furnace – running continuously high temperatures and in harsh environmental conditions –, many components the furnace – running continuously at high temperatures and cannot be visually inspected, thus an intelligent maintenance in harsh environmental conditions –, components cannot be visually inspected, thus an intelligent in environmental conditions –, many many maintenance componentsa tool,harsh featuring real-time monitoring capabilities, represents cannot be visually inspected, thus an intelligent maintenance tool, featuring real-time monitoring capabilities, represents a cannot be visually inspected, thus an intelligent maintenance solution to increase visibility on, andcapabilities, keep under represents control, thea tool, featuring real-time monitoring solution to increase visibility on, andcapabilities, keep under represents control, thea tool, featuring real-time monitoring asset health status. The implementation pathunder of thecontrol, intelligent solution to increase visibility on, and keep the asset health status. The implementation pathunder of thecontrol, intelligent solution to increase visibility on, and keep the maintenance tool and the results achieved by its use in the asset health status. The implementation path of the intelligent maintenance tool and the results achieved by its use in the asset health status. The implementation path of the intelligent plant are herein presented. maintenance tool and plant are herein presented. maintenance tool and the the results results achieved achieved by by its its use use in in the the plant are herein presented. plant presented. Safetyare is herein claimed as an important issue in process design and Safety is claimed as an important issue in process design and in the operations in an theimportant chemicalissue process industrydesign (Zhaoand et Safety is as in process in the operations in an theimportant chemicalissue process industrydesign (Zhaoand et Safety is claimed claimed as in must process al.,the 2009), and the risks for workers be carefully in operations in the chemical process industry (Zhao al., 2009), and the risks for workers must be carefully in the operations in etthe process industry (Zhao et et considered (Barreto al.,chemical 1997). To this end, maintenance is al., 2009), and the risks for workers must be carefully considered (Barreto etrisks al., 1997). To this end, maintenance is al., 2009), and the for workers must be carefully charged with a high et level of responsibility andmaintenance organizational considered (Barreto al., 1997). To this end, is charged with a high et level of responsibility andmaintenance organizational considered (Barreto al., 1997). To this end, is involvement, encompassing also issues related to the: i) asset charged with a high level of responsibility and organizational involvement, encompassing also issues related to the: i) asset charged with a high level of responsibility and organizational lifecycle optimization, ii) also safety management and iii) involvement, encompassing issues related i) lifecycle optimization, ii) also safety iii) involvement, encompassing issuesmanagement related to to the: the:and i) asset asset environmental management. lifecycle optimization, ii) environmental management. lifecycle optimization, ii) safety safety management management and and iii) iii) environmental environmental management. management.
Today, these organizational requirements can be supported Today, these organizational requirements can be supported by a mature state of the art favouring the implementation of Today, these organizational requirements can supported by a mature of the art favouring the implementation of Today, thesestate organizational requirements can be be supported CBM (Condition Based Maintenance) programs as well as by by a mature state of the art favouring the implementation of CBM (Condition Based Maintenance) programs as well as by by a mature state of the art favouring the implementation of the novel capabilities promised by the transformation brought CBM (Condition Based Maintenance) programs as well as by the novel capabilities promised by the transformation brought CBM (Condition Based Maintenance) programs as well as by about by capabilities the Industry 4.0 paradigm. Indeed, “twobrought of the the novel promised by the transformation about by capabilities the Industry 4.0 paradigm. Indeed, “twobrought of the the novel promised by 4.0 the are transformation characteristic features of Industry computerization by about by the Industry paradigm. “two of the characteristic features of4.0 Industry 4.0 areIndeed, computerization by about by the Industry 4.0 paradigm. Indeed, “two of(…). the utilising cyber-physical systems and intelligent factories characteristic features of Industry 4.0 are computerization by utilising cyber-physical systems and intelligent factories (…). characteristic features of Industry 4.0 are computerization by Maintenance is one of systems the application areas, referred to as utilising cyber-physical and intelligent factories (…). Maintenance is one of the application areas, referred to as utilising cyber-physical systems and intelligent factories (…). Maintenance is 4.0” (Kans etapplication al., 2015).areas, In other scientific Maintenance one of the referred to Maintenance is 4.0” (Kans et al., 2015). In other scientific one of4.0 the isapplication referred to as as context, Maintenance related to areas, what is defined as Maintenance 4.0” (Kans et al., 2015). In other scientific context, Maintenance 4.0 et is al., related to what is defined as Maintenance 4.0” (Kans 2015). In other scientific Smart Maintenance (Lee etisal.,related 2006).toCall thisisevolution as context, Maintenance 4.0 what defined as Smart Maintenance (Lee etisal.,related 2006).toCall thisisevolution context, Maintenance 4.0 what defined as you want, the new Maintenance 4.0 solutions areevolution worth to be Smart Maintenance (Lee et al., 2006). Call this as you want, the new Maintenance 4.0 solutions areevolution worth to be Smart Maintenance (Lee al., 2006). Call this as considered as new a lever to et foster the achievement of relevant you want, the Maintenance 4.0 solutions are worth to be considered as new a lever to foster the achievement of relevant you want, the Maintenance 4.0 solutions are worth to be results in asset and maintenance management (Deloux et al., considered as lever to foster the achievement of results in asset maintenance management (Deloux et al., considered asetaaand lever to and foster the achievement of relevant relevant 2008; Huynh al., 2012 Djurdjanovic et al., 2003). results in maintenance management (Deloux 2008; al., 2012 and Djurdjanovic et al., 2003). et resultsHuynh in asset assetet and and maintenance management (Deloux et al., al., 2008; Huynh et al., 2012 and Djurdjanovic et al., 2003). 2008; Huynh al., 2012 Djurdjanovic et al., 2003). Various workset have beenand presented, and scientific literature Various works have been presented, and scientific literature is plentyworks of innovative approaches, implicitly or explicitly Various have presented, and scientific literature is plentyworks of innovative implicitly or explicitly Various have been beenapproaches, presented, scientific literature referred toofas innovative Maintenance 4.0, Smart and Manufacturing, or with is plenty approaches, implicitly or explicitly referred toofas innovative Maintenance 4.0, Smart implicitly Manufacturing, or with is plenty approaches, or explicitly the mosttotraditional idea of4.0, e-Maintenance (Lee et al., 2006 referred as Maintenance Smart Manufacturing, or with the mosttotraditional idea of4.0, e-Maintenance (Lee et al., 2006 referred as Maintenance Smart Manufacturing, orMany with and Muller et al., idea 2008, Guillén et al.,(Lee 2016). the most traditional of e-Maintenance et al., 2006 and most Muller et al., idea 2008, Guillén et al.,(Lee 2016). Many the traditional of e-Maintenance et al., 2006 innovative solutions about prognostics (e.g. Abichou et al., and Muller et 2008, Guillén al., 2016). Many innovative solutions about prognostics (e.g. Abichou et al., and Muller et al., al.,(e.g. 2008, Guillénet et et al., 2016). Many 2012), diagnostics Fumagalli al., 2011) and mixed innovative solutions about prognostics (e.g. Abichou et al., 2012), diagnostics (e.g. Fumagalli et al., 2011) and mixed innovative solutions about prognostics (e.g. Abichou et al., reality tools (e.g. Espindola et al., 2011 and2011) Espindola et al., 2012), diagnostics (e.g. Fumagalli et al., and mixed reality tools (e.g. Espindola et al., 2011 and Espindola et al., 2012), diagnostics (e.g. Fumagalli et al., 2011) and mixed reality reality tools tools (e.g. (e.g. Espindola Espindola et et al., al., 2011 2011 and and Espindola Espindola et et al., al.,
Copyright@ 2016 IFAC 27 2405-8963 © 2016 2016, IFAC IFAC (International Federation of Automatic Control) Copyright@ 27 Hosting by Elsevier Ltd. All rights reserved. Peer review under responsibility of International Federation of Automatic Copyright@ 2016 IFAC 27 Control. Copyright@ 2016 IFAC 27 10.1016/j.ifacol.2016.12.155
2017 IFAC IMS 20 December 5-7, 2016. Austin, TX, USA
Luca Fumagalli et al. / IFAC-PapersOnLine 49-31 (2016) 19–24
2013) are indeed available, and such smart maintenance systems are expected to progressively enter in the industry in various sectors.
transformation affecting the asset degradation (i.e. cooling panels of the furnace). From a methodological perspective, we can say that the Smart Water Monitoring leads to envision information about the furnace available in the tool. It is the result of a first step towards building the CPS perspective: it can be asserted that the actual implementation reflects, at least, part of the implementation path required for CPS in manufacturing. Using, as a reference model, the 5C architecture for implementation envisioned by (Lee et al., 2015), we may assert that the implementation path has been initiated through the first two levels, i.e. the Smart Connection level and the Data-to-Information Conversion level. This paper especially concentrates on the Data-to-Information Conversion level, presenting the novel approach resulting in it.
As stated by Warren Bennis, “the factory of the future will have only two employees, a man and a dog. The man will be there to feed the dog. The dog will be there to keep the man from touching the equipment” (Fisher 1991). Being quite far away from this long-term vision, the focus nowadays should be on the practical application of a few but effective technological solutions to pave the way towards the next generation of manufacturing systems. In other words, the ambitious aim of the factory of the future still stands. It can be also agreed in steel making with a special emphasis on the safe operations of the production assets: keeping the human as far as possible from inappropriate interventions on the machines/plants is an imperative, eventually leading to keep the optimal asset use and to avoid introducing risks for the operators.
The approach used to develop and implement the condition monitoring tool is summarized in three main steps: i)
Fault tree analysis, to understand how the system under study can fail and the related critical scenarios to identify the key measurable parameters (i.e. key variables) to monitor and map the asset behaviour; ii) Engineering and deployment of the State Detection and Health Assessment tools (part of the overall Smart Maintenance tool); iii) Implementation and integration of the tools in the existing ICT architecture of the plant. The steps enabled to create the analytics now running the Smart Water Monitoring. They are discussed through the following structure of the paper. Section 2 introduces the furnace and its key elements. Section 3 summarizes the use of fault tree analysis, to identify the deviations in the process parameters and, thus, to select the signals needed in order to represent the asset behaviour in its degradation. The signals are then the relevant input of the tool. Section 4 presents the development of the tools for state detection and for health assessment. Section 5 illustrates the implementation of the Smart Maintenance tool in ICT architecture. Section 6, through three real cases, describes the rich output information for the process supervisor and for the maintenance engineer provided by the tools. Eventually, Section 7 provides conclusions, envisioning some future steps towards the implementation of CPS in EAFs.
To move towards this end, a Smart Maintenance tool should provide as much as possible a complete visibility of the asset health status, thus avoiding a need of maintenance operators’ intervention in the nearby of the asset. This research aims at presenting a proof of usefulness of such a type of tool, by discussing a practical implementation in the steel making industry as well as by providing evidences of how the data coming from the plant floor level can be transformed into relevant information, namely relevant indicators, elaborated and displayed at an upper level for managing decisions. The paper grounds on some first information provided by Barreto et al., 1997; Colace et al., 2013 and Colace et al., 2015. The tool has been developed in an industrial environment, at Tenaris Dalmine plant, with the purpose of its practical use in the day-by-day operations. The measurable parameters are the relevant ones for applying a condition-based operation of the furnace: if operating conditions are approaching risky situations, they can be identified thanks to the tool, revealing the asset health status, eventually leading to the subsequent operational decisions by the operator. More specifically, the tool aims at monitoring the parameters of the furnace on the cooling panels and on the refractory bottom/hearth. To this regard, in this paper we will refer to the tool also as a Smart Water Monitoring according to its primary functionality. This communicates the value of the functionality: the tool helps managing the information gathered through the analysis of data resulting from the monitoring of the cooling water of the EAF; therefore, building on such a monitoring, information extraction leads to a “smart water” that conveys the relevant indicators for the asset health status. This type of analysis enables to make evident how the concept of digital twin, that is typical of Cyber-Physical Systems (CPS), is actually being implemented in the case. More specifically, it is remarkable a reduction of distance between the physical transformation – happening in the cooling water – and even physical-chemical transformation – occurring in the furnace and influent for the physical transformation of the water – and the cyber level implementing the analytics: it could be asserted that the monitored water, used as the conveying entity, is acting as a digital twin indirectly linking to the physical and chemical
2. FURNACE DESCRIPTION The EAF is constituted by a lower sheet metal keel (lowerpart) coated with refractory bricks, which serves to contain the liquid steel, and is cooled by a cage, with a structural function, that supports panels cooled by a water circuit (Ferro et al., 2007). In the heating process, the panels cooled by a water circuit could be stressed out, and therefore fail with a leak of water in the furnace, which is finally leading to risky situation. The panels are particularly relevant for the present research. In Tenaris Dalmine plant, there are two electric furnaces with the following characteristics: capacity 105 t, diameter 6.1 m, weight of the structure (considering also refractory bricks) 230 t. The characteristics of each furnace allow to obtain 98 t 28
2017 IFAC IMS December 5-7, 2016. Austin, TX, USA
Luca Fumagalli et al. / IFAC-PapersOnLine 49-31 (2016) 19–24
of liquid steel, with an average power request of 67 MW per cast, and a maximum power peak of 89 MW. The casting cycle is around 38.5 min and this corresponds to a productivity above 150 t/h. The main contribution of energy needed for melting the scrap is supplied from the electricity for about 2/3 and for 1/3 by chemical energy supplied by the lances and the burners on board. In particular: melting 100 t of steel requires approximately 40 MWh of electricity; the chemical process is instead controlled by a system that monitors gases exiting the furnace and optimizes the injection by lances, providing oxygen; more specifically, the chemical system is a set of devices that provides multipoint injection of oxygen, methane and carbon. It can be easily understood that a lot of energy is provided to melt the steel and this impacts on the stress that the panels can suffer.
21
graphical representation of the logical path that connects the unwanted event – namely the “root” of the tree, also said “top event” – with the events that are its root causes. The technique follows a hierarchical methodology which first connects the top event with the events that are directly intended as its root causes; such events (intermediate events) are in turn related to other events that are the underlying causes. The hierarchical tree construction is completed when it has reached the so-called events / basic causes – that is to say, the "leaves" of the tree. For the proposed case, FTA was built relying both on experience of the industrial experts and theoretical issues analysed by academic experts, involved in the research. The identified top events are the holing of the panels and the presence of the water in the bottom/hearth. In figure 3 and 4 an extracted part of the FTA related to the two top events identified is presented. Some boxes of the FTA are provided with notes, helpful in order to document the experience and knowledge synthesized during the first project meetings.
Fig. 1. Rendering of the EAF
Fig. 3. Extract of fault tree analysis: holing of the panels
The panels are positioned around the furnace circumference on two different rows. To this regard, they are defined by the Tenaris practitioners as “down panels”, when they are in the first row close to the hearth/bottom, and “up panels” when they are in the second row, close to the furnace roof. According to the fault tree analysis, “down panels” reveal to be less stressed since they are most of the time covered by the steel slag. On the other hand, the “up panels” are quite stressed, and they may be broken by different mechanisms, namely: suddenly broken by a splash, progressively degraded to the exposition when not covered by the steel slag, stressed by irradiation by electric arc.
Fig. 4. Extract of fault tree analysis: presence of water in the bottom/hearth For each critical scenario related to a top event, the key measurable parameters – i.e. the key parameters that best describe the behaviour of the assets – are identified. The objective is to define a limited group of variables, in the remainder key variables, highly representative of the asset / scenario or some of its features. The key variables identification is a relevant phase of the State Detection and Health Assessment tools deployment. In fact, the cooling panels’ behaviour is not easy to be modelled. Sensors can be hardly installed due to harsh conditions, thus few information about water temperature, flow and pressure can be collected. Moreover, within the panel, the water goes under a change of physical status, creating also areas in which there is water vapour, making very difficult to deploy a physical model of the panel itself. For these reasons, in this industrial case the identification of key variables related to an anomalous behaviour of the panels is an important phase to approach problem solving. Once identified the key variables and completed the feasibility study, the monitoring tool needs to be engineered with specific algorithms in order to assess in which state the furnace is working. The FTA clearly leads to the decision of developing two different part of the tool,
Fig. 2. Rendering of the furnace with panels 3. FAULT TREE ANALYSIS AND KEY VARIABLES RELATED TO CRITICAL SCENARIOS The Fault Tree Analysis (FTA) has been developed by the U.S. Air Force and then used in various sectors in order to support the reliability analysis of a technical system. The analysis according to the FTA technique is based on the development of a tree, called Fault Tree, which allows a 29
2017 IFAC IMS 22 December 5-7, 2016. Austin, TX, USA
Luca Fumagalli et al. / IFAC-PapersOnLine 49-31 (2016) 19–24
during the heating cycles due to steel melting, the reference feature of the key variable indicator is defined; in this way, the expected values of the key variable indicator at the generic time t are known and it is possible to build a control chart with upper and lower thresholds. Subsequently, for this part only SD model is developed, as the alarm is enough to take a decision, i.e. to stop the EAF because exceeding thresholds means the presence of water in the bottom/hearth (which is clearly a risky situation).
inclusive of their respective algorithms: i) Smart Water Monitoring for panels; ii) Smart Water Monitoring for refractory hearth. 4. ENGINEERING AND DEPLOYMENT OF THE STATE DETECTION AND HEALTH ASSESSMENT TOOLS For the top event “Holing of the panel”, the Smart Water Monitoring for panels tool has been developed. In particular, State Detection (SD) and Health Assessment (HA) models are developed for this first tool: their main characteristics are summarized in the remainder. We start with the SD model. 1. An alarm threshold model has been defined with a SD aim. In particular, a control chart has been deployed to take under control the key measurable parameter for each panel. 2. All the critical events that go out of the control chart are counted in a “critical events counter”. The alarm is then displayed when such a counter overcomes the defined threshold. 3. In order to fine tune the SD model, the alarm is also parametrized according to the number of critical events within a certain timeframe. Starting from the variables of the SD model for panels, it is then possible to engineer the HA model. 1. The model is based, for each panel, on the overlapping area between two distributions of a key feature extracted by the analysis of variables related with cooling water of the panels. It enables matching the most recent features (used as process signatures resulting from smart water monitoring) with those features observed during normal and faulty asset behaviour. 2. Variables have been manipulated with data analysis approaches in order to obtain a Gaussian distribution. Thus, once obtained the distributions, it is possible to sketch on a graph two curves: the first one as reference feature/signature, coming from data of not stressed operating conditions of the panel; the second one as feature/signature representative of the actual asset health status due to the operating conditions. 3. For each panel, the construction of the reference feature and the actual feature are needed in order to compute the overlapping area, finally corresponding to an indicator that is used to measure the actual overlap. The tool also compares the value of the average key variables in the period of analysis with its reference average, being the maximum difference an empirical value. The set of information from the two different algorithms delineates a matrix built upon four quadrants with corresponding scenarios/alert levels. Regarding the top event “presence of water in the bottom/hearth” a SD model is engineered as Smart Water Monitoring for refractory hearth/bottom. In particular, it is worth remarking that, in the feasibility study, it has been identified that the presence of water in the bottom of furnace (i.e. on the hearth/bottom) determines an anomalous trend of a key variable indicator. Therefore, based on historical data,
5. IMPLEMENTATION AND INTEGRATION OF THE TOOLS IN THE EXISTING ICT ARCHITECTURE The state detection and health assessment models engineered are then coded into the Smart Maintenance tool, and the output must be shown to the process operator/maintenance engineer in the control room. Therefore, the Smart Water Monitoring is developed in such a way that is accessible as Level 2 tool in the ICT architecture (see figure 5).
Fig. 5. Architecture of Tenaris Dalmine existing ICT systems The tool is web-based and fully integrated with the suite used by plant operations in Tenaris Dalmine. This fully realizes the paradigm postulated since many years about integration of Condition Based Maintenance solutions within the information system for maintenance management (Fumagalli et al., 2009). Moreover, the Smart Water Monitoring has been developed as highly parametrized tool: it is thus possible to apply it on other similar furnaces, even with differences in configurations. It would be relevant for further developments to other EAFs, beyond Tenaris Dalmine. 6. TESTING PHASE: REAL CASE ANALYSIS For a more complete and clear illustration of the tools and its rich output information three real cases are presented. In a first case, during the testing phase of the developed tools, there were some water tests to identify the presence of water in the hearth/bottom. During the heating cycles of testing days there are abnormal features of key variables. The tests started in the heating cycle number 16 (“campagna 16” in the legenda of Fig. 6), where the values of the key 30
2017 IFAC IMS December 5-7, 2016. Austin, TX, USA
Luca Fumagalli et al. / IFAC-PapersOnLine 49-31 (2016) 19–24
variable of castings during the cycle remained lower than the expected values.
23
variables bigger than the respective thresholds, there is an alert level 4 (i.e. the 4th quadrant of the risk matrix). This is the worst scenario in which the actual process is different from the reference process, and the actual feature is moving to the right, i.e. this represents a deviation leading towards more degradation. In other words, the process seems to stress the panel more than the normal operating conditions. This arises attention: the operator should be aware of the fact that that the actual maintenance policy would be no longer suited; indeed, it is recommended increasing inspections to control/prevent any possible damage.
Fig. 6. State detection tool for hearth/bottom. This can be highlighted by comparison with the reference values and the past heating cycle (number 12 – “campagna 12” in the legenda). At the beginning of the next campaign (number 17 – “campagna 17” in the legenda), the values of key variable have an anomalous feature compared to the reference values. It is due to the water presence in the hearth/bottom, until all the water vaporizes and the key variable reaches the reference value. In a second case, during a planned inspection, two holes of small dimensions were discovered in a panel.
Fig. 9. State detection tool for panels.
Fig. 7. Picture of the panel took in the inspection Looking in the days prior to the inspection, the SD tool for hearth/bottom shows how the key variable is lower than the expected value of the reference feature, over the minimum threshold. After some time, the key variable returns to the expected values because, probably, the holes were covered and clogged by dross, thus there were no more water leakage.
Fig. 10. Health assessment tool for panels. 7. CONCLUSIONS Production process in steel making industry involves many variables and operators cope with the tasks of monitoring, controlling and diagnosing the health status of the production assets. This often results in high difficulties to effectively analyse the current assets’ behaviour, to detect and diagnose process anomalies. This eventually impacts on the capability to quickly take appropriate control actions. The experience and knowledge gained by Tenaris Dalmine process and maintenance staff gave the foundations for initiating a novel approach, finally introducing the herein presented Smart Maintenance tool. The tool enables to detect incipient failures through monitoring of furnace panels and hearth/bottom, before causing a downtime that may bring disastrous consequences on people safety, integrity of assets and non-continuity of the production cycle. The fact that the ICT infrastructure was pre-existent, as well as the importance of safety issues in such an asset operation, represented good reasons for investing in the tool as solution for the company, bringing innovation for maintenance. Today, the tool is
Fig. 8. Zoom of state detection tool for hearth/bottom To better understand this case, it is interesting to evaluate the behaviour of the SD tool for the panels for the same event: figure 9 shows the register of the critical events counter with some key variables bigger than the upper threshold in the days before the inspection, i.e. castings which have stressed out the panel and probably determined the holes. It is evident that the panels appeared to be subject to stress in these days (signal in Fig. 8 out of control band). In the third case, the panel of this real case seems to change its actual feature. Fig. 10 shows the health assessment tool in which the actual Gaussian feature (red line) and the reference feature (black line) are compared. As the rate of overlapping area is low and the difference between the average of the key 31
2017 IFAC IMS 24 December 5-7, 2016. Austin, TX, USA
Luca Fumagalli et al. / IFAC-PapersOnLine 49-31 (2016) 19–24
running and the aim is to correlate the critical events counter and the overlapping areas (i.e. the indicators provided by the tool) to the useful life of the panels. In the Smart Water Monitoring for Panels, the validated HA and SD algorithms are the starting point in order to continue towards this direction. To this end, it is necessary to continue the testing phase in order to collect the data needed to characterize the life of the panels up to their replacement/repair. The tracking of the panels life will then serve to map the information of Overlapping Areas (resulting from Health Assessment model within Smart Water Monitoring for Panels tool) and the critical events counter (resulting from State Detection model within Smart Water Monitoring for Panels tool) with the actual degradation occurred/observed in the panel. As a vision for the future, the final aim will be the building of an acceptable degradation function of the panels. This would be the basis for the definition of a reliability model of the panels themselves – to express the probability not to degrade before a time t –, thus for the computation of the subsequent risks. On the whole, it could be eventually possible to obtain (by a computational model) a prognostic risk prediction. A further extension of the Smart Maintenance tool could be built by reflecting on the vision of the social network of machines suggested by Lee et al., 2015. An extension of the analytics approach would be helpful to provide EAFs with self-comparison ability. This will be useful for performance rating among the fleet owned by the company, and for a similarity identification between the performance and assets operating condition, eventually useful to develop a powerful predictive analytics of the remaining useful life based on the different operating conditions. On the scientific level, it would enable continuing the implementation path towards building cyber-physical systems in production.
Djurdjanovic D, Lee J and Ni J. (2003) Watchdog agent – an infotronics-based prognostics approach for product performance degradation assessment and prediction. Adv Eng Inform; 17: 109–125. Espíndola D., Fumagalli L., Garetti M., Botelho S., Pereira C., (2011). Adaption of OSA-CBM Architecture for Human-Computer Interaction through Mixed Interface, INDIN 11 conference, Caparica, Lisbon, Portugal, 26-29 July 2011, 485 - 490. Espíndola D., Fumagalli L., Garetti M., Pereira C.E., Botelho S., Ventura Henriques R. (2013) A model-based approach for data integration to improve maintenance management by mixed reality. Computers in Industry, Volume 64, Issue 4, May 2013, 376-391. Ferro L., Giugliano P., Galbiati P., Memoli F., Giavani C., Maiolo J. (2007). The electric arc furnace of Tenaris Dalmine: from the application of the new technologies of digital electrode regulation and multipoint injection to the dynamic control of the process. 16th IAS Steelmaking Conference, 6th – 8th November 2007, Rosario, Argentina, 59-72. Fumagalli L., Macchi M., Rapaccini M. (2009). Computerized maintenance management systems in SMEs: A survey in Italy and some remarks for the implementation of Condition Based Maintenance, IFAC Proceedings Volumes (IFAC-PapersOnline), 13 (PART 1), pp. 1615-1619. Fumagalli L., Ierace S., Dovere E., Macchi M., Cavalieri S., Garetti M. (2011). Agile diagnostic tool based on electrical signature analysis, IFAC Proceedings Volumes (IFAC-PapersOnline), 18 (PART 1), pp. 14067-14072. Fisher M. (1991) The millionaire's book of quotations. p. 15 Guillén A.J., Crespo A., Macchi M. & Gómez J. (2016). On the role of Prognostics and Health Management in advanced maintenance systems. Production Planning & Control, Huynh KT, Anne B and Christophe B. (2012). Adaptive condition-based maintenance decision framework for deteriorating systems operating under variable environment and uncertain condition monitoring. Proc IMechE, Part O: J Risk and Reliability. Epub ahead of print 7 November 2012. Kans M, Galar D and Thaduri A. (2015). Maintenance 4.0 in Railway Transportation Industry. Proceedings of the 10th World Congress on Engineering Asset Management (WCEAM 2015) Springer International Publishing, page 317-331, isbn 978-3-319-27064-7. Lee J, Behrad Bagheri, Hung-An Kao. (2105). A CyberPhysical Systems architecture for Industry 4.0-based manufacturing systems, Manufacturing Letters, Volume 3, January 2015, Pages 18-23, ISSN 2213-8463. Lee J, Ni J, Djurdjanovic D, et al. (2006). Intelligent prognostics tools and e-maintenance. Comput Ind 2006; 57(6): 476–489. Muller A, Crespo Marquez A and Iung B. (2008). On the concept of e-maintenance review and current research. Reliab Eng Syst Safe, 93(8): 1165–1187. Zhao J, Cui L, Zhao L, et al. (2009). Learning HAZOP expert system by case-based reasoning and ontology. Comput Chem Eng, 33: 371–378.
REFERENCES Abichou B., Voisin A., Iung B. (2012). Choquet integral parameters inference for health indicators fusion within multi-levels industrial systems: application to components in series. 2nd IFAC Workshop on Advanced Maintenance Engineering, Service and Technology, AMest'12, Sevilla, Spain Barreto SM, Swerdlow AJ, Smith PG, et al. (1997). A nested case-control study of fatal work relate injuries among Brazilian steel workers. Occup Environ Med; 54:599– 604. Colace C., Fumagalli L., Pala S., et al. (2013). An intelligent maintenance system to improve safety of operations of an electric furnace in the steel making industry. Chem Eng, 33: 397–402. Colace C., Fumagalli, L., Pala, S., Macchi, M., Matarazzo, N. R., & Rondi, M. (2015). Implementation of a condition monitoring system on an electric arc furnace through a risk-based methodology. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability, 1748006X15576441. Deloux E, Castanier B and Berenguer C. (2008). Maintenance policy for a deteriorating system evolving in a stressful environment. Proc IMechE, Part O: J Risk and Reliability; 222: 613–622. 32