kth.sePublications
12345673 of 18
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Data-driven modeling for online predictions in steelmaking: To optimize calcium additions and castability in low alloyed liquid steels
KTH, School of Industrial Engineering and Management (ITM), Materials Science and Engineering, Process. KTH Royal Institute of Technology.ORCID iD: 0000-0003-2076-7228
2024 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

In this work, innovative data-driven process models were designed, developed, and examined online for their application in the steel industry. The objective was to help operators in decision making during calcium additions in liquid steel and casting of steel grades which are susceptible to submerged nozzle (SEN) clogging. The developed data-driven process models were examined in a real industrial environment to demonstrate the use of supervised machine learning (ML) and deep learning (DL) algorithms for online predictions of specific process parameters. The adaptation of two novel data-driven decision support systems in production helped steelmakers to address the critical challenge of minimizing costly production stops caused by SEN clogging. Calcium (Ca) additions during secondary steelmaking is a well-adopted practice to transform solid oxide non-metallic inclusions (NMIs) into globular shaped liquid oxides. This practice of Ca treatment helps to avoid SEN clogging. Hence, operators in steel plants follow standard operating procedures (SOP) that detail the use of static quantity of Ca wire additions. However, these SOPs, providing a baseline for production, do not account for the dynamic nature of steelmaking processes and the varying chemistry of NMIs for each 'heat' produced. To address this gap, the impact of varying CaSi wire additions, beyond the SOP's guidelines, on the transformation and behavior of NMIs in low-alloyed steel grades was explored by developing the 'ClogCalc' decision support system. The aim was to establish a more reliable and responsive approach to Ca treatment, potentially leading to more effective control in preventing SEN clogging. The implementation of 'ClogCalc' has demonstrated a significant 30% reduction in SENclogging, as evidenced by industrial trials at Voestalpine steel plant.

Recognizing 'castability' of steel by monitoring the conditions of SEN clogging is essential for uninterrupted continuous casting. With this information prior to casting, operators can take preventive action against possible SEN clogging occurrences, thus reducing unplanned downtimes. In response to the severe implications of SEN clogging, a novel approach to forecast 'castability' of steels was introduced by developing the 'Checkcast' decision support system. The adaptive neuro fuzzy inference system (ANFIS) and long short-term memory (LSTM) network model was used as a base algorithm for estimating 'castability' of steel grade. The output of 'Checkcast' helps to anticipate the event of clogging by analyzing both the past condition of the SENs and changes in the steel chemistry during the transport of the steel from refining ladle to the tundish. The verification was conducted at the Sidenor steel plant. While the primary focus of this study was to develop data-driven process models, efforts were also made to study the foundational principles governing the SEN clogging and evolution of NMIs in studied steel grade. Laboratory experiments were also conducted on liquid steel samples obtained from steel plants. Parametric liquid windows were derived using systematic thermodynamic assessments using FactSage software. In addition, to understand the wetting characteristics at interfaces of NMIs and the SEN refractory, interfacial properties were calculated using ab-initio calculations.

Abstract [sv]

I den här studien utvecklades och undersöktes innovativa datadrivna processmodeller för en integreradanvändning inom stålindustrin. Målet var att underlätta för operatörer i stålverken att bestämmakalciumtillsatser i flytande stål, samt att ge indikationer om gjutbarheten hos stålsorter som är känsligaför ingensättning i de gjutrör (SEN) som leder stålet till kokillen. De utvecklade processmodellernaimplementerades i en verklig industriell miljö utan att förändra den befintliga IT-infrastrukturen istålverket i syfte att visa användningsområden för övervakad maskininlärning (ML) och djupinlärning(DL) av integrerad bevakning och förutsägelser. Anpassningen av dessa datadrivna modeller iproduktionen hjälpte ståltillverkarna att hantera det kritiska problemet med att minimera kostsammaproduktionsstopp orsakade av SEN-igensättning. Kalciumtillsättning under sekundär ståltillverkningär en väl vedertaget metod för att omvandla icke-metalliska inklusioner (NMIs) i fasta oxider tillglobulär-formade flytande oxider. Detta förfaringssätt hjälper till att undvika att SEN-röret täpps igen.Därför följer operatörerna i stålverket de standard operative procedurer (SOP) som föreskriveranvändningen av bestämd mängd av Ca-tillsatser för specifika stålsorter. Dessa SOP utgör en baslinjeför produktionen, men de tar inte hänsyn till ståltillverkningsprocessens dynamiska karaktär eller denvarierande kemin hos NMIs för varje ‘värme’ som produceras. För att ta itu med denna kunskapsbristutforskades i denna studie effekten av att variera Ca-tillsatser, utöver de standardiserade SOPriktlinjerna,på omvandlingen och beteendet hos NMIs i låglegerade stålsorter genom att utveckla dendatadrivna processmodellen ‘ClogCalc’. Syftet var att etablera en mer tillförlitlig och responsiv metodför kalciumbehandling, vilket potentiellt leder till mer effektiv kontroll för att förhindra SENigensättning.Resultaten efter implementeringen av ‘ClogCalc’-modellen har betydande konsekvenseri att reducera SEN-igensättning med upp till 30%.

Att känna igen gjutbarheten hos stål genom att övervaka tillståndet hos SEN är också väsentligt för enoavbruten gjutning. Med denna information före gjutning kan operatörer vidta förebyggande åtgärdermot möjliga igensättningshändelser, vilket minskar oplanerade driftstopp. Som svar på de allvarligakonsekvenserna av SEN-igensättning introducerades ett nytt tillvägagångssätt för att förutsegjutbarheten hos stål genom att utveckla modellen ‘Checkcast’. ANFIS (från engelska ‘AdaptiveNeuro-Fuzzy Inference System’) och nätverk med LSTM (‘Long Short-Term Memory’) användes sombasalgoritmer för denna modell. Utdata från denna processmodell hjälpte till att förutse händelsen avigensättning genom att analysera både det tidigare tillståndet hos SEN och förändringarna i stålkeminunder transporten av stålskänken från raffinering till gjutningsprocessen. Medan den primära fokusen idenna studie var att utveckla datadrivna processmodeller, gjordes även ansträngningar för att studerade grundläggande principerna som styr SEN-igensättning och förloppet av icke-metalliskakontaminationer (NMIs) i de olika stålsorterna. Laboratorieexperiment utfördes på flytande stålproverfrån stålverk. Det parametriska vätskefönstret härleddes genom systematiska termodynamiska analysermed hjälp av FactSage-programvaran. För att förstå agglomerations- och vätningsegenskaperna vidgränssnitten mellan NMI och stålsmältan, beräknades materialens gränsytegenskaper med hjälp av en ab-initio-metod.

Place, publisher, year, edition, pages
Stockholm: KTH Royal Institute of Technology, 2024. , p. 155
Series
TRITA-ITM-AVL ; 2024:23
Keywords [en]
calcium treatment, castability, clogging, data-driven, steelmaking
National Category
Metallurgy and Metallic Materials
Research subject
Materials Science and Engineering
Identifiers
URN: urn:nbn:se:kth:diva-354824ISBN: 978-91-8106-081-2 (print)OAI: oai:DiVA.org:kth-354824DiVA, id: diva2:1905433
Public defence
2024-11-08, F3 / https://kth-se.zoom.us/j/62350179538, Lindstedtsvägen 26, Stockholm, 10:00 (English)
Opponent
Supervisors
Note

Principal supervisor: Assoc. Prof. Björn Glaser, KTH

Co-supervisor: Docent Dr. Andrey Karasev, KTH

Chair at the defense: Prof. Mikael Ersson, KTH

Opponent: Prof. Qifeng Shu, University of Oulu, Finland 

Members of the grading committee         

Dr. Dmitrij Ladutkin, Saarstahl AG, Tyskland

Assoc. Prof. Valentina Colla, Scuola Superiore Sant'Anna, Italien

Prof. Dr. Stefan Sandfeld, Jülich Forschungszentrum, Tyskland

Substitute: Assoc. Prof. Stefan Schönecker, Materialvetenskap, KTH

Available from: 2024-10-15 Created: 2024-10-14 Last updated: 2024-10-31Bibliographically approved
List of papers
1. Online Supervisory System for In-Process Optimization of Calcium Additions by Continuously Monitoring the State of Non-metallic Inclusions Inside Low-Alloyed Liquid Steels
Open this publication in new window or tab >>Online Supervisory System for In-Process Optimization of Calcium Additions by Continuously Monitoring the State of Non-metallic Inclusions Inside Low-Alloyed Liquid Steels
2024 (English)In: Metallurgical and materials transactions. B, process metallurgy and materials processing science, ISSN 1073-5615, E-ISSN 1543-1916, Vol. 55, no 3, p. 1395-1413Article in journal (Refereed) Published
Abstract [en]

A decision support system was developed using supervised machine learning (ML) approach for optimization of calcium (Ca) additions by continuously monitoring the physical state of non-metallic inclusions (NMIs) inside low-alloyed liquid steels. In this work, two instances were considered to design the base algorithm for the proposed supervisory system: (1) Clogging of submerged entry nozzle (SEN) during continuous casting of steels due to accumulation of solid oxide non-metallic inclusions (NMIs) and ( 2) Ca treatment during secondary steelmaking for modification of oxide NMIs from solid to liquid state to avoid SEN clogging. At first, experimental investigations were carried out on liquid steel samples from three low-alloyed Ca-treated steel grades from the same steel family to evaluate the characteristics of solid oxide NMIs that cause SEN clogging. In the next step, data-driven models were developed using an in-house ML algorithm trained primarily with process data for calculating the value of the newly proposed dummy parameter 'Clog.' These models, after testing, were architected to develop a supervisory system based on experimental investigations and data-driven models. The objective of this proposed supervisory system was to predict the optimum quantity of Ca needed for successful modification of NMIs from solid to liquid state to avoid SEN clogging based on the forecasted 'Clog' value. Finally, industrial data from ~ 3000 heats were tested to verify the results obtained from the developed supervisory system. The results confirmed that this novel supervisory system could predict the optimum class of Ca for all studied steel grades with 95 to 98 pct accuracy. The integration of this online supervisory system in steel production is expected to minimize operators' corrective actions in achieving realistic control of Ca additions.

Keywords
Calcium treatment, Clogging, Machine learning, Online monitoring, Submerged Entry Nozzle
National Category
Metallurgy and Metallic Materials
Research subject
Metallurgical process science
Identifiers
urn:nbn:se:kth:diva-354804 (URN)10.1007/s11663-024-03035-z (DOI)
Funder
KTH Royal Institute of TechnologyEU, Horizon 2020, 869815
Note

QC 20241021

Available from: 2024-10-14 Created: 2024-10-14 Last updated: 2024-10-21Bibliographically approved
2. Adaptive Neuro‐Fuzzy Inference System‐Long Short‐Term Memory Hybrid Model to Forecast Castability of Al‐Killed Steel Prior to Continuous Casting
Open this publication in new window or tab >>Adaptive Neuro‐Fuzzy Inference System‐Long Short‐Term Memory Hybrid Model to Forecast Castability of Al‐Killed Steel Prior to Continuous Casting
2024 (English)In: Steel Research International, ISSN 1611-3683, E-ISSN 1869-344XArticle in journal (Refereed) Published
Abstract [en]

Continuous casting of aluminum (Al) deoxidized steels demands careful inspection due to the occurrence of submerged entry nozzle (SEN) clogging, leading to unexpected production stops. Recognizing the castability of aspecific “cast” by monitoring the condition of the SEN is essential for uninterrupted casting. With this information prior to casting, operators can take preventive action against possible clogging occurrences, thus reducing unplanned downtimes. In response to the severe implications of SEN clogging, this work introduces a novel way to forecast castability of Al-killed steels. A hybrid model is proposed that integrates the adaptive neuro-fuzzy inference system (ANFIS) and long short-term memory (LSTM) networks. The output of the model helps to anticipate the event of clogging by analyzing both the past condition of the SEN and changes in the steel chemistry during the transport of the steel ladle from refining to the casting process. A comprehensive analysis of 150 casts helped to build the ANFIS algorithm for estimating the castability index (CI) parameter from steel chemistry. LSTM algorithm is used as asubsequent step to forecast castability in the next 20–25 min. Discrepancies between the predictive response and the actual conditions are reported. Although the real-time implementation of the proposed model is the ultimate goal, the focus of this work was to present the methodology and demonstrate its potential.

Keywords
artificial intelligence, castability, clogging, forecasting, submerged entry nozzle
National Category
Metallurgy and Metallic Materials
Identifiers
urn:nbn:se:kth:diva-354805 (URN)10.1002/srin.202400220 (DOI)
Funder
EU, Horizon 2020, 869815
Note

QC 20241021

Available from: 2024-10-14 Created: 2024-10-14 Last updated: 2024-10-21Bibliographically approved
3. Practical Implications of Using an Online Data-Driven Optimizer for Calcium-Treated Steels
Open this publication in new window or tab >>Practical Implications of Using an Online Data-Driven Optimizer for Calcium-Treated Steels
2024 (English)In: Metallurgical and materials transactions. B, process metallurgy and materials processing science, ISSN 1073-5615, E-ISSN 1543-1916, Vol. 55, no 5, p. 3923-3937Article in journal (Refereed) Published
Abstract [en]

Calcium (Ca) additions during secondary steelmaking are a well-adopted practice to transform solid oxide non-metallic inclusions (NMIs) into globular-shaped liquid oxides. The claimed hypothesis that liquid NMIs reduce SEN clogging has been proven in the past by researchers. However, the exact quantity of Ca needed to transform the physical state of NMIs during steelmaking remains uncertain. Operators in the steel plant use a consistent quantity of Ca additions for specific steel grades, but this approach does not account for the varying physical states and evolving dynamics of NMI's characteristics in each 'heat'. To overcome this, a study was conducted to explore the impact of varying Ca additions on the transformation and behavior of NMIs in low-alloyed Ca-treated steel grades. The aim was to establish a more reliable and responsive approach to Ca treatment, potentially leading to more effective control in preventing submerged entry nozzle (SEN) clogging. The proposed methodology involved online monitoring of NMIs state coupled with controlled variations in Ca addition, deviating from fixed quantity, to observe its effects on NMIs state transformations. Through careful analysis of collected data and the implementation of a data-driven optimizer, this study reports the practical implications of using optimal amounts of Ca during secondary steelmaking. The resulting change due to dynamic calcium silicide (CaSi)-cored wire additions and their impact on SEN clogging were evaluated. The findings reveal the significant role of optimal CaSi wire additions, leading to improved steel castability and a notable 30 pct reduction in SEN clogging tendencies. The results obtained after the implementation of the data-driven optimizer 'ClogCalc' have significant implications for steel manufacturers, offering new insights into enhancing Ca treatment efficiency.

Keywords
calcium treatment, clogging, ladle refining, machine learning, steelmaking
National Category
Metallurgy and Metallic Materials
Identifiers
urn:nbn:se:kth:diva-354818 (URN)10.1007/s11663-024-03226-8 (DOI)
Funder
EU, Horizon 2020, 869815
Note

QC 20241021

Available from: 2024-10-14 Created: 2024-10-14 Last updated: 2024-10-21Bibliographically approved
4. Physics-informed data-driven model to estimate submerged entry clogging possibility with the aid of ab-initio repository
Open this publication in new window or tab >>Physics-informed data-driven model to estimate submerged entry clogging possibility with the aid of ab-initio repository
(English)In: Steel Research International, ISSN 1611-3683, E-ISSN 1869-344XArticle in journal (Refereed) Submitted
National Category
Metallurgy and Metallic Materials
Identifiers
urn:nbn:se:kth:diva-354819 (URN)
Funder
EU, Horizon 2020, 869815
Note

QC 20241015

Available from: 2024-10-14 Created: 2024-10-14 Last updated: 2024-10-15Bibliographically approved

Open Access in DiVA

Doctoral thesis_sudhanshu kuthe(16355 kB)305 downloads
File information
File name FULLTEXT01.pdfFile size 16355 kBChecksum SHA-512
0bc224eef417fbc766687ee38b5a262740035ced8c40ed5a3791881ae4f950deab0a3412db45dd0e9b210cd3e5da105717105db333fa8f99ff7e4f4915bf1486
Type fulltextMimetype application/pdf

Authority records

Kuthe, Sudhanshu

Search in DiVA

By author/editor
Kuthe, Sudhanshu
By organisation
Process
Metallurgy and Metallic Materials

Search outside of DiVA

GoogleGoogle Scholar
Total: 305 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

isbn
urn-nbn

Altmetric score

isbn
urn-nbn
Total: 830 hits
12345673 of 18
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf