From Data to Decisions: Unlocking Advanced Analytics for Optimal Steel Ladle Management
Understand the different decision-making levels and how advanced analytics can be leveraged for more efficient management of steel ladles.
About this article
I am a PhD student contributing to the Work Package 4 (WP4) of MSCA-DN CESAREF. My project is developed in cooperation with Tata Steel Nederland and TU Wien.
I study the application of advanced analytics techniques to improve the energy efficiency of steel ladle logistics. This article aims to provide a brief introduction to this subject and its relevance.
Introduction
Where does steel come from?
Although we don’t notice, steel is everywhere. Without it, you couldn’t read this article on your phone or laptop! It is so present in almost every aspect of our lives that 1951 million tonnes of crude steel were produced in 2021 alone.
There are two main ways to produce this large amount of steel: the blast furnace (BF) and electric arc furnace (EAF) steelmaking routes. In a few words, the first creates new steel from iron ore and fuel, while the latter is mainly used to recycle scrap. All the processes involved can be seen in this nice diagram from EUROFER.
Each route has four main operations: ironmaking, steelmaking, secondary metallurgy, and casting. I will provide a brief description of one of these steps.
After casting, several mechanical processes are applied to shape the steel, depending on its application. You can learn more about it here, for example.
Ironmaking
Ironmaking relies on a chemical process called reduction. Applying hot gas to a mix of fuel, iron ore, and flux makes it possible to separate the carbon from the iron ore (iron oxides) and produce the hot metal. Nowadays, most of the production comes from Blast Furnace using coke as fuel.
You have probably heard that this process is very intensive regarding CO2 emissions. For this reason, several European steelmakers are committed to substituting it with cleaner technologies that use Hydrogen as a fuel in the near future. Isn’t that cool?! You can read more about this technology here.
Steelmaking
Opposed to the ironmaking process, steelmaking relies on the oxidation processes. In general, oxygen is blown to reduce the carbon content and increase the molten steel temperature. This can be done with two main pieces of equipment: the BOF (basic oxygen furnace) or the EAF.
The BOF is charged with hot metal produced in the ironmaking step and a small amount of scrap. On the other hand, the EAF is mainly charged with solid steel scrap. For this reason, the EAF is also equipped with electrodes to melt the scrap by applying electrical energy. The output of both processes is molten steel with low carbon content and high temperature. You can read more about this process here.
Secondary metallurgy
Can we use the same steel to produce car parts and construction beams? No, because they need steel with different mechanical properties! This is why the secondary metallurgy is so essential. The steel temperature and composition are precisely controlled to produce the desired steel grade through various processes and the addition of alloying materials.
Another exciting aspect of this step is that it is performed in a vessel called a steel ladle. This essential equipment can hold and transport steel at very high temperatures. This is achieved by constructing it with different layers of refractory materials. Look what it looks like! Impressive, right?!
Casting
Finally, the casting process is reached. The molten steel is loaded into a mold and then solidified into slabs at this step. Nowadays, most steelmakers use continuous casting machines. This is a fascinating process, and I encourage you to read more about it!
The decision-making levels in a steel plant
Now that you know the basics about how steel is made, let’s understand how to connect all these processes to produce it efficiently.
The decision-making process within an industrial plant can be split into five levels based on the impact, time scale, and uncertainty involved [1]. This can be seen in the image below, but let’s also describe what each one is doing.
Planning
At this level, forecasts are used to build a high-level production plan on a scale of weeks to months. For example, this entails estimating how much steel customers will purchase next year. The output is the amount and which steel grades should be produced, such that inventory restrictions are met, and customers’ orders are fulfilled.
Programming
Given the existing plan, one must next define how to program the different processes involved to produce the correct steel grade amount at the right time. Therefore, based on available unit processes, high-level mathematical models of their behavior, and operational restrictions, a schedule defines when each activity should start to achieve the production plan.
Real-time optimization
The real-time optimization level estimates the reference values, or set points, necessary to execute the proposed schedule. It can involve sophisticated models for each process and the several operational restrictions that must be respected to achieve the scheduled production.
Supervisory and regulatory control
Finally, supervisory and regulatory control are implemented to ensure set points are applied. The supervisory level usually involves human-machine interfaces and control rules for the desired process set points. The regulatory layer operates near real-time and communicates directly with sensors and actuators, ensuring that physical equipment is correctly and safely operated.
Improving steel ladle operations with advanced analytics
Advanced analytics applications can be categorized into descriptive, diagnostic, prescriptive, and predictive analytics. This is summarized in the figure below, the so-called analytics maturity curve.
Advanced analytics applications are mainly applied to the first four levels of the hierarchy. Regulatory control should rely on other techniques that provide explicit rules and stability guarantees. Ensuring safety is the number one rule in all industrial operations!
An end-to-end application is now discussed to exemplify how advanced analytics can be applied in the steelmaking industry. To make things more interesting, I will use an application to steel ladle management.
As discussed before, steel ladles are essential to achieve increasingly challenging production goals. Predicting the remaining lifetime of the refractory material is a crucial aspect of the process. By doing so, it is possible to improve the planning of the ladle repairs and reduce the refractory material consumption by maximizing its usage.
We will now answer each question presented in the figure below to help achieve our goal of improving steel ladle management.
What happened?
Understanding what causes the degradation of the refractory material is the first step of this process. Historical data containing details about all the operations applied and the measured wear levels during its lifetime are then collected. Finally, this data is cleaned, transformed, and key variables are presented in a friendly dashboard to enable the first insights.
Measuring the refractory wear is quite a complex procedure. It is nowadays mainly achieved with laser scanners. You can see an example here.
Why did it happen?
Using the data collected previously, an statistical analysis is performed to understand the distribution of each dependent variable. The correlation with the wear can also be evaluated to identify the variables that affect it the most.
Another approach is to use explainable models, such as linear regression and random forests, to start quantifying the effect of each variable. This enables us to identify which variables have the most potential for predicting refractory wear.
What will happen?
A model that can accurately predict refractory wear can now be constructed. For example, a neural network could be employed for this task. As a final result, a simple web application can be built around this model to enable stakeholders to provide the selected input variables and obtain the predicted wear.
What should we do?
This is where we can start integrating the prediction with the different decision-making levels. The application from the previous example enables predictions, but it’s still not connected to the action that will be performed with that information.
This predictive model can be part of a prescriptive maintenance tool in a planning application. The accurate prediction lets us know the best moment to repair the ladle, possibly reducing maintenance costs.
It can also be used to improve the production schedule. Knowing the ladle’s remaining lifetime enables a more efficient assignment of ladles regarding the steel weight, for example. This can lead to a higher production throughput by knowing beforehand how much steel can be produced at the refining stage.
The wear prediction can also be used in real-time as an input to a process optimization system and improve the steel temperature prediction. For example, the operations at the ladle furnace can be adjusted to consider the actual refractory wear, leading to safer and optimized operating practices.
Conclusions
This article presents basic concepts regarding the steelmaking processes and how decision-making is organized in a typical industrial plant. Different levels of decisions are possible, which are classified on the scale, uncertainty, and impact involved.
An application of advanced analytics to steel ladle management is also presented. It illustrated the connection to the decision-making levels and potential applications that can be explored.
Next steps
As for the next steps, stay tuned for more detailed technical articles covering other applications and new results from the CESAREF consortium!
References
[1] Mark L. Darby, Michael Nikolaou, James Jones, Doug Nicholson. RTO: An overview and assessment of current practice, Journal of Process Control, Volume 21, Issue 6, 2011. https://doi.org/10.1016/j.jprocont.2011.03.009.
This article is part of the dissemination activities of the CESAREF project. This project has received funding from the European Union’s Horizon Europe research and innovation program under grant agreement no.101072625.
Funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or European Research Executive Agency (REA). Neither the European Union nor the granting authority can be held responsible for them.
