This article highlights use cases of advanced analytics (i.e. machine learning, mathematical programming, and simulation) in the steel production value chain. Steel products play an important role in our daily life, ranging from real estate construction, agricultural equipment, automotive parts and frames, home appliances, and also e.g. aerospace structural components.
Steel products are the results of a long and comprehensive value chain, starting from the iron ore extracted in iron ore mines. The essential steps of the iron-to-steel value chain are:
- Iron pellet production
- Pig iron production
- Steel production
In this article, I focus on flat steel products, i.e. steel coils, slit coils or sheet metal stacks for further processing in e.g. automotive industry. Flat steel product is flat rolled steel. Flat steel product ships in the form of packaged coils, slit coils, or sheet metal stacks. The transformation from iron ore to finished product (packaged coils or slit coils) is illustrated below.
There are many other variations to the steel value chain, and many other types of final steel products.
Steel value chain, from iron ore to final product
As mentioned the essential steps of the flat steel product production process are iron pelletizing, pig iron casting, and rolled steel production. These major steps of the steel value chain take place sequentially.
Magnetite pelletizing process – from ore to pellet
Iron pelletizing is the process of producing magnetite pellets from iron ore. The general production process is illustrated below.
The iron ore, e.g. in the form of rock debris from blasting operations, first undergoes crushing. Crushing is performed through a series of sequential crushing machines. Fine granularity increases between each crushing stage. The maximum rock debris size allowed to pass on to the next crushing stage is defined, and debris is cycled for another crushing cycle in a given stage until its rock size is below the defined maximum allowed size.
After crushing, grinders reduce rock size further. The resulting debris then go to a bedding and blending yard, where a bedding and blending process takes place. The purpose is to make mixed ore properties. Mixed ores are more convenient for production, improving production efficiency and reducing production costs. Another objective of the blending process is to ensure uniform ore feed quality during further ore processing and refinement. This contributes to a normal, stabile production and a consistent quality.
After bedding and blending, ore debris undergo further milling stages. Separators are deployed between each milling stage, filtering the increasingly fine granular powder by its magnetic property. Impurities and magnetic iron ore content are separated in this way, and purity thus increases after each stage. Eventually, once milling has been completed, the powder mixture is forwarded to froth flotation. Here, a mixture of water, chemicals that aid separation processes, and air bubbles remove impurities such as silicates from the resulting sludge. Meanwhile magnetic filtering and separation between flotation bath increases purity step by step.
The resulting magnetite sludge product is stored in storage tanks, and waste sludge is pumped out of the system for waste handling in e.g. tailings. Larger waste, such as rough granular rock debris, are transported to disposal sites with trucks, rail, or by barge or boat transport.
The valuable magnetite sludge stored in storage tanks is now processed into pellets. First, grounding drums, also referred to as pelletizing drums, are used to produce pellets from dried sludge. The raw pellets then undergo preheating, cooking, and rotary cooling. This is usually done in a kiln.
The final product of this process is the magnetite pellet, which is stored in closed stock areas and later transported by rail or barge transport.
Pig iron production – from magnetite pellet to raw iron cast
Pig iron, also referred to as raw iron casting, raw iron, crude iron, or wrought iron, is produced from magnetite pellets. This raw form of iron is used for further processing into steel and steel alloys.
Pig iron has a high carbon content. It is produced from magnetite pellets and functions as an intermediate good for further processing in steel industry. It is produced by smelting magnetite pellets in a blast furnace, which is fed magnetite pellets (i.e. refined iron ore) and coke (a refined form of coal). The liquid iron is casted into ingots.
Pig iron has a carbon content of somewhere around or above 4%. This makes it very brittle. Since it is brittle its applications are limited and it is mainly produced for downstream steel making. It is also used for ductile iron production, which also has a high carbon content but is ductile and thus less brittle.
Steel making – from pig iron to flat steel product
Flat steel products are sold in the form of coils, slit coils, or sheet stacks. The production starts at the furnace, e.g. electric arc furnace. The general production process is summarized in below illustration.
The furnace is used to melt scrap steel or pig iron pellets into liquid form. Depending on the desired material properties additional alloys are added. Stainless steel e.g. contains chromium. Oxygen supply reduces carbon content of the liquid steel, and calcium is added to remove sulphides and oxides.
The liquid steel is then casted into solid yet still very hot state, in e.g. a continuous stream that is then cut into slabs. Depending on the production program, slabs are stored in a slab yard, or forwarded directly to hot rolling. The first scenario is referred to as cold charging, and the latter is referred to as direct hot charging. Direct hot charging is beneficial as it eliminates the need of reheating slabs for hot rolling. This is done for cold slabs that are sourced in the slab yard, and is performed in a reheating furnace ahead of the hot rolling process.
The hot rolling mill rolls slabs down in thickness. Throughout the rolling process the rolls of the rolling mill experience heavy wear and tear. Usually, slabs are thus rolled in a coffin-shaped slab rolling program. Here slabs with small widths are rolled first, and slab width increases slab by slab until a certain point. After that slab widths decrease again. This is out of roll wear purposes, allowing the rolls to heat up (lower to higher slab widths), and then stabilize roll heat beyond a certain point (decreasing slab widths). Due to the heavy wear and tear rolls of the rolling mill are exchanged frequently, nevertheless. These roll changes must be planned and are an important schedule optimization lever, alongside direct hot charging.
After hot rolling, slabs have been reduced in thickness and are now longer. At the beginning, their temperature was about 2,000C. Cooling from such temperatures is difficult to control and is not uniform. The surface of hot rolled coils is thus rough. The shapes are rounded and not perfect. This is why some hot rolled coils are later-on e.g. surface-treated in pickling, sand blasting, acid-bathing, and grinding lines.
Hot rolled steel is e.g. used in real estate construction work, agricultural equipment, metal buildings, stampings, and automotive frames. But other applications require for additional rolling steps, namely cold rolling.
Cold rolling reduces sheet thickness further, and improves surface quality as well as the strength of the steel. Cold rolling is performed below recrystallization temperature. The rolling process creates small errors and ruptures in the steel, and changes its structure. This improves the mechanical properties of the steel. It can also make the steel less elastic, which can be treated with annealing processes that allow for recrystallization after cold rolling runs. In sum, this creates harder, stronger steel that is also more flexible and ductile and has a smoother, more even and more defined surface. This type of steel is e.g. used in metal furniture, home appliances, and automotive parts.
In addition to cold rolling mills and annealing lines, steel can be cold-treated in turning, grinding and polishing lines as well as it e.g. can be trimmed and coated. Eventually, it is then, if required by the customer, slit into slit coils, and packaged. This is done on slitting and packaging lines.
Analytics in magnetite pellet production
Simulation is the main analytics application that I want to point out for magnetite pellet production. Using simulation the layout and process of existing or planned facilities is modeled in a virtual environment. Models are used for experiments – aiming at e.g. proof of concept, variant comparison (e.g. comparing two layout variants against each other), bottleneck search or capacity planning. A similar example for coal mining has been shared on SCDA in another blog. Please see the link below.
A simulation has multiple advantages when compared to static calculus or norm-based estimations:
- A simulation considers system dynamics, i.e. the actual chain of events in the production plant as they occur throughout time.
- A simulation considers interactions and interdependencies. E.g. a crane must wait for trucks before it can unload its cargo onto the truck.
- A simulation considers actual process models and schedules, e.g. some processes operate around the clock while other processes operate in a single or double shift model.
- A simulation considers stochastic system behavior, as it unfolds along the timeline. For example, machine failures, scrap rates, bad batches, variances in ore content etc.
Once a verified simulation model has been developed and calibrated with actual or assumed data, it can be used to answer questions as follows:
- Are the intended capacities sufficient for dynamic production execution? Is resource utilization efficient and well-balanced?
- Is the blending yard sufficiently large?
- Are there enough vehicles for internal transports, e.g. between crushers and/or grinders? Are there to many vehicles?
- Are there enough maintenance resource units to ensure production stability, considering blending issues and variances in ore quality and content?
Another good example of applied analytics in magnetite pelletizing is mathematical programming. It is e.g. applicable to the blending problem in the blending yard. As the yard contains fine granular ore debris with increased but varying iron content, mathematical programs are applied. These programs are optimization models that aim at optimizing one or several objectives. One common objective is to maintain uniform iron contents over time. This is because uniform contents reduce production costs and ensure consistent quality further downstream.
Flat product steel production and analytics
One important analytics aspect in flat steel product production is direct hot charging. Direct hot charging reduces energy consumption and thus variable production costs, directly adding to operational margins. However, only a fraction of slabs rolled in the hot rolling mill are direct hot charging. The reasons for this are various operational constraints. Some of them are:
- Different shift models at the hot rolling mill and casters
- Different peak throughput capacities at hot rolling mill and casters
- Unexpected customer orders, i.e. ad hoc production orders
- Coffin shaped rolling program at hot strip mill
- Capacity constraints downstream, e.g. in buffers, cold-finishing lines or other post-processing lines
Considering all of this various scheduling problems are solved in steel rolling. First, a master production schedule must be established and maintained. It needs to align customer demand with slab yard capacity (cold charging) and caster production schedules (direct hot charging). Furthermore, it must be aligned with all coupled processes downstream. Second, short-term job sequencing must be established and maintained for all relevant units. The approach to this is complex and depends on the provider. A mixture of optimization modeling and heuristic simulation-based scheduling can be applied since mathematical scheduling is very computation-heavy.
Another problem which is approached using analytics is slab sourcing from the slab yard itself. Based on the hot rolling mill production schedule specific, sometimes unique, slabs must be delivered by the slab yard just-in-sequence. In the slab yard slabs are stacked in high stacks. This means that a large amount of restack movements must be conducted by the yard cranes. These movements are unproductive moves and have significant, negative impact on yard throughput. Mathematical optimization modeling is thus applied to optimize job sequences performed by the yard cranes. These models reduce the amount of unproductive restack movements and ensure just-in-sequence slab delivery to the hot rolling mill.
Concluding remarks on analytics in steel production
For the scope of this article, I considered the following tools, methods, and techniques as examples of “analytics” in ore and steel production:
- Discrete-event simulation for material flow and process simulation.
- Mathematical programming for e.g. scheduling and sequencing.
I provided examples of such analytics applications in iron ore pelletizing and steel rolling. Application examples include:
- Iron ore blending in blender yards, for reduced production costs and uniform input quality throughout the steel value chain.
- Process simulation of existing production facilities to detect bottlenecks in the facility, e.g. in internal transport system, processing capacities, etc.
- Scheduling of jobs for cost reduction, facility utilization, and/or lead time reduction.
There are other types of analytics methods and techniques that can be applied to iron pellet and steel production. This includes such that are generalizable and applied in unrelated industries. For example, forecasting applications for predictive maintenance and guided demand planning. Or constraint programming for shift scheduling for work forces and other types of shared resource pools. I, instead, highlighted examples that are more industry-specific.
If you are interested in reading more examples of applied analytics, in mining, aluminum rolling, or other industries, then the following articles are a good place for you to start:
- Link: Optimization via master production scheduling
- Link: PUSH production and analytics
- Link: Production scheduling in SAP ERP system
- Link: Optimized machine setup sequence
- Link: Manufacturing MRP planning software
- Link: A method of ore blending based on the quality of benefication
Mining engineer with industrial iron ore mining and pig iron production experience. Mining operations simulation expert.