Steel is produced in basic oxygen furnaces  (BOF) and in electric arc furnaces (EAF). Steel casting is done through  continuous casters and by ingot teeming. Depending on method of casting, profiles,  and sizes, semi-products (ingots, blooms, billets) are transported downstream  through production supply chain. As could be seen from the scheme the melting  and casting are the key stages of production. That is why melt shop and casting  scheduling is so important for the company.
The old scheduling process was arranged  in the following way. Input data was prepared in manual mode by 5 specialists,  who prepared requirements for casting based on planned orders of rolling mills.  Based on the requirements, specialists created melt shop and caster schedules. These  schedules had sufficient accuracy on the horizon of 4 – 7 days. Beyond this horizon the accuracy significantly deteriorated.  Scheduling took much time and efforts.   Rolling mills requirements changes or deviations from the schedule in  many cases lead to rescheduling. Often because of high laboriousness  rescheduling was done after the end of working days or on weekends. Absence of rolling  mills requirements beyond 1-2 week horizon significantly reduces efficiency of  decision making of planners. Also data quality needed to be improved.  Manufacturing execution systems monitored only general volumes of production  without accounting for heats, heats sequencing, and without information about  allocated orders.
As a result the company had high  level of work in process and low due date performance. It was logical that the  issue of rearranging of scheduling process and improving its efficiency was  open.

General task formulation of melt shop and caster scheduling

Melt shop and caster schedules are  normally generated at least once a day—sometimes more often. Thousands of  orders must be combined and sequenced for production through iron making, steelmaking,  and continuous casting.  Each time this  occurs, the planner tries to balance multiple business objectives, which exist  in a tradeoff relationship.
The first goal of melt shop and caster  scheduling is to produce orders on time, that is, to minimize order earliness  and order lateness .  The  desired production time for each order at casting is usually given by some  higher level production planning function, which balances order loads across  the entire plant.  Planners try to  respect this requirement by scheduling each order for production on the day  that it was requested.  If orders are  produced too early, unnecessary inventory may result.  If orders are produced late, customer  delivery performance suffers and unnecessary expediting costs may be incurred.
The second goal of melt shop and caster  scheduling is to maximize tundish utilization .  For each steelmaking grade, there is some  maximum number of heats that can be cast using a single tundish.  When the maximum is reached, the tundish must  be switched out and relined with new refractory.  Maximizing tundish utilization reduces  operating cost per ton, minimizes the number of “top” and “bottom” slabs (or  blooms, or billets) as a percent of total production, and helps keep liquid  steel flowing for as much calendar time as possible.  The Czech plant has several casters, and some  portion of the order book is often considered to be “swing orders,” able to be  produced at alternative casters.  Tundish  utilization can normally be improved by intelligent allocation of swing orders  to casters during the caster scheduling process.
The third goal is to minimize grade and width  transitions within the tundish.   Minimizing grade transitions improves prime yield, as material from the  chemical transition zone may have limited usefulness for customer orders.  Reducing the number and  severity of slab width transitions is important for flat products producers, as  it reduces wear on both casting and hot rolling equipment, reduces breakout  risk, and can improve prime yield.
The fourth goal is to minimize the  amount of stock within the melt shop and caster schedule, and focus production  on current demand.  As manufacturing  orders are combined into heats, and heats into tundish sequences, stock (or  “open order”) material must sometimes be inserted, to fill out a heat or extend  tundish length.   Production of stock is usually a waste of precious manufacturing time,  since it cannot be converted immediately into revenue, but lingers in  inventory.
The fifth goal of melt shop and  caster scheduling is to minimize over-grading. Each customer  order has alloy requirements that must be satisfied.  In order to fill out heats, extend tundish  length or avoid grade transitions, orders are sometimes “over-graded,” that is,  a more expensive chemistry is used to satisfy an order than is actually  required by the customer.  Thus minimization  of over-grading reduces alloy costs.
For integrated steelmaking plants—those with  blast furnace operations—there is a sixth goal: control of liquid iron  inventory.  In every such plant, the  fleet of torpedo cars that transport liquid iron from the blast furnace to  steelmaking (see Figure 7) is finite: there are only a certain number of cars  available at any given time.  Ironmaking  is the input to this liquid iron inventory, and steelmaking consumes it.  Steelmaking production volumes are driven by  caster schedules—wide schedules consume more iron, narrow schedules consume  less.  Since blast furnace production is  difficult to modulate rapidly, caster schedules must be continuously adjusted  to control the amount of liquid iron “on wheels.”  This helps avoid costly or environmentally  irresponsible methods of disposing of excess iron.
For those steel plants that are  capable of significant amounts of hot charge rolling, two additional goals help  define the quality of integrated casting and rolling schedules: material charge  temperature into the reheat furnace, and hot mill asset utilization.  Retaining as much heat as possible in the  semi-finished product reduces fuel costs per ton, and increases throughput for  furnace limited mills.  Hot mill  utilization improves when rolling campaigns are planned at the maximum length  for each campaign type, minimizing the number of required roll changes.
All of these goals exist in a complex tradeoff  relationship (see Figure 2).