Raw materials purchase & mix of steel grades portfolio optimization

Back to list


A steelmaking plant in Brazil with annual production of 5 million rolled products


Develop an application to optimize the purchase of raw materials and the mix of steel grades.



Collaboration with N-SIDE began during the crisis period in 2008. At the time, reducing production cost had become the steel industry’s main concern.

The main optimization lever in this direction was raw materials purchasing, which constitutes 75% of total steel production costs. However, minimizing production cost alone soon became insufficient. Questions of profit optimization and then of determining an optimal production level rapidly came to the fore.

The order book of a steel-making plant is filled with several thousand rolled products. Therefore, determining optimal production level must necessarily pass through an optimization of the order book, highlighting the most profitable products.

Optimal production level and an ideal mix of raw materials are strongly linked with:

  • process constraints specific to the plant;
  • market context for raw materials (purchasing price and availability);
  • market context for end products (selling price and market demand).

Existing models were only able to optimize profit on a plant by plant basis. Such a solution might be very different from the optimal solution for an integrated process (coke plant, sinter plant, blast furnace, steel-making shop, cold rolling mill and power plant). The steelmaker therefore needed a tool able to optimize the global profit of their plants and respond rapidly to changes in market context.


N-SIDE customized its tool SCOOP to match the client’s processes. The core of the model is a non-linear optimization model, which includes thousands of variables and constraints. The objective is to maximize the profit of the integrated plant.

The application has been developed in two phases:

  • In 2009, the first part of the project entailed calibrating the basic version of SCOOP with some customization. At that time, the model was only able to minimize production cost, as total production was part of the model’s fixed input.
  • In 2010, an updated version was deployed with variable production level. In order to provide feasible solutions, a full productivity model was developed and calibrated. The steel grades and their end products were gathered respectively into two families, one of about 30 and the other of 200 products, to maintain production while shrinking optimization time.

Since releasing the application and training its users, N-SIDE has continued collaborating with the client as part of our business support contract. This involves help for simulations and development of small custom enhancements for the model. This collaboration enables a time saving of up to 80% when defining new scenarios.


The following benefits were quickly highlighted by production managers:

  1. Fast reaction to market context: The plant was able to react faster during the crisis period of 2009. Re-firing of blast furnaces was scheduled wisely, taking into account the simulations performed with SCOOP, and they were the first to come back on the market once conditions improved.
  2. Optimal purchase of raw materials: Like many steel companies in Brazil, this client has their own mine. SCOOP helps them find the best solution, between consuming their own ore or selling it outside and consuming a different one for their process, depending on current market context.
  3. Identification of bottleneck equipment: The productivity model points out bottleneck equipment. By increasing the capacity of such equipment and comparing scenarios, they are able to evaluate the ROI of a small investment project.
  4. Ranking of end products by profitability: Marginal production cost is a strong tool for evaluating the profitability of each grade according to ranking, thereby justifying the optimal grades mix provided by SCOOP.

Beside the economic benefits, SCOOP also brings the notable advantage of gathering people from different plants to discuss common problems. This kind of meeting is held each week. The small optimization time of SCOOP (1 to 2 minutes) enables live simulations during meetings.

Print Friendly
Back to list