UniSoma – CST (Arcelor Mittal) – Supply Chain Strategic Planning Optimizer
Based on a MIP model and integrated to SAP APO (SNP Module of SAP), an AIMMS component was developed for CST´s Supply Chain Network Planning. This Optimizer will support the strategic decision process of obtaining the production mix optimization considering customer profitability as well the production and logistics capacities for CST and its subsidiary Vega do Sul.
In August of 2004 CST – Arcelor Brasil S.A. started its production capacity expansion from 5 million to 7.5 million tons of steel per year, including slabs and hot-rolled coils. The new level will be obtained through the construction of new productive units, such as a new Blast Furnace , a Coke Plant (1,5 million tons of coke per year) and a Continuous Casting Machine.
The corresponding investment of approximately US$ 1 billion aims at retaking the slabs international market – since October of 2002 part of the slabs production has been used in a Hot Strip Mill of 2 million tons per year.
In the beginning of 2006, after a consulting analysis, conducted by IBM Global Business Services, CST chose SAP – APO as its strategic planning tool. At that time, the fundamental question was how to sell the planned expansion of capacity in an optimized way and which customers (and at what level of delivery) should be selected to maximize CST´s profitability given the increase of availability in the different production areas.
CST has adopted SAP – APO due to its native integration with SAP ERP System as well its transactional functionalities like workflow and built-in middleware. However, CST also realized that some problem requirements could not be represented inside Supply Network Planning (SNP), one of the APO Modules selected for implementation. The mixture optimization in the converter and the customer list selection based on profitability could not be represented. Therefore, UniSoma was invited to build an AIMMS based optimization model/component, to be integrated with SNP, in order to accomplish the requirements of all areas involved with the supply chain processes and to replace the standard APO optimizer.
A generic MIP model was developed to support two similar applications. Both of these contain a network representation of CST´s Supply Chain, including all locations, like plants (production resources), distribution centers, customers and transport lanes connecting these locations.
The first use of the application (“Step 0”) is designed for CST’s Sales Department. It is, in fact, a Sales and Operations Planning tool (S&OP) which helps CST to select the optimal customer list, based on a profitability criteria. The model considers constraints like production capacities and customer mixture of products.
The second use of the application (“Step 1”) is designed for CST’s Production Planning Department. It supports the generation of an optimized Entrepreneurial Planning on monthly basis for several years.
The Steps are started on user demand from SNP: input data is passed from SNP to the Optimizer and output data is sent back to SNP from the model. The interfacing between SNP and AIMMS is done through SAP Exchange Infrastructure or XI, the Net Weaver tool for Process Integration. The technological integration between SAP – APO and the AIMMS-based Optimizer was defined according to the following basic schema:
As the diagram generically shows, the AIMMS-based Optimizer is utilized as an external optimization component (running on an independent optimization server). Data is transferred through a “volatile” Oracle database. “Calling Message” is a Web Service Call through which SAP – XI invokes the Optimizer. “A” is a Web Service that receives the Calling Message and calls AIMMS through a COM object (“B”). After the run has succeeded, a status notification message is sent back to XI. In fact, the solution is based on the AIMMS Agent Technology in combination with the AIMMS Web Service functionality to guarantee and allow simultaneous running of Step 0 and Step 1.
Another important project integration aspect relates to the mapping between SNP and the Optimizer data models (data model integration). The Optimizer data model was developed according to the SNP data model since SNP has a closed data model. For most of the model input parameters a direct correspondence inside SNP data model was found (i.e. a SNP parameter with the same indexes). Exceptions were treated through SAP customization resources like Z tables and planning books.
How did AIMMS add value?
The following AIMMS functionalities were essential to guarantee project development and implementation on-time:
- The AIMMS algebraic language allowed for the development of a generic and unique MIP model for Step 0 and Step 1
- Thanks to AIMMS Script Language it was possible to write a heuristic to the non-linear calculation of the yield in the converter
- The AIMMS Web Service functionality provided a fast and robust interface implementation between SAP – XI and the Optimizer
- Data integration to the Oracle database was easy through the standard ODBC connectivity in AIMMS
- The AIMMS Multi-Agent Technology made it possible to have a ready to go control of simultaneous executions of Step 0 and Step 1
- The native and advanced integration of AIMMS with the CPLEX solver made it possible to solve the large optimization problems in less than 20 minutes with a maximum gap of 3% (typical scenarios of Step 1 (a monthly planning over 1 year ahead) generate MIP problem sizes of 160,000 constraints and 235,000 variables with 4,000 binaries)
“The AIMMS based Optimizer developed by UniSoma is the ‘heart’ of the System and the most important component of the project. It optimizes demand based on customer profitability – a request which we did not find in any ‘package’ – indicating the most lucrative flat steel customers for CST and Vega do Sul”.
|Marcia Schwarz de Assis
CST´s Project Manager
Information Technology Department of CST
About Arcelor Brasil S.A. and CST
In 2005 Arcelor Brasil S.A. (an Arcelor Mittal subsidiary) was founded by the union of three Brazilian companies with a strong presence in the steel market: Companhia Siderúrgica Belgo-Mineira, Companhia Siderúrgica de Tubarão (CST) and Vega do Sul. The holding has an annual installed production capacity of 11 million tons of flat and long steel/year. It is among the largest steel companies operating in Latin America.
CST is a world leader in the steel slab market, responsible for 20% of worldwide sales volume and one of the largest companies in Brazil. Strategically located in the city of Vitória, the capital of Espírito Santos State, in eastern Brazil, CST has established itself as a supplier of high-quality semi-finished steel products (slabs and hot-rolled coils) for the automotive industry. To guarantee CST´S production capacity expansion from the current 5 million tons to 7.5 million tons of steel plates/year a new coke plant (called Sol Coqueria Tubarão) has been constructed with a capacity to produce 1.5 million tons of coke/year.