The system was developed in C++ and it uses memory resident architecture which means that the solution doesn’t use hard drive during calculations and thus it has high performance. The system uses for modeling constraint satisfaction programming algorithms including linear and mixed integer programming, constraint-based search, heuristics and other OR algorithms (Taha H., 2005).
The key results of the implementation of the developed system are: significant rise in speed of calculation (it takes minutes not hours now); improvement of schedule quality thanks to detail modeling of steel making rules, application of contemporary optimization algorithm and friendly user interface for manual changes of a schedule.
Results and discussion
The process of melt shop and caster scheduling was completely renewed at Trinecke Zelezarny thanks to implementation of the new solution. Today casting requirements are generated automatically based on a valid calendar production plan. This allows a planner to work on the complete calendar production plan horizon – up to 4 months. Currently melt shop and caster schedules are created for 40 days. The first 14 days of the horizon are planned in detail. The schedule is updated regularly based on real situation in production. The visibility of schedules and wide planning horizon significantly improved economic and technological reasonableness of planners’ decisions.
The result of new process implementation was efficiency improvement not only in melting and casting but also at the company in general. One of the key performance indicators of steelmaking is the level of wasted tundish life at continuous casting. An improvement of this indicator enables a decrease in the costs of production, with simultaneous increase of the total performance of the steel plant. The achievement of a good level for this indicator is especially difficult due to the following factors: necessity to produce a large variety of steel grades, taking into account special requirements of individual customer orders; high share of small orders in proportion to one heat, let alone to the minimal size of a tundish lot. The level of wasted tundish life at continuous casting is evaluated as the difference between the theoretical maximum and the planned sequence length for individual steel grades, where the theoretical sequence length derives from the maximum length of sequence given by the technology of production and the planned sequence length derives from the casting schedule at the continuous casters. Statistics of average tundish continuous casting is provided on Figure 6.