A combination of design of experiments and genetic algorithms should be used to investigate the impact of possible variations of the ingot body, its head and insulation. Also, the layout of the steel die was open to modification, to support optimal robustness of the casting process and meet the new quality demands.
First, the engineers from TŽ analyzed the current production process to identify the root causes for the observed macroshrinkage. One of the ingots was sectioned and the porosity was analyzed. Subsequently, the steel die was modified to position five thermocouples along the height of one of the four ingots, allowing for precise temperature measurements during the casting process. These temperature recordings were then used to obtain optimal height-dependent heat transfer coefficients for MAGMASOFT®, matching the measured and calculated temperature curves.
Based on the good agreement between the simulation and reality, the next step was to set up an autonomous optimization following multiple objectives. To identify the best set of process parameters and geometrical features, two objectives were defined:
1.) minimum shrinkage porosity,
2.) minimum centerline microporosity.
Since a casting free of macroscopic shrinkage was required according to the new specifications, the allowed porosity in each of the evaluation areas of the ingot body was set to 0 %. In total, 11 independent casting and geometric design variables were used in the optimization procedure, covering the flow profile during teeming, the dimensions of the ingot, its head, the insulating sleeve as well as the steel die and the pouring temperature. MAGMASOFT® was then used to autonomously generate a first set of designs and eight following generations. The created and tested variations were comprehensive, and they covered options not exploited by the engineers by practical trials in the past.
The engineers evaluated the analyzed designs in the assessment perspective. According to the ranking, the top five designs led to similar values regarding the objectives, offering the engineers the chance to select the best design out of those five in terms of costs of production.
The evaluation of the main effect diagrams revealed a strong correlation between the height of the ingot head with the centerline and macroporosity.
Below a critical height, considerable macroscopic shrinkage porosity and centerline shrinkage form in the ingot body. These designs were unacceptable for TŽ and therefore excluded from further assessment.