Abstract
Not only the strength of steel but also the variance of strength depends on composition and thermomechanical treatments. Information about the variance of strength is important when composition and thermomechanical treatments are planned to fulfill the strength requirements. In this work, heteroscedastic linear models were used for joint modeling of the mean and dispersion of strength. The models were used as tools in the optimization of the strength margins for steel plate products. The models were created by using industrial process data, and they were successfully used in a steel plate mill.
Keywords:
- Conditional distribution function
- Expected squared prediction error
- Heteroscedastic regression
- Joint modeling of mean and dispersion
- Large industrial data set
- Planning of process settings
- Probability prediction
- Standardized residual
- Steel plate mill
- Strength margin
- Strength of steel
- Tensile test
- Uncertainty about input variables
- Variance modeling
- Variance of strength
ACKNOWLEDGMENTS
The authors are grateful to Ruuki Steel Plate Mill for the data, advice, and funding that made this research possible.