The process chains for metal production (e.g. steel, aluminium) are complex and involve numerous metallurgical and physical sub-processes. The aim is to produce products fulfilling several quality targets in a reliable and reproducible manner under minimised energy and resource effort. To achieve this target, a thorough knowledge of the process behaviour is required, which should be reflected within process and sensor data analysis tools and predictive process models, to be used within on-line monitoring, control and decision support systems.
Within the SPIRE project “Optimization and performance improving in metal industry by digital technologies” (INEVITABLE) such on-line monitoring, control and decision support systems have been developed and applied in various metal making processes, as primary and secondary steelmaking, cold rolling and aluminium investment casting. Models and solutions based on them were developed to provide several cognitive solutions for:
Process & Equipment monitoring, where digital twins for parallel simulation and soft virtual sensors to estimate unmeasured process behaviour were developed.
Decision Support Systems to support process operation optimization and enable better final product quality prediction.
The solutions adopted different approaches to develop digital entities of specific segments of the metal production chain:
First principle models of a process segments with well-known physics (use of algebraic and differential equations)
Statistical & AI-based modelling (e.g. image and frequency analysis, feature extraction, cause-effect relationship analysis, data-based models like multi-variate regression models, ANN, Neuro-Fuzzy-based models, Gaussian processes, Bayesian networks) for modelling the process behaviour for known operating conditions and for improving the 1st principles models
INEVITABLE project provides several best practices of cognitive digital based solutions where modelling approaches were applied within the process chain of metal production.