Speaker
Description
Knowing the material properties seamlessly across the full product length is a prerequisite for enhanced product safety. Primetals Technology created a customizable IT solution based on artificial intelligence and machine learning algorithms which calculates the mechanical properties of a hot rolled or annealed or galvanized steel strip just in time and almost without additional equipment.
Chemical composition and measured process parameters, which are provided from a mill’s data storage and processing systems, serve as input for purpose-built ML models which are able to calculate yield strength, tensile strength and fracture elongation over the whole strip length or at a certain position where normally test specimen are taken from.
Our solution, called Quality Guard, is designed to send back the predictions within a few seconds to the data storage and the results are available to react in time on deviations. For example, the estimated values of the models can be used to decide whether a material is well within a specified tolerance range or outside the tolerance and therefore it needs to be blocked and re-tested conventionally. The accurate predictions of this tool can also be used for automatic grading without physical testing.
In addition to this basic quality functionality we adapted our solution for product developers, process engineers and quality departments. This full featured solution, called Quality Lab, is a lab, where you can modify, like in a real lab, the relevant process parameter and the chemical analysis of your product by moving sliders or by direct numerical input. The built-in partial dependence plots provide structured visualization where graphs show how parameter variation is influencing the result. Analyzing various influences on actual product or optimizing or designing completely new grades becomes much more reliable and faster with this solution.