3.1.1. Steel Plant Scrap Yard System

The steel plant uses an internal coding system for the raw material types, which follow a logic that is based on where the raw material is sourced from, its quality, composition, and shape. The majority of the raw material types represent steel scrap. Although nickel granulates and anthracite, i.e., coal, have distinct coding, these raw materials will be referred to as scrap types well knowing that they are not scrap. The main reason is that the focus of this study is on the effect of steel scrap on the EE consumption of the EAF. The effect of alloying elements is not of interest. Furthermore, nickel granulates and anthracite represent a small fraction of the total amount of charged raw materials for all steel grades.

The composition of the scrap can vary from low-to high alloying elements such as carbon, chromium, nickel, and molybdenum. Furthermore, quality-hampering impurities such as copper and tin are prevalent in some scrap types that are considered of low quality. The scrap quality is also dependent on the confidence in the actual content of the scrap. In general, more reliability is put on quality from internal sourced scrap and scrap from other steel plants than on purchased scrap from municipal waste incineration plants. The shapes of scrap can vary from thin plates to bulky residuals from foundries and own arising scrap.

14 of the 33 scrap types were charged in less than 10% of the heats represented by the data. To reduce the number of variables in the numerical experiments, a variable was created as the sum of the charged weights of these scrap types. This variable is called *SCRAggr*, see also Figure 3.

The goal of including this representation in the numerical experiments is to investigate if the most granular representation of the available scrap, in the steel plant of study, is the most optimal with respect to the accuracy of the EE consumption prediction models.

### 3.1.2. Visual Categorization

The melting rate of scrap in the EAF is predominantly dependent on the surface area available to heat transferring media such as the hot heel and radiation from the arc plasma. By categorizing the scrap types according to their distinct physical shape, two benefits can be achieved. First, scrap types with similar shapes are expected to have closely related melting performance in the EAF due to their similar surface-area-to-volume ratios. Second, the number of variables will be reduced which is beneficial from a statistical modeling point of view.

The difference between the categories Internal1 and Internal2 is based on the apparent density and the difference between the categories Plate1 and Plate2 is based on the thickness of the plates. They were separated because a difference in apparent density and plate thickness will affect the melting time of the scrap pieces.

4 of 14 visual categories were charged in less than 10% of the heats represented by the data. As with the steel plant scrap representation, a variable was created as the sum of the charged weight of these scrap types to reduce the number of variables. This variable is called *Aggregate*.

### 3.1.3. Density Categorization

All scrap types have an estimated apparent density conducted by either established standards or by the steel plants engineers. The main goal of this scrap representation is to categorize each scrap type into four categories, which are based on the estimated apparent density, i.e., density interval. All scrap types have an estimated apparent density range or a lower bound apparent density value.

The following categories were created, based on ranges of reported apparent densities:


Intuitively, scrap types with lower apparent densities should melt faster than scrap types with higher apparent densities. Since the density of the metal pieces can be assumed to be the same, the variation in apparent density is a measure of "porosity" of the scrap types. Higher "porosity" leads to more surface area available for heat transfer, which in its turn leads to higher melting rates and consequently shorted melting times.
