The AFS cupola process model: a computer tool for foundries.

After 10 years of research and development, a model has been developed to integrate numerous variables and guide cupola operations toward optimal performance.

Cupolas produce 70% of the world's iron for castings by melting iron and steel scrap and using energy developed from the combustion of coke. Cupola energy costs are low, and they utilize wider ranges of scrap than other melting processes. As a result, the cupola is generally the lowest cost producer of molten iron, particularly when iron requirements exceed 10 tons/hr.

In addition to cost benefits, the environment inside the cupola is oxidizing - making it less susceptible to tramp element contamination - and its elaborate emission systems permit melting of zinc-coated scrap.

The cupola process has a down side as well. The oxidizing environment in the cupola leads to a significant loss of alloy elements, and the wide range of charge materials being melted makes composition control more difficult. In addition, the relative simplicity of cupola construction belies the complexity of the processes going on inside. Improvements in these areas would make the cupola even more cost-effective and efficient than it is today.

A Cupola Operation Solution

It was envisioned that problem areas in cupola operation could be significantly reduced if a computerized process model were developed that integrated the complex interactions to predict cupola output based on a knowledge of the inputs. To this end, the development of the AFS Cupola Process Model was undertaken in 1989, jointly sponsored with the Dept. of Energy and 18 corporations.

The work was carried out by the Univ. of Michigan, Massachusetts Institute of Technology, Academy of Sciences of the Czech Republic and General Motors. In addition, numerous individuals and organizations contributed to the development. Version 1 of the model will be completed soon and introduced for sale.

After more than 9 years of development, this PC-based computer program helps foundries improve cupola productivity and operating costs. The program operator inputs the properties of the cupola, blast and charge, and the output enables foundries to explore and improve current operations. For example, the model can predict the effects of charging and operating conditions on melt rate, cost, chemistry and temperature of the final iron - a useful tool for "what if" studies.

The model also is a powerful tool for evaluating the economics associated with cupola melting as it considers the effects that operating parameters such as alloy loss, carbon (C) pickup, combustion efficiency and thermal loss will have on the cost of producing hot metal. In terms of long-term applications, the model can assist in developing the cost and operational benefits of major cupola modifications, such as adding a refractory lining or increasing the blast temperature capability.

The cupola model is a collection of mathematical equations that describes the important chemical and heat-transfer processes in the cupola. The program divides the cupola into more than 1000 levels, stretching from charge door to the iron dam, and computes the conditions that exist at each level. As a result, the model not only provides output data but also provides the opportunity to visualize what happens inside the cupola.

Figure 1 illustrates how the model allows users to visualize the cupola's internal conditions. Plotted are the temperatures of the solid charge materials, liquid iron and blast, extending from the charge door (zero level) to the tuyere region (22-ft level). The model provides the long-awaited answers to such universal questions as "I wonder where my coke bed is?" or "How wide is the melt zone?" Many other conditions can be examined, such as pinpointing regions where energy generation or alloy loss occur.

The plotting routine also can combine data from different runs on the same graph, helping the operator to understand how changes in the inputs produced the observed results. The model developers considered plotting as a key tool for the improvement of cupola operation as it enables operators to develop a keener sense of the connection between their actions and the internal working of the cupola. An additional feature of the model is a routine that enables comparisons of input and output of several model runs in tabular form.

The utility of the model stems from its ability to integrate the effects of many variables. A list of the key cupola variables required by the model is given in Table 1. To facilitate customizing the model for a particular cupola, data on charge materials and cupola design are stored by the model for easy retrieval. Default values also are provided in many cases. Although the computer code is highly complex, the dialogue between the operator and the computer is simplified by convenient menu windows. Computation time on a PC is generally less than 1 min.

Model Accuracy

The cupola model has been applied to a wide range of cupola operating conditions with good results. The model accurately predicts changes in output, going from one input condition to another. The quality of prediction of absolute values of the outputs is constantly being upgraded. The current level is illustrated in Table 2 for four diverse cupola operations. The range of conditions represented by the data is: hot blast temperature - 225-1200F (107-647C); percentage steel - 0-70%; and percentage coke - 9-12%. Both refractory-lined and unlined cupolas are represented. In general, agreement between the model and measured [TABULAR DATA FOR TABLE 1 OMITTED] [TABULAR DATA FOR TABLE 2 OMITTED] data is better than 10% for melt rate and 0.25 absolute percentage for C and silicon (Si). As seen in the last column, iron temperature is difficult to predict accurately, however, trends are accurately portrayed.

Customized Control Charts

Foundries frequently change input conditions (blast rate, iron-to-coke ratio, etc.) to produce desired changes in the output iron. Using data from the model, a foundryman can create customized charts, detailing how iron temperature, chemistry and melt rate are affected by these key input variables. The charts can remove the guesswork out of making the changes needed for effective operation.

An illustration of a customized control chart is given in Fig. 2. It shows how changes in blast rate and iron-to-coke ratio will affect iron temperature and melt rate (referred to as a net diagram). The data points represent nine individual model runs. Many more runs could have been made to provide a much finer grid needed for practical cupola operation. To avoid confusion, the iron compositions provided by the model for each data point were omitted. However, the chart indicated significant trends for both C pickup and Si recovery.

Most cupolas today enrich the blast with oxygen ([O.sub.2]). To capture the effects of changing [O.sub.2] enrichment levels, the calculations for Fig. 2a (which were generated assuming 1% [O.sub.2] enrichment) were repeated at different [O.sub.2] levels. From this data, a series of charts were generated that covered changes in blast rate, iron-to-coke ratio and [O.sub.2] enrichment. An overlay of the charts covering [O.sub.2] enrichment levels of 0-4% is given in Fig. 2b.

Economic Analyses

Every cupola operator knows that scrap quality affects alloy loss and C pickup. But even an experienced foundryman will find it difficult to accurately determine how a new charge material will perform. In this respect, the model offers assistance as it can compute the effects of scrap thickness and composition on alloy yield, ?? pickup and iron temperature. This type of information makes it easier for foundries to make informed scrap buying decisions. For example, you could evaluate whether low-cost scrap is worth the additional alloy loss it may cause.

Figure 3 illustrates how the model provides key purchasing information. Plotted are the effects of steel scrap thickness (36% of the charge weight) on the temperature and C and Si content of the iron. In this particular case, it is seen that steel thicknesses of 0.17-0.33 in. produce the optimum temperature and alloy recovery results - that very thin and very thick steel significantly reduce iron temperatures, and thin steel can seriously reduce C pickup and increase Si loss. Finally, through multiple runs, model computations can be used to figure the needed additions of coke, oxygen and ferroalloy to compensate for the negative effects of non-ideal scrap.

Cost/Benefit Considerations

The ability of the model to project beyond current conditions makes it a useful tool in decision making for major cupola improvements. This is illustrated in Table 3, which examines the cost savings that could be expected from an increase in blast temperature capability from 850 (447C) to 1200F (667C). Using a customized control chart for a 1200F hot blast temperature, two conditions were found where iron temperature, melt rate and final percent C were the same as for the base case. In the first case, the [O.sub.2] enrichment was eliminated. In the second case, both oxygen and coke requirements were lowered below those of the base case. Applying nominal costs to oxygen and coke of $0.40/ccf and $175/ton, respectively, the anticipated savings were calculated. Weighing these savings against the capital, installation and upkeep costs of a higher temperature blast system enables the foundry to reach a rational decision as to the cost/benefits of the improvement.

Problem Solving 

A good example of problem solving is the model's ability to deal with the effects of humidity in blast air. It is well recognized that humidity has a negative impact on cupola operation, particularly on metal temperature and C pickup. The model was challenged to determine the levels of [O.sub.2] needed to counteract the effects of rising humidity.

To accomplish this, the impact of humidity on metal temperature at different [O.sub.2] enrichment levels was determined using the model. This data is plotted in Fig. 4 as a function of the relative humidity of 80F (27C) ambient [TABULAR DATA FOR TABLE 3 OMITTED] air. The relative humidity scale is 0-200%, with the highest value used to represent fog. A similar plot (not shown) of the relationship between the final C content of iron and relative humidity also was produced. From the slopes of the lines in both plots, the ability of [O.sub.2] to counteract the effects of humidity was determined. For example, an addition of 2% [O.sub.2] was required to counteract the effects of a sudden change in relative humidity from 50% to 100%. Data also was developed to compensate the effects of humidity over a range of ambient temperatures.

The Future

Extension and improvement of the AFS Cupola Process Model will be an ongoing process with direction largely dictated by the users. In the immediate future, efforts will concentrate on improving model accuracy and the addition of features that will expand the software's capabilities. For instance, the first release will have limited ability to predict the effects of SiC usage and will not output a slag composition. This type of functionality already is being worked on for inclusion in the next generation of the software.

Considerable development has already been carried out to have the model serve as a teacher to a type of artificial intelligence called neural networks. Once the neural network is "trained," it can provide data for "what if" economic and operational optimizations nearly instantaneously. Also, work is underway to utilize the model predictions as part of a cupola control system that will integrate sensor input to act as an online advisor to foundrymen making real-time decisions in cupola melt shops.

This article was adapted from a presentation at the 1998 AFS 2nd International Cupola Conference. Conference proceedings are available from AFS Publications at 800/537-4237.