Experiences in Ductile Iron Production
Greenfield Start-up Cupola vs. Coreless Electric Melting
Ductile Iron Society 1998 Annual Meeting, June 17-19, 1998
By Joel W. Yates, Briggs & Stratton Corporation

Abstract
This paper will be discussing two different topics that are similar in context. Both are based on the start-up of an electric melting ductile iron facility. The first topic will discuss the greenfield start-up and the second will compare this foundry to another foundry in the same company. Keep in mind that both foundries are identical in every way except for the melting process. This is very important to some of the conclusions being made. Also, the paper is based on metallurgical aspects of a start-up situation.
 
1. Greenfield Start-up of a Ductile Iron Foundry
The start-up of a greenfield operation involves many different areas of normal foundry practices. Well over 95% of the effort is placed on the physical construction of the facility. The rest involves the training of new people and designing the overall process of melting, molding and finishing the castings. The topic of this paper is to focus on the metallurgical aspects of starting a new facility.
 
Melting
The foundry being discussed does all melting using three (3) line frequency, 20 metric ton coreless electric furnaces. These offer approximately 21 tons per hour of melting capacity. All charge materials are pre-weighed and preheated by natural gas direct flame. This is done due to the nature of the furnaces being used. The melting practice used is what is referred to as "heel melting." Of the 40,000 pounds of metal in the furnace, 4500 pounds is tapped and 4500 charged for melting. The entire process of tapping, charging, sampling and melting takes approximately 20 minutes.
 
There are only three charge materials being used, pig iron, fragmentized black steel scrap and returns from operations. Cleanliness and size are very critical due to slag amounts and density respectively. Since the furnaces do not have a back-tilt option for slagging, all slag must be removed manually using steel "spoons." Therefore, the amount of generated slag becomes an important factor. Also, this is why as much sand is removed from returns as possible using a media drum. The size of the charge materials are monitored closely for density so that the entire charge of 4500 pounds will fit easily into the charging bucket following preheating operations.
 
There are just two alloying materials used in the melting operation . The first is graphite, which serves as both a carbon additive and as a nucleator of the base iron. The second is silicon carbide. This serves both as a silicon additive and also as a deoxidizer of the base iron. Both of these alloys are added to the furnace after tapping and before charging. In a heel melting operation, the addition of these materials before charging allows for a larger recovery rate of the alloys due to its submersion in the bath by the charge. The amounts of each alloy needed in each heat is dependent on the carbon and silicon thermal analysis results obtained from the furnace on the previous heat.
 
In a start-up situation, it will need to be determined how much iron should be tapped. This will depend on the most even combination between the needs of the molding department and the capabilities/efficiencies, of the melting department. This will be different in every facility. Also, it is best to start with the highest quality of charge materials and alloys possible. This is one variable of quality that you do not want to have to worry about when operations are starting. After a benchmark is set, experimentation into other materials can safely be done.
 
Treatment Method
The equipment chosen for ductile treatment are "modified" tundish ladles. Basically, they are 5000 lb. capacity double-spouted teapot ladles with an alloy pocket wall designed in. The purpose of this ladle is to prevent the initial molten iron from contacting the alloy until enough iron has entered the ladle as to maximize magnesium recoveries. Since magnesium is mostly lost due to oxidation, it is in our best interest to submerge the alloy under the molten bath for as long as possible.
 
Since different grades of iron will be produced, it was necessary to devise a way of making copper and ferro-manganese additions. What we came up with was a semi-computerized way to make such additions. Since there are two molding lines, it was necessary to be able to alloy for both. At the initial treatment stage, the ladle operator will alloy up to the lower alloy grade of the two lines. The metal delivered to the molding lines, therefore, is ready for one of the lines, but needs to be alloyed up for the second grade. The computers are used by showing base iron chemistry data to the ladle operator and letting them determine the pounds of alloy needed. The final chemistry targets for each job being run are shown to the operator and a conversion chart is also provided. An example of the computer screen the operator sees and the conversion chart are provided in figure 1 and figure 2.
 
Click here for figure 1
 
Click here for figure 2
 
As you can see, by taking the base iron copper and manganese levels and the corresponding targets from the screen, the operator can determine the alloy needed, in pounds, on the conversion chart on the right. This chart is used at the treatment vessel where 4500 lbs. is being treated. A similar chart is used at the molding lines and is based on either 2250 lbs., half a full heat, or 1500 lbs., a third of a heat. Either may be necessary depending on the iron needs of each line. Also, the same type of conversion chart is used for ferro-manganese additions.
 
This method of adding alloy was extremely helpful in the start-up due to the fact that the work force was completely green. In other words, they had no foundry experience at all. The next step in this process would be to have the computer automatically figure out the pounds needed on its own. All the operator would have to do would be to tell the computer which furnace they were going to tap from. This can be very important, especially on Monday morning after a furnace reline when the chemistries in each furnace are different.
 
Molding
The pouring method used is an unheated "bottom pour" ladle. The stopper rod is operated by an automatic laser guided pouring system. Alloys used for inoculation include 75% ferro-silicon and some instream inoculant. The ferro-silicon is added to the ladle upon receipt of iron and the instream is blown into the stream as the molds are being poured.
 
Determining the frequency of nodularity and chemistry sampling will be important when considering the type of parts being produced in the future. It may be a good idea to go ahead and sample every heat from all molding lines. This will be a good practice to be in use, even though it can be costly, especially if you plan on making "safety critical" parts in the future.
 
Historically, nodularity was determined using a rating chart that had examples of different percentage levels of nodularity, pearlite and carbides. A good example of this is the "Ductile Iron Microstructures Rating Chart" that AFS publishes. This can be seen in just about every ductile iron foundry in the Midwest. When going through a greenfield start-up, this can pose a rather large problem. Since there is no experience in reading nodularity to draw from and the fact that everyone reads a visual chart differently, it is to one’s advantage to look into imaging software. This software will allow for a totally subjective recording of nodularity, allowing no room for interpretation of results. This will be helpful when going through a start-up and experiencing the "growing pains" of learning the process’ capabilities. Figure 3 and figure 4 are examples of such software and what it can do.
 
% Nodularity Testing
As you can see, it is able to measure nodularity and pearlite by measuring contrast in light. Carbide percentage is determined using the same method.
 
The data obtained from this analysis can then be used to predict how long it takes before the iron has faded to the point of being unusable. Using this type of unpressurized ladle, the magnesium fades fairly quickly depending on time, temperature and surface contact area of the iron with the ambient air. To do a prediction such as this, and it is highly recommended that a new facility does, it will be necessary to compile historical time study data comparing the nodularity at different time intervals. This data can then be analyzed through a regression software to predict, historically, what the length of time it will take before the iron is at or below 80% nodularity. An example of such a regression is shown in figure 5 and figure 6 (below) along with the data used.
 
Even though, by examining the data, you know the iron can go longer than the predicted time, on average it cannot. This is possibly a result of how fast the ladle was emptied. The longer the ladle can stay full, the slower the iron will "fade." This is a result of the amount of surface area in contact with the air.
 
Laboratory
When setting up the laboratory, make sure the testing equipment is user friendly. It will be probable that the people hired will not have any experience running a tensile machine or spectrometer and will not have an in-depth knowledge of personal computers. Research this carefully before expending large amounts of resources.
 
It is a good idea to monitor results of mechanical properties testing using some form of statistical software. A new facility may see a lot of interest from its customers regarding their capabilities in meeting physical property specifications. Remember, the customer has never purchased castings from you before and will be cautious at first. Following on the next page are some examples of how these physical properties can be monitored.
 
The "R Squared" value represents the correlation of the regression between the two results being compared. The closer this number is to 1, the more consistent the equipment and the iron is. Also, it allows the customer to be able to predict the physical properties just as we did for the nodularity predicting mentioned in figure 6.
 
Figure 6