In search of a cleal steel heat.
For the past four years, the Steel Founders' Society of America and the U.S. Dept. of Energy have sponsored a Clean Cast Steel Technology Program at the Univ. of Alabama-Birmingham. The program's objective is to improve casting product quality by removing or minimizing surface oxide macroinclusions that will allow the production of higher integrity castings that require less weld repair time.
Over the course of the program, several in-plant designed experiments defined and evaluated practical methods for improving steel cleanliness. Two designed experiments examined the effect of bottom pour nozzle configuration and filtration on steel casting quality. In both studies, heat-to-heat variations were found to be important in determining the overall casting quality.
Heat-to-heat variations, and their causes, have troubled steel foundrymen for some time. Therefore, experiments were conducted at four foundries to determine if a relationship exists between the variables in the melting and pouring practices on average casting quality. In describing the experiments, this article examines the influence of melting and pouring practices on casting quality and, subsequently, how to reduce heat-to-heat variations.
At least 30 heats of steel were produced with data being collected on melting conditions, casting conditions and casting cleanliness. These results were statistically analyzed and discussed with foundry personnel to identify variables that significantly affected heat quality. Confirmation heats were then poured at the foundries with variables at the optimum levels to verify improvements in heat quality.
Each foundry had to meet several requirements in order to participate in the trials. First, it had to provide typical defects for examination. Figure 1 illustrates a typical cope surface macroinclusion from a production casting. This oxide was a complex mixture of alumina, silica and manganese oxide that was produced from a reaction of the oxidizable elements in steel with oxygen.
[Figure 1 ILLUSTRATION OMITTED]
Second, the foundry had to have a casting quality problem that was running at a reasonable production rate. Also, casting production documentation of at least 30 heats was required so that enough data existed for a good statistical analysis.
If a large number of castings were poured from a single heat, at least five castings from the heat were examined to provide a basis for evaluating within heat variations. If fewer than five castings were poured from a heat, the casting(s) should represent a significant portion of the heat. Each foundry determined the method of measuring casting quality.
Participating foundries were asked to supply as much information as possible along with any other documentation that might be available, such as heat sheets. In most cases, a significant amount of data was already available. In general, the foundries provided only information already existing in the foundry.
The largest workload was in collecting the data, keeping track of the castings through the foundry and rating the castings for quality. The slag from the furnace and the pouring ladle were sampled along with the metal for further analysis.
After data was collected, the variables from each heat were keyed into a statistical analysis program. Normality tests were conducted on the quality data to ensure that the statistical analysis techniques being used were valid.
The data was statistically analyzed using two methods. First, the variables were analyzed individually using a simple regression analysis and then ranked according to the highest [R.sup.2] the highest statistical significance. [R.sup.2] is the amount of variation in the dependent variable (in this case, metal quality) that could be explained by a variable or a set of variables.
Second, a Rest was used to detect statistically significant differences of the variables between the cleanest and dirtiest heats. The variables from the cleanest heats were compared to the variables from the dirtiest heats and tested to see if a statistical difference existed. Although the order changed slightly on occasion, both methods usually picked the same variables.
The quality ratings were expressed in different terms at the foundries because the castings produced and methods used to rate heat quality were different. The quality ratings were standardized using a z-score to account for these differences. The z-score is the heat quality rating minus the average for the population divided by the standard deviation. Basically, it is the number of standard deviations that the particular score is from the arithmetic mean of the population.
There were usually 10-12 variables correlating to heat quality. The list of variables was then reviewed with foundry personnel. Usually, the list was narrowed to 4-5 variables that strongly affected heat quality. These variables were usually associated with the melting or pouring practice, and reasonable engineering solutions are generally available for improving the heat quality. A subsequent trial was conducted at the participating foundries to verify the influence of the variables selected.
One acid practice foundry and three basic practice foundries participated in the trials.
Acid Foundry 1 melted 1025 carbon steel in a 20-ton, acid electric arc furnace. Mold pour weight was about 2250 lb with a casting weight of 1711 lb. Four to five castings were poured from each heat. Data for 35 heats was collected for the initial analysis.
Basic Foundry A melted 1025 carbon steel in a 10-ton, basic electric arc furnace. A total of 25 heats were produced and analyzed.
Basic Foundry B melted 1025 carbon steel in a 25-ton electric arc furnace. Mold pour weight was about 7200 lb with a casting weight of about 4000 lb. Only one mold out of a total of 3 to 6 for the heat was included in the trial. Data from 30 heats was collected for the initial analysis.
Basic Foundry C melted high manganese carbon steel in a 10-ton, basic electric arc furnace. The molds were poured from a 10-ton, bottom-pour ladle. The molds had a pour weight of about 300 lb and a net weight of 150 lb. A total of 27 heats were used in the initial analysis.
Acid Foundry 1
Over the span of 12 months, 35 heats were collected. About 150 variables were recorded or calculated for each heat and analyzed. The foundry selected six variables from a list of 34 that had a reasonable probability of influencing the quality of the heat.
An additional 24 heats were poured with the six variables set at values expected to produce clean heats. The average heat quality rating for the 24 heats improved from about 32.6 in. of surface inclusions per casting to an average of 18.1 in.--a decrease of about 44%.
Subsequent foundry analyses indicated that three variables had the greatest influence on casting quality: metal head height, pouring temperature and silicon concentration after the oxygen blow.
Basic Foundry A
About 80 variables were recorded or calculated for each heat and analyzed. Initial data analysis indicated that 10 variables influenced heat quality. From this list, six of the variables involved carbon and manganese concentrations in the metal.
After discussions with the foundry, an L4 designed experiment was conducted to evaluate the effect of meltdown manganese and carbon concentrations on heat quality. Eight heats were poured in the L4 with two heats at each condition (low carbon/low manganese; low carbon/high manganese; high carbon/low manganese; and high carbon/high manganese).
Subsequent analysis revealed that meltdown carbon didn't significantly affect heat quality. The manganese concentration at meltdown was, however, important with low meltdown manganese producing cleaner heats. The heat quality improved from an average of about 4 to less than 2--for an improvement of roughly 50%.
Analysis of the data also revealed a positive correlation (48%) between the concentration of manganese and carbon at meltdown. In effect, the manganese and carbon concentrations follow each other. There was also a positive correlation (75%) between meltdown carbon concentration and oxygen blow time. The heats with lower meltdown manganese concentrations had less oxygen blown into the metal.
More manganese was available for reaction with oxygen in the clean heats. This indicated that less oxygen was available to react with the manganese.
Over two months, 31 heats were poured and evaluated. About 140 variables were recorded or calculated for each heat. Initial review of the data showed that one heat had an inclusion surface area four times higher than the next largest value. The data also failed all tests for normality. This heat was removed from the analysis. The normality for the remaining heats was adequate to proceed with further analysis.
Initial analysis indicated that four variables strongly influenced heat quality. In both tests, the manganese recovered from the slag after tapping had the strongest correlation to casting quality, followed by the angle of the oxygen lance, number of heats on the ladle lining and tap rate.
The initial results were discussed with foundry personnel to determine the next course of action. The manganese recovery indicated that the after-tap slag was important. The amount of manganese recovered from the slag is influenced by the volume of slag tapped on the metal and the concentration of manganese oxide in the slag.
The higher tap rate for improved metal quality was due to reduced exposure of the metal to atmospheric oxygen. The number of heats on the ladle lining indicated that the lining must be baked out to remove all oxygen sources. The lance angle's affect on metal quality was unusual. Quantifying the angle of the lance is subjective, since the operator is moving the lance to remove slag as well as to boil off carbon. But lance angle was a strong indicator in both analyses.
A second set of castings was poured with the four variables set at their optimum values to demonstrate that these variables do, in fact, influence metal and casting quality. The tap rate was increased as much as possible. Oxygen lance angle was kept to about 45 [degrees]. The number of heats on a ladle lining wasn't a practical control variable.
Seven additional heats were poured under these conditions. The average area of cope surface inclusions decreased by about 51%, from 75.4 to 37.2 sq in. The variation in casting quality decreased by 76%, from 53.8 to 13 sq in. Additionally, the foundry is now conducting a designed experiment to determine which variable is the most important and which variable could be used to increase or decrease metal and casting quality.
Basic Foundry C
Initial analysis of the foundry data indicated that four variables influenced heat quality. The amount of aluminum added to the ladle strongly affected metal quality. The variations in the amount of aluminum added, however, were small so the effect might not have been as strong as the analysis indicated.
Slag samples were taken from the furnace after block, from the ladle in the pit after deoxidation, and from the ladle at the end of the heat. The color of the slag was rated from 1 to 5 with 1 being light gray and 5 being black. The numbers for the end of heat slag ranged from 1 to 3.5 with only a few ladle slags at the higher end of the number scale. There was a strong correlation between slag color and metal quality.
Charge material and final temperature appeared to playa role in metal quality. The final temperature varied about 100F over the range of castings poured.
The variables indicated by the analysis provided a starting point for discussions with the foundry to identify factors that are logical in controlling metal quality. Once a reasonable list of factors has been identified, subsequent trials will be conducted at the foundry to verify the effect of the factors on metal quality.
Three of the four foundries showed significant improvement in casting quality by manipulating the melting practice. Improvements ranged from 40-50%. The fourth foundry has not yet completed the confirmation experiments.
The important variables from these trials can be grouped into two separate categories.
The first category is the pouring and mold filling practice. Metallostatic head pressure and tapping rate were important in three of the four foundries. Lower head pressures reduced the velocity of the metal exiting the nozzle and throttling of the ladle. Head height was not a variable in Basic Foundry A because the entire heat was used to pour the selected casting.
Figure 2 illustrates the effect of head height on heat casting quality at Basic Foundry A. As the mold number increased, the head height decreased. As shown, the first 2-3 castings poured from the heat are significantly dirtier than the last 2-3 castings poured (fewer in. sq of surface inclusions). High metal velocities and throttling were shown to increase air entrapment and reoxidation of the metal.
[Figure 2 ILLUSTRATION OMITTED]
The second category concerns the concentration of oxidizable elements contained in the steel. Silicon, manganese and aluminum concentrations were important factors in all four foundries. These elements can be indicators of the oxygen concentration contained in the steel. Carbon is typically used as an indicator of oxygen concentration in the steel bath. Research has shown, however, that other oxidizable elements also can influence the oxygen content of the steel and the availability of the oxygen to form reoxidation inclusions in steel castings.
By using the oxidizable elements as indicators and manipulating the melting practice to optimize the concentration of these elements, improvements in casting quality can be observed.
Clean heats can consistently be produced if the oxygen concentration in the steel bath can be minimized through improved melting practice and reducing exposure of the steel to atmospheric oxygen during pouring and filling.
This article was adapted from a paper presented at the 1997 Steel Founders' Society of America T&O Conference.
|Printer friendly Cite/link Email Feedback|
|Date:||Feb 1, 1998|
|Previous Article:||On the inside track of the custom rim industry with Ultra Wheel.|
|Next Article:||Gain a competitive edge with TQM.|