Controlling the production of compacted graphite iron.

Refined control technologies can now allow foundries to meet the ongoing demand for higher performance engines through CGI castings.

Although compacted graphite iron (CGI) has existed for more than 40 years, its applications have been limited to simple, low-volume components with wide microstructural tolerances such as flywheels, exhaust manifolds, pump housings and brackets. Recently, the increasing demands for improved performance and power-to-weight ratios in gas and diesel engines have forced the car and truck OEMs to evaluate CGI for the next generation of engine blocks, heads and bedplates. These applications, however, require a quantum leap in foundry process capability in CGI as the production volume is high, the geometric complexity and shrink tendencies are high, the microstructure specification is narrow, and the risk tolerance for out-of-spec castings is zero.

These production and performance requirements of automotive blocks, heads and bedplates serve as the basis for defining a product specification. For proper castability, machinability and performance, a 0-20% nodularity range is required in all performance-critical sections. More importantly, flake-type graphite must be avoided to prevent local weaknesses that could lead to failure.

For optimal machinability, carbides must be avoided and titanium (Ti) (which is used for nodular control) must not be used. The foundry needs a reliable process that minimizes metallurgical scrap rates, and the automaker requires a quality assurance (QA) certification that 100% of the castings satisfy the microstructure specification. Reliable series production requires an accurate analysis of the molten iron followed by an on-line control action to eliminate process variation before casting.

This article describes how one process control system is being used to meet CGI's stringent production requirements.

CGI Stability

The primary reason that CGI has not previously been adopted for series production of complex components is that the stable CGI range is too narrow to ensure risk-free production. Although the actual magnesium (Mg) spec is different for each product, it generally spans a range of 0.008%. In practice, however, the usable Mg range is even smaller. Because active Mg fades at a rate of 0.001% every 5 min, the initial start-cast composition of the iron must be sufficiently far away from the abrupt CGI-to-gray iron transition at the low end of the Mg range. This buffer is necessary to ensure that flake-type graphite does not form before the end of pouring, which may be as long as 15 min after the initial Mg addition. Simultaneously, the Mg starting point must not be too close to the upper limit in order to minimize the formation of nodular graphite in the faster-cooling thin sections.

A second consideration is that the Mg target composition is not constant. If the active oxygen (o) and/or sulfur (s) contents are high, they will consume active Mg and shift its range toward higher values. Conversely, if the O or S levels are relatively low, the Mg range will shift toward lower values. For these reasons, variations in the charge material and melting history make it impossible to define a fixed chemistry specification for CGI.

Although the transition from CGI-to-gray iron occurs over as little as 0.001% active Mg, the natural Mg fading does not initially result in a fully flake-type microstructure. In the absence of sufficient Mg, the graphite begins to grow with a flake morphology. As the eutectic solidification proceeds radially outward, the Mg segregates ahead of the solid/liquid interface and, depending on the initial Mg content, compacted graphite may become stable near the perimeter of the eutectic cell. The ultimate result is that flake-type graphite first appears in CGI microstructures as discrete flake patches.

The abruptness of the CGI-to-gray iron transition is illustrated in Fig. 1, which shows that a flake patch structure (ultimate tensile strength [UTS]: 300 MPa) in a 25 mm diameter test bar can be converted to a fully compacted microstructure (UTS: 450 MPa) with the addition of only 10 g of Mg in a 1-metric ton ladle.

The stable CGI range also is sensitive to the addition of inoculant. Higher inoculation levels provide more nuclei, thus favoring nodular graphite. Higher inoculation levels must be compensated with lower Mg additions while lower inoculation levels require high Mg additions to ensure stable CGI growth. Factors such as furnace superheat, holding time, charge composition and type and amount of inoculant therefore influence the Mg requirements. The sensitivity of CGI to inoculation is illustrated in Fig. 2, which shows that the addition of 750 g of inoculant to a 1-metric ton ladle can change the nodularity from 3% to 21% in a 25 mm diameter test bar.

Thermal Analysis 

The basis of any process control technology must be an accurate analysis of the solidification behavior of the molten iron because you can't control what you can't measure. The reliable control of CGI requires simultaneous measurement of the Mg behavior, the proximity of the treated iron to the abrupt gray iron transition, the Mg fading and the inoculation level. One such system for meeting those requirements is outlined here.

The 200 g thermal analysis sample is obtained by immersing SinterCast's patented sampling probe into the molten iron after the Mg and inoculant addition has been made. During the 3-sec immersion time, the walls of the sampling cup are heated to thermal equilibrium with the molten iron. The thin-wall immersion sampler ensures a constant sample volume, prevents oxidation of the iron during pour-in filling and provides an accurate measurement of undercooling because of the elimination of chill-solidification.

In order to simulate the fading of Mg that occurs naturally in the ladle and in the castings, the walls of the sample vessel are coated with a reactive material that consumes active Mg. The convection currents circulate the sampled iron along the reactive walls and cause a low-Mg iron to accumulate in the flow-separated region at the base of the probe. In the simplest sense, one thermocouple monitors the non-reacted bulk iron and thus determines the start-of-casting behavior while another thermocouple predicts the end-of-casting solidification behavior.

The reactive coating is designed such that the iron that accumulates in the flow-separated region contains 0.003% less active Mg than in the bulk iron. Therefore, if the initial Mg content of the iron is too close to the CGI-to-gray iron transition, the flow-separated area will solidify as gray iron, yielding a gray-like cooling curve from the bottom thermocouple. In contrast, if the Mg start point is sufficiently high to prevent flake graphite formation during the pouring time, the bottom thermocouple will yield a good CGI curve.

An etched cross-section of a solidified probe [ILLUSTRATION FOR FIGURE 3 OMITTED] shows the separated flow region, the bulk iron and the protective thermocouple tube. The loss of 0.003% active Mg in the separated-flow region has resulted in the formation of undercooled D-type flake graphite and, due to reduced diffusion distances, a ferritic matrix. The size of the flake-type graphite zone in the bottom of the probe is directly proportional to the initial Mg content of the bulk iron and can be calculated as the time-integrated heat release from the cooling curve produced by the bottom thermocouple.

The r-squared correlation coefficient for the relationship between the calculated heat release and the size of the flake zone is in excess of 0.90. This simulation of the after-fading end-of-pour behavior of the iron allows corrective Mg additions to be made before casting begins.

Production Considerations

Unlike gray and ductile irons, the sensitivity of CGI to Mg and inoculant additions prevents foundries from adopting the traditionally conservative philosophy of over-treatment. The sensitivity of the CGI microstructure to both Mg and inoculant means that CGI is actually stable within a four-sided window and not within a simple Mg range. Reliable CGI production therefore requires simultaneous control of the Mg and inoculant from the start until the end of casting in order to stay within the microstructure specification.

Despite all efforts and discipline, variation in the base treatment result is inevitable. Variations in parameters such as charge mix, furnace temperature, furnace holding time, ladle preheating, tapping rate, tap stream impact location (inside or outside of alloy pocket), tap weight, condition of the alloy pocket, actual Mg content in the FeSiMg alloy, layering of the alloy sandwich and amount of steel cover all affect the Mg recovery.

In addition to these variables, the active O and S content of the base iron changes the size and location of the CGI window while variations in ladle holding, transport and pouring times change the time available for fading.

Last, the most unpredictable source of variation can be operator error, or the difference in work habits between different operators. During large volume series production, the only certain way to eliminate process variation is to evaluate the solidification behavior of the iron after the base treatment has been made. In this way, all of the variables influencing alloy recovery and the size and location of the CGI window are exhausted. Corrective additions of Mg and/or inoculant can be made, if necessary, to bring each melt to the desired solidification behavior before the start of casting. A two-step measure-and-correct control strategy minimizes variation and eliminates the risk of flake-type graphite appearing in the final product.

Process Control for CGI Production

As shown in Fig. 4, the control process begins by obtaining a thermal analysis sample of the Mg and inoculant base treated. Depending on the result of the thermal analysis, the software determines the necessary amount of Mg and inoculant cored wires and the operator is prompted to start the wirefeeder. At the conclusion of wirefeeding, the ladle is transported to the molding line to begin casting. The entire on-line measure-and-correct process requires about 3 min and is conducted in parallel with deslagging and ladle transport.

When the sample volume in the probe has solidified, the cooling curves are analyzed and the results are presented as dimensionless inoculant and Mg indices. As shown in Fig. 5, these indices can be plotted on a Mg-inoculant matrix together with the CGI specification window for the product to be cast. Although the exact size and location of the window is different for each product, the production strategy is to always start casting in the upper-right corner of the specification window. When every ladle is brought to these precise coordinates before the start of casting, the fading of Mg and inoculant will result in neither flake patches nor carbides before the end of casting.

From an operational point-of-view, the base iron held in the furnace has no Mg and relatively little inoculating potency. During base treatment, the Mg and inoculant additions first consume the active O and S and the iron "jumps" upward on the chessboard. In the present example, the after-base treatment Mg and inoculant indices (dimensionless) are 65 and 45, respectively.

Because the most significant process variables manifest themselves as variations in the base treatment result and are exhausted, the thermal analysis result obtained after base treatment indicates the state of a static and stable melt. It only remains to add extra Mg and/or inoculant to move from the base treatment coordinates to the desired start-casting coordinates.

In the example of Fig. 5, an amount of Mg cored wire equivalent to 7 Mg index units is added to the melt followed by an equivalent inoculant wire addition of 33 inoculant index units. The correlation between index units and meters of cored wire are calibrated for each foundry (ladle size, temperature, etc.) and product (solidification rate, time to empty ladle, etc.) and programmed in the correction algorithm.

According to the results of more than 100,000 CGI castings, the average corrective addition of Mg wire is only 5m/1-metric ton ladle which, at 12 g Mg/m of wire and a recovery of 50%, corresponds to just 30 g Mg/metric ton of iron. Because the corrective additions are very small, and because the process variables are eliminated during the base treatment operation, the recovery during correction is reliable. It is not necessary to obtain a second thermal analysis sample before pouring.

The objective of the base treatment operation is to intentionally undertreat the iron relative to the start-cast coordinates. The base treatment target is chosen so that even if all production variables combined to achieve the highest possible Mg and inoculant recovery, the iron would only arrive at the startcast coordinates.

To optimize the efficiency of the foundry process, the Mg and inoculant indices for each ladle are displayed as histogram run charts. Depending on the trend in the run chart, the amounts of Mg and inoculant added to subsequent base treatments can be changed by the operator according to the foundry operating procedure. The measured carbon equivalent value also is displayed. This data is compiled in daily production summaries to satisfy traceability and QA.

Although the thermal analysis and subsequent control strategy can achieve reliable production, an optional end-of-pour QA sample may be obtained by immersing the probe directly in the pouring basin of the last mold. In comparison to conventional QA tests, the thermal analysis result provides a real-time evaluation of the iron which, if necessary, allows for early segregation of castings. The Mg fade simulation in the probe makes this non-destructive testing QA technique more effective than metallographic or ultrasonic inspection as it inspects a larger sample volume and also shows how the iron would solidify if pouring were to continue for an additional 10-15 min.

The measure-and-correct process control strategy also can be applied to CGI production from pressurized pouring furnaces.

Process Control System Requirements 

The requirements of a process control system are defined by the needs of the product and the foundry production process. In the case of CGI engine blocks, the control system must be able to simultaneously control the Mg and inoculant within a narrow specification window and maintain the window from the start until the end of casting. The nodularity of all performance-critical sections must be kept within the 0-20% range to optimize castability, machinability and operational performance, and flake graphite must be avoided.

The starting point of any process control system is an accurate and comprehensive measurement of the molten iron. The metallurgical information, however, only indicates the behavior of the iron during solidification. The reliable high-volume production of CGI requires that the metallurgical understanding be developed into control actions that eliminate process variation and preclude operator error.

The most effective way to eliminate process variation is to evaluate the molten iron after base treatment and make corrective additions of Mg and inoculant prior to casting. Such an on-line measure-and-correct control strategy ensures consistency at the molding line and eliminates the risk associated with CGI production.

Another Way of Doing It

Editor's Note: The above-described method and other CGI process control technologies were detailed during a cast iron div. panel at the 1998 AFS Casting Congress. In another 1998 paper, Ford Motor Co. presented its research (98-095) on the development of a high-volume process for CGI cylinder blocks. C.R. Reese and W.J. Evans shared the results of Ford's project to develop an in-the-mold treatment process in an automotive production environment at its Cleveland Casting facility. The main components of an in-the-mold process include a reaction chamber and running system design, treatment alloy chemistry, base S and Ti, pouring rate and iron chemistry.

This type of process was considered for its high Mg recovery without treatment fading, no treated metal holding requirement, interface with autopouring, easy adjustment from downtime and minimal economic impact across high-volume operations. The method's key concerns included: the base S level control, the potential to generate dross in the mold, pouring rate consistency, late treatment's effect on nodularity control, and reduced casting yield and space on the pattern.

After assessing sensitivity to critical parameters, Ford defined a process window compatible with production capabilities that provided target block structures and properties. This process window was used for a 30-piece run on the 2-liter block to gauge process capability.

The researchers concluded that a properly designed in-the-mold treatment process is capable of producing CGI structures and properties suitable for cylinder blocks. Among their conclusions was that process consistency is highest with a MgFeSi alloy containing about 4% Mg and that the alloy should contain 0.5-1% Ce and be controlled to a specific mesh size and freshness. In addition, Ti levels of 0.15-0.22% improve graphite uniformity throughout the block and markedly expand the range of residual Mg, producing a suitable vermicular graphite morphology.