Optimized Cooling Rates of Microalloyed Steels (Part II)
The first installment of this article covered the problem of properly cooling forged microalloyed steels. In this second and final article we discuss details on heat-transfer coefficients, experimentation, production simulation and final results enabling more extensive use of this class of steels.
The first part of this article ran in the December 2019 issue. It gave an overview of a challenge facing forgers of microalloyed steels – controlled cooling! Now that the problem of controlled cooling of forged microalloys has been modeled and solved, other challenges are clear and present for the forging industry. But before tallying those opportunities, let’s present the data and the solutions to the problem at hand – controlled cooling of microalloyed steels.
Transient heat-transfer processes can easily be simulated using implicit, non-linear finite element models (FEMs) to predict a part’s thermal response to heating and cooling. Such models represent a routine use of a very mature technology. At the conclusion of a hot forging simulation, the temperature of each node in the workpiece is known. The cooling rate after forging can be predicted by simulation if the cooling process is known.
For a microalloyed steel forging, controlled, forced-air cooling is typical. Hot forgings are cooled as they are transported on a shrouded conveyor through cooling zone(s) with one or more cooling fans. It is not practical to experimentally measure the cooling rate in a subsurface region of these forgings. However, an FEM model can easily predict these cooling rates if the heat-transfer coefficient (HTC) or convection coefficient is known.
Typical HTC values are available in literature, although these are approximate values at best. They are typically estimated to the nearest order of magnitude. It is also uncommon to see any temperature dependence included in the literature. So, the public data are reasonable for an initial guess, but not for an accurate model. To obtain an accurate model, the forger needs to identify the heat-transfer coefficients for continuous cooling.
Computational fluid dynamics (CFD) codes are very sophisticated and can predict thermal response in highly transient systems. CFD is mature but extremely tedious and time-consuming. Using CFD to model a simple fan-cooling application is probably excessive, with a high cost in time and data.
A practical engineering approach to extracting the surface convection coefficient from a real process involves a mix of experimental data and simulations. A controlled-cooling experiment is conducted with thermocouples embedded close to the surface of a workpiece. The output of the experiment is a temperature-versus-time dataset for each thermocouple. An approximate guess for the HTC is provided by an initial computer model, which is subsequently run to match the experiment. Point tracking is used to extract a temperature-versus-time curve for each thermocouple.
After the model is completed, the curves are compared and differences reconciled. Using an optimization routine, a subsequent simulation is run with the updated HTC data, which is a function of temperature and location. After several iterations, the solution converges when the time-temperature prediction from the model matches the experiment. This is an empirical model, which is practical in a forge-shop environment.
Jernberg Industries funded experimental work to determine such HTCs. An automotive wheel-hub forging made from microalloyed steel was fitted with six thermocouples at specific subsurface locations. A photo of the forging and thermocouples is shown in Figure 5. The forging was then heated in a gas box furnace to a temperature just below 1093˚C (2000˚F), as shown in Figure 6. It was then cooled by a single fan similar to the fans used in the actual production line.
The data was processed by SFTC using the inverse HTC module from the DEFORM system. The outputs of this work were surface HTCs as a function of temperature. Unique coefficients were defined for each of three regions of the forging. Cooling rates were associated with portions or zones of the forging (i.e., fast, intermediate and slow rates). With the HTC data available, SFTC ran a model to compare the time-temperature result from the model and experiment. Figure 7 compares the two results at five locations.
The correlation was excellent and validated the reverse-engineered HTCs for the sample fan. The expectation was that these HTCs would be adequate for predicting the cooling rate of a forging, with a known temperature profile, in a production cooling setup.
Companies have simulated their hot forging processes for decades with excellent success. Forging models typically include heat transfer to the extent that it influenced the forging process. Forging simulations involving multiple operations have accounted for thermal response prior to forging, during transfers and when resting on tools. A similar multiple-operation simulation approach was used to model the post-forge cooling process for one of Jernberg’s microalloyed steel production parts.
To validate modeling, Jernberg captured and compiled production data related to the cooling behavior of 15V30M production forgings. Pyrometer measurements were taken at five locations (A-E) immediately after forging and at various instances along the conveyor cooling system (Fig. 8). Each data point represented the average of 10-14 individual surface temperature measurements collected across different parts.
Post-forge cooling simulations were run by SFTC using HTC data from the earlier fan study. Like the experimental forging, the production forging was assumed to have three HTC zones. These regions accounted for different rates of cooling (Fig. 9). The schematic shown in Figure 10 depicts the modeled conveyor setup. No fans were present in conveyors #1 and #3. Fans were mounted at the entry to conveyor #2 and the exit tunnel after conveyor #3. The schematic indicates the locations along the conveyor (T1-T8) where the temperature readings were taken (Fig. 8).
The initial, average cooling results looked reasonable (Fig. 11), but there was concern with the slope of the cooling curve after transformation. The slope represented the cooling rate, which was the primary focus of the study. The team then decided to re-evaluate the model setup and assumptions.
Air-flow differences were identified between the experimental fan setup and certain sections of the production cooling system. Engineering judgment was used to adjust the HTCs for conveyor #2 and tunnel sections. The final cooling model (Fig. 12) obtained a better overall fit to the experimental data average.
The final cooling model reasonably matched production observations. Figure 13 shows heat profiles on the model and production part at three points along the conveyor. Figure 14 compares production measurements to simulated temperatures at five locations on the part. There was a high level of confidence in the cooling-rate predictions. Hardness was measured at two locations of the production forging: the stem and center (Fig. 15). Predicted cooling rates were extracted from the same locations in the forging model. The measured hardness and predicted cooling rates were then plotted together on the 15V30M critical cooling-rate graph from CSM (Fig. 16).
While a wider range of process conditions might diverge, the production “sweet spot” matched the lab experiments well. Both the CSM study and Jernberg’s production experience identified the cooling rates necessary to obtain an ideal hardness.
Future Research Opportunities
This project is the first published trial of its kind. Future improvements were discussed by the project team, some of which are summarized here. Additional developments will allow the test methods to be applied to a wider range of applications.
Testing the HTC for various fan configurations as a function of location, temperature and fan speed would provide a more accurate post-forge cooling model. Variations could include shrouded-versus-ducted conveyors, fan speed and average air velocity along the conveyor. Knowing the effect of these factors on the cooling rate will enable the forger to further fine-tune cooling for optimal microstructure and properties.
Developing a database of cooling-rate-versus-hardness data for a wider set of alloys is desirable. It could be coupled with data analytics tools to refine the target cooling rate and the probability of meeting mechanical-property requirements. Further refinement may be possible with the inclusion of chemistry and the material supplier.
As a reader of this article, you can also participate in future projects! Contact Jim Miller of SFTC regarding our company’s interest to continually improve your processes for your customers requiring microalloyed steels.
The authors thank Jernberg Industries for their contributions in the experiments, production testing and project reporting involved in these efforts. Markus Knoerr, Jernberg’s VP of engineering at the time, is credited with the original FIERF project concept. The team also thanks the Forging Industry Education Research Foundation (FIERF) for supporting CSM’s efforts and supporting technology transfer across the industry as it delivers affordable, high-quality parts for demanding applications.
Co-author James Miller is Director of Sales and Marketing and frequent FORGE contributor John Walters is Vice President for Scientific Forming Technologies Corp. (SFTC), Columbus, Ohio. Jim can be reached at firstname.lastname@example.org or at 614-451-8330. Co-author and frequent FORGE contributor Chet van Tyne is Professor Emeritus, Metallurgical and Materials Engineering, Colorado School of Mines, Golden, Colorado. He can be reached at email@example.com.