As forgers are pressured to remain competitive, they must act on opportunities to reduce costs, eliminate scrap and improve operational efficiency. Process simulation and advanced testing methodologies have permitted engineers to gain deep insights into their processes, allowing many to modify old processes or launch new ones that are optimized and even robust.

Traditional Optimization

Many successful optimization projects involved the manual setup and simulation of process and/or design variations. Individual iterations often represented a single change in geometry, process condition or material. Results traditionally involved A-vs.-B comparisons. Multiple comparisons were often referenced as engineers converged on a preferred solution. The scope of each study was dictated by computing speed, setup time, manpower and other constraints.

Such an example of process optimization was published in a thesis by Edgar Espinoza.^[1]  It studied a wrench forging produced by Green Bay Drop Forge that was in need of improvement. The existing process required many hammer blows, produced significant flash and was sensitive to billet positioning. Tooling modifications were designed in Unigraphics NX. The existing process and alternate workflows were simulated in the DEFORM system. A number of potential solutions were evaluated during the study. New iterations were based on information gained from prior iterations.

The final proposal was tested on the shop floor and compared to the original design, as shown in Figure 1. The redesign resulted in 22% fewer hammer blows and produced less flash. The flash was also more uniform, which reduced the process’s sensitivity to billet positioning.

A PRO-FAST project involving Delfasco Forge, SFTC, the Forging Defense Manufacturing Consortium (FDMC) and the Defense Logistics Agency (DLA) represented a similar application.^[2] A forging process had 11% scrap despite the special attention paid to each production run. DEFORM was used to simulate the as-is process while ensuring results matched production. It was then utilized to test numerous die and process modifications.

Simulation revealed that the imprecise roll-gathering operation led to inconsistent finish forgings. A preforming operation was added to provide more control over material flow. After 16,000 new production forgings, scrap had been reduced to 3% and savings exceeded $100,000. The original and modified die designs are compared in Figure 2.

Such basic optimization workflows are easily employed during development. They are mature and practical approaches for many companies today. The most significant constraints are the time required to manually process iterations and the limited comparisons available from a typically small sample set. 

Integrated Methodologies

Over the last decade, design of experiments (DOE) and optimization methods have been integrated into process-simulation software. The advanced technology has allowed large sets of simulations to be defined, automatically processed and efficiently analyzed using powerful statistical tools. The fundamental concepts and simulation-based implementations of these methods were summarized in previous articles. (See “Design of Experiments and Optimization in Process Simulation,” FORGE, August and October 2015).

A recent case study demonstrated ways in which DOE technology let users better understand and improve processes. The PRO-FAST project involved Cornell Forge, SFTC, FDMC and DLA.^[3] The goal of the study was to improve a steel gear-blank forging process.  This was accomplished using DOE tools integrated into DEFORM. The existing process experienced high scrap rates due to underfill on the long end of the forging (Fig. 3). Excessive production times resulted from the need for frequent troubleshooting and operator adjustments.

The first step of the project was to characterize the existing process conditions. This ensured that the simulation model accurately reflected reality. DOE tools were used to reverse-engineer process conditions that were otherwise difficult to measure on the shop floor. Tornado charts revealed that the process was predominantly sensitive to friction. Surface-response plots verified what the operators and managers already knew – that the suitable processing window was narrow.

With proper inputs established, simulation-based DOE was used to refine the process. This step was much more advanced than the manual A-vs.-B comparisons used in traditional development. Integrated DOE tools automated hundreds of simulations. They evaluated numerous process variations across the entire production process. Several studies were used to find the ideal workpiece geometry, tool geometries and process conditions. The ideal process redesign was predicted to evenly and completely fill both ends of the forging.

Robustness

Optimization does not ensure robustness. A process that never has success is poorly designed. A process operating in normal ranges with frequent failures is not robust. A robust process provides regular success when operating conditions are within expected ranges. The criteria that define a robust process may vary by application.

Precision cold-forming and critical aerospace forgings are often held to tighter process controls than open-die or hammer forging. The concept of forging robustness is graphically illustrated in Figure 4. The response plots’ axes represent the normal operating ranges for billet temperature and length. Nominal forging runs are indicated by the green dot. Some forging runs may deviate from the nominal.

The forging process that lacks robustness produces bad results even though the process conditions stay within expected ranges. The robust forging process provides good results across the entire range of normal operating conditions.

Returning to the gear-blank case, the third stage of the PRO-FAST project was to ensure the redesign provided a robust production process. A final DOE study evaluated the optimized design against the expected variation in process inputs. Figure 5 compares surface-response plots obtained from DOE studies on the original and modified processes. The plots measure the amount of die fill. Red regions indicated scenarios where the forging did not fill.

The original forging process only worked in a narrow processing range. This range is indicated by the small green circle. It is bounded by a line indicating the maximum allowable forging temperature. In contrast, the redesigned process was broadly robust. This was the case across multiple variables, including transfer time, forging temperature, cutoff length and friction. The nominal run is indicated by the red dot. In some cases, robustness extended far beyond the normal operating ranges (dashed red lines).

The effectiveness of the optimization and robustness efforts was proven out during shop trials. The trial run produced 352 forgings. Zero forgings exhibited non-fill. The production rate was also faster than before. This might have been due to the robustness of the process, which was designed to handle any expected process variations. A photo of the redesigned forging progression is shown in Figure 6.

Conclusion

These examples illustrate the advantages of forging development through computer simulation. Optimization efforts may involve simple comparisons or advanced test methodologies. DOE facilitates reverse engineering of input parameters and identification of an ideal process. Robustness can be designed into a process. This ensures that production is successful not only when conditions are ideal but also when they are not. Users can thus avoid the need to later fix a job on the shop floor, where changes are more costly and difficult.

For additional information visit www.deform.com.