May 2023 Volume 5

AUTOMATION

The Application of Digital Process Twins (DPTs) in Forging The physics-informed and data-driven Digital Process Twin approach is particularly relevant in the context of forging processes, as the behavior of metallic materials under the unique thermo mechanical loads of forging often results in a material response that differs substantially from ‘handbook’ data. For example, this may involve data from a Gleeble System or similar process-specific characterization setup. By relying on process parameter/load/ structure/performance relationships, rather than one-dimensional constitutive material response under idealized testing conditions, the DPT approach allows manufacturing-induced material changes to be optimized to deliver location-specific properties. Combining in-situ process sensing and limited validation trials, fast semi-analytical modeling tools can be used to quickly establish reliable process regimes, reducing the likelihood of process induced anomalies. These methods have already allowed DPTs to proactively demonstrate examples of location-specific control over critical, process-induced material properties (e.g., residual stress or recrystallized surface layers generated during machining of Ni based superalloys). Importantly, in addition to microstructure/ quality metrics, DPTs also consider cost and throughput metrics that are critical to any industrial operation. The IMPI approach optimizes for ‘dollars and cents’, rather than stopping at ‘stress and microstructure’, as is the case with more academic approaches. Current forging process models, typically based on finite element analysis (FEA), require complex simulations to be set up and analyzed by modeling experts, while relying on simplifying continuum assumptions and constitutive material models as inputs. It is not unusual for FEA to take days or weeks to set up and run, which may be appropriate during product development, but not during production or for short-run components. Meanwhile, DPTs of forging processes operate in real-time by focusing on process specific material behavior and leveraging data-driven, physics

informed, and machining-learning (ML) paradigms to improve computational efficiency by more than an order of magnitude over existing modeling approaches. To realize this improved model efficiency without sacrificing required predictive accuracy, DPTs take in the process-specific material response to the unique loads of a specific forge press to define “well-centered” processing windows that are robust against changes in process conditions that can occur due to varying demands of part geometry, and other influences that affect the thermal history of the workpiece. Based on these insights, more detailed FEA analysis and experimental trials may be conducted for additional optimization when justified. To enhance the quality and relevance of data obtained from the forging process, we implement a novel approach: model-informed sensor selection. Just as we have set aside ‘universal’ material property data, likewise we reject generic process sensors; indeed, we have observed countless examples of non-value-added sensor deployment in the field due to one-size-fits-all approaches. A specific example of applying model-informed sensor selection to forged components can be found in an Air Force program, which sought to improve the forging and machining of aluminum bulkheads for the F-35. Fit-for-purpose process modeling combined with high-throughput, residual stress measurements enabled Howmet Aerospace to implement a proprietary coldworking process that mitigated the negative effects of forging-induced residual stress (RS) during post-process machining and use (distortion, reduced fatigue life, etc.). By identifying the relevant process metric (pre machining RS), along with relevant model-informed sensors (contour method RS measurement), the project team generated pedigreed process data that was subsequently used to both update the DPT model and implement quality control protocols during production. Figure 2 illustrates the key aspects of how a version of the DPT paradigmwas deployed during the program.

Figure 2: Implementation of model-informed sensor selection in forged aluminum bulkheads. (Courtesy of Hill Engineering LLC)

FIA MAGAZINE | MAY 2023 37