2024 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2024): Special Session: Feedforward and Feedback Control in AM II
Program Organizers: Joseph Beaman, University of Texas at Austin
Wednesday 8:00 AM
August 14, 2024
Room: 400/402
Location: Hilton Austin
Session Chair: Sneha Prabha Narra, Carnegie Mellon University
8:00 AM
Proposal for How to Treat L-PBF Like Any Other Plant: State Feedback Control Design and Computational Issues: David Hoelzle1; Nathaniel Wood2; 1Ohio State University; 2University of Michigan
Laser Powder Bed Fusion (L-PBF) is a manufacturing process, and just like the multitude of manufacturing processes in which the process can be drastically improved by applying modern feedback control methods, L-PBF is in need of process control. However, our group and others have previously shown that the large, multi-dimensional system scale and fast time scale of L-PBF make advanced control a formidable challenge that has precluded step-changes in capabilities in the past. However, if we rethink the process and controller design, we believe that L-PBF quality and performance can be improved in a materials and part geometry agonistic manner. This paper will present current efforts in state estimator and state feedback design, which we have shown to work experimentally and in simulation, respectively, and provide a proposal for improving the simplicity and computational efficiency of these methods to make real-time feedback of temperature in L-PBF a reality.
8:20 AM
Toward SmartScan 2.0 - An Intelligent Scanning Sequence Optimization Method for Reduced Part Deformation and Residual Stress in LPBF Based on Thermo-Mechanical Models: Chuan He1; Chinedum Okwudire1; 1University of Michigan
Laser Powder Bed Fusion is an additive manufacturing process increasingly utilized across various industries. Despite its advantages, LPBF often results in undesirable part deformation and residual stress. To overcome these challenges, the authors’ previous research introduced SmartScan 1.0, a method that determines the laser scanning sequence by using thermal models to optimize for thermal uniformity. This paper presents preliminary work toward SmartScan 2.0, a new version of SmartScan that is derived based on a thermomechanical instead of a purely thermal models. The thermomechanical model used in SmartScan 2.0 consists of a thermal finite difference model combined with a mechanical finite element model. The objective function directly minimizes part deformation instead of using thermal uniformity as a proxy. SmartScan 2.0 is compared against SmartScan 1.0 in experiments involving the laser marking of beams and plates. The results show large reductions in part deformation using SmartScan 2.0 compared to SmartScan1.0.
8:40 AM
Vector-Level Feedforward Control of LPBF Melt Pool Area Using a Physics-based Thermal Model: Nathaniel Wood1; Nicholas Kirschbaum2; Chang-Eun Kim3; Thejaswi Tumkur3; Chinedum Okwudire2; 1Air Force Research Laboratory; 2University of Michigan; 3Lawrence Livermore National Laboratory
Effective, in-situ process control is a longstanding goal of Laser Powder Bed Fusion (LPBF), and there exists a wealth of research on modulating the laser power in response to measured process signatures like the melt pool area. Ongoing challenges in these efforts include difficulty in making the underlying models flexible, difficulty in preemptively responding to sudden changes in the processing environment, and difficulty accommodating process inputs besides the power. This work presents a novel power-to-meltpool-area controller that addresses these challenges in turn by: (1) Basing it on an accurate and lightweight conduction-based thermal model, which is calibrated irrespective of part geometry by leveraging how it computes the area from the underlying substrate temperature, (2) letting modeled area forecasts preemptively modulate the power instead of merely reacting to existing disturbances, and (3) assigning a uniform signal for each scan vector, which will accommodate inputs like the speed which aren’t continuously-variable.
9:00 AM
Demonstration of Numerical Optimal Control for Multiple Fusion-based Additive Manufacturing Processes: Mikhail Khrenov1; William Frieden Templeton1; Sneha Narra1; 1Carnegie Mellon University
Spatially varying heat input in AM processes provides both opportunities and challenges for process planning. While heuristic methods and grid-based process mapping have been useful guides for parameter choice, they struggle with highly transient scenarios, varying part geometries, and are not suited to directly optimizing outcomes. In this talk, we examine an alternative approach based on formulating AM processes in state-space and solving for process plans using numerical trajectory optimization.We demonstrate this approach in two different AM processes: electron beam powder bed fusion (EB-PBF) and wire arc additive manufacturing (WAAM). In the EB-PBF case, we seek to minimize thermal variance over a layer. In the WAAM case, we seek to drive the part’s microhardness to follow a desired distribution. We solve both cases using a common set of tools for additive manufacturing optimal control (ADDOPT) that we have developed, and demonstrate the solutions with experimental validation.
9:20 AM
Dual-laser L-PBF Microstructural Tuning via Physics-based Feedforward Control: Nathaniel Wood1; Andrew Gillman1; Edwin Schwalbach1; Sean Donegan1; Chinedum Okwudire2; 1Air Force Research Laboratory; 2University of Michigan
A major deficiency of Laser Powder Bed Fusion (PBF) is its inability to sculpt engineering features like microstructure throughout the build like electron beam (EB) PBF, which is because the EB can realize scans the laser cannot. In this work, we mimic an EB scan in a dual-laser PBF system to constrain temperatures in the layer to be within the phase transition region of 316L SS, thus promoting an equiaxed structure that is normally achievable only by post-process heat treatments. Our dual-laser scan is based on one laser (the melting laser) following the nominal, slicer-derived scan path and the second (the heating laser) rapidly moving behind it to regularize temperatures and slow down cooling. The heating laser’s movements are scheduled by a greedy algorithm that uses a fast and accurate thermal model to choose the move which drives forecasted temperatures the closest to the middle of the transition region.
9:40 AM
Achieving Consistent Microstructure through Feedforward Temperature Control in Laser Powder Bed Fusion: Shawn Hinnebusch1; William Templeton2; Praveen Vulimiri1; Alaa Olleak1; Florian Dugast1; Sneha Narra2; Albert To1; 1University of Pittsburgh; 2Carnegie Mellon University
Part qualification is a critical step in advancing additive manufacturing, essential for achieving consistent microstructure and build quality. Heat accumulation can lead to part failure and irregular microstructure, emphasizing the need for precise temperature control. This study utilizes a layerwise thermal process simulation model to determine additional cooling time required to maintain a specified temperature. Leveraging a surrogate model, a true thickness layerwise model with powder is developed and used to determine the cooling time it takes the part to cool below the specified temperature. Calibration and validation are performed utilizing an infrared camera (IR) to capture the interpass temperatures. A recoater blade crash that occurred during the original build is rectified with feedforward temperature control. A large variation in hardness in the original build has been observed in the high heat accumulation areas compared to other regions; in contrast, the variation is significantly reduced with the feedforward control.
10:00 AM
Statistical Process Control for WAAM for Productivity and Quality Improvements: Austen Thien1; Christopher Saldana1; 1Georgia Institute of Technology
Wire-arc additive manufacturing (WAAM) is a wire-fed welding-based metal additive manufacturing process where the thermal management greatly influences build quality. The interlayer dwell time is a critical parameter that determines the process condition stability, the geometric uniformity and the overall production time of the build. Thus, control strategies are needed for the WAAM process to produce stable geometry and deposition conditions while minimizing production times. In this study, a closed-loop control technique uses cumulative summation (CUSUM) to detect in-situ statistical deviations of contact tip to workpiece distance (CTWD) and current data caused by short interlayer dwell times and then implements corrective process parameters to compensate. This CUSUM control loop is applied to single-bead 20-layer WAAM builds using mild steel (ER70S-6) material. The effects of input data stream choice and corrective parameter value on the controller performance are determined by evaluating production metrics, statistical distribution of process data, and response time.
10:20 AM Break
10:40 AM
Melt Pool Temperature Measurement Across Multiple Metal AM Processes through Application of Two-Color Thermal Imaging: Jack Beuth1; Alexander Myers1; Jonathan Malen1; 1Carnegie Mellon University
Over the past two years, the authors have developed two-color thermal imaging methods which yield high resolution thermal field measurements that are independent of material emissivity. This talk will review multiple applications of the method that include laser powder bed fusion, electron beam powder bed fusion, and powder fed directed energy deposition. In the powder bed processes high speed imaging is used and a careful balance must be struck between long frame exposure times capturing sufficient emissions and short frame exposure times needed to capture rapid melt pool movement. In DED processes, conventional speed color imaging is possible and real-time monitoring and control is feasible. In all processes, melt pool surface temperature measurements are allowing accurate specification of molten metal properties needed to accurately model melt pool thermal fluid fields, which is of critical concern in the application of advanced multi-physics melt pool modeling.
11:00 AM
In-Situ Photodiode Collection and Trends Across Feed Forward Parts: Gabe Guss1; Saad Khairallah1; Amit Kumar1; Justin Patridge1; Steven Hoover1; CE Kim1; Ava Ashby1; Ibo Matthews1; 1LLNL
Intelligent Feed Forward (IFF) is a data driven approach to optimizing input parameters to laser powder bed fusion. At its core is predicting parameters that results in a target value of an in-situ monitored photodiode signal. This presentation will cover collection and trends of this signal while printing varied geometries on a GE M2 and SLM280 machine. A workflow where data from each machine is brought into a common format, plotted in 3D, and analyzed using statistics will be presented.
11:20 AM
Iterative Learning for Efficient Additive Mass Production: Christos Margadji1; Douglas Brion2; Sebastian Pattinson1; 1University of Cambridge; 2Matta Labs
Material extrusion could enable on-demand production of complex and personalised parts but is limited by low reliability, particularly in higher-volume production. Machine learning-based methods may enhance reliability, but are often themselves insufficiently reliable for use in production. Foundation artificial intelligence models have enabled significant improvements in performance across many tasks. Here, a vision-based control system is reported, coupling active learning and uncertainty awareness with a foundation model to continually learn to build a specific part better. The resulting framework is called Iterative Learning, as it improves performance by learning from its own errors during repeated build cycles of the same part. The iterative learning approach is shown to enable robust error detection and correction while being more space, time and computationally efficient compared to a naïve fine-tuning approach. This provides a path showing how foundation models may be adapted to enhance reliability across a wider range of additive manufacturing processes.