Research Areas

Smart Manufacturing and Data Analytics: Our research group focuses on smart manufacturing and data analytics, such as industrial automation and robotics, additive manufacturing (AM), data analytics, and decision support. Our current primary research is in AM technologies, specifically the wire + arc additive manufacturing (WAAM). Despite significant progress in the AM field, a number of technical challenges remain, such as lack of standards/guidelines; modeling and simulation tools; AM design tools; data information management; limited number of available materials; and build capacity, processing time, certification, and qualification. Among those challenges, we focus on four main specific topics 1) modeling and data analytics in AM, 2) big data management and data mining techniques in AM, 3) real time monitoring and quality control, 4) issues of quality assurance and standard efforts. The resarch team will provide innovative and transformative knowledge and methodology, leading to greater efficiency and improved profitability in the aerospace, defense, and automotive manufacturing industries

Wire+Arc Additive Manufacturing (WAAM) Systems and Processes: The WAAM system was developed with several considerations: (1) inexpensive system setup costs, (2) fabrication of multi-functional parts, (3) wide selection of wire fillers, (4) sustainability in terms of energy efficiency and less spatter, (5) near-net shape fabrication, and (6) high deposition rate (near 10 kg/h). There are two independent systems in the setup. As shown in the real photo of the setup, the left system consists of a six-axis robot arm (i.e., Fanuc ArcMate 100 iB), a welding power source (i.e., Fronius Cold Metal Transfer Advanced 4,000 MV R), a VR7000 wire feeder, and a controller (i.e., Fanuc R-J3iB). The right system consists of a six-axis robot arm (i.e., Fanuc ArcMate 100 iB), a welding power source (i.e., Lincoln Electric, Model: PowerWave F355i), and a controller (i.e., Fanuc R-J3iB). Both systems can be run separately or simultaneously based on the requirement. Currently, the research team is installing five more robots (ArcMate 120iC) in a new space (1,000 ft²) for research and education enhancement. From these system setups, the research team has been successfully depositing a wide range of functional and geometric parsts. The deposted materials include steel and aluminum alloys, H-13, Inconel 625, Titanium grade 5, high-entropy alloy, Mo, and TZM alloys. Currently, the research team is focusing on deposition of refractory alloys, such as Tungsten/Molybdenum/Niobium and refractory high entropy alloys and its materials characterization at high temperatures.

Multi-scale, In-situ Materials Characterization: The main undesirable feature of the metal AM process is the non-equilibrium thermal cycles, consisting of the solid-melting crystallization and solid-remelting recrystallization under fast heating and cooling conditions, which generate anisotropic microstructures and defects. These heterogeneous and defects-inducible structures can be categorized into macro-, micro-, and nano-scales. The macro-scale defects and inhomogeneities can include deformation due to residual stress, volume shrinkage, segregations, oxide formation, inclusions (e.g., slags, voids, and cracks), and surface irregularity due to the balling effect and spatter. The micro-scale defects cover micro-segregations, grain morphologies, intermetallics, microvoids, and microcracks. The nano-scale inhomogeneities can include the grain boundary (GB), twin boundary (TB), and dislocation. Understanding the root-causes of defects and inhomogeneities is one of the main factors from this multi-scale, in-situ materials characterizaiton. In order to tackle these issues, multi-scale, in-situ materials characterization techniques are used, inlcuding in-situ neutron diffraction.

Multi-modal Process Signatures-based Machine Learning: The area where the arc source interacts with metal wire is called the “weld-pool” or “heat-affected zone.” It is a complex physical phenomenon, since it is related to physical properties in the arc energy source (e.g., power, mode, and wavelength) and wire material (e.g., thermal conductivity, emissivity, and diameter). The characterization of the weld-pool features (e.g., maximum temperature, size, length, and depth) is very important for validating the integrity of the melt-pool and layer, detecting defects (e.g., balls, cracks), and analyzing part properties (e.g., microstructure analysis and tensile strength) for process verification and part validation (quality assurance). Based on measurements (i.e., process signatures, microstructures, and properties), the research team is relating them by utilizing several data analytics techniques (e.g., machine learning and statistical approach). For example, the research team will generate machine-learning models as follows: (1) prepare the data sets (1D/2D signatures, microstructures, and properties); (2) refine and extract the features from the data sets for the training; (3) label and train them via machine learning techniques (e.g., neural network, CNN, ensemble learning); and (4) develop and validate the machine learning models.

Process Optimization with Uncertainty Quantification: The goal is to develop decision support framework of WAAM with uncertainty consideration. The specific goals are as follows: 1) development of a generic WAAM framework that consists of robot-based WAAM system, 2) establishment of its decision support framework which can manage the unwanted features in multi-criteria decision making (MCDM) problems, 3) integration of WAAM framework and quality information framework (QIF) for inspecting the part quality, and 4) its sustainability assessment. To address the research issues, this research characterizes the relationships between process parameters, its microstructure, mechanical properties of an AM part, and its process performance. Based on these relationships, we will develop a surrogate model–based, multi-criteria decision-making methodology (MCDM) of the WAAM process which can control the process performance, the part quality, and its sustainability for mass customization. It can reduce manufacturing cost, time, waste, and defects and address in situ process problems with satisfying the part quality requirements.

Multi-scale, Multi-physics Computational Modeling: It is challenging to explore the underlying physics and develop the design rules using only trial-and-error or design of experiment approaches, since AM process and refractory alloys requires significant resources. Coupling experiments with physics-based computational modeling will lead to the foundational understanding of defect formations, solidification, and deformation behaviors in an effective way. This integration of experimental-characterization and theoretical-modeling efforts will provide unique, transformative, and efficient opportunities to synergistically understand and elucidate the defect formation and deformation behaviors of additively-manufactured refractory alloy structures. Moreover, realistic predictive computational models will serve as powerful guides to experiments, thereby reducing a number of them. There are not many studies on this physics-informed, data-driven approach to advance the knowledge and understand the underlying physics of the additive manufacturing process for the refractory alloys. The goal is to link process-structure-property-performance for design rule establishment using the multi-scale, multi-physics computgational modeling approach.

Physics-informed, Data-Driven Digital Twin Non-Destructive Evaluation: The specific goal of this research is to investigate the digital inspection method using the process signature measurements, which can analyze process signature measurements digitally to validate a part instead of conducting physical tests, one of non-destructive evaluation methods. Part properties can be validated/inspected by analyzing extracted features from the AM process signature. For example, geometric accuracy, the residual stress of a part, microstructure analysis, geometric distortion, and other properties (e.g., tensile strength and surface roughness) can be theoretically estimated by analyzing process signature measurements. Using an image-based pyrometer (thermal imaging system), process signatures are measured, and image stacks of melt-pool/HAZ/layer characteristics must include important information (e.g., thermal history and melt pool shape, size, and temperature) for thermal analysis (e.g., cooling rate and cooling direction, solidification rate), which can be effectively used for microstructure analysis.