A D V A N C E D M A T E R I A L S & P R O C E S S E S | J A N U A R Y / F E B R U A R Y 2 0 2 3 3 0 forward for the benefit of the community as a whole. Process Monitoring and Machine Learning in Additive Manufacturing Prof. Jack Beuth Carnegie Mellon University Process monitoring must be viewed in terms of the physical phenomena being observed. With laser powder bed fusion (LPBF) processes, some physical phenomena of interest include (i) melt pool spatter, (ii) pore defects, (iii) re-coater blade collisions, (iv) thermal nonuniformity and thermal drift, and (v) registering data to the location in the build. Spatter detection in LPBF is important to the understanding of defect evolution. This can be done with high-speed video cameras in combination with technologies that can automatically track particle trajectories. Some observations from work done at Carnegie Mellon University are (i) lack of fusion can induce more spatter; (ii) keyholing radically increases spatter; (iii) build rates that are too fast or too slow can increase spatter; and (iv) scan strategies with more starts and stops can increase spatter. For directed energy deposition wire fed processes, machine learning of melt pool video data has been used to identify irregularities with minimal training sets. ML and artificial intelligence are needed for analysis because of the huge data sets available. ML and AI are useful in making design choices and for parameter selection, for process monitoring and control, and for the analysis of images and materials characterization data. The integration of ML tools across the additive manufacturing data ecosystem is of potential high value. The AM ecosystem may be divided into preprocessing, processing, and post processing. The preprocessing phase involves the design and development of new materials and macrostructures. In this phase, process design optimization, powder characterization, process map optimization, and the development of multimaterial structures and novel lattice materials can be accomplished. During the processing phase, in situ sensor data can be analyzed and used for process monitoring and control, as well as for flaw detection. Post process applications include defect and surface finish characterization. Challenges with Adopting Digital Manufacturing Technologies Mr. Mick Maher Maher & Associates LLC While the benefits of utilizing digital manufacturing technologies such as Internet of Things (IoT), AI/ML, and computational modeling and simulation are numerous, many companies are struggling to adopt them. Incorporating digital based systems into traditional manufacturing operations is extremely disruptive. It requires thorough planning and comprehensive buy-in from across the organization to be successful. To do this well, companies need to be aware of the challenges that will be faced. In adopting digital manufacturing technologies, the challenges most likely to be faced can generally be related to strategy, business case, infrastructure, resources, data management, and business practices. Companies need a holistic strategy to guide the adoption of digital manufacturing. Common mistakes range from having no strategy to failing to understand their internal operations and the extent of effort required for full utilization. The strategy needs to tie overarching business goals to the solutions sought. In addition to having a clear understanding of the value that “digital” brings to their operations, management should be keenly aware of the task at hand and the effort that will be required. Change management is key to preparing the workforce and being able to maintain momentum through the successes and failures that occur in any change. Adoption of digital manufacturing technologies can be difficult, but a solid business case helps focus the effort. Business cases tend to focus on incremental changes. To fully realize the benefits touted for digital manufacturing technologies, the business cases need to look at the global changes brought about due to the pervasive impacts these technologies have on the organization. Business cases can often underestimate the effort required, which undermines the adoption when overly optimistic goals are missed. At the same time, secondary and tertiary impacts are often excluded from the analysis. Companies need to assess their current infrastructure and resources. Connectivity is a key requirement for digital manufacturing technologies. Without a clear overarching strategy for implementation, the company can become burdened with and overwhelmed by multiple, standalone computational systems requiring their own specialists and support. Understanding the existing workforce and infrastructure helps identify training and talent acquisition needs, as well as areas of investment to ease the transition to an optimum digital state for manufacturing operations. Implementations are often very site-centered, but to fully take advantage of the digital environment, the entire supply chain should be considered a resource that needs to be connected into the system. Cyber-physical security of the infrastructure and data need to be addressed to minimize the risks inherent in the use of digital manufacturing technologies. In fact, data management is a challenge that is often overlooked until the problem becomes overwhelming. In adopting these technologies, a clear policy is required that not only addresses security but includes data volume, relevance, storage, and accessibility. A company’s business practices can become a challenge. The impact of tech data package ownership, qualification, and certification, in addition to intellectual property as related to adoption of digital manufacturing technologies and changes to these practices, need to be considered. In a similar vein, companies have not adapted to an ever-shrinking technology cycle. There is an assumption that implementing these technologies is a one-time event, and this perspective does not consider
RkJQdWJsaXNoZXIy MTYyMzk3NQ==