Abstract
Modern technological demands have led to the development of, and necessity for, increased capability in specialized metal production. These demands include the desire to produce bespoke components designed with complex geometries and internal structures, difficult to achieve via traditional manufacturing methods. Metal additive manufacturing (AM) has burgeoned around these needs and has become increasingly accepted in the 21st century. Modern metal AM is capable of producing high tolerance components in a variety of material systems including aluminum, steel and titanium alloys. Of the metal AM processes available powder bed fusion (PBF) methods, and in particular selective laser melting (SLM), has become the most widely adopted method and has become increasingly commonplace in aerospace, automotive and medical applications among others for complex component production.
However, SLM produced components frequently carry undesirable effects inherent to the process, including the development of large and anisotropic residual stresses (RS) and process-induced defects such as lack of fusion pores within parts. These common SLM issues result in the reduction of assurance in mechanical properties of produced components, particularly in fatigue, inhibiting further widespread process acceptance. These issues frequently necessitate increased post-processing heat treatments and enhanced quality control inspections, increasing production time and cost for SLM. Process monitoring methods to ensure part quality and improve production have been studied widely but have been largely limited in providing direct connections to final resulting component defects from non-contact in situ measurements.
This dissertation elucidates these issues by developing novel techniques to detect, monitor and predict both RS and defect development in SLM components through non-contact measurements, with direct links to final component characteristics. This work is approached in three thrusts. First, a methodology was developed utilizing three-dimensional digital image correlation (3D-DIC) to estimate residual stress magnitude and distribution from surface curvature measurements in SLM parts using an analytical approach. Residual stress estimates from 3D-DIC curvature are verified using X-ray diffraction measurements of the components. Second, a method to detect and predict subsurface porosity in SLM components during production was developed using full-field infrared (IR) imaging, diverging from common IR techniques focused on localized melt-pool monitoring. Defect formation identified by IR is verified using scanning electron microscopy. Finally, a method was developed to predict the likelihood of defect formation from irregularities in powder bed spreading using 3D-DIC and machine learning classification of the errors.
Through this work, new methodologies for effective non-contact process monitoring and quality control were developed for SLM production, which provide direct links between in-process measurements and ultimate part characteristics. Key findings relating to the developed methods include (i) the relationship between surface geometry/curvature developed during the production process and final resulting residual stresses, (ii) the relationship between thermal irregularities (size and magnitude) measured during SLM processing and the physical microstructural defects that they result in, and (iii) the development of a predictive tool able to identify whether or not an observed powder error during SLM processing is likely to result in a defect. By developing improved monitoring techniques along with an enhanced understanding of the relationship between in-process observations and final component characteristics, for both residual stresses and internal defect formation, the developed methods can enable enhanced quality control for SLM components, provide a means for more efficient component certification, and give guidance for identifying RS and defects during production. These monitoring methods and analysis procedures may lead to direct real-time process corrections during production when errors occur, which can aid in reducing costly and time-consuming post-process treatments and inspections, thereby further expanding the applicability and financial feasibility of SLM manufacturing. The developed in-process monitoring and analysis methods may also be invaluable during the development of processing parameter sets for new SLM alloys in development to achieve faster optimization and certification.
