Prediction and Control of Thermal History in Laser Powder Bed Fusion
Files
TR Number
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The long-term research goal of this dissertation is to enable flaw-free production of metal parts using the laser powder bed fusion (LPBF) additive manufacturing (AM) process. As a step towards this long-term goal, the goal of this work is to predict and control the thermal history of an LPBF part. The thermal history is the spatiotemporal distribution of temperature in an LPBF part as it is created layer by layer. Thermal history is the primary cause of flaw formation in LPBF. To realize this goal, the objective of this dissertation is to establish and advance a novel thermal modeling method based on the concept of spectral graph theory, which is more than 10 times faster than existing finite element-based methods for the same level of accuracy. The central hypothesis is that physics-guided prediction, optimization, and control of thermal history mitigates flaw formation and enhances functionally critical properties of LPBF-processed parts when compared to parts produced without control of thermal history. The practical rationale and need for this work are as follows. LPBF is becoming increasingly prevalent due to its ability to fabricate complex structures that would otherwise be impossible with traditional subtractive and formative manufacturing processes. The freedom of geometry afforded by AM processes such as LPBF enables designers to place a stronger emphasis on design efficiency rather than the manufacturability of components. It also facilitates greater supply chain flexibility, reducing part lead times and costs. For example, making an aerospace part weighing just one kilogram with traditional subtractive and formative techniques requires processing 20 kilograms of raw material—a buy-to-fly ratio of 20:1—and lead times for new parts are often several months long. LPBF reduces the buy-to-fly ratio to less than 5:1, and the lead time is just a few weeks. Despite these advantages, LPBF has seen limited industry adoption and use, especially in safety-critical applications, due to the tendency of the process to form flaws. Approximately one in three parts are affected by flaws. Flaw formation leads to inconsistent part properties and can cause catastrophic failures in safety-critical aerospace, defense, and biomedical applications. Flaw formation in LPBF parts is mainly attributed to the thermal history. Thermal history, in turn, is influenced by complex design-process-material-machine interactions that require mathematical modeling. Rapid and accurate prediction of the thermal history can enable practitioners to avoid flaw formation and achieve desired part properties by optimizing the part design and process parameters before the part is printed. Previous research in graph theory computational thermal modeling of LPBF has demonstrated that it is approximately 10 to 20 times faster than existing finite element-based techniques, and the prediction error has been experimentally validated to be less than 5%. This dissertation leverages the graph theory approach to address the burgeoning practical need for a rapid, accurate, and experimentally validated physics-based approach for mitigating flaw formation and ensuring part quality in LPBF. Specific outcomes are as follows.
- In Chapter 2, the graph theory modeling approach was used for feedforward control of processing parameters to mitigate thermal-induced flaw formation in LPBF. The key idea was to avert heat buildup in an LPBF part before it is printed by adapting process parameters layer by layer based on thermal history predictions from the graph theory model. This replaces cumbersome empirical parameter optimization with a physics-guided strategy. The effectiveness of the approach was demonstrated with Inconel 718 parts. Compared to parts produced with empirically optimized processing parameters, parts produced with parameters obtained from the graph theory thermal history predictions showed superior geometric accuracy, finer grain size, increased microhardness, and improved surface roughness.
- In Chapter 3, the graph theory thermal modeling approach was combined with data-driven machine learning to predict two important microstructure-related characteristics, namely meltpool depth and primary dendritic arm spacing in LPBF-processed Inconel 718 parts. These characteristics are critical determinants of functional physical properties, e.g., yield strength and fatigue life. Rapid and accurate models for predicting microstructure evolution are therefore valuable to reduce process development time and achieve consistent properties. However, owing to their computational complexity, existing models for predicting microstructure evolution are limited to a few layers. This work addresses the aforementioned research gap via a novel physics and data-integrated modeling approach. This approach enabled prediction of primary dendritic arm spacing with a root mean square error of ≈ 110 nm.
- Chapter 4 advances a physics-guided model predictive control approach to autonomously optimize the processing parameters for an LPBF part before it is printed. Currently, LPBF processing parameters are optimized through empirical studies based on simple test coupons. However, processing parameters optimized for test coupons seldom transfer to complex, practical parts, necessitating further rounds of testing. This is because the thermal history, which is the root cause of flaw formation in LPBF, is unique to each part shape. This model predictive control strategy was applied to stainless steel 316L parts. It eliminated support structures, resulting in significant material savings, improved geometric accuracy, and enhanced microstructure homogeneity.
- Chapter 5 further advances model predictive control of thermal history. It ascertains the effect of four different thermal history control strategies on geometric integrity, surface finish, and microstructure evolution in LPBF-processed stainless steel 316L parts. The four thermal history strategies tested were (i) constant end-of-cycle temperature, (ii) minimization of thermal gradients, (iii) control of cooling rate, and (iv) constant inter-layer time. Controlling the cooling rate and minimizing the thermal gradients were found to be more effective at controlling microstructure heterogeneity.