Monitoring, Modeling, and Statistical Analysis in Metal Additive Manufacturing: A Review

Abstract

Despite the significant advances made involving the additive manufacturing (AM) of metals, including those related to both materials and processes, challenges remain in regard to the rapid qualification and insertion of such materials into applications. In general, understanding the process-microstructure-property interrelationships is essential. To successfully understand these interrelationships on a process-by-process basis and exploit such knowledge in practice, leveraging monitoring, modeling, and statistical analysis is necessary. Monitoring allows for the identification and measurement of parameters and features associated with important physical processes that may vary spatially and temporally during the AM processes that will influence part properties, including spatial variations within a single part and part-to-part variability, and, ultimately, quality. Modeling allows for the prediction of physical processes, material states, and properties of future builds by creating material state abstractions that can then be tested or evolved virtually. Statistical analysis permits the data from monitoring to inform modeling, and vice versa, under the added consideration that physical measurements and mathematical abstractions contain uncertainties. Throughout this review, the feedstock, energy source, melt pool, defects, compositional distribution, microstructure, texture, residual stresses, and mechanical properties are examined from the points of view of monitoring, modeling, and statistical analysis. As with most active research subjects, there remain both possibilities and limitations, and these will be considered and discussed as appropriate.

Keywords: additive manufacturing; modeling; monitoring; statistics.

Figure 1

Schematic of an additive manufacturing process, separated into (a) processing and (b) material state, showing a—feedstock (powder, wire), b—energy source (laser or electron beam), c—melt pool, spattering and vapor plume, d—defects (spherical porosity, lack of fusion defects), e—compositional distribution, f—microstructure and texture, and g—residual stresses and distortion.

Figure 2

Schematic of an additive manufacturing process, separated into (a) processing and (b) material state, with an emphasis on monitoring, showing a—feedstock (powder, wire), b—energy source (laser or electron beam), c—melt pool, spattering and vapor plume, d—defects (spherical porosity, lack of fusion defects), e—compositional distribution, f—microstructure and texture, and g—residual stresses and distortion.

Figure 3

Schematic of an additive manufacturing process, separated into (a) processing and (b) material state, with an emphasis on modeling, showing a—feedstock (powder, wire), b—energy source (laser or electron beam), c—melt pool, spattering and vapor plume, d—defects (spherical porosity, lack of fusion defects), e—compositional distribution, f—microstructure and texture, and g—residual stresses and distortion.

Figure 4

Schematic of an additive manufacturing process, separated into (a) processing and (b) material state, with an emphasis on statistics (or availability of data for statistical analysis), showing a—feedstock (powder, wire), b—energy source (laser or electron beam), c—melt pool, spattering and vapor plume, d—defects (spherical porosity, lack of fusion defects), e—compositional distribution, f—microstructure and texture, and g—residual stresses and distortion.

Figure 5

Discrete element model (DEM) simulation of additive manufacturing powder rake over time. Reprinted with permission from [141].

Figure 6

Ashby-like diagram of volumetric energy density versus characteristic material temperature for various material types categorized by color. Recreated with permission from [181].

Figure 7

Examples of monitoring melt pool and vapor plume. (a) Vision-based in situ monitoring results of melt pool detection in LBPF, (b) effect of scan speed on vapor plume in LBPF, the scan direction is indicated by the red arrow. Figure 7a is reprinted from [66] under Creative Commons Attribution License (CC BY); Figure 7b is reprinted from [188] under Creative Commons Attribution License (CC BY).

Figure 8

Integrated thermal profile (left column) and solidification (right column) model at 327 μs (a,b), 674 μs (c,d), and 967 μs (e,f). Reprinted with permission from [18].

Figure 9

Hierarchy of mechanical properties with associated critical microstructural features (based on Maslow’s hierarchy of needs [223]).

Figure 10

In situ X-ray imaging showing defect formation and melt pool outline detection in LPBF. Figure is reprinted from [237] under Creative Commons Attribution License (CC BY).

Figure 11

Finite volume method (FVM) model of melt pool with keyholing and pore formation over time (a) t = 2.395 ms, (b) t = 2.4 ms, (c) t = 2.415 ms, (d) t = 2.43 ms, (e) t = 2.445 ms, (f) t = 2.45 ms, (g) t = 2.455, (h) t = 2.47 ms, and (i) t = 2.495 ms. Figure is reprinted from [155] under Creative Commons Attribution License (CC BY).

Figure 12

Schematic of laser-induced breakdown spectroscopy (LIBS) to optimize data collection of composition and temperature. Figure is reprinted from [82] under Creative Commons Attribution License (CC BY).

Figure 13

Experimental distortion detection setup used in a laser AM setup. Figure is reprinted from [120] under Creative Commons Attribution License (CC BY).

Figure 14

Modeling of residual stress in an additively manufactured sample at different layers. Figure is reprinted from [227] under Creative Commons Attribution License (CC BY).