PresentationPDF Available

State-of-the-art of in-process measurement methods for defect identification in metal powder bed fusion

Authors:

Abstract

A review and classification of in-process measurement methods used in powder bed fusion by their observable process signatures, and a global summary of the achieved results in the literature.
In-process
measurement
methods for defect
identification in metal
powder bed fusion
Afaf Remani
September 22nd, 2021
University of Nottingham
Framework
What is this SOTA about? 3
Termino l o gy
In-process: taken during the process or between manufacturing steps
In-line: taken on separate systems along the production line
In-situ: taken directly from the location of the process
On-machine: taken on the machine where the process is occurring
Off-machine: taken outside the location where the process
Off-line: taken after process completion and outside the process environment
What this SOTA is
A review of the in-process measurement methods used in powder bed fusion
A classification of methods by their measurable observable process signatures
A summary of achieved measurement results
What this SOTA is not
A comparative study of system performances
A classification of in-process methods measuring derived process signatures
Leach R K and Carmignato S 2020 Precision Metal Additive Manufacturing CRC Press.
Why this SOTA? 4
PBF processes are currently the AM processes with
the highest industrial maturity
The use of the most advanced industrial
measurement tools has been centered around PBF
Interesting research proposing practical solutions to
defect detection and quality control
Fast evolution of the state of the art
First studies
Feasibility analysis of measurement methods
Limited attempts to automatically detect
defects
Current studies
Generation of massive amounts of data
Usage of multiple sensors, toolkits and
machine learning
The number of papers published nearly doubled
between 2018 and now. 20% of the studies
published since 2018 presented a machine
learning tool to analyse in-process data
Grasso M, Remani A, Dickins A, Colosimo B and Leach R K 2021 In-situ measurement and monitoring methods for metal powder bed fusionan updated review Meas. Sci. Technol. 32 112001.
State of the art
Classification of measurable levels in PBF 6
Process signature: a quantity that can be measured during the
manufacturing process and that helps provide information about
process stability, underlying physical phenomena and onset defects.
Observable signatures
Derived signatures
Directly acquired in-process
Estimated via modelling
We classify in-process measurement levels in PBF as follows:
Level 0
Level 1
Level 2
Level 3
Level 4
Signals from sensors embedded into the AM
system (chamber control, automation, …)
Powder bed & printed slice (powder bed
inhomogeneity, geometrical distortions,
surface patterns, …)
Scan track (process by-products: spatters &
plume, hot/cold spots, heating gradients, …)
Melt pool (stability of size, shape, intensity,
temperature distribution, …)
Under the layer (melt pool depth, sub-
surface pore and crack formation, …)
Grasso M, Remani A, Dickins A, Colosimo B and Leach R K 2021 In-situ measurement and monitoring methods for metal powder bed fusionan updated review Meas. Sci. Technol. 32 112001.
Level 0 Process conditions 7
Level 0 measurements are taken via software that monitor chamber log signals.
Chamber log signals represent current and voltage levels, temperature and pressure signals,
powder dosing, etc
Variations in these log signals indicate changes in chamber conditions, which could be
correlated with the occurrence of defects
Pore occurrence varies systematically
with build height log signals
Abrupt/unexpected changes in log signals
indicate defects in the printed region
It is possible to extract rake sensor data
indicative of incorrect powder spreading
conditions
“Falcon” is a tool that analyses large multivariable
datasets generated by log signals
Steed C, Halsey W, Dehoff R, Yoder S, Paquit V and Powers S 2017 Falcon: Visual analysis of large, irregularly sampled, and multivariate time series data in additive manufacturing Comput. Graph. 63 50-64.
Layer time
Plate temperature
Power supply
Build state
Level 1 Powder bed and printed slice (1) 8
Level 1 measurements are taken using sensors that
cover the entire powder bed once or multiple
times per layer.
The expected data is a 2D/3D representation of
the powder bed and the printed slice
Level 1
In-line
coherent
imaging
Level 1 Blade-mounted scanners
Sensors from flatbed document scanners mounted on re-
coaters to collect powder bed full-field 2D scans
No distortions related to perspective or lighting
Due to their small depth of field, linear sensors cannot
detect variations in bed thickness
Tan Phuc L and Seita M 2019 A high-resolution and large field-of-view scanner for in-line characterization of powder bed defects during additive manufacturing Mater. Des. 164 107562.
Examples of irregularities detected using a flatbed scanner reconstruction
using focus level mapping
Blade-
mounted
scanners
Off-axis
imaging
Fringe
projection
0
Level 1 Powder bed and printed slice (2) 9
Level 1 Off-axis imaging
In the visible range
Performance was improved by varying lighting
Feature extraction and slice contouring were used
to classify irregularities
In-process images using different lighting conditions to show powder bed irregularities
In the near infra-red (NIR)/ infra-red (IR)
Commonly used in EB-PBF
NIR video frames transformed into layer images
Local pixel intensity variations were used as proxies
for defect presence
Examples of optical
tomography images of (a)
multiple cubic samples and
(b) a cylinder produced under
gas flow variation
Gobert C, Reutzel E, Petrich J, Nassar A and Phoha S 2018 Application of supervised machine learning for defect detection during metallic powder bed fusion additive manufacturing using high resolution imaging Addit. Manuf. 21 517-528.
Bamberg J, Zenzinger G and Ladewig A 2016 In-process control of selective laser melting by quantitative optical tomography, 19th World Conference on Non-Destructive Testing, 2016.
ab
Level 1 Powder bed and printed slice (3) 10
Level 1 Fringe projection
Single-view and multi-view setups
Height map resolution is determined by the spatial
frequency of the fringes and the geometric
arrangement of the hardware Height maps of samples with changing height wavelengths, acquired using a multi-view
fringe projection setup (FOV = (250 ×250) mm)
Development of a multi-view fringe projection system for in-process measurements
inside a Renishaw AM250 machine (FOV = (250 x 250) mm)
Example of a height map and a height surface profile gathered using multi-view fringe
projection (FOV = (150 ×150) mm)
Kalms M, Narita R, Thomy C, Vollertsen F and Bergmann R 2019 New approach to evaluate 3D laser printed parts in powder bed fusion-based additive manufacturing in-line within closed space Addit. Manuf. 26 161-165.
Dickins A, Widjanarko T, Sims-Waterhouse D, Thompson A, Lawes S and Leach R K 2020 Multi-view fringe projection system for surface topography measurement during metal powder bed fusion J Opt Soc Am A 9 B93-B105.
Remani A, Williams R, Thompson A, Dardis J, Jones N, Hooper P and Leach R K 2021 Design of a multi-sensor measurement system for in-situ defect identification in metal additive manufacturing. Proc. Euspen SIG: Adv. Prec. in Addit. Manuf.
Level 2 Scan track (1) 11
Level 2 measurements are taken at high temporal resolutions
to observe interactions between the laser beam and the
material (plume emissions), process by-products (e.g. spatter)
and the overall thermal history.
Level 2
Spatter captured using high-speed X-ray video imaging (top) and
plume emissions captured with high-speed IR video imaging
Guo Q, Zhao C, Escano L, Young Z, Xiong L, Fezzaa K and Chen L 2018 Transient dynamics of powder spattering in laser powder bed fusion additive manufacturing process revealed by in-situ high-speed high-energy X-ray imaging Acta Materialia 151 169-180.
Grasso M, Colosimo B 2019 A Statistical Learning Method for Image-based Monitoring of the Plume Signature in Laser Powder Bed Fusion Robot. Comput. Integ. Manuf. 57 103-115.
Level 2 Measurement of process by-products
Interactions between the plume and gas flow were observed
Spatter was tracked along its trajectory (position and velocity)
Light sources were used to cope with limited sensor sensitivity
Measurement
of process
heatmaps
Measurement
of process
by-products
Level 2 Scan track (2) 12
Level 2 Measurement of process heatmaps
Short wave IR imaging
Minimises emissivity-dependent measurement errors
Captures thermal features and signature variations
Medium and long wave IR imaging
High sensitivity even at very high temperatures
Captures transient and fast phenomena
Visible imaging
Does not allow actual temperature measurements
Pixel intensity values used as proxies for thermal gradients
Reconstruction of local cooling profiles, with (a) an IR image of the powder
bed, (b) the time over a threshold index computation and (c) the cooling
profile
2D map (left) and 3D reconstruction (right) of hotspots detected using
pixel intensity description and visible imaging during an EB-PBF build
Paulson N, Gould B, Wolff S, Stan M and Greco A 2020 Correlations between thermal history and keyhole porosity in laser powder bed fusion Addit. Manuf. 101213.
Grasso M, Valsecchi G and Colosimo B 2020 Powder bed irregularity and hot-spot detection in Electron Beam Melting by means of in-situ video imaging Manuf. Lett. 24 47-51.
Level 3 Melt pool (1) 13
Level 3 measurements are taken at very high resolutions
and at the highest level of detail to observe the melt pool
(also called a “weld pool”).
Level 3
High-speed video
imaging
Pyrometry
Level 3 Pyrometry
Employs pyrometry by means of one or multiple photodiodes
Suitable to measure melt pool radiation intensity
2D maps of melt pool intensities were constructed by
synchronising photodiode signals with laser spot coordinates
Example of a thermal distribution diagram captured using a dual-
wavelength pyrometer and functional data analysis to predict porosity
Khanzadeh M, Chowdhury S, Marufuzzaman M, Tschopp M and Bian L 2018 Porosity prediction: Supervised-learning of thermal history for direct laser deposition J. Manuf. Syst. 47 69-82.
Level 3 Melt pool (2) 14
Level 3 High-speed video imaging
Wide usage of co-axial NIR video imaging to capture and
filter out melt pool emissions
Off-axis video imaging was combined with high
magnification optics to view the melt pool
Melt pool metrics and cooling rates were estimated
a) Co-axial melt pool video frames in the visible range, b) melt pool surface temperature estimation via dual wavelength co-axial video imaging, c) melt pool
surface temperature estimation via off-axis video imaging for different process parameters
Yuan B, Giera B, Guss G, Matthews I and Mcmains S 2019 Semi-supervised convolutional neural networks for in-situ video monitoring of selective laser melting Proc: 2019 IEEE Winter Conference on Applications of Computer Vision (WACV) (pp. 744-753).
Hooper P 2018 Melt pool temperature and cooling rates in laser powder bed fusion Addit. Manuf. 22 548-559.
Zhirnov I, Mekhontsev S, Lane B, Grantham S and Bura N 2020 Accurate determination of laser spot position during laser powder bed fusion process thermography Manuf. Lett. 23 49-52.
Level 4 Under the layer 15
Level 4 measurements are taken with ad-hoc machine configurations to examine elastic
and plastic deformations in the solidified material.
X-ray video imaging was used to measure the
cross-section of the melt pool
Subsurface melt pool dynamics and pore
formation were observed
Acoustic emissions were used to measure elastic
energy releases (cracks and delamination)Apparatus of an in-process X-ray imaging setup
Example of an in-process X-ray video frame
Paulson N, Gould B, Wolff S, Stan M and Greco A 2020 Correlations between thermal history and keyhole porosity in laser powder bed fusion Addit. Manuf. 101213.
Guo Q, Zhao C, Qu M, Xiong L, Escano L, Hojjatzadeh S and Chen L 2019 In-situ characterization and quantification of melt pool variation under constant input energy density in laser powder bed fusion additive manufacturing process Addit. Manuf. 28 600-609.
In-process defect detectability 16
Single track defects
Microstructural
irregularities
Porosity and lack of
fusion
Geometrical distortions
Residual stresses,
cracks and delamination
Spatter ejections
Unexpected changes in plume salient features
Abnormal changes in the layer thermal gradient
Increased or decreased solid-liquid interface velocity
Voids and discontinuities of different sizes
Abrupt changes in pixel intensity or thermal values
Deviations in the slice from the nominal contour
Local mismatches in the monitored region of interest
Differences between sub-layer stress vectors
Sudden phase changes and energy releases
Reconstructed anomaly detection using
layerwise imaging in the visible range and
a convolutional neural network (CNN)
Scime L and Beuth J 2018 A multi-scale convolutional neural network for autonomous anomaly detection and classification in a laser powder bed fusion additive manufacturing process Addit Manuf. 24 273-286.
Level 2
+
Level 3
Level 2
+
Level 3
Level 1
+
Level 3
Level 1
+
Level 2
Level 4
+
Level 1
Process control 17
The repeatability of PBF processes is
dependent on possible levels of process
control to improve part quality:
For predictable defects
Pre-compensate for geometrical distortions
in the CAD model
Set locally varying process parameters and
scan strategies in the build file
For non-predictable defects
Combine in-process measurements and real-
time closed-loop adaptations of process
parameters at melt pool or track level
Correct or remove in-process defects after
they have been detected
Liu C, Le Roux L, Ji Z, Kerfriden P, Lacan F and Bigot S 2020 Machine Learning-enabled feedback loops for metal powder bed fusion additive manufacturing Procedia Comput. Sci. 176 2586-2595.
Conclusions
Open issues and future research directions 19
Challenges Research directions
Most findings were reported from the measurement
of seeded defects
More focus should be put on realistic scenarios
during which pores naturally form
Lack of methods suitable to detect local flaws with
acceptable false alarm rates
The root causes of false alarms (e.g. partial
remelting) need more attention
The majority of methods tackled single track
experiments or the production of simple specimens
Defect mechanisms in complex shapes should be
investigated along with methods to transfer this
knowledge to in-situ monitoring systems
Extensive usage of supervised learning paradigms,
that require large amounts of data and time-
consuming data labelling
Need for training procedures that are easily
implementable in industry, solutions for big data
mining and more adaptive “self
-
learning” paradigms
The majority of in-process data analysis was not
applied in real-time
Test in g ne ed s to o cc ur o n-
line and computational
efficiency needs improving
Many in-process sensing methods are difficult or
impossible to install on industrial machines
Need for more in-house developed systems that
reflect similar capabilities to those tested in in-
process systems
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Size and shape of a melt pool play a critical role in determining the microstructure in additively manufactured metals. However, it is very challenging to directly characterize the size and shape of the melt pool beneath the surface of the melt pool during the additive manufacturing process. Here, we report the direct observation and quantification of melt pool variation during the laser powder-bed fusion additive manufacturing process under constant input energy density by in-situ high-speed high-energy x-ray imaging. We show that the melt pool can undergo different melting regimes and both the melt pool dimension and melt pool volume can have orders-of-magnitude change under a constant input energy density. Our analysis shows that the significant melt pool variation cannot be solely explained by the energy dissipation rate proposed previously. We found that energy absorption changes significantly under a constant input energy density, which is another important cause of melt pool variation. Our further analysis reveals that the significant change in energy absorption originates from the separate roles of laser power and scan speed in depression zone development. The results reported here are important for understanding the laser powder-bed fusion additive manufacturing process and guiding the development of better metrics for processing parameter design.
Article
Full-text available
In laser powder bed fusion, melt pool dynamics and stability are driven by the temperature field in the melt pool. If the temperature field is unfavourable defects are likely to form. The localised and rapid heating and cooling in the process presents a challenge for the experimental methods used to measure temperature. As a result, understanding of these process fundamentals is limited. In this paper a method is developed that uses coaxial imaging with high-speed cameras to give both the spatial and temporal resolution necessary to resolve the surface temperature of the melt pool. A two wavelength imaging setup is used to account for changes in emissivity. Temperature fields are captured at 100 kHz with a resolution of 20 μm during the processing of a simple Ti6Al4V component. Thermal gradients in the range 5–20 K/μm and cooling rates in range 1–40 K/μs are measured. The results presented give new insight into the effect of parameters, geometry and scan path on the melt pool temperature and cooling rates. The method developed here provides a new tool to assist in optimising scan strategies and parameters, identifying the causes of defect prone locations and controlling cooling rates for local microstructure development.
Level 4 measurements are taken with ad-hoc machine configurations to examine elastic and plastic deformations in the solidified material
  • I Zhirnov
  • S Mekhontsev
  • B Lane
  • S Grantham
  • N Bura
Zhirnov I, Mekhontsev S, Lane B, Grantham S and Bura N 2020 Accurate determination of laser spot position during laser powder bed fusion process thermography Manuf. Lett. 23 49-52. Level 4 measurements are taken with ad-hoc machine configurations to examine elastic and plastic deformations in the solidified material.