Powder Characterization for Additive Manufacturing Processes Lisa Markusson Materials Engineering, masters level 2017 Luleå University of Technology Department of Engineering Sciences and Mathematics
Powder Characterization for Additive
Manufacturing Processes
Lisa Markusson
Materials Engineering, masters level
2017
Luleå University of Technology
Department of Engineering Sciences and Mathematics
i
Preface
This master thesis work has been carried out at GKN Aerospace Engine Systems Sweden at
the Department of Process Engineering in Trollhättan, Sweden. This is a degree project in
engineering materials performed as the final part of my Masters of Materials Sciences and
Engineering degree.
ii
To my grandparents
iii
Abstract
The aim of this master thesis project was to statistically correlate various powder
characteristics to the quality of additively manufactured parts. An additional goal of this
project was to find a potential second source supplier of powder for GKN Aerospace Sweden
in Trollhättan. Five Inconel® alloy 718 powders from four individual powder suppliers have
been analyzed in this project regarding powder characteristics such as: morphology, porosity,
size distribution, flowability and bulk properties. One powder out of the five, Powder C, is
currently used in production at GKN and functions as a reference. The five powders were
additively manufactured by the process of laser metal deposition according to a pre-
programmed model utilized at GKN Aerospace Sweden in Trollhättan. Five plates were
produced per powder and each cut to obtain three area sections to analyze, giving a total of
fifteen area sections per powder. The quality of deposited parts was assessed by means of
their porosity content, powder efficiency, geometry and microstructure. The final step was to
statistically evaluate the results through the analysis methods of Analysis of Variance
(ANOVA) and simple linear regression with the software Minitab.
The method of ANOVA found a statistical significant difference between the five powders
regarding their experimental results. This made it possible to compare the five powders
against each other. Statistical correlations by simple linear regression analysis were found
between various powder characteristics and quality of deposited part. This led to the
conclusion that GKN should consider additions to current powder material specification by
powder characteristics such as: particle morphology, powder porosity and flowability
measurements by a rheometer.
One powder was found to have the potential of becoming a second source supplier to GKN,
namely Powder A. Powder A had overall good powder properties such as smooth and
spherical particles, high particle density at 99,94% and good flowability. The deposited parts
with Powder A also showed the lowest amount of pores compared to Powder C, a total of 78
in all five plates, and sufficient powder efficiency at 81,6%.
Keywords: Powder Characteristics, Inconel 718, Additive Manufacturing, Laser Metal
Deposition, ANOVA, Regression Analysis
iv
Sammanfattning
Syftet med detta examensarbete var att statistiskt korrelera olika pulveregenskaper med
kvaliteten på additivt tillverkade delar. Ett vidare syfte med projektet var att finna en
potentiell andrahandsleverantör av pulver för GKN Aerospace Sweden i Trollhättan. Fem
pulver av materialet Inconel® 718 från fyra individuella pulverleverantörer har analyserats i
detta projekt gällande pulveregenskaper såsom: morfologi, porositet, storleksfördelning,
flytbarhet och bulkegenskaper. Ett pulver av de fem, Pulver C, används för nuvarande
produktion på GKN och fungerar som en referens. De fem pulvren har additivt tillverkats
enligt en förprogrammerad modell, framtagen på GKN Aerospace Sweden i Trollhättan,
genom processen ’laser metal deposition’. Fem provplåtar producerades per pulver och
provbereddes för att erhålla tre ytor för vidare analys, totalt femton ytor per pulver. Kvaliteten
på deponerat material bedömdes utifrån dess porositet, verkningsgrad av pulver, geometri och
mikrostruktur. Det slutliga steget var en statistisk analys av resultaten genom metoderna
Analysis of Variance (ANOVA) och enkel linjär regression med mjukvaran Minitab.
Metoden enligt ANOVA fann en statistisk signifikant skillnad mellan de fem pulvren
gällande dess egenskaper och experimentella resultat. Detta gjorde det möjligt att kunna
jämföra de fem pulvren mot varandra. Statistiska samband utifrån en enkel linjär
regressionsanalys erhölls mellan olika pulveregenskaper och kvalitativa resultat av deponerat
material. Detta ledde till slutsatsen att GKN bör överväga tillägg till nuvarande
pulvermaterialspecifikation med pulveregenskaper såsom: partikelmorfologi, pulverporositet
och flytbarhet genom mätningar av en reometer.
Ett pulver bedöms ha potential att bli en andrahandsleverantör till GKN Aerospace, nämligen
Powder A. Pulver A hade övergripande goda pulveregenskaper såsom jämna och sfäriska
partiklar, hög partikeltäthet på 99,94% och god flytbarhet. De deponerade proverna med
Pulver A uppvisade även lägst antal porer i jämförelse med Powder C, totalt 78 porer i alla
fem provplåtar, och godkänd pulververkningsgrad på 81,6%.
v
Acknowledgements
I would like to give my deepest gratitude to my supervisors during this master thesis project;
Lars Östergren at GKN Aerospace Sweden in Trollhättan and Marta-Lena Antti at Luleå
University of Technology. Their great knowledge and guidance have brought this project to
its best. Special thanks also goes to Jimmy Johansson at GKN Aerospace Sweden for sharing
his expertise throughout this master thesis project. The operators at GKN who produced the
experimental samples must not be forgotten and I thank you for the help and support. A final
acknowledgement for the staff at the Department of Process Engineering at GKN for making
these past months filled with laughter and new experience. Thank you.
Trollhättan, March 2017
vi
Table of Contents
1 Introduction..................................................................................................................... 1
1.1 Collaborate Company Presentation .......................................................................... 1
1.2 Background .............................................................................................................. 1
1.3 Goal .......................................................................................................................... 2
1.4 Scope ........................................................................................................................ 2
2 Literature Review ........................................................................................................... 3
2.1 Additive Manufacturing Processes .......................................................................... 3
2.1.1 Process Classifications ........................................................................................ 3
2.1.2 Advantages & Disadvantages .............................................................................. 4
2.2 Laser Metal Deposition with Powder ....................................................................... 6
2.2.1 Basis of Deposition Process ................................................................................ 6
2.2.2 Powder Nozzles & Feeder System ....................................................................... 7
2.2.3 Basic Deposition Geometry ................................................................................. 9
2.2.4 Process Parameters & Their Effect on Deposited Geometry ............................ 10
2.2.5 Heat Transfer, Solidification & Microstructure Characteristics ...................... 10
2.3 Nickel Based Superalloys ...................................................................................... 12
2.3.1 Inconel 718 ........................................................................................................ 13
2.4 Powder Manufacturing Processes .......................................................................... 14
2.4.1 Gas Atomization ................................................................................................ 15
2.4.2 Plasma Atomization ........................................................................................... 16
2.4.3 Plasma Rotation Electrode Process .................................................................. 16
2.5 Powder Characteristics ........................................................................................... 16
2.5.1 Morphology ....................................................................................................... 17
2.5.2 Porosity .............................................................................................................. 17
2.5.3 Size & Size Distribution ..................................................................................... 18
2.5.4 Rheology ............................................................................................................ 19
2.5.5 Bulk Properties .................................................................................................. 21
vii
2.5.6 Quality Assessment of Powder ........................................................................... 22
2.6 Statistical Significance ........................................................................................... 22
2.6.1 Analysis of Variance .......................................................................................... 22
2.6.2 Regression Analysis ........................................................................................... 24
3 Materials & Methods .................................................................................................... 25
3.1 Powder & Sheet Material ....................................................................................... 25
3.2 Powder Characterization ........................................................................................ 26
3.2.1 Morphology ....................................................................................................... 26
3.2.2 Porosity .............................................................................................................. 27
3.2.3 Particle Size Distribution .................................................................................. 29
3.2.4 Rheology ............................................................................................................ 30
3.2.5 Bulk Properties .................................................................................................. 32
3.3 Laser Metal Deposition .......................................................................................... 33
3.4 Deposit Evaluation ................................................................................................. 34
3.4.1 Sample Preparation ........................................................................................... 34
3.4.2 Defects ............................................................................................................... 34
3.4.3 Geometry ........................................................................................................... 34
3.4.4 Microstructure ................................................................................................... 35
4 Results & Discussions ................................................................................................... 36
4.1 Powder Characterization ........................................................................................ 36
4.1.1 Morphology ....................................................................................................... 36
4.1.2 Porosity .............................................................................................................. 42
4.1.3 Particle Size Distribution .................................................................................. 44
4.1.4 Rheology ............................................................................................................ 49
4.1.5 Bulk Properties .................................................................................................. 53
4.2 Deposit Evaluation ................................................................................................. 54
4.2.1 Powder Efficiency .............................................................................................. 54
4.2.2 Geometry ........................................................................................................... 56
viii
4.2.3 Defects ............................................................................................................... 59
4.2.4 Microstructure ................................................................................................... 62
4.3 Statistical Evaluation.............................................................................................. 68
4.4 Statistical Correlation ............................................................................................. 76
4.4.1 Median Particle Size – Basic Flow Energy ....................................................... 76
4.4.2 Median Particle Size – Powder Efficiency ........................................................ 79
4.4.3 ShapeFactor – Basic Flow Energy .................................................................... 81
4.4.4 ShapeFactor – Powder Efficiency ..................................................................... 83
4.4.5 Basic Flow Energy – Powder Efficiency ........................................................... 85
4.4.6 Basic Flow Energy – Multi-bead Height ........................................................... 87
4.4.7 Particle Pore Frequency – Deposit Pore Frequency ........................................ 89
4.4.8 Particle Pore Size – Deposit Pore Size .............................................................. 91
5 Conclusions .................................................................................................................... 98
6 Future Work................................................................................................................ 100
7 Bibliography ................................................................................................................ 101
8 Appendix ...................................................................................................................... 106
8.1 Particle Pixel Area Fraction ................................................................................. 106
8.2 Particle Pore Diameter ......................................................................................... 107
8.3 Summary of Pore Data in Powder & Deposited Part ........................................... 108
8.4 Morphology .......................................................................................................... 109
8.5 Rheometer ............................................................................................................ 109
8.6 Part Geometry ...................................................................................................... 110
8.7 Statistical Evaluation............................................................................................ 111
ix
Abbreviations & Nomenclature
Name Abbreviation
Additive manufacturing AM
Aluminum Al
American society for testing and materials ASTM
Analysis of variance ANOVA
Chromium Cr
Electrode induction melting gas atomization EIGA
Face-centered cubic FCC
Gas atomization GA
General electric GE
Hot isostatic pressing HIP
Iron Fe
Laser metal deposition LMD
Laser metal deposition with powder LMD-p
Nickel Ni
Niobium Nb
Optical microscope OM
Particle size distribution PSD
Plasma Atomization PA
Plasma rotation electrode process PREP
Scanning Electron Microscopy SEM
Tantalum Ta
Three-dimensional 3D
Titanium Ti
Vacuum Inert Gas Atomization VIGA
x
Parameter Symbol Unit
Apparent density g/cm3
Aeration energy mJ
Aeration ratio -
Basic flowability energy mJ
Bead area mm2
Conditioned bulk density g/cm3
Consolidation energy mJ
Consolidation index -
Depth of penetration mm
Deposition height mm
Deposition width mm
Flow rate index FRI -
Hall Flow rate s/50g
Heat flux W/m2
Laser power W
Powder efficiency %
Powder mass delivered g
Powder mass deposited g
Powder mass flow rate g/min
Root angle α °
Scanning speed mm/min
Specific energy mJ/g
Stability index SI -
Tap density g/cm3
1
1 Introduction
The worldwide success of the emerging technology additive manufacturing (AM) is due to
the exceptional opportunity to produce complex near-net-shapes in a single process. By
adding material layer upon layer, a preprogrammed three-dimensional (3D) model is formed
without extensive subtracting methods associated with conventional production. Additive
manufacturing has great future promises and comprehensive research is put in to the
technology. The key to the success of AM will be to understand the relationship between
process variables, material properties and final structure.
1.1 Collaborate Company Presentation
GKN Aerospace is one out of four divisions of GKN PLC, a British company in the forefront
of global technology. GKN has a long industrial heritage which can be traced back to late 18th
century to a small iron work on the hillside of Welsh. For the last two-and-a-half centuries
they have been one of the leaders in industrial evolution.
GKN Aerospace is one of the world’s largest independent suppliers to the aviation industry,
both commercial and military. With three main product areas; Aerostructures, Engine
components/sub-systems and Special products, they are the global market frontier in all three
areas. GKN Aerospace Engine Systems is a subdivision to GKN Aerospace and is located
over four continents with its headquarter in Trollhättan, Sweden. Today, 90% of all
commercial flights take off every day with technology supplied from the division of Engine
Systems (GKN Aerospace, n.d.).
1.2 Background
Additive manufacturing is a suitable process to restore and strengthen components where the
technology of laser metal deposition (LMD) is used in the facilities at GKN Aerospace
Engine Systems Sweden in Trollhättan. In order to produce structures by laser metal
deposition of sufficient standard, the initial powder need to meet proper quality. It is therefore
of importance to understand the impact of various powder characteristics on the final
deposited structure. The knowledge of powder characteristics and their correlation to
deposited structures is of importance to GKN’s further development of the technology in
order to enhance future production capabilities. Nickel based superalloys are commonly used
materials for high-performance aerospace applications which will be the material of interest
in this project.
2
1.3 Goal
The goal of this project is to find a powder characteristics guideline, based on statistical
evaluation, to review on powder batches to appraise an output of certain quality. The goal is
also to find a potential second source powder supplier. The main two questions to answer
during this project are:
Do some powder characteristics have a statistical significant impact on part
quality?
Is there a powder supplier that shows the potential of becoming a second source to
GKN?
1.4 Scope
The initial scope of this project is to distinguish and evaluate powder characteristics of
interest. This follows by laser metal deposition of standardized structures and evaluation of
their final quality. The aim is to identify and discuss any statistical evaluation and correlation
between the characteristics of powder and final part quality.
3
2 Literature Review
This chapter will provide the theory necessary to set a foundation to the project. An
introduction to additive manufacturing processes is the first area in subject during this
literature review. This follows by a deeper look into the specific additive manufacturing
process for this project, laser metal deposition with powder (LMD-p). A section on the
material used in this project, a nickel based superalloy, is also presented. A review on powder
characteristics and quality assessment of powders with the LMD process in mind follows. The
statistical importance and appropriate means of assessment is presented as a final part.
2.1 Additive Manufacturing Processes
Additive manufacturing is a collective term for various fabrication techniques whereby
material is joined with a layer-on-layer approach to produce a preprogrammed 3D data model.
Other well-known terms to the common mass are 3D printing, additive layer manufacturing,
rapid prototyping and freeform fabrication. The term of 3D printing is more common in
commercial contexts whereas additive manufacturing is more referred to in industry. This is a
technology that has attained a lot of research and development within numerous sectors such
as aerospace, automotive, medical and consumer goods (EPMA, 2015).
2.1.1 Process Classifications
The standard definition of additive manufacturing established by the American society for
testing and materials (ASTM) F42 Technical Committee on Additive Manufacturing
Technologies is the ‘process of joining materials to make objects from 3D model data, usually
layer upon layer, as opposed to subtractive manufacturing methodologies’. The ASTM F42
committee has also classified these methods by means of the process baseline where seven
major processes have been identified. Three of the seven categories are applicable to construct
parts by metal in a single-step AM process which is mainly of interest in this project, see
Figure 1. Level one in the diagram shows the three process categories: Powder bed fusion,
Direct energy deposition and Sheet lamination. Level two specifies the material distribution
type followed by the source of fusion in level three. The fourth and final level specifies the
type of material used as feedstock (ASTM International, 2015).
The boxes in blue highlights the process classification that is of interest in this project; Direct
energy deposition also called laser metal deposition (LMD). The process of laser metal
deposition with powder (LMD-p) will be further explained in section 2.2. The two processes
that implement powder as a raw material are Powder bed fusion and Direct energy deposition.
The fundamental difference between these two methods is the way in which the powder is
introduced. For Powder bed fusion a powder bed is pre-laid over the substrate whereby the
4
laser beam moves over the surface and fuses the two together. With Direct energy deposition
the powder is fed directly onto the surface whereby the laser beam simultaneously melts the
powder and surface layer and creates a metallurgical bond. Both methods create thin tracks of
rapidly solidified material, also referred to as beads, resulting in high-density structures close
to 100% (Gibson, Rosen & Stucker, 2010).
Figure 1. An overview of the various single-step AM processes for metallic materials (modified from
ASTM International, 2015).
2.1.2 Advantages & Disadvantages
The many advantages of additive manufacturing are the reasons for this growing technology
within industrial sectors worldwide. Some general advantages of AM are the following:
Less Material Waste
This is a great advantage for the aerospace industry where the volumes of high-cost
materials, such as titanium and superalloys, will drastically be reduced. This is simply
due to the AM approach of bottom-up manufacturing rather than a production of
subtracting nature from large billets, which reduces material use and waste
substantially (Royal Academy of Engineering, 2013). Up to 60% can be saved by
LMD by reduced material waste (Ford & Despeisse, 2016). For parts with a high buy
to fly ratio, i.e. the weight ratio between initial billet and finished part, AM will play a
major role in reducing material uses and costs (Allen, 2006).
Shorter Lead-Times
Another ground for cost savings compared to conventional manufacturing is the basis
of the single–step process and its resulting shorter production lead-time, from design
5
to finished product (Ford & Despeisse, 2016). The lead-time is believed to be reduced
by up to 80% compared to conventional production methods within the aerospace
industry (SmarTech Market Publishing, 2014).
Low-Volume Customized Production
For customized metal parts in small or medium size batches, AM is a suitable
production method. Production of complex shaped parts that requires extensive and
difficult machining will profit from AM (Baumers, 2012). The initial investment cost
is large in terms of the actual machinery, but the lack of moulds and additional tooling
is a large beneficiary. Due to the simplicity of the AM process, it is also easier to
make design changes or produce different products with the same equipment, simply
change the input 3D data model. This may boost a more product innovative business
environment within a company. With conventional methods a change in product
design might require new moulds and/or tooling, an expensive adjustment (Beiker
Kair & Sofos, 2014).
According to General Electric (GE) Aviation, the production of fuel nozzles by AM will
reduce production cost with up to 75%. This cost reduction is thanks to the advantage of non
extensive assembly, i.e. a single-step process with less material used. There are big
expectations and believes that additive manufacturing will reach a more standardized product
market with an economy of scale that can compete against conventional manufacturing
(D’Aveni, 2015). But the challenges today are many such as:
Low Deposition Rates
When it comes to the production rate for high-volume production, AM does not
challenge conventional methods today. To allure industry sectors with larger
production volumes, deposition rates of AM need to increase (Baumers, Dickens,
Tuck & Hague, 2016).
Feedstock Material
The building materials available on the market are still somewhat limited and research
need to be put in to develop and standardize new materials of sufficient quality (Ford
& Despeisse, 2016). Even though less material is used, the price of raw material is still
high. But as the technology enhances and the market for AM grows, both machinery
and raw material prices will hopefully drop (SmarTech Market Publishing, 2014).
6
Reliability & Quality
AM is still a novel and emerging process where the reliability and applicability to
produce parts in high-demanding industry sectors needs research. Research is needed to
understand material behavior, process liability, structure properties and the relationship
between these areas (Chen, He, Yang, Niu & Ren, 2016). This also goes hand in hand
by standardizing materials and various technologies to ensure the quality and reliability
of AM. This is important for already investing business sectors and to appeal others to
invest and implement (Royal Academy of Engineering, 2013).
2.2 Laser Metal Deposition with Powder
2.2.1 Basis of Deposition Process
The process of laser metal deposition creates surface layers or near-net-shaped 3D structures
by laser fusion of powder and substrate. The powder material is fed onto the surface through a
nozzle by an inert carrier gas, normally argon or nitrogen, and is completely melted by the
laser beam, see Figure 2. The laser beam simultaneously melts a thin surface layer and a
metallurgical fused bond is formed between the two (Fraunhofer ILT, 2012). One of the great
advantages of laser metal deposition is a small heat affected zone due to the low heat input.
The system of feeding powder to the laser beam focus is either through a coaxial or off-axis
nozzle (Hauser, 2014). The complete system of powder nozzle, laser system and inert gas
tubes are referred to as the deposition head. The relative movement of the deposition head and
work piece is performed by a multi-axis robot or gantry system. How the work piece and
deposition head moves relatively to each other can vary; the substrate and deposition head
may move simultaneously or the substrate moves and tilts with the deposition head at a more
fixed position and vise versa (Gibson, Rosen & Stucker, 2010).
Figure 2. Schematic illustration of LMD-p with a coaxial powder nozzle (Frank, 2016).
7
2.2.2 Powder Nozzles & Feeder System
With a coaxial powder nozzle the powder flows directly into the laser beam assisted and
protected by the inert carrier gas. The powder stream flows through a conically shaped outlet
with an annular gap, Figure 3, or multiple outlets, Figure 4. A conically shaped nozzle is
constructed by a concentrically mounted inner and outer cone creating a defined offset
between the two. In case of multiple outlets, the powder flows through channels inside the
nozzle.
Figure 3. Coaxial powder nozzle with a conformed
annular gap (Fraunhofer ILT, 2014a).
Figure 4. Coaxial powder nozzle with multiple
outlets (Fraunhofer ILT, 2014b).
The off-axis system feeds the powder in a lateral position to the laser beam. The powder
efficiency of the system highly depends on the angle and distance between the nozzle and
work piece as well as the relative movement of powder stream and work piece. This nozzle is
therefore more suitable for surface cladding and not high precision printing (Poprawe, 2011).
A simple comparison of the presented nozzle systems are seen in Figure 5. A higher precision
is achieved with an annular gap but this comes with the cost of a lower deposition rate.
Figure 5. Comparison of the different powder nozzle systems (modified from Hauser, 2014).
A common powder feeding system is based on the principal of a rotating feed disk with a
rectangular annular groove, see Figure 6 for illustration. The powder is contained in the so
called hopper with an incorporated stirrer which rotates during feeding to ensure a continuous
flow. The powder is delivered from the hopper down to underlying container onto the rotating
8
powder disk. The annular groove is filled with powder in a controlled and consistent manner
by the spreader unit under simultaneous rotation of the disk. Under rotation the powder is
transported from the spreader to the opposite side of the disk where a suction unit is placed.
This nozzle-shaped suction unit is attached to a hose that enables the delivery of powder from
the feeding unit to the deposition nozzle. The powder is emptied from the annular groove, and
its incorporated suction unit, by suction with carrier gas. The carrier gas is introduced to the
feeding system from the bottom whereby the container is slightly pressurized during
operation. This slight pressure forces the carrier gas to be exhausted from the suction unit,
with powder, to the delivery hose. During one full disk rotation the filled groove is emptied of
powder. The powder mass flow rate is then consequently controlled by the number of
rotations per minute of the disk. A volumetric method is used to control an accurate and
stable amount of delivered powder to the system. This means that a scale is incorporated with
the powder hopper to measure the continuous weight loss of powder (GTV, n.d.).
Figure 6. Illustrations of a powder feeding system with a rotating feed disk (top) as well as a closer
view of its suction and spreader units (bottom) (Oerlikon Metco, 2016).
9
2.2.3 Basic Deposition Geometry
The deposition of a single track, also referred to as a bead, is seen in Figure 7. The bead
height (h), width (w), root angle (α) and bead area (A) is illustrated (Zhang, Li & Deceuster,
2011). When depositing multi-bead tracks the beads can have varying degrees of overlap, see
Figure 8 for illustration. The centre distance between adjacent beads (d), and consequently its
degree of overlap, is of importance when considering the area of valley generated between the
beads and layers. An increase in centre distance between two beads will increase the area of
valley and thereby generate a less smooth and flat surface and potential lack of fusion.
Deposition of a multi-bead and multi-layer structure can also be deposited with various
degree of overlap between individual beads and subsequent layers. An example of a multi-
layer deposit with subsequent layers positioned with an overlap close to 0% is seen in Figure
9. The subsequent layers can also be deposited with a 50% overlap as seen in Figure 10
(Ding, Pan, Cuiuri & Li, 2015).
Figure 7. Illustration of single-bead
geometry (Zhang, Li & Deceuster, 2011).
Figure 8. Schematic illustration of basic overlap between
multi-bead deposits (Ding, Pan, Cuiuri & Li, 2015).
Figure 9. Experimental result of multi-bead and multi-layer wire deposition (Ding, Pan, Cuiuri & Li,
2015).
Figure 10. 3D model (left) and illustration (right) of a multi-layered structure with a 50% longitudinal
and 50% transverse overlap (Zhang, Li & Deceuster, 2011).
10
2.2.4 Process Parameters & Their Effect on Deposited Geometry
The final geometry of a deposited structure is highly dependent on various input process
parameters such as:
Laser power ( )
Scanning speed (
Powder mass flow rate ( )
Zhong, Biermann, Gasser and Poprawe (2015) have shown that an increase in scanning speed
decreases the single track width and height, due to the smaller amount of powder fed per unit
length of deposit. The same result in track height and width is seen if only the laser power is
increased. On the other hand, if powder mass flow is increased the track height instead
increases, along with the root angle. Consequently, by increasing powder mass flow, a higher
amount of laser is absorbed and a smaller melt pool is created. This dependency of powder
mass flow on track width and height was also showed by Ahsan, Pinkerton, Moat and
Shackleton (2011) but with a more pronounced decrease in track width, i.e. more increased
root angle.
Powder efficiency ( , i.e. the percentage of blown powder actually deposited, has been seen
to increase with increasing laser power. If more power is put into the system, more powder is
able to melt. However, if the laser power is too large, the amount of energy will melt deeper
into the substrate causing dilution of material and thereby lower the deposition height, with
respect to a constant powder mass flow (Mahamood, Akinlabi, Shukla & Pitvana, 2013).
2.2.5 Heat Transfer, Solidification & Microstructure Characteristics
The molten pool created during deposition conveys its heat to the substrate, the build material
and through the shield gas. The kinetics of heat transfer of the material determines the grain
growth morphology and orientation. The solidification kinetics is dependent on the melt pool
geometry which is influenced by the speed and power of the laser beam (Sames, List,
Pannala, Dehoff & Babu, 2016). The heat flux of the molten pool ( ) is the resultant of the
horizontal ( ) and vertical heat ( ) fluxes, see Figure 11. The vertical heat flux is generated
by the heat loss to the substrate whereby the horizontal flux is generated by the moving laser
beam. Grain growth is then evidently in the opposite direction of the resultant heat flux ( ).
This generally generates epitaxial columnar grains grown from the substrate parallel to the
scanning direction. The primary dendritic spacing is also proportional to the cooling rate
where a finer microstructure is obtained with a higher cooling rate (Zhong et al., 2016).
11
Figure 11. Illustration of longitudinal section of a deposited track and the resultant heat fluxes (Zhong
et al., 2016).
The grain orientation is also dependent on the scanning path of a multi-layer deposit. Typical
scanning paths are unidirectional and bi-directional as seen in Figure 12. The section where
the beads starts and ends is referred to as start-and-stop in the process. These deposition paths
may create different amounts of time for cooling between the layers. A forced time for
cooling occurs with a unidirectional scanning path when the laser focus moves back to the
starting point before depositing the following layer. This forced cooling time is lost for bi-
directional paths as the deposition head continuously build the next layer. Parimi,
AswathanarayanaSwamy, Clark and Attallah (2014) studied the influence on grain orientation
of the two different deposition strategies. It was clearly shown that the grain orientation aligns
with the moving energy input, giving an alternating orientation whit a bi-directional, see
Figure 13(a,b). The unidirectional layers were oriented at an angle of 50-60° with respect to
the substrate. In case of bi-directional layers, they were oriented at an angle of 45-50° with
respect to the substrate. In both cases the grains re-nucleate from previous layers. This shows
that the orientation is not only dependent on the vertical and horizontal heat fluxes but also on
the orientation of previous layer. Similar grain orientation and angles between adjacent layers
were found in the work by Wei, Mazumder & DebRoy (2015) with 60° and 45° between
layers with respect to the horizontal plane. Parimi et al. (2014) also showed the presence of a
banded structure, indicated by white arrows in Figure 13(a,b), where a finer equiaxed grain
zone is found between the layers. This banded structure is more evident for unidirectional
scanning than bi-directional due to a lower temperature between the layers, as explained with
the forced cooling time. These fine grained zones also minimize and finally disappear when
moving further away from the substrate, i.e. the heat sink effect is subsequently lost. The
grain morphology development is also greatly affected by laser power. Parimi et al. (2014)
also increased the laser power, from 390 W to 910 W, with a bi-directional scanning. This
resulted in long columnar grains continuously growing from previous layer with an angle of
80°, see Figure 13c. The higher energy input gave no inter-layer fine grained zones due to
insufficient cooling rates, but a clear demarcation was found at the top of the deposit with
small columnar grains.
12
Figure 12. Typical scanning paths a) unidirectional and b) bi-directional (Parimi,
AswathanarayanaSwamy, Clark & Attallah, 2014).
Figure 13. OM images showing the grain orientation between layers for a) unidirectional b) bi-
directional and c) bi-directional scanning with high laser power (Parimi, AswathanarayanaSwamy,
Clark & Attallah, 2014).
2.3 Nickel Based Superalloys
Due to the demanding environment of many aerospace components, mainly the gas turbine
engine, the aerospace sector is the main market for these nickel based superalloys. Aerospace
engines have high demands on mechanical properties at elevated temperatures, generally
speaking above 70% of a materials melting temperature. In the hottest regions of the jet
engine, an operating temperature up to 1300°C can be reached. Nickel (Ni) based superalloys
13
are rare materials which can match these demands with their high-temperature creep strength
as well as a high oxidation and corrosion resistance during long time periods. The main
contribution to the hot corrosion resistance is the high levels of chromium (Cr) added to
superalloys with help by titanium (Ti). Due to the high market price of these alloys, the
possibility to lower material usage by AM is of particular interest. The parts of a jet engine
which are exposed to the highest temperatures such as the combustion chamber, turbine
blades and exhaust nozzle, are dependent on these high-performing superalloys.
The microstructure of a general Ni-based superalloy is mainly constituted by two phases, a
face-centered cubic (FCC) -nickel matrix and ordered distributed phase, i.e
with a FCC crystal structure, see Figure 14. Formation of these precipitates are the key
strengthening mechanism for many superalloys and the addition of tantalum (Ta), titanium
and niobium (Nb) promotes this phase. A Ni-based superalloy generally has a composition
of 40-50 wt% Ni along with other various alloying elements (Mouritz, 2012).
Figure 14. Schematic illustrations of the two main crystal structures in a nickel based super alloy: FCC
γ-nickel (left) and (right) (modified from Aveson, 2011).
2.3.1 Inconel 718
One of the most commercially known superalloys is Inconel 718 which is mainly used at
lower service temperatures up to 750°C. Niobium is added whereby the formation of body-
centered tetragonal ordered precipitates are the main strengthening mechanism.
Strengthening by formation of is also of importance. However, the
strengthening phase is only stable up to 649°C and a longtime exposure above this
temperature may transform the phase to a more unfavorable -phase, orthorhombic , as
a result of overaging. A smaller amount of this phase is beneficial due to grain refinement
effect and control. Carbides are also important secondary phases in terms of strengthening
(Donachie & Donachie, 2002). The so called intermetallic Laves phases may also form which
are hexagonally close-packed phases with a general form of
and is detrimental in larger amounts to the structure. These intermetallic Laves phases form
Ni
Ni
Al, (Al,Ti)
14
by segregation of Nb in the material and is reported to be visual as bright inter-dendritic
phases (Radhakrishna & Rao Prasad, 1997).
The microstructure of a 718 alloy deposited by laser differs compared to conventionally
casted parts with subsequent heat treatment due to different thermal history of the material. A
casted alloy 718 with subsequent hot isostatic pressing treatment (HIP) is viewed in Figure 15
(Lee, Chang, Tang, Ho & Chen, 2006). The microstructure of a laser metal deposited alloy
718 is seen in Figure 16. This image shows a resultant dendritic structure for a single track
with long columnar grains. The red marks in the image displays how the geometry of a
single-bead is measured (Zhong et al., 2016).
Figure 15. OM image (left) and SEM image (right) of a casted and HIP treated alloy 718 viewing (Lee,
Chang, Tang, Ho & Chen, 2006).
Figure 16. Viewed etched microstructure of a single track deposited alloy 718 (Zhong et al., 2016).
2.4 Powder Manufacturing Processes
How the powder will behave and spread under deposition is very much influenced by the
manufacturing process of the raw powder. This is influenced by the quality and size
distribution of the powder. There are various techniques to produce metal powder, but the
most common methods to produce high quality powders suitable for LMD are gas
15
atomization (GA), plasma atomization (PA) and plasma rotating electrode process (PREP). In
common for all these methods is that the particle size distribution (PSD) is usually under
control by the manufacturer through process control and sieving as a final part of the process
(Dawes, Bowerman & Trepleton, 2015).
2.4.1 Gas Atomization
Gas atomization is a common and well established method for producing desired spherical
particles. A raw material is placed in a top chamber, a furnace, where the material is melted.
Molten metal then enters the chamber below, either through a tundish or directly, where a
high-pressure jet stream of inert gas atomizes the melt. The metal droplets solidify and are
collected in a bottom chamber, see Figure 17. Due to the high demands on powder purity in
aerospace, melting is usually performed by vacuum induction or electrode induction melting
furnaces, see Figure 18 and Figure 19. Vacuum inert gas atomization (VIGA) produces
refined and degassed melts where the melt pours through a tundish nozzle to the atomization
chamber. Electrode induction melting gas atomization (EIGA) uses raw materials in terms of
rods. The rod is fed, rotated and melted by an induction coil. The melt directly enters the
atomization chamber with no use of a tundish. This process reinsures no contact with a
melting crucible minimizing the risk of contamination for reactive alloys (ALD Vacuum
Technologies, n.d.). This results in spherical particles with a wide particle size distribution of
0-500 µm. This PSD can be in a more narrow range by controlling the gas flow. However,
producing powder with GA normally generates so called satellites. Theses satellites are
surface irregularities consisting of smaller particles adhered to the surface of larger ones, a
less attractive characteristic (Dawes, Bowerman & Trepleton, 2015).
Figure 17. Basic illustration of the gas atomization process (Dawes, Bowerman & Trepleton, 2015).
16
Figure 18. Illustration of a tilting crucible (left) and a
bottom pouring crucible (right) used for a vacuum
induction melting furnace (ALD Vacuum Technologies,
n.d.).
Figure 19. Illustration of the induction
coil used for an electrode induction
melting furnace (ALD Vacuum
Technologies, n.d.).
2.4.2 Plasma Atomization
Plasma atomization is a process that produces highly spherical particles of good quality by
raw material with a high melting point. A feedstock in terms of wire is fed into a chamber
whereby it is melted by a plasma torch. The melt simultaneously atomizes in a low vacuum,
inert gas environment inside the chamber. The metal droplets then solidify when freely falling
in the chamber. This process also minimizes potential impurities of the material since the
atomized droplets never comes in contact with any solid surface during solidification. The
range of particle size distribution produced is between 0-250 µm with the greater part of
particles ranging between 0-106 µm (AP&C, 2015).
2.4.3 Plasma Rotation Electrode Process
This process is similar to plasma atomization and shows comparable powder quality. The
difference between them two is that PREP uses a feedstock bar whereby the material rotates
as it is melted. This leads to that the molten material atomizes and rapidly solidifies by
influence of centrifugal forces. This process is also performed in an inert gas environment
inside the chamber minimizing contamination and oxidation. The range of particle sizes
produced is between 0-100 µm (Dawes, Bowerman & Trepleton, 2015).
2.5 Powder Characteristics
There is a need to understand and assess more knowledge about powder characteristics and its
impact on the AM process. Powder characteristics such as morphology, particle size and
distribution, flowability, bulk properties and porosity are important to assess in order to
ensure powder performance in the machine and final structures. Evenly spreading of powder
during deposition is crucial to deposit uniform layers (Zlotwinski & Garboczi, 2015). ASTM
17
International has provided a technical standard guide to characterize metal powder in purpose
for additive manufacturing (ASTM International, 2014).
2.5.1 Morphology
The particle morphology is an important characteristic that influence laser metal deposition in
terms of flowability and packing density. This characteristic is dependent on the
manufacturing process of powders where gas atomization, plasma atomization and plasma
rotation electrode process are common methods as described in previous section 2.4. These
processes produce mainly spherically shaped particles. In general PA and PREP
manufactured powders show a higher quality in terms of a highly spherical shape and fine
surfaces with less shape irregularities. These surface irregularities are referred to as satellites
which are more widely seen for GA manufactured particles (Ahsan, Pinkerton, Moat &
Shackleton, 2011). These satellites are formed due to a difference in solidification rate
between smaller molten particles that adheres to partially molten particles of a larger size
(Zhong, Biermann, Gasser & Poprawe, 2015).
Particle morphology is easily studied through investigation by scanning electron microscopy
(SEM). A clear comparison between GA and PREP manufactured powders are seen in Figure
20.
Figure 20. SEM images showing comparison in particle morphology of GA (left) and PREP (right)
produced Ti6A4V powders (Ahsan, Pinkerton, Moat & Shackleton, 2011).
2.5.2 Porosity
The porosity of powders, both internal and on particle surfaces is an unwanted characteristic
that will affect the degree of porosity on deposited structures. This is greatly influenced by the
chosen manufacturing method. GA produced powders have been seen to have a higher
porosity than equivalent powder manufactured by PREP. This higher degree of porosity can
be explained by entrapped gas inside the particle due to the nature of the GA process. The
18
surface pores possibly generated by GA, seen in Figure 20, are usually non-existing for the
PREP powder seen in the same figure. The internal porosity can be viewed by optical
microscopy (OM) on samples with a polished cross-section, see Figure 21, where it is
apparent that GA powders experience a lower quality (Zhong et al., 2016).
Figure 21. OM image of polished cross-sections of GA (left) and PREP (right) produced Inconel 718
powders viewing internal porosity (Zhong et al., 2016).
A non-destructive imaging technique available for studying internal porosities is micro
computed tomography which 3D scans the sample by means of x-ray to detect any pores. This
method can be implemented to study both raw powder (Ahsan, Bradley & Pinkerton, 2011)
and deposited structures (Khademzadeh, Carmignato, Parvin, Zanini & Bariani, 2015).
2.5.3 Size & Size Distribution
Produced powder will have various size ranges dependent on the method of production as
mentioned in section 2.4. The desired powder size distribution depends on the process of
application whereby the general desired range for LMD processes is between 50-150 µm i.e.
particle diameter. The particle size will affect the powders ability for flow, spread and
packing during deposition (EPMA, 2015).
Powder size distribution can also affect the powder efficiency as showed by Kong, Carroll,
Brown & Scudamore (2007). They deposited Inconel 625 and found the highest powder
efficiency, and evidently highest layer deposition, with a size range of 44-88 µm and median
value around 74 µm. The use of smaller particles showed a lower efficiency most likely due
to coagulation inside the nozzle, disturbing the mass flow. A decrease in efficiency was also
shown for too large particles due to a resulting powder beam focus larger than laser beam
focus, consequentially not able to melt all powder.
There are several techniques to find the size distribution such as dynamic image analysis,
laser diffraction and dry sieving. The shape of the particles to investigate may affect the
19
choice of method. Many of the available techniques operate with the physical assumption of
spherical particles, such as laser diffraction and sieving. The particle size distribution is often
viewed as a histogram with the particle diameter on the x-axis, given in bin ranges, with the
frequency and cumulative percentages of the sizes on the y-axis (Horiba Instruments, 2016).
2.5.3.1 Laser Diffraction
Laser diffraction results are often viewed as a volume distribution where the distribution
width on the x-axis shows the values, Dv10, Dv50 and Dv90, where the lowered v simply
stands for volume distribution. Dv50 is the particle diameter where 50% of the population lies
below this value and the other 50% above, i.e. the median of a distribution. Dv10 and Dv90 is
the particle diameter under which 10 and 90% of the population lies. In general the median
value of a volume distribution, i.e. Dv50, is the most reported value when a single value of the
size distribution is described. A measure of the distribution width is often given by the span
and is defined as (Horiba Instruments, 2016):
2.5.3.2 Image Analysis
Image analysis is a tool that measures each individual particle giving a number distribution of
the sample, which can be performed on a static or dynamic basis. Values of Dn10, Dn50 and
Dn90 along with the mean are also often reported with image analysis, where the lowered n
stands for number distribution. This number distribution can often be converted to a volume
distribution in order to compare with other techniques, which is recommended if possible. In
general the median value of a number distribution, i.e. Dn50, is the most reported single value
of a size distribution (Horiba Instruments, 2016).
2.5.4 Rheology
A powder’s ability to flow is an important property for additive manufacturing since it affects
powder supply through the powder feeder and nozzle and ultimately production rate,
spreading and packing ability. Flowability is very much affected by the particle size and
distribution since the basis of powder flow is due to surface friction between particles. This
gives that a finer powder has more apparent surface area and resultantly higher interparticle
friction and lower flow characteristics. This also includes the presence of satellites; a higher
degree of satellites may have an effect on the flow due to mechanical interlocking between
surface irregularities. Another aspect that affects this interparticle friction is the moisture
content of the powder; higher moisture content increases friction and may result in larger
particle agglomerates and lowers the flowability (Slotwinski & Garboczi, 2015).
20
2.5.4.1 Hall Flowmeter
One standardized method of measuring powder flowability is based on the time it takes for a
particular mass to freely flow through a so called Hall Flowmeter funnel. A mass of 50 g is
placed in the funnel, a cone-shaped tool, whereby the powder is released, timed and captured
in a container below. The Hall Flow rate ( ) is given in the units of time per sample mass,
whereby the release of powder can be on a static or dynamic basis (ASTM International,
2013a). The level of sensitivity of this method is low, but it might give an indication of
flowability if powders are compared to each other.
2.5.4.2 Rheometer
Another method to describe the flowability is by a FT4 Freeman rheometer. It can be used to
measure dynamic flow and shear properties giving valuable result of internal particle friction
of powder in motion.
It has been shown that the flowability of powders from the same supplier with similar size
distribution and Hall Flowmeter results, may result in various degrees of resistance to flow
measured by a rheometer. Freeman (2007) showed that the internal friction and cohesive
forces between particles are not simply explained by one method alone and that the Hall Flow
measurement may not be sensitive enough to detect differences among various powders.
Clayton, Millington-Smith & Armstrong (2015) also demonstrated the variation in flowability
of powders with similar size distributions from various suppliers and manufacturing
processes. The use of a rheometer detected the impact of manufacturing method and also the
variance between suppliers that deliver powder with similar specifications.
Flowability of a powder is measured by the energy required to establish flow when a rotating
blade moves either downwards or upwards through the powder, as seen in Figure 22. The
definition of a confined test, i.e. downward motion, is basic flowability energy (BFE) in units
of mJ. This motion generates a state of a relatively high stress mode and compression in the
powder. The basic flowability energy is affected by many factors such as, morphology, size
distribution, texture and cohesivity. The factor that influences the flowability to a larger
degree is the particle size distribution. Therefore, if various powders with similar size
distributions result in different BFE values one can suspect that physical factors such as
surface area, texture and morphology are more likely to be the reason for variations. The
definition of an unconfined test, i.e. upward clockwise motion, is specific energy (SE) and is
normalized by the powder mass given in unit mJ/g (Freeman, 2007).
21
Figure 22. Illustration of the confined, i.e. downward (left) and unconfined, i.e. upward (right) testing
by a FT4 Freeman rheometer (Freeman Technology, n.d.).
The flowability can also be measured by influence of external factors such as air or
consolidation. This is of interest due to external influence by carrier and protection gas as
well as potential vibrations that might pack the powder in feeder. An aeration test is used to
measure the influence of air to the resistance to flow defined as aeration energy (AE). This
aeration test gives an indirect measure of the cohesive strength between particles. Low
cohesive strength will result in well aerated powders and lower flow energies. The reduction
in flow energy from a non-aerated state, i.e. simply the BFE value, may be given as the ratio
between the two measurements defined as aeration ratio (AR) = BFE/AE. A consolidation test
may be performed to examine how the powder will flow after a powder mass has been
subjected to a number of taps to the vessel, defined as consolidation energy (CE). The flow
energy increase between the un-consolidated and consolidated state can now be given as the
ratio between the two, defined as the consolidation index (CI) = CE/BFE (Freeman
Technology, n.d.).
2.5.5 Bulk Properties
Apparent density ( ), also known as bulk density, is a measured physical characteristic
related to a powder’s ability to pack. This is a property of importance concerning die packing
and powder feeding during the deposition process. The apparent density of a powder shows
the correlation of a powders mass to freely fill a hollow space. This property may be
measured by the use of a Hall Flowmeter funnel where the powder flows freely through an
orifice and fills a container placed below (ASTM International, 2013b).
The apparent density may also be measured using a rheometer, also referred to as conditioned
bulk density (CBD). The principal method is similar with the addition of an initial
conditioning sequence with the rheometer where a rotating blade goes through the powder
22
filled container to remove any possible air pockets and thus homogenizing the powder mass
(Freeman Technology, n.d.).
Tap density ( ) is a measure of a powder’s mass ability to fill a vessel after a defined
number of taps to the vessel. This will consolidate the powder giving an increased density
compared to a powder’s bulk density. This test will simulate vibrations that may occur during
the process, handling and transporting. The difference between the two values will indicate
how sensitive a powder is to possible vibrations (Freeman Technology, n.d.).
2.5.6 Quality Assessment of Powder
To summarize, the powder characteristics that on hand indicates a powder of high quality for
laser metal deposition are:
Highly spherical particles
Fine surfaces
Few satellites
Low internal porosity
Few surface pores
Narrow size distribution
High purity
The various techniques to identify and verify these characteristics are summarized in Table 1.
Table 1. Summary of powder characteristics and their assessment techniques.
Characteristic Techniques
Morphology SEM/OM
Porosity Particle polishing+OM/Micro computed tomopgraphy
Particle size distribution Dry sieving/laser diffraction/dynamic image analysis
Flowability Hall Flowmeter/rheometer
Bulk properties Hall Flowmeter/rheometer
2.6 Statistical Significance
2.6.1 Analysis of Variance
It is of value to use basic statistics to compare sample data in order to assess whether group
populations differ from one another. One can compare data between two or multiple groups,
independently or dependently. A simple test to analyze data from multiple independent
23
groups, three or more, is analysis of variance (ANOVA). To clarify further reading in this
section, the independent groups of analysis in this project are the various powder suppliers.
ANOVA is based on a so called F-test which is the ratio of two variances with accounts for
the number of degrees of freedom. The variance (σ2) of sample data is the square of the
standard deviation and describes the dispersion of data from the population mean. The
variance is given by:
where is the observed value, is the mean value for the data series and is the number of
observations. The F-value of the F-test is given as:
The numerator of Variation between sample mean is calculated by the variation of each
individual group mean against the overall mean of the groups. This means that if the
individual group means are close to the overall mean, the variance is low. The denominator of
Variation within the samples is calculated as the sum of each observation variance from its
own group mean divided by the error degrees of freedom. This means that if the group
populations are substantially different and one can compare them against each other, the F-
value need to be high. ANOVA tests the groups against a null hypothesis which states that the
group population means are equal. If the null hypothesis is not statistically true, the mean of
two or more populations are different. It is important to statistically prove the difference
between group data to be able to compare the groups against each other. If the ANOVA test
would be repeated by picking random data from each population and plot the given F-values,
a probability distribution would result known as the F-distribution, see Figure 23 as an
example. This F-distribution is plotted with the assumption of a true null hypothesis. If the F-
value from a single ANOVA test, for example 3,3 in Figure 23, is placed in the F-distribution
plot, the probability of receiving a value at least as high as the value of 3,3 from the sample
data is 3,1%. The probability of receiving a certain F-value from the distribution is known as
the p-value. This p-value is compared to a pre set significance level of the analysis which in
most cases is 0,05 or 5%. This consequently means that if the p-value is lower or equal to
0,05 the null hypothesis can be rejected. This p-value is what determines the statistical
significance against the null hypothesis (Frost, 2016). However, this p-value should not be
interpreted as the probability of mistakenly rejecting the null hypothesis. It is simply a
probability value of observing an F-value due to sampling error, not a proof that the null
hypothesis is false (Frost, 2014).
24
Figure 23. Example of an F-distribution and resulting probability of an F-value, i.e. the p-value (Frost,
2016).
2.6.2 Regression Analysis
A regression analysis is a statistical approach to describe the relationship between input data,
known as the predictors, to an output data, known as response. The regression can be
performed by a linear or nonlinear model to describe this relationship. In this project a simple
linear regression analysis is performed whereby one predictor is linearly described by one
response. The regression analysis will results in a linear model which best explain the
relationship between predictor and response, commonly noted as:
where is the predicted response, is the intercept value, is the slope coefficient and
is the predictor value. This linear model is based on the least-squares approach, i.e. the
analysis finds a line that makes the sum of squared prediction errors as small as possible. This
also means to find the values of and that minimizes the sum of squared prediction
errors, noted Q in this example as (Pennsylvania State University, 2017):
The analysis also gives an R-sq value which is a statistical measurement of how well the
regression model describes the data. This value is given as a percentage of the response
variation that is described by the linear model. The higher the R-sq value, the more data fits
the regression model. The regression analysis can visually be presented by a fitted line plot as
seen in Figure 24 with the resultant regression model and corresponding coefficients at the
top of the graph. However, a high R-sq value does not always indicate a good fit to the model
25
which can be viewed by a residual plot. The residual is simply the difference between the
observed and modeled value, i.e. a positive or negative response error to the model. The
residual plot reveals how the residuals are scattered in reference to the fitted values and
should be in a random and unpredicted pattern. This means that if the residuals show a pattern
of negative or positive residuals, the error to the model is biased, i.e. the error is predictable.
For a biased pattern the interpretation of a high and good R-sq value may be false (Frost,
2013a).
Figure 24. Example of a fitted line plot viewing the resulting regression model and R-sq value (Frost,
2013b).
3 Materials & Methods
In this section the powders in subject of investigation in this project are presented and their
characterization techniques. The methodology of laser metal deposition is explained, followed
by the quality assessment methods for deposited structures.
3.1 Powder & Sheet Material
The materials used in this project are Inconel 718 alloy powders aimed for additive
manufacturing processes. The powders are dry with no flow additives. The substrate material
available for deposition was three mm thick Inconel 718 plates. The standard composition of
alloy 718 is seen in Table 2 (Special Metals Corporation, 2007). Studies have been made on
five different powders from four individual commercial powder suppliers. These powders
have been named A, B, C, D and E and will be referred to as so from now on forward. Each
powder and their main applicable process and respective manufacturing method is seen in
Table 3. The current in-house material specification at GKN of Inconel 718 powder for laser
26
metal deposition specifies the required composition, manufacturing method, Hall Flow rate
and particle size distribution.
The powder used at GKN in Trollhättan today is Powder C. Powder C is therefore included in
this project as a reference powder. The obtained results in terms of powder characteristics and
quality in the final part will be evaluated with Powder C in mind as the reference.
Table 2. Standard composition of Inconel alloy 718 given in wt%.
INCONEL ® Alloy 718 Standard Composition
Ni Cr Fe Nb+Ta Mo Ti Al Co
50,0-55,0 17,0-21,0 Bal. 4,75-5,50 2,80-3,30 0,65-1,15 0,20-0,80 1,00 max
C Mn Si P S B Cu
0,08 max 0,35 max 0,35 max 0,015 max 0,015 max 0,006 max 0,30 max
Table 3. Information of powders in subject of investigation in this project.
Investigated Powders
Powder Applicable process Manufacturing method
A Electron beam melting PA
B Plasma spray GA
C - EIGA
D Powder bed fusion/selective laser melting VIGA
E Laser Sintering VIGA
3.2 Powder Characterization
3.2.1 Morphology
The shape of powder particles was qualitatively analyzed by SEM, a Hitachi TM3000. Each
powder was studied in order to visually evaluate the quality in terms of satellite content and
sphericity. Evaluations were made with aims to rank the powders in comparison to each other.
The powder morphology was also quantitatively analyzed by image analysis. A very thin
layer of powder was adhered to a strip of crystal clear tape. The powder covered tape was
attached onto a sturdy plastic slide and placed in the filmstrip holder. The film scanner, a
CanoScan FS4000US Film Scanner, works with a fluorescent light source which illuminates
and scans the tape with a scanning resolution of 4000 dpi. Three plastic slides were prepared
for each powder. The scanned images were analyzed in the image analysis software
PowderShape. Random areas were selected over the three slides resulting in five
measurements per powder. A detailed statistical analysis was done per area giving a
parameter denoted as ShapeFactor which is related to an ideal shape, in this case a circle. This
27
shape factor compares the measured surface of the object (PO) to a surface equivalent circle
(PC) by the convex perimeter as:
A ShapeFactor of 1 gives that the analyzed particle is an ideal circle.
3.2.2 Porosity
3.2.2.1 Sample Preparation
In order to qualitatively and quantitatively establish the porosity of the powders, they were
hot mounted and further grinded and polished to obtain a cross-section. The machine used for
hot mounting was a Buehler SIMPLIMET™ 2000. The machines used for grinding and
polishing were a Buehler MOTOPOL™ 2000 and an AUTOMET™ 300 respectively.
A thin layer of powder was placed onto the bottom flat surface of the hot mounting cylinder
where high-density bakelite was added to enclose the metal powder sufficiently. Once the
powder was mounted the specimen was gently grinded with a 1200 and further 2500 grit size
SiC paper. This grinding scheme of each specimen was done in small steps, with monitoring
in OM in between the steps, in order to assess when an approximate cross-section of the
particles was reached. As a final step the specimen was polished with 3 µm diamond paste
slurry.
3.2.2.2 Image Analysis
There is no available standard to quantitatively measure the porosity of powder. In this
project, a free image analysis software has been used, ImageJ, to quantitatively measure the
area fraction of pores.
Numerous optical microscope images of random areas with overall representation of the
specimen were taken using Olympus BX60M with a magnification of 50x. These images
were processed and analyzed in ImageJ by first transforming the image to a grayscale 8-bit
image followed by making the image binary, i.e. in only a black and white scale. The
software was then set to measure the area fraction of black in the image. The apparent pores
were then colored and filled manually with white, see Figure 25. The area fraction was
measured once again and a difference in the amount of black was obtained. This difference
gives a pixel area fraction of pores in each image. The average area fraction was taken over
all images analyzed.
28
Figure 25. Binary image viewing apparent pores (left) and manually filled pores (right).
For a highly spherical powder with low satellite content, internal and closed pores are easy to
identify. Difficulty arises for judgment of surface connected pores, mainly for particles with a
more irregular shape and a lot of satellites. The identification is very much based on
individual judgment. To ensure that equal judgment is made for each image and powder, own
decision grounds are set. As a general rule the pore need to be closed of their
circumferential to be considered as a pore. These pores are then manually drawn and closed to
the most likable shape and filled in. Those particles who do not meet these grounds are
simply considered to be of an abnormal shape. Some examples of both cases of judgment are
given below. A good example of a clear open pore is viewed in Figure 26. A case not as
obvious is seen in Figure 27, but where the hollowness is considered as a pore due to its
closed nature and dept into the particle. Two judgment calls for non-pores are seen in Figure
28.
Figure 26. Clear example of a surface connected
pore.
Figure 27. Example of an open pore (red arrow)
with grounds of judgment.
29
Figure 28. Example of a two cavities (red arrows) that do not count as pores.
The pore sizes were also of interest and measured on ten frames taken with a magnification of
50x in the optical microscope, Olympus BX60M, for each powder. Measurements were done
from the same images taken for pixel area fraction measurement in ImageJ. For pores that
were non-spherical it was the largest axis that was measured. All pores from the size of 5 µm
were noted and averaged giving a mean pore size of the powder. Limitations to this step were
the resolution of the image when small particles were measured. This gives that pore sizes
measuring around 5-8 µm should be viewed with some uncertainty.
3.2.3 Particle Size Distribution
3.2.3.1 Image Analysis
The particle size distribution of the powders was measured by image analysis. Plastics slides
with powder were prepared and scanned as described above in section 3.2.1. Three plastic
slides were prepared for each powder. The scanned images were analyzed in the image
analysis software PowderShape. Three areas were randomly selected and averaged per slide,
giving three measurements per powder. A statistical analysis was done per image giving a
volume distribution of the particle size for each powder.
3.2.3.2 Laser Diffraction
The particle size distribution of the powders was measured by laser diffraction using a
Mastersizer 2000. A laser beam illuminated a dry powder sample whereby detectors
measured the intensity of light scattered. A volume based particle size distribution of the
sample sphere diameter is obtained. Each powder was tested three times with an equal
powder mass for each new batch. The sample mass ranged from 15-18 g for the five powders.
30
3.2.4 Rheology
3.2.4.1 Hall Flow Rate
The Hall Flow rate was measured by a static flow method. A schematic illustration of the test
setup is seen in Figure 29. The bottom of the orifice was blocked by a finger whereby powder
was poured into the center of the funnel. The orifice was then unblocked and the timing
device was manually and simultaneously started. The powder mass of 50 g flowed unaided
through the funnel into a bottom container. Three measurements of each powder,
continuously fresh samples batches, were recorded and averaged according to ASTM
standard (ASTM International, 2013a). All measurements were performed by one operator
only due to human errors. Protective gloves were also used to avoid moisture from the finger
blocking the orifice. These tests were conducted in order for the same operator to obtain a
value for all powders.
Figure 29. Schematic illustration of Hall Flowmeter setup according to ASTM standard. Setup stand
and bottom scale (left) and funnel (right) are the setup basis of the method. Modified from: (ASTM
International, 2013a)
3.2.4.2 Rheometer
A FT4 Freeman rheometer was used when measuring various rheological parameters. The
equipment consisted of a test vessel with a defined volume placed onto a bottom scale with a
precision blade able to move downward and upward the vessel. The test vessel had a
removable top part making it possible to perform a splitting action to level of excessive
powder giving a precise volume, see Figure 30 for illustration.
Funnel
holder
Scale
Funnel
31
Figure 30. Illustration of the splitting action of test vessel (Freeman Technology, n.d.).
All powders were dried at 70°C over night prior testing to ensure the same starting condition
concerning any disturbing moisture content. The powders were then cooled down to room
temperature in a desiccator before testing.
A stability and variable flow rate test method was performed on each powder twice. The
stability test sequence is a general basic flowability energy test performed seven times. The
BFE test sequence consisted of a conditioning cycle, splitting action and test cycle. The vessel
was firstly excessively filled with powder followed by a conditioning cycle. Conditioning
gently displaces the powder by a clockwise rotating action of the precision blade downward
through the vessel. This will remove possible air pocket, consolidation during filling and
create a homogenized powder sample that is slightly aerated. After conditioning of powder,
the vessel was split to give a precise powder volume. The precision blade then rotated anti-
clockwise downwards through the powder with a blade tip speed of 100 mm/s. During this
cycle the force, torque and height were measured parameters giving resulting flow energy
diagram from where the flow energy was calculated, see Figure 31. This test cycle was
repeated seven times where conditioning was performed before each new test. The variable
flow rate was performed in the same manner but with a decreasing blade tip speed starting
from 100 mm/s decreasing to 70, 40 and finally 10 mm/s. The stability and variable flow rate
test methods were performed in one combined program. Each powder was tested two times
with a new powder lot for the second test. If one value is to be reported as the BFE it is the
seventh test cycle from the stability test that is noted according to manufacturer’s guidance.
32
Figure 31. Illustration of test vessel viewing with measured parameters (left) and the resulting energy
gradient diagram (Freeman Technology, n.d.).
The influence of external factors was also tested. An aeration test sequence measured the
change in flow energy with increasing air supply from the bottom of the vessel. The flow
energy was initially measured with no air supply, a general BFE test but with two
conditioning cycles prior to test. The flow energy was further measured with air evenly
supplied through a grid net at a speed of 1 mm/s, also with two conditioning cycles prior to
test. The flow energy was then measured with a stepwise increase of air supply of 2, 3, 4, 5, 6,
8, and finally 10 mm/s. One conditioning cycle was performed in between these following
test cycles. One aeration test sequence was performed per powder.
A consolidation test was also performed whereby the vessel is manually tapped fifty times
prior to testing. A conditioning cycle was performed before tapping and the splitting action
occurs after tapping. The test cycle performed was a general anti-clockwise downward
motion of the precision blade. Each powder was tested one time.
3.2.5 Bulk Properties
3.2.5.1 Apparent Density
Apparent density ( ) is measured using a density cup with a volume of 25,01 cm3
accompanied by the Hall Flowmeter setup. The cup was weighed empty and placed below the
funnel whereby a volume of powder was poured into the center of the funnel. The powder
flowed freely into the cup below until powder overflowed the cup. Excessive powder on top
was leveled off by using a nonmagnetic blade. The density cup was then weighed once again
and the difference between weights was recorded as the powder mass. This powder mass was
divided by the cup volume and reported as g/cm3. One measurement for each powder was
done according to ASTM standard (ASTM International, 2013a).
33
3.2.5.2 Conditioned Bulk Density and Tap Density
The conditioned bulk density (CBD) was measured with the use of Freeman rheometer
equipment. The cup with a defined volume was placed onto the bottom scale and excessively
filled. The powder underwent a conditioning sequence whereby the powder sample gently
displaced into a more homogenized state. The filled vessel was then split giving a precise
volume in the cup. The mass was now measured and a resulting conditioned bulk density was
reported. One measurement was performed for each powder.
The tap density was performed with the same equipment and manner but with the addition of
fifty manual taps to the vessel before splitting. The powder mass was measured and the tap
density recorded ( ). One measurement was performed for each powder.
3.3 Laser Metal Deposition
Inconel 718 powder was deposited by a TRUMPF TruLaser Cell 7020 with a 4 kW disc laser.
Scanning movements were controlled by a 5-axis gantry system. The powder was fed by a
rotating disc feeder system, GTV PF 2/2, carried with argon gas. The powder was deposited
onto the substrate with a coaxial annular nozzle from Fraunhofer ILT. An argon gas shielding
was also introduced through the central passage on the nozzle in order to protect the molten
pool from oxidation as well as the laser optics. Each powder was deposited by a pre-defined
standard test utilized at GKN facilities in Trollhättan. The standard test consists of a single-
bead and multi-bead build-up on Inconel 718 plates deposited with a bi-directional scanning
path. The first layer onto the substrate of the multi-bead deposit is sixteen beads wide with a
50% overlap between beads. Following layer decreases by one bead in width giving a 50%
overlap between beads of subsequent layer. Five layers were deposited for each multi-bead
build-up giving a width of twelve beads in the top layer, see Figure 32. Five plates were
produced per powder. Each plate was weighed previous to and after deposition to be able to
calculate the powder efficiency.
The process parameters used were fixed and equal for all powders. The parameters are
however optimized for the reference powder, Powder C, and should be kept in mind.
Figure 32. Illustration of a standard deposit cross-section (not to scale).
34
3.4 Deposit Evaluation
3.4.1 Sample Preparation
All deposits were equally prepared for subsequent evaluation of their geometry,
microstructure and porosity content, both for multi and single-bead. The deposited parts were
sliced at four positions along the transverse direction creating three area sections for
evaluation. Cutting was performed by a Struers Secotom-10 using aluminum oxide cut wafers
to obtain a narrow cut and fine surfaces. The three surfaces of interest for each multi and
single-bead, see Figure 33, were indicated by arrows and named A, B and C. The cut surfaces
were hot mounted, grinded and polished. The machine used for hot mounting was Buehler
SIMPLIMET™ 2000. Grinding and polishing were performed using Buehler MOTOPOL™
2000 and AUTOMET™ 300 respectively. The specimen was grinded with a 600 and 1200
grit size SiC paper. As a final step the specimen was polished subsequently with 9 and 3 µm
diamond paste slurry and finalized with a Mastermet solution. Each mounted surface was
labeled according to the example shown below.
Example: A1A
A(B,C,D,E) = Powder supplier 1(2,3,4,5) = Plate No. A(B,C) = Area section
Figure 33. Image of a deposited standard test viewing marked cutting lines, cross-sections of interest
(arrows) and labeling.
3.4.2 Defects
Each polished section was inspected in an optical microscope, an Olympus BX60M, for
defects such as pores, micro-cracks and potential lack of fusion. All detected pores above the
size of 10 µm were measured and noted.
3.4.3 Geometry
In order to evaluate the deposited geometry of the cross-sections, the polished surfaces were
electrolytic etched using Oxalic acid 3,0 V for approximately five seconds. A
35
stereomicroscope, an Olympus SZX9, was used to measure the deposits. The measured
dimensions were width (w), height (h) as well as the minimum and maximum depth of
penetration ( ) into the substrate. The multi and single-bead width and height was measured
as seen in Figure 34. The width was firstly measured from edge to edge in line with the
substrate, which also makes a guideline for further measurement of height and penetration.
The height was measured approximately at the center if the deposit. The chosen minimum
and maximum penetration depth measured for the multi-bead was by visual estimation. The
measured penetration and height of the single-bead was made approximately in the middle,
see Figure 34.
Figure 34. Image showing how the geometry of a multi- and single-bead is measured.
3.4.4 Microstructure
The microstructure of deposits was evaluated in an optical microscope, an Olympus BX60M.
Etching of the samples was done with Kalling’s agent for approximately eight seconds. Plate
number 3 and corresponding area section B were chosen for each powder to be observed.
36
4 Results & Discussions
The results from various powder characterization techniques are presented, followed by the
assessed quality of deposited structures. The statistical evaluation of powder characteristics
to quality of deposited part is presented as a final part in this section.
4.1 Powder Characterization
4.1.1 Morphology
The morphology of all five samples was evaluated with the ambition to rank them against one
another according to apparent shape abnormalities. The results are given in Table 4 where the
powders are valued from 1-7, where 1 indicates the lowest level of irregularities and 7 the
highest. Grounds of judgment were from observations and images taken in SEM. Images that
are overall representative for each powder are seen in Figure 35 to Figure 39 below.
The powder with least satellite content and shape irregularities is Powder A, given a value of
1. Powder B is on the opposite side of the scale with a high satellite content and shape
irregularities, given a value of 7. In between these two ends of the scale are three powders
that are somewhat similar. Powder D and E are given a value of 5 with a quality far lower
than Powder A but still not as much irregularities as Powder B. Powder C is judged to have a
slightly higher quality than D and E, given a value of 4. This result was somewhat expected
considering the different manufacturing techniques. Powder A is produced by plasma spray
which is known to deliver spherical particles with a low amount of satellites, as presented in
section 2.5.1. The remaining powders are all produced by the same technique, gas
atomization, which in general generates more satellites than plasma spray. This also shows
that powder quality, in terms of satellite content, is greatly influenced by powder supplier
alone.
Table 4. Qualitative ranking of powders according to their morphology.
Powder Morphology
Value Powder
1 A
2 -
3 -
4 C
5 D,E
6 -
7 B
37
Powder A
Figure 35. SEM images of Powder A with 250x (top), 500x (bottom left) and 1000x magnifications
(bottom right).
38
Powder B
Figure 36. SEM images of Powder B with 250x (top), 500x (bottom left) and 1000x magnifications
(bottom right).
39
Powder C
Figure 37. SEM images of Powder C with 250x (top), 500x (bottom left) and 1000x magnifications
(bottom right).
40
Powder D
Figure 38. SEM images of Powder D with 250x (top), 500x (bottom left) and 1000x magnifications
(bottom right).
41
Powder E
Figure 39. SEM images of Powder E with 250x (top), 500x (bottom left) and 1000x magnifications
(bottom right).
The quantitative results from image analysis measurements are seen in Table 5 below. A
ShapeFactor of 1 gives a particle mean shape equivalent to the ideal circle. The reported
values are average of minimum 600 particles. The total number of particles analyzed varied
for each powder, from 5200 to 18600 particles.
The powder with ShapeFactor closest to 1 is A followed by Powder C. The powder with
highest values is B. The descending order of ShapeFactor values corresponds to the
observations in SEM and final qualitative ranking. What can be thought to be a small
42
difference between the values of Powder A and B show a big difference in qualitative
observations.
Table 5. Average particle ShapeFactor for all five powders obtained from image analysis.
Particle ShapeFactor
A B C D E
Mean StDev Mean StDev Mean StDev Mean StDev Mean StDev
1,055 0,001 1,072 0,002 1,059 0,001 1,064 0,001 1,070 0,001
4.1.2 Porosity
The porosity content given in pixel area fraction is plotted for each analyzed image taken for
each powder. The pixel area fraction is averaged over all images as values are added. This
enables to see how many images are needed to reach a stable average, i.e. when an additional
measurement is added the average does not fluctuate. This gives that the number of
measurements, or images, are not equal for each powder since the average is stabilized at
different levels. The measured fraction of each image and powder is seen in appendix 8.1.
A summarized histogram of the resulting mean porosity and particle density is given in Figure
40. The image analysis results give that Powder B has the largest mean pixel area fraction of
0,207% with Powder E not far behind that value of 0,174%. Powder A and C have the lowest
mean values of only 0,060% and 0,063% respectively. Powder D shows a value close to
Powder A and C of 0,078%.
Particle Density
Powder Density (%)
A 99,940
C 99,937
D 99,922
E 99,826
B 99,793
Figure 40. Histogram of the averaged pixel area fraction for each powder (left) and the resulting
particle density ranked (right).
The pore sizes were also measured from ten images for each powder to obtain data to plot as a
size frequency distribution. The number of images is equal in order to have the same area
analyzed for all powders. The pores are measured and plotted into four bin ranges: 5-10, 11-
A B C D E
Mean 0,060 0,207 0,064 0,078 0,174
0,000
0,100
0,200
Pix
el A
rea
Fra
ctio
n (
%)
Porosity
43
25, 26-50 and above 50 µm. A frequency distribution is plotted in Figure 41 and data are seen
in Table 6. Results show that Powder B has a substantially higher number of pores than the
other powders. On the other hand, the pores in B are in the lower size ranges if compared to
the other powders. Powder A and C distinguish from the others with the lowest number of
pores, only 22 and 26 respectively, where C shows the largest mean size of 22 µm. The
measured pores for each image and powder are seen in appendix 8.2.
Table 6. Data from pore diameter measurements.
Particle Pore Diameter
Powder Mean (µm) Max (µm) No.
A 17 42 22
B 13 46 122
C 22 48 26
D 15 40 71
E 18 54 81
Figure 41. Frequency distribution of pore sizes for all five powders.
0
20
40
60
5 10 25 50 More
Fre
quen
cy
Bin range (µm)
A - Pore diameter
0
20
40
60
5 10 25 50 More
Fre
quen
cy
Bin range (µm)
B - Pore diameter
0
20
40
60
5 10 25 50 More
Fre
quen
cy
Bin range (µm)
C - Pore diameter
0
20
40
60
5 10 25 50 More
Fre
quen
cy
Bin range (µm)
D - Pore diameter
0
20
40
60
5 10 25 50 More
Fre
quen
cy
Bin range (µm)
E - Pore diameter
44
There is a large difference in pore content even among the powders that are produced by the
same method. This shows a clear supplier dependency. On the other hand Powder A and C
show similar results despite the different production methods. Powder A was expected to
have the highest quality since it is produced by plasma atomization compared to Powder C
and its gas atomization technique, as presented by Zhong et al. (2016). This shows that it is
possible to produce a powder by GA with an equivalent quality as PA or PREP by what is
believed to be process optimization.
4.1.3 Particle Size Distribution
4.1.3.1 Image Analysis
The image analysis results are summarized and averaged for all three slides that were
prepared for each powder in Figure 43. The results are given as a volume distribution of the
particle sizes. Individual distribution result for each powder is also viewed where statistical
values of Dv90, Dv50 and Dv10 are given in Figure 42. The distribution span is also calculated
for all powders and is given by:
It is seen from resulting statistical values that B, C, D and E have a similar distribution width.
The median value of A,B and E is similar at ranging from 74,4-76,1 µm. Powder C and D are
close at a value of approximately 69 µm. Powder A differentiate from the other powders with
a wider span which is clearly seen in Figure 43. All powders have a PSD within material
specifications.
A – Statistics
D10 D50 D90 Span
58,7 76,1 99,3 0,533
0
10
20
30
40
0 50 100 150
rela
tive
vo
l%
Bin range (µm)
A - IA
A - averaged
45
B – Statistics
D10 D50 D90 Span
63,1 74,4 87,5 0,329
C – Statistics
D10 D50 D90 Span
60,9 69,8 79,5 0,266
D – Statistics
D10 D50 D90 Span
59,1 69,4 80,6 0,310
E – Statistics
D10 D50 D90 Span
66,1 75,4 84,7 0,247
Figure 42. Particle size distribution for all five powders from image analysis measurements.
0
10
20
30
40
0 50 100 150
Rel
ativ
e vo
l%
Bin range (µm)
B - IA
B - averaged
0
10
20
30
40
0 50 100 150
Rel
ativ
e vo
l%
Bin range (µm)
C - IA
C - averaged
0
10
20
30
40
0 50 100 150
Rel
atie
vo
l%
Bin range (µm)
D - IA
D - averaged
0
10
20
30
40
0 50 100 150
Rel
ativ
e vo
l%
Bin range (µm)
E - IA
E - averaged
46
Figure 43. Particle size distribution from image analysis measurements.
4.1.3.2 Laser Diffraction
Results from laser diffraction measurements are given below where the average size
distribution is plotted in Figure 44. In Table 7 the values of Dv90, Dv50 and Dv10 in µm are
seen for all three tests performed per powder. The distribution span is also calculated for all
powders.
It is seen that Powder C and D have a similar distribution in terms of span width and a lowest
PSD. The same similarities follow for Powder B and E and a resulting higher PSD. A slightly
wider distribution width is seen with Powder A which has a similar D10 as B and E but a
higher D90 value, also seen in Figure 44. All powders are within GKN material
specifications.
Figure 44. Particle size distribution results from laser diffraction measurements.
0
10
20
30
40
0 50 100 150
Rel
ativ
e vo
l%
Bin range (µm)
PSD - Image analysis
A - averaged
B - averaged
C - averaged
D - averaged
E - averaged
0
5
10
15
20
0 50 100 150
Rel
ativ
e vo
l%
Bin range (µm)
PSD - Laser diffraction
A - averaged
B - averaged
C - averaged
D - averaged
E - averaged
47
Table 7. Summary of statistical result from laser diffraction measurements.
Particle Size Distribution
Powder D10 (µm) D50 (µm) D90 (µm) Span
A1 55,0 75,8 104,2
A2 54,9 75,6 104,1
A3 55,1 75,8 104,4
A - averaged 55,0 75,7 104,2 0,650
B1 53,7 72,9 99,7
B2 53,7 72,9 99,6
B3 53,7 73,0 99,7
B - averaged 53,7 72,9 99,7 0,630
C1 48,6 66,2 90,1
C2 48,8 66,4 90,2
C3 48,8 66,4 90,3
C - averaged 48,7 66,3 90,2 0,626
D1 48,1 65,4 89,4
D2 48,1 65,4 89,5
D3 48,1 65,5 89,6
D - averaged 48,1 65,4 89,5 0,633
E1 53,7 72,6 98,7
E2 53,9 72,8 98,9
E3 53,8 72,7 98,8
E - averaged 53,8 72,7 98,8 0,619
4.1.3.3 Comparison of Image Analysis & Laser Diffraction
If the results from image analysis and laser diffraction are compared to each other, it is clear
that the two methods do not produce equivalent results. The distribution span calculated for
laser diffraction is much wider than for image analysis, with the exception of Powder A
which is slightly closer, see Figure 45. The median value also shows a slightly higher value
by image analysis measurement. This shows a limitation of the image analysis equipment
used, which was somewhat expected. The image analysis method present at GKN is a cheap
and simple method to investigate the PSD, but used with knowledge of possible limited
accuracy. The method of laser diffraction was then performed, by an external part, to
investigate the differences. By the results one can say that image analysis equipment at GKN
should not be used to obtain an accurate PSD but can give an approximate value of the
median size. To be mentioned is also the number of particle analyzed by each method. The
number of particles analyzed by image analysis varied from 5200 particles for Powder B to
18000 particles for Powder A. This can be compared to laser diffraction where a countless
number of particles were analyzed due to a much larger sample size.
48
A
D10 D50 D90 Span
IA 58,7 76,1 99,3 0,533
LD 55,0 75,7 104,2 0,650
B
D10 D50 D90 Span
IA 63,1 74,4 87,5 0,329
LD 53,7 72,9 99,7 0,630
C
D10 D50 D90 Span
IA 60,9 69,8 79,5 0,266
LD 48,7 66,3 90,2 0,626
D
D10 D50 D90 Span
IA 59,1 69,4 80,6 0,310
LD 48,1 65,4 89,5 0,633
0
10
20
30
40
0 50 100 150
Rel
ativ
e vo
l%
Bin range (µm)
A
A - IA
A - LD
0
10
20
30
40
0 50 100 150
Rel
ativ
e vo
l%
Bin range (µm)
B
B - IA
B - LD
0
10
20
30
40
0 50 100 150
Rel
ativ
e vo
l%
Bin range (µm)
C
C - IA
C - LD
0
10
20
30
40
0 50 100 150
Rel
ativ
e vo
l%
Bin range (µm)
D
D - IA
D- LD
49
E
D10 D50 D90 Span
IA 66,1 75,4 84,7 0,247
LD 53,8 72,7 98,8 0,619
Figure 45. PSD comparison between image analysis and laser diffraction for all five powders.
4.1.4 Rheology
4.1.4.1 Hall Flow Rate
The Hall Flow rate measurements for all powders are seen in Table 8 below. The averaged
flow rate measured shows that Powder A and C have the lowest value, 13 s/50g, and thus the
best flow rate. Powder B and E show the highest value of 16 s/50g with Powder D next in
ranking with 15 s/50g. But overall one can view the results as two groupings. All powders are
within GKN material specification.
Table 8. Hall Flow rate from Hall Flowmeter measurements.
Hall Flow Rate
Powder Test 1 Test 2 Test 3 FRH (s/50g)
A 12,74 12,70 12,67 13
B 16,35 16,59 16,54 16
C 13,21 13,33 13,12 13
D 15,13 15,02 15,04 15
E 15,73 15,55 15,49 16
4.1.4.2 Stability, Variable Flow & Consolidation
The stability and variable flow energy test are results obtained from the confined test of the
FT4 Freeman rheometer. The basic flow energy results from all five powders are plotted and
seen in Figure 46. The exact values obtained are found in appendix 8.4. A summary of the
various test results and calculated indexes are seen in Table 9. It is the seventh and last basic
flow energy value from the stability test that is given in Table 9.
The stability test, i.e. test cycle 1-7, show a small deviation between the first and last test
cycle for all five powders and thus good results. This is also shown from the calculated
0
10
20
30
40
0 50 100 150
Rel
ativ
e vo
l%
Bin range (µm)
E
E - IA
E - LD
50
stability index (SI) as it is close to one. The lowest flow energies and thus the best result are
found for Powder A and C. Powder E show the highest flow energies and thus the worst
result. The results for Powder B and D have similar results fall into an intermediate group of
ranking.
The results of the unconfined test, i.e. the resulting specific energy (SE), are also seen in
Table 9. The specific energy results for Powder A and C are the lowest which gives the
lowest resistance to unconfined flow. The result for Powder E is the highest and thus the
highest resistance to flow. In the intermediate level are B and D with similar SE values. The
specific energy is calculated as:
The variable flow energy test results are also seen in Figure 46 as test cycle 8-11. From this
figure and Table 9 it is seen that the BFE value of Powder A continuously increases with a
decreasing blade tip speed. A small difference is seen for Powder D, C and E where the BFE
value increases with speed of 70 mm/s but then decreases again as the speed is decreased.
Powder C on the other hand shows a small increase at 70 mm/s but then remains at a stable
value. However, the increases and decreases are overall small which is shown by the
calculated flow rate index (FRI) in Table 9. The FRI is the ratio between test cycle 8 and 11
where all powders have a ratio close to 1. The consolidation test and the resulting
consolidation indexes (CI) are also seen in Table 9. From the CI values it is seen that Powder
A have the lowest flow energy increase by 31% when subjected to simulated vibrations.
Powder B and D reviews the highest flow energy increase close to 70%. Powder C and E
have results in between these.
Table 9. Summary of rheometer measurement results.
Summarized Flow Data
Powder BFE (mJ) SI FRI SE (mJ/g) CEtap50 (mJ) CI
A 575 1,03 1,07 1,62 754 1,31
B 609 0,93 1,01 2,05 1032 1,69
C 435 0,96 1,01 1,61 608 1,40
D 644 1,04 1,01 1,96 1070 1,66
E 891 1,10 0,96 2,63 1299 1,46
51
Figure 46. Stability and variable flow energy results.
These flowability results show a limitation of Hall Flow measurements. The confined test and
resulting basic flow energies identify two powders that distinguishes as two extremes from
the others, which is not identified by Hall Flowmeter. This shows the lack of sensitivity by
the Hall Flowmeter which corresponds to previous observations by Clayton, Millington-
Smith & Armstrong (2015).
4.1.4.3 Aeration Test
The basic flow energy was measured by the influence of air flow supplied to the powder
sample. The measured flow energies for each powder are seen in Table 10 and plotted in
Figure 47. From the graph it is seen that Powder E shows no decrease in flow energy until an
air velocity of 4 mm/s is supplied and any larger decrease is not seen until an air velocity of 8
mm/s. Powder A and D show a decrease after 2 mm/s and thus have less cohesive forces
between particles than E. The best result is observed for C which has the largest decrease in
flow energy as the air velocity increases.
Figure 47. Flow energies plotted with increasing air velocity supplied.
0
200
400
600
800
1000
0 1 2 3 4 5 6 7 8 9 10 11
Bas
ic fl
ow
ener
gy (
mJ)
Test cycle
Stability & variable flow energy test
A
B
C
D
E
0
100
200
300
400
500
600
700
0 1 2 3 4 5 6 7 8 9 10
Bas
ic f
low
ener
gy (
mJ)
Air velocity (mm/s)
Aeration test
A
B
C
D
E
52
The aeration ratio (AR) is another measure of this which makes is easy to compare the
powders against each other and given as:
The ratio is calculated for values obtained at 5 mm/s and 10 mm/s noted as AR5 and AR10
respectively in Table 10. The general guideline of a cohesive or non-cohesive powder is by
the value of the aeration ratio given as:
AR ≈ 1 2 < AR < 20 AR >> 20
The powder is not
sensitive to aeration.
Usually a very cohesive
powder.
Average sensitivity to
aeration. Most common
range value for
powders.
Highly sensitive to
aeration. Very low
cohesive strength and
likely in a fluidized
state.
Table 10. Summary of results from aeration test.
Aeration Test
Powder AR0 (mJ) AE5 (mJ) AR5 AE10 (mJ) AR10
A 553 391 1,41 137 4,04
B 615 390 1,58 185 3,32
C 443 244 1,82 53 8,29
D 505 365 1,38 95 5,29
E 597 548 1,09 280 2,13
By the aeration energy at a velocity of 5 mm/s it is seen that none of the powders are
considered as sensitive to aeration. Powder E have a value close to 1 and can be considered as
a very cohesive powder at this velocity. On the other hand Powder C has the value closest to
2 and thus the most sensitive powder at 5 mm/s. If the air velocity is increased to 10 mm/s all
powders are above the value of 2 and are considered to have an average sensitivity to
aeration. However, Powder E still has the lowest value and is in the very bottom scale of the
range and thus considered to show the worst result. Powder C still shows the best result with a
value over 8 and can be concluded to have the lowest cohesive forces between particles. The
ranking of Powder A, B and D is harder to establish by certainty. All measured flow energies
at various air velocities are seen in appendix 8.5.
53
4.1.5 Bulk Properties
4.1.5.1 Apparent Density
The measured apparent density for all five powders is found in Table 11. The results show
that Powder A has the largest apparent density, i.e. the best ability to freely pack a defined
volume. The powder with the lowest ability to freely pack, i.e. lowest apparent density, is B.
The other three powders show more similar intermediate results.
Table 11. Apparent density results from Hall Flowmeter measurements.
Apparent Density
Powder Powder mass (g) Cup volume (cm3) )
A 116,8 25,01 4,67
B 103,3 25,01 4,13
C 111,5 25,01 4,46
D 109,4 25,01 4,37
E 106,9 25,01 4,27
4.1.5.2 Conditioned Bulk Density & Tap Density
The measured conditioned bulk density (CBD) from the Freeman rheometer equipment, see
Table 12, all show a higher value if compared to the apparent density which was expected.
The conditioning sequence removes possible air pockets and homogenizing the powder which
on a positive response should increase the density of the powder. The density increase is
similar for all powder aside from Powder A with a slightly smaller increase.
The conditioned bulk density is also compared to the tap density where a density difference is
given in Table 12. A lower density difference gives a powder less sensitive to vibrations. The
best results are shown by Powder A followed by Powder C. The remaining powders show
very similar results with a slightly larger difference in density. The level of density increase is
somewhat reflected on the resulting consolidation energy and consolidation index as viewed
in Table 9. The smallest density increase by A and C can be reflected to the lowest
consolidation indexes from rheological measurements.
54
Table 12. Density results from FT4 Freeman rheometer measurements.
Conditioned Bulk Density & Tap Density
Powder CBD (g/cm3) (g/cm
3) Diff.
A 4,88
5,12 0,24
4,84 0,28
B 4,38
4,80 0,42
4,39 0,41
C 4,75
5,08 0,33
4,74 0,34
D 4,66
5,06 0,40
4,63 0,43
E 4,56
4,98 0,42
4,55 0,43
4.2 Deposit Evaluation
4.2.1 Powder Efficiency
The powder efficiency (µ) during deposition is calculated from the ratio of the mass actually
deposited ( ) and the powder delivered during deposition ( ). The actual mass of
deposited powder is given from the difference in sheet weight before and after deposition.
The powder mass delivered during time of deposition is calculated by the total length of
deposited powder multiplied by the powder mass flow rate ( ) all divided by the scanning
speed ( ):
In total there are 71 beads deposited with a single length of 30 mm on each plate. This gives a
delivered powder mass of 5,11 g. The powder efficiency for each powder can now be
calculated and is seen in Table 13. The lowest efficiency is given for Powder C at 73,6% and
the highest for Powder E at 88,6%. The remaining powders show comparable results in
between C and E. However, the standard deviation of the deposited powder mass is also
calculated in Table 13 where a larger scatter in results is found for Powder E. This shows a
less repeatable process for E in terms of deposited powder mass. However, all powders show
an efficiency above 70% which is considered a well approved result.
55
Table 13. Powder efficiency during deposition.
Powder Efficiency
Powder Plate mD(g) Mean mD (g) StDev (g) (g) µ (%) Mean µ (%)
A
1 4,15
4,17 0,02 5,11
81,2
81,6
2 4,16 81,4
3 4,19 82,0
4 4,19 82,0
5 4,17 81,6
B
1 4,18
4,22 0,04 5,11
81,8
82,6
2 4,20 82,2
3 4,20 82,2
4 4,26 83,3
5 4,26 83,4
C
1 3,80
3,76 0,04 5,11
74,4
73,6
2 3,80 74,4
3 3,72 72,8
4 3,74 73,2
5 3,76 73,6
D
1 4,33
4,32 0,04 5,11
84,7
84,5
2 4,36 85,3
3 4,25 83,2
4 4,35 85,1
5 4,31 84,3
E
1 4,41
4,53 0,11 5,11
86,3
88,6
2 4,55 89,0
3 4,58 89,6
4 4,68 91,6
5 4,42 86,5
The difference between the smallest and largest mass of deposited powder is believed to be
linked to a powder’s ability to flow or be aerated. Powder E shows the highest deposited mass
with a mean value of 4,53 g, close to a 0,8 g difference from the lowest value of 3,76 g by
Powder C. This is a big variation considering the small deposited volume. The BFE values of
Powder C and E show the highest and lowest flowability respectively. This is also true for the
aeration ratio measured for 5 and 10 mm/s. One theory is that Powder E has a larger
difference in dispersion of the powder in carrier gas. In Figure 48 there are illustrations of a
well aerated powder and less aerated, perhaps the slight difference between Powder C and E.
A well dispersed and flowable powder, as Powder C, might be lost to a greater extent when it
exits the nozzle by the carrier gas. This correlation is hard to prove since the process today
has no ways available to measure or detect these differences of a powder feeding through the
hose and the nozzle outlet. There is also the possibility of a powder to attach to the inner
walls of the hose and nozzle, which is difficult to perceive.
56
Figure 48. Example of a well aerated (left) and less aerated powder (right) (Freeman
Technology, n.d.).
The work done by Kong, Carroll, Brown & Scudamore (2007), as presented in section 2.5.3,
showed a relationship between PSD and powder efficiency where a powder of similar PSD as
this project produced the highest efficiency. A powder with smaller PSD appeared to pulse,
most likely clogging inside the nozzle whereas a larger PSD created a too large powder beam
spot size. The PSD differences between the powders in this project are not as big as for Kong
et al., but the reason for differences is believed to be somewhat similar in terms of clogging.
4.2.2 Geometry
The average geometries for the single- and multi-beads are seen in Figure 49 to Figure 54
below. The five obtained values for each section and powder are averaged and seen in
appendix 8.6.
The results show that Powder C differentiates from the others with the lowest multi-bead
height, seen in Figure 49. It can also be seen that Powder D and E have similar height. The
same similarity is found for A and B. The standard deviation for each powder is small which
shows a repeatable process. A small decrease in height is seen for section B, which is
considered as the stable deposition area. The measured multi-bead width shows comparable
results for all five powders.
The penetration depths are important where Powder A shows the best results for both the
minimum and maximum values. The corresponding lowest penetration is found for Powder E.
The minimum required penetration is however 0,1 mm which all powders have exceeded.
The differences in single-bead height and width between powders are small. One observation
is however that Powder C floats out slightly more, i.e. a lower and wider single- and multi-
bead in comparison to the others.
57
Figure 49. Average multi-bead height for all powders.
Figure 50. Average multi-bead width for all powders.
Figure 51. Average multi-bead minimum and maximum penetration depth of all powders.
1,50
1,60
1,70
1,80
1,90
A B C
Hei
ght
(mm
)
Area section
Multi-bead: Height
A
B
C
D
E
11,35
11,40
11,45
11,50
11,55
11,60
A B C
Wid
th (
mm
)
Section area
Multi-bead: Width
A
B
C
D
E
0,10
0,15
0,20
0,25
0,30
0,35
A B C
Pen
etra
tio
n (m
m)
Area section
Multi-bead: Min. and Max. Penetration
Amax
Amin
Bmax
Bmin
Cmax
Cmin
Dmax
Dmin
Emax
Emin
58
Figure 52. Average single-bead height for all powders.
Figure 53. Average single-bead width for all powders.
Figure 54. Average single-bead penetration depth of all powders.
0,15
0,20
0,25
0,30
A B C
Hei
ght
(mm
)
Area section
Single-bead: Height
A
B
C
D
E
1,45
1,50
1,55
1,60
A B C
Wid
th (
mm
)
Area section
Single-bead: Width
A
B
C
D
E
0,20
0,22
0,24
0,26
0,28
0,30
A B C
Pen
etra
tio
n (
mm
)
Area section
Single-bead: Penetration
A
B
C
D
E
59
4.2.3 Defects
All detected and measured pores for all five plates and three section areas, in total 15 surfaces
analyzed per powder, are found in Table 14. No lack of fusion or micro-cracks was detected
in the samples.
The powders with lowest part quality, in terms of apparent pores, are found to be B and E
with the highest number of pores counted, over 200 in total. Powder A and C have the lowest
number of pores counted, 78 and 31 pores respectively. The measured pores were also sorted
into bin ranges 10, 25, 50 and above 50 µm and plotted as a frequency distribution, see Figure
55 to Figure 59. The frequency distributions show that B has the largest number of pores but
the most in a smaller size range. On the opposite side of the scale is C with lowest number of
pores and which many are in a larger size range in comparison. Powder A shows the second
lowest number of pores with more in a larger size range as C. It is also interesting to compare
detected pores in the single-beads. The result for Powder B and E shows the largest number
of pores, 16 and 11 respectively, which is a great amount for the small area of a single-bead.
This show the substantially lower quality of E and B compared to A and C which had no
pores.
The pore frequency in the three area sections, A, B and C, was also investigated in order to
see if there could be a difference, see Figure 55 to Figure 59. The results for each powder are
seen in Table 15 to Table 19. Since section A and C are cuts before and/or after start-and-
stop, i.e. the turn of path for each bead and layer, it was believed that more pores could be
detected here. This is true for section A which show a higher number of pores compared to
section B for all five powders. However, this is not a general trend for section C and the
differences between the more process-stable part, i.e. section B, are not always big.
These results raise the question whether few but larger pores are better or worse than many
but smaller in this project. However, when concluding on part quality the amount of pores is
considered the most importance factor in comparison to powder efficiency or geometry. The
present implementation of laser metal deposition in production at GKN, have fairly low
deposition volumes which make the efficiency a secondary quality factor. If the deposition
volumes would increase the powder efficiency would of course be more important. Powder C
is then concluded to have the highest part quality and Powder A as secondary best.
60
Table 14. Summary of data of apparent pores in all powders.
Summarized Part Pore Diameter
Multi-bead (µm) Single-bead (µm)
Powder Max size Median size No. Max size Median size No.
A 128 23 78 - - -
B 63 16 222 40 15 16
C 53 22 31 - - -
D 57 19 130 20 - 2
E 69 21 247 39 17 11
Table 15. Summary of section data of apparent pores in Powder A.
Powder A
Multi-bead (µm) Single-bead (µm)
Section Max size Median size No. Max size Median size No.
A 75 23 38 - - -
B 128 28 17 - - -
C 90 22 23 - - -
All 128 24 78 - - -
Figure 55. Frequency distribution of pores for Powder A in all 5 plates (left) and sorted by area section
(right).
Table 16. Summary of section data of apparent pores in Powder B.
Powder B
Multi-bead (µm) Single-bead (µm)
Section Max size Median size No. Max size Median size No.
A 56 16 94 26 16 5
B 48 15 76 40 17 7
C 63 17 52 35 12 4
All 63 16 222 40 15 16
0
50
100
150
10 25 50 More
Fre
quen
cy
Bin range (µm)
A - Multi-bead
0
20
40
60
80
100
A B C
Fre
quen
cy
A - Multi-bead
61
Figure 56. Frequency distribution of pores for Powder B in all 5 plates (left) and sorted by area section
(right).
Table 17. Summary of section data of apparent pores in Powder C.
Powder C
Multi-bead (µm) Single-bead (µm)
Section Max size Median size No. Max size Median size No.
A 56 22 18 - - -
B 43 21 6 - - -
C 36 25 10 - - -
All 56 22 34 - - -
Figure 57. Frequency distribution of pores for Powder C in all 5 plates (left) and sorted by area section
(right).
Table 18. Summary of section data of apparent pores in Powder D.
Powder D
Multi-bead (µm) Single-bead (µm)
Section Max size Median size No. Max size Median size No.
A 57 20 45 - - -
B 47 19 42 14 - 1
C 48 20 4 20 - 1
All 57 19 129 20 - 2
0
50
100
150
10 25 50 More
Fre
quen
cy
Bin range (µm)
B - Multi-bead
0
20
40
60
80
100
A B C
Fre
quen
cy
B - Multi-bead
0
50
100
150
10 25 50 More
Fre
quen
cy
Bin range (µm)
C - Multi-bead
0
20
40
60
80
100
A B C
Fre
quen
cy
C - Multi bead
62
Figure 58. Frequency distribution of pores for Powder D in all 5 plates (left) and sorted by area section
(right).
Table 19. Summary of section data of apparent pores in Powder E.
Powder E
Multi-bead (µm) Single-bead (µm)
Section Max size Median size No. Max size Median size No.
A 68 22 89 35 22 4
B 69 20 75 14 14 3
C 54 20 79 39 17 5
All 69 21 243 Sum 17 12
Figure 59. Frequency distribution of pores for Powder E in all 5 plates (left) and sorted by area section
(right).
4.2.4 Microstructure
The microstructure for each powder is viewed is with an optical microscopy image taken for
the multi- and single-bead, see Figure 60 and Figure 61 respectively. It is seen in the multi-
bead build up that a finer columnar structure is achieved in the first layer to the substrate. This
correlates to the heat sinking effect explained in section 2.2.5. The first layer has a higher
cooling rate and thus results in a finer microstructure. The subsequent layers show larger
grains that grow from previous layer often over multiple layers which are especially evident
for Powder C.
0
50
100
150
10 25 50 More
Fre
quen
cy
Bin range (µm)
D - Multi-bead
0
20
40
60
80
100
A B C
Fre
quen
cy
D - Multi-bead
0
50
100
150
10 25 50 More
Fre
quen
cy
Bin range (µm)
E - Multi-bead
0
20
40
60
80
100
A B C
Fre
quen
cy
E - Multi-bead
63
Powder A
Powder B
64
Powder C
Powder D
65
Figure 60. OM images showing the microstructure of a multi-bead for all five powders.
Powder A
Powder E
66
Powder B
Powder C
67
Figure 61. OM images showing the microstructure of a single-bead for all five powders.
Powder D
Powder E
68
4.3 Statistical Evaluation
A statistical evaluation is made for various powder evaluations as well as properties of the
deposited material. This evaluation is done by the statistical method Analysis of Variance
(ANOVA) with the software Minitab. As described in section 2.6.1, ANOVA is a method to
determine whether the mean differs between two or more groups. If the P-value is less or
equal to the significance level one can, with a statistical significance, reject the null
hypothesis:
H0: All population means are equal
A high F-value in terms leads to a low P-value. The multiple groups that are the factors of the
null hypothesis is the various powder suppliers, i.e. A, B, C, D and E. If the P-value for each
test is below 0,05 there is statistical significant evidence that the two or more groups mean are
not from the same population. If the P-value is above 0,05 one cannot with statistical
significant certainty discard the null hypothesis that the groups come from the same large
population. This also means that these factors cannot be further used for any statistical
correlation, due to lack of statistical evidence that the powders are different. For each test the
groups are tested against various responses, i.e. characterization or experimental results. For
the measured multi-bead and single-bead geometry, noted MB and SB respectively, the three
area sections are analyzed individually.
All tests performed are summarized in Table 20. It is seen that the ANOVA test for multi-
bead width for all area sections have a P-value larger than the significance level. The same
result is given for the width of single-bead area section A and B. This gives that the groups
are not statistically significantly different. All the remaining responses show a P-value lower
than 0,05 whereby the null hypothesis can, by a statistical significance, be rejected. A boxplot
is also reported for responses that show a p-value below 0,05. These boxplots gives a
possibility to see the distribution of data and a trend between different responses. The
responses that show a p-value below 0,05 and are of interest for the subsequent statistical
correlation are shown below. Additional boxplots are seen in appendix 8.7.
69
Table 20. Summary of results from ANOVA tests.
ANOVA
Response F-value P-value Response F-value P-value
Hall Flow rate 807,87 0,000 Basic flow energy 487,70 0,000
Particle porosity 26,34 0,000 Deposit pore freq. 20,60 0,000
Particle pore freq. 9,89 0,000 Deposit pore dia. 11,64 0,000
Particle pore dia. 6,65 0,000 Powder efficiency 109,56 0,000
ShapeFactor 110,79 0,000
IA – D10 4764 0,000 LD – D10 8343,88 0,000
IA – D50 9,80 0,002 LD – D50 10547,83 0,000
IA – D90 99,50 0,000 LD – D90 13460,25 0,000
MB height – A 60,04 0,000 SB height – A 15,57 0,000
MB height – B 62,87 0,000 SB height – B 4,87 0,007
MB height – C 81,44 0,000 SB height – C 9,32 0,000
MB width – A 1,12 0,374 SB width – A 1,11 0,380
MB width – B 1,21 0,339 SB width – B 2,17 0,110
MB width – C 0,55 0,698 SB width – C 6,82 0,001
MB Pmax – A 8,26 0,000 SB Pmax – A 4,39 0,010
MB Pmax – B 7,06 0,001 SB Pmax – B 4,82 0,007
MB Pmax – C 2,91 0,048 SB Pmax – C 3,92 0,016
MB Pmin – A 6,99 0,001
MB Pmin – B 4,38 0,011
MB Pmin – C 5,28 0,005
The particle size distribution obtained from laser diffraction measurements, noted as LD in
boxplot, is seen in Figure 62. All three boxplots of D10, D50 and D90 show the same
response trend between suppliers. The boxes itself are also very narrow which shows a low
standard deviation between individual measurements.
70
EDCBA
56
55
54
53
52
51
50
49
48
Dia
mete
r (u
m)
LD - D10
EDCBA
75,0
72,5
70,0
67,5
65,0
Dia
mete
r (u
m)
LD - D50
EDCBA
106
104
102
100
98
96
94
92
90
Dia
mete
r (u
m)
LD - D90
Figure 62. Boxplot of D10, D50 and D90 values obtained by laser diffraction measurements for all
five powders.
71
When looking at the boxplot of basic flowability energy and powder efficiency, see Figure 63
and Figure 64, a response trend can be spotted between them two. The two responses seem to
follow the same pattern for the suppliers. This indicates that the basic flowability influence
the resulting powder efficiency during deposition. The multi-bead height also follows the
same response trend as BFE and efficiency between the suppliers, seen in Figure 69.
EDCBA
900
800
700
600
500
400
Basic
flow
energ
y (
mJ)
Basic flowability energy
Figure 63. Boxplot of the basic flowability energy obtained by rheometer measurements for all five
powders.
EDCBA
90
85
80
75
70
Eff
eciency
(%
)
Powder efficiency
Figure 64. Boxplot of measured powder efficiency for all five powders.
72
The particle porosity in terms of pixel area fraction is seen in Figure 65 and the measured
ShapeFactor by image analysis is seen in Figure 66.
EDCBA
0,5
0,4
0,3
0,2
0,1
0,0
Pix
el a
rea f
ract
ion (
%)
Particle porosity
Figure 65. Boxplot of the measured pixel area fraction for all five powders.
EDCBA
1,075
1,070
1,065
1,060
1,055
1,050
ShapeFact
or
ShapeFactor
Figure 66. Boxplot of measured ShapeFactor for all five powders.
73
A clear response trend is seen in the boxplot between the particle pore frequency and deposit
pore frequency, see Figure 67. A response trend is also seen between the particle pore
diameter and deposit pore diameter even if it is more subtle, see Figure 68.
EDCBA
35
30
25
20
15
10
5
0
Fre
quency
Particle pore frequency
EDCBA
60
50
40
30
20
10
0
Fre
quency
Deposit pore frequency
Figure 67. Boxplot of measured particle and deposit pore frequency for all five powders.
74
EDCBA
60
50
40
30
20
10
0
Dia
mete
r (u
m)
Particle pore diameter
EDCBA
140
120
100
80
60
40
20
0
Dia
mete
r (u
m)
Deposit pore diameter
Figure 68. Boxplot of the measured particle and deposit pore diameter for all five powders.
The multi-bead height for all three area sections are seen in Figure 69. All three area sections
follow the same response trend. A response trend is also found between the multi-bead
heights and the boxplots of basic flow energy and powder efficiency in Figure 63 and Figure
64.
75
E-AD-AC-AB-AA-A
1,9
1,8
1,7
1,6
1,5
Dista
nce
(m
m)
MB height - Section A
E-BD-BC-BB-BA-B
1,9
1,8
1,7
1,6
1,5
Dista
nce
(m
m)
MB height - Section B
E-CD-CC-CB-CA-C
1,9
1,8
1,7
1,6
1,5
Dista
nce
(m
m)
MB height - Section C
Figure 69. Boxplot of the measured multi-bead width for all area sections and powders.
76
4.4 Statistical Correlation
A regression analysis was performed with the software Minitab in order to statistically
correlate the input data, known as predictors, to the output data, known as responses. The
predictor and response also need to have an equal sample size. The analysis will show how
well one can describe and predict a certain output from the input data by a linear regression
model. This fit of data to the model can also be graphically viewed by a Fitted Line Plot. The
result gives a value of R-sq which is a statistical measure of how well the regression model
describes the data. This value is given as a percentage of the response variation that is
described by the linear model. The higher the R-sq value the more data fits the regression
model. A general rule of thumb for a high enough R-sq value is 60%.
Due to the time restriction of this project only three or five predictors and responses are
analyzed, which is in the very bottom of accepted sample sizes. The suggested sample size to
ensure a precise estimate of model strength is 40 samples or more. With this is mind, a
response fit to the model should perhaps be considered more as a statistical trend than a
precise prediction for high R-sq values. The same goes for lower values around 50% which
may view a good trend considering the small sample size. This also gives that a residual
pattern is hard to identify, whereby no residual plots are reported.
4.4.1 Median Particle Size – Basic Flow Energy
A regression analysis is made between the median particle size obtained from laser diffraction
measurements and the basic flowability energy. Since the laser diffraction measurements
were performed three times on each powder and basic flowability seven times, a selection of
flowability measurements had to be made to ensure equal sample sizes. For each powder the
flowability measurement 1, 4 and 7 was chosen.
A fitted line plot for the analysis is seen in Figure 70. The resulting analysis showed an R-sq
value above 60% for all powders with the exception of Powder C. Powder B, D and E show
an R-sq value above 90% and thus a model that describes the responses very well. Powder C
has a value at 42,9% which indicate that the model cannot statistically predict the basic
flowability by the measured median particle size. However, with only three measured values
the fit of 42,9% is not necessarily bad and could indicate a statistical correlation. These
results statistically prove what has been presented from literature and in house-knowledge
that the particle size affects the flowability of the powder.
If one look at the regression coefficients at each boxplot, it is clear that a small increase in
median particle size shows a large increase in BFE. One remark is given to the median values.
The repeatability of laser diffraction gives that there are very small differences between the
77
three measurements. The difference between the three measured flow energies is larger which
reflects on the regression coefficient. Powder E shows this clearly with a very small
difference between LD measurements and a much larger difference between flow energies,
giving the largest coefficient.
75,8475,7875,7275,6675,60
575
570
565
560
555
D50 (um)
Basi
c flow
energ
y (m
J)
S 7,44670
R-Sq 64,7%
R-Sq(adj) 29,3%
A - D50 & BFEBFE - A = - 4145 + 62,20 A - D50
72,94872,93672,92472,91272,900
650
640
630
620
610
D50 (um)
Basi
c flow
energ
y (m
J)
S 9,34084
R-Sq 90,8%
R-Sq(adj) 81,6%
B - D50 & BFEBFE - B = - 52037 + 722,2 B - D50
78
66,3666,3266,2866,2466,20
452,5
450,0
447,5
445,0
442,5
440,0
437,5
435,0
D50 (um)
Basi
c flow
energ
y (m
J)
S 9,35311
R-Sq 42,9%
R-Sq(adj) 0,0%
C - D50 & BFEBFE - C = - 4007 + 67,10 C - D50
65,4665,4465,4265,4065,38
645
640
635
630
625
620
D50 (um)
Basi
c flow
energ
y (m
J)
S 3,43253
R-Sq 97,0%
R-Sq(adj) 94,0%
D - D50 & BFEBFE - D = - 19757 + 311,6 D - D50
72,8072,7672,7272,6872,64
890
880
870
860
850
840
830
820
810
800
D50 (um)
Basi
c flow
energ
y (m
J)
S 4,90294
R-Sq 99,3%
R-Sq(adj) 98,5%
E - D50 & BFEBFE - E = - 37816 + 531,8 E - D50
Figure 70. Fitted line plot of the median particle size to the basic flow energy for all five powders.
79
4.4.2 Median Particle Size – Powder Efficiency
A regression analysis is made between the median particle size and deposition efficiency.
Since we have three values for each particle size value, the calculated efficiency of plate 1, 3
and 5 was used in this analysis.
A fitted line plot for the analysis is seen in Figure 71. All powders show a value above 60 %.
This gives that there is a statistical correlation between median particle size and powder
efficiency. But if one reflects over the actual values of efficiency, despite some large
coefficients, the very small increase in PSD values does not greatly affect efficiency. The
exception is found for Powder E since the standard deviation for efficiency is much larger.
This in combination with the same small differences in median size also gives a larger slope
coefficient for Powder E than the others.
75,8475,8075,7675,7275,6875,6475,60
82,0
81,9
81,8
81,7
81,6
81,5
81,4
81,3
81,2
81,1
D50 (um)
Effic
iency
(%
)
S 0,0739440
R-Sq 98,2%
R-Sq(adj) 96,4%
A - D50 & EfficiencyEffic. - A = - 174,9 + 3,387 A - D50
72,9572,9472,9372,9272,9172,9072,89
83,5
83,0
82,5
82,0
81,5
D50 (um)
Effic
iency
(%
)
S 0,678545
R-Sq 65,3%
R-Sq(adj) 30,6%
B - D50 & Efficiency Effic. - B = - 1589 + 22,92 B - D50
80
66,35066,32566,30066,27566,25066,22566,200
74,5
74,0
73,5
73,0
D50 (um)
Effic
iency
(%
)
S 0,553508
R-Sq 75,0%
R-Sq(adj) 50,0%
C - D50 & EfficiencyEffic. - C = - 452,4 + 7,934 C - D50
65,47565,46065,44565,43065,41565,40065,385
85,5
85,0
84,5
84,0
83,5
83,0
D50 (um)
Effic
iency
(%
)
S 0,826947
R-Sq 72,4%
R-Sq(adj) 44,8%
D - D50 & Efficiency Effic. - D = - 1309 + 21,30 D - D50
72,77072,74572,72072,69572,67072,64572,620
92
91
90
89
88
87
86
D50 (um)
Effic
iency
(%
)
S 0,136642
R-Sq 99,9%
R-Sq(adj) 99,7%
E - D50 & Efficiency Effic. - E = - 2437 + 34,74 E - D50
Figure 71. Fitted line plot of the median particle size to the powder efficiency for all five powders
81
4.4.3 ShapeFactor – Basic Flow Energy
A regression analysis is made between a particle shape factor, denoted ShapeFactor, and the
basic flowability energy. The average ShapeFactor was calculated five times per powder
giving that five basic flowability values are needed. Measurement 1,3,5,6 and 7 was chosen
for this analysis.
A fitted line plot for the analysis is seen in Figure 72. A correlation between ShapeFactor and
BFE can be made for Powder B, D and E due to a resulting R-sq value above 60%. Powder A
gives a value of 47% which could indicate a statistical correlation due to the low sample
sizes. The lowest R-sq value of 35,5% is found for Powder C which gives that the
ShapeFactor and flow energy cannot be correlated.
1,05551,05501,05451,05401,05351,05301,0525
575
570
565
560
555
ShapeFactor
Basi
c flow
energ
y (m
J)
S 5,76124
R-Sq 47,0%
R-Sq(adj) 29,3%
A - ShapeFactor & BFEBFE - A = - 3745 + 4086 ShapeFactor-A
1,0741,0731,0721,0711,0701,0691,068
650
640
630
620
610
600
ShapeFactor
Basi
c flow
energ
y (m
J)
S 11,2749
R-Sq 68,7%
R-Sq(adj) 58,3%
B - ShapeFactor & BFEBFE - B = - 5784 + 5985 ShapeFactor-B
82
1,06081,06021,05961,05901,05841,05781,0572
450
445
440
435
ShapeFactor
Basi
c flow
energ
y (m
J)
S 6,31292
R-Sq 35,5%
R-Sq(adj) 14,0%
C - ShapeFactor & BFEBFE - C = - 2509 + 2783 ShapeFactor-C
1,06561,06521,06481,06441,06401,06361,0632
645
640
635
630
625
620
615
610
ShapeFactor
Basi
c flow
energ
y (m
J)
S 4,36098
R-Sq 88,9%
R-Sq(adj) 85,3%
D - ShapeFactor & BFEBFE - D = - 12473 + 12309 ShapeFactor-D
1,07101,07051,07001,06951,06901,06851,0680
900
880
860
840
820
800
ShapeFactor
Basi
c flow
energ
y (m
J)
S 15,7474
R-Sq 84,9%
R-Sq(adj) 79,8%
E - ShapeFactor & BFEBFE - E = - 29825 + 28679 ShapeFactor-E
Figure 72. Fitted line plot of the particle ShapeFactor to the basic flow energy for all five powders
83
4.4.4 ShapeFactor – Powder Efficiency
A regression analysis is made between a particle shape factor, denoted ShapeFactor, and the
powder efficiency. The powder efficiency was calculated for all five plates giving that five
basic flowability values are needed. Measurement 1,3,5,6 and 7 was chosen for this analysis.
A fitted line plot for the analysis is seen in Figure 73. A correlation between ShapeFactor and
powder efficiency can be made for Powder A, D and E due to a resulting R-sq value above
60%. Powder B resulted in a value of 57,4% which is close to the threshold and could
indicate a statistical correlation due to the low sample sizes. The lowest R-sq value of 15% is
found for Powder C which gives that the ShapeFactor and powder efficiency cannot be
correlated.
1,05551,05501,05451,05401,05351,05301,0525
82,0
81,8
81,6
81,4
81,2
81,0
ShapeFactor
Effic
iency
(%
)
S 0,201057
R-Sq 75,3%
R-Sq(adj) 67,0%
A - ShapeFactor & EfficiencyEffic. - A = - 196,8 + 264,0 ShapeFactor-A
1,0741,0731,0721,0711,0701,0691,068
83,5
83,0
82,5
82,0
81,5
ShapeFactor
Effic
iency
(%
)
S 0,551907
R-Sq 57,4%
R-Sq(adj) 43,2%
B - ShapeFactor & EfficiencyEffic. - B = - 163,4 + 229,5 ShapeFactor-B
84
1,06081,06021,05961,05901,05841,05781,0572
74,5
74,0
73,5
73,0
ShapeFactor
Effic
iency
(%
)
S 0,745286
R-Sq 15,0%
R-Sq(adj) 0,0%
C - ShapeFactor & EfficiencyEffic. - C = - 123,6 + 186,2 ShapeFactor-C
1,06561,06521,06481,06441,06401,06361,0632
85,5
85,0
84,5
84,0
83,5
83,0
ShapeFactor
Effic
iency
(%
)
S 0,363360
R-Sq 86,4%
R-Sq(adj) 81,9%
D - ShapeFactor & EfficiencyEffic. - D = - 885,2 + 911,3 ShapeFactor-D
1,07101,07051,07001,06951,06901,06851,0680
92
91
90
89
88
87
86
85
ShapeFactor
Effic
iency
(%
)
S 0,888972
R-Sq 88,1%
R-Sq(adj) 84,1%
E - ShapeFactor & EfficiencyEffic. - E = - 1899 + 1858 ShapeFactor-E
Figure 73. Fitted line plot of the particle ShapeFactor to the powder efficiency for all five powders
85
4.4.5 Basic Flow Energy – Powder Efficiency
A regression analysis is made between the basic flowability and powder efficiency. The
powder efficiency was calculated for all five plates giving that five basic flowability values
are needed. Measurement 1,3,5,6 and 7 was chosen for this analysis.
A fitted line plot for the analysis is seen in Figure 74. All five results show an R-sq value
above 60%. Powder E has the best fit with an value of 94,2%. This concludes that there is a
correlation between flow energy and powder efficiency. However, when looking at the slope
coefficients it is seen that a large increase in BFE does not largely affect the efficiency.
575570565560555
82,1
82,0
81,9
81,8
81,7
81,6
81,5
81,4
81,3
81,2
Basic flow energy (mJ)
Effic
iency
(%
)
S 0,230778
R-Sq 67,4%
R-Sq(adj) 56,5%
A - BFE & Efficiency Effic. - A = 58,00 + 0,04193 BFE - A
650640630620610
83,6
83,2
82,8
82,4
82,0
Basic flow energy (mJ)
Effic
iency
(%
)
S 0,473346
R-Sq 68,7%
R-Sq(adj) 58,2%
B - BFE & EfficiencyEffic. - B = 60,74 + 0,03477 BFE - B
86
452448444440436
74,5
74,0
73,5
73,0
Basic flow energy (mJ)
Effic
iency
(%
)
S 0,507029
R-Sq 60,7%
R-Sq(adj) 47,6%
C - BFE & Efficiency Effic. - C = 38,44 + 0,08011 BFE - C
648640632624616
86,0
85,5
85,0
84,5
84,0
83,5
83,0
Basic flow energy (mJ)
Effic
iency
(%
)
S 0,535104
R-Sq 70,5%
R-Sq(adj) 60,6%
D - BFE & EfficinecyEffic. - D = 45,04 + 0,06306 BFE - D
880860840820800
92
91
90
89
88
87
86
Basic flow energy (mJ)
Effic
iency
(%
)
S 0,617168
R-Sq 94,2%
R-Sq(adj) 92,3%
E - BFE & Efficiency Effic. - E = 35,97 + 0,06174 BFE - E
Figure 74. Fitted line plot of the basic flow energy to the powder efficiency for all five powders
87
4.4.6 Basic Flow Energy – Multi-bead Height
A regression analysis is made between the BFE and multi-bead height. Five multi-bead
heights are available which gives that five basic flowability values are needed. Measurement
1,3,5,6 and 7 was chosen for this analysis.
A fitted line plot for the analysis is seen in Figure 75. All resulting R-sq values are above
60% with the exception of Powder C. A value of fit at 57,5% gives that the model cannot
statistically predict the multi-bead height by the basic flow energy. However, with only five
values a fit of 57,5% could indicate a statistical correlation for Powder C.
575570565560555
1,79
1,78
1,77
1,76
1,75
1,74
Basic flow energy (mJ)
Heig
ht (m
m)
S 0,0117927
R-Sq 67,4%
R-Sq(adj) 56,5%
A - BFE & Heighth - A = 0,5539 + 0,002143 BFE - A
650640630620610
1,79
1,78
1,77
1,76
1,75
1,74
1,73
1,72
Basic flow energy (mJ)
Heig
ht (m
m)
S 0,0088168
R-Sq 88,8%
R-Sq(adj) 85,1%
B - BFE & Heighth - B = 0,9786 + 0,001231 BFE - B
88
452448444440436
1,61
1,60
1,59
1,58
1,57
1,56
1,55
1,54
1,53
1,52
Basic flow energy (mJ)
Heig
ht (m
m)
S 0,0234375
R-Sq 57,5%
R-Sq(adj) 43,4%
C - BFE & Heighth - C = 0,0424 + 0,003470 BFE - C
648640632624616
1,84
1,83
1,82
1,81
1,80
1,79
1,78
Basic flow energy (mJ)
Heig
ht (m
m)
S 0,0132929
R-Sq 80,2%
R-Sq(adj) 73,6%
D - BFE & Heighth - D = 0,5232 + 0,002041 BFE - D
880860840820800
1,875
1,850
1,825
1,800
1,775
1,750
Basic flow energy (mJ)
Heig
ht (m
m)
S 0,0111899
R-Sq 93,7%
R-Sq(adj) 91,5%
E - BFE & Height h - E = 0,9101 + 0,001062 BFE - E
Figure 75. Fitted line plot of the basic flow energy to the multi-bead height for all five powders.
89
4.4.7 Particle Pore Frequency – Deposit Pore Frequency
A regression analysis is made for the particle pore frequency and deposit pore frequency. The
pores counted in each area section and plate is summarized giving one total pore frequency
value for each one of the five plates. To obtain an equal sample size for the powder
frequency, the counted pores in the powder needed some adjustment. The pores were initially
measured and counted on ten images making it easy to combine the results for every other
two images into one, resulting in five frequency values. Keep in mind that the number of
pores counted in the powder does not correspond to the same area of analyzed material in the
part, i.e. the area of powder analyzed is not normalized or extrapolated to the part area.
A fitted line plot for the analysis is seen in Figure 76. The analysis shows a good correlation
between initial and resulting pore frequency for all five powders. All powders have an R-sq
value above 60%. Powder A, D and E show a strong correlation with a value above 90%.
When viewing the regression coefficients for each powder, the level of increase is the lowest
for powder C and the highest for powder E. Powder E has a high regression constant of 36
which influences the larger pore frequency increase. When looking at these results it should
be remembered that the process parameters are optimized for the reference powder, Powder
C. The various values in regressions coefficients between powders indicate that process
parameters need individual optimization for each powder.
6,05,55,04,54,03,53,0
25,0
22,5
20,0
17,5
15,0
12,5
10,0
Powder frequency
Part
fre
quency
S 1,75046
R-Sq 93,3%
R-Sq(adj) 91,1%
A - Powder & PartA-part = - 6,231 + 4,962 A-powder
90
36333027242118
70
60
50
40
30
Powder frequency
Part
fre
quency
S 7,87883
R-Sq 78,1%
R-Sq(adj) 70,8%
B - Powder & PartB-part = - 2,60 + 1,926 B-powder
7,26,45,64,84,03,22,4
10
9
8
7
6
5
4
Powder frequency
Part
fre
quency
S 1,29921
R-Sq 75,7%
R-Sq(adj) 67,5%
C - Powder & PartC-part = 2,043 + 0,9149 C-powder
282420161284
45
40
35
30
25
20
15
Powder frequency
Part
fre
quency
S 2,45463
R-Sq 96,0%
R-Sq(adj) 94,6%
D - Powder & PartD-part = 8,316 + 1,231 D-powder
91
282420161284
58
56
54
52
50
48
46
44
42
40
Powder frequency
Part
fre
quency
S 0,875022
R-Sq 98,1%
R-Sq(adj) 97,5%
E - Powder & PartE-part = 36,39 + 0,7538 E-powder
Figure 76. Fitted line plot of the powder pore frequency to the part pore frequency for all five powders
4.4.8 Particle Pore Size – Deposit Pore Size
A regression analysis is also interesting for the particle pores sizes and deposit pore sizes. For
this analysis the total number of pores in the five multi-beads was sorted by bin ranges 10, 25
and 50 µm. To obtain an equal sample size per bin range, the measured and counted pores in
the powder needed some adjustment as in section 4.4.7. The measured pores were combined
for every other two images into one and sorted into bin ranges. This gives five frequency
values for each bin range. The particle frequency for powder is noted as P and deposited part
as D with corresponding bin range in combination. Keep in mind that the number of pores
counted in the powder do not correspond to the same area of analyzed material in the part, i.e.
the area of powder analyzed is not normalized or extrapolated to the part area. The
summarized data is found in appendix 8.3.
Fitted line plots for the analyses are seen in Figure 77 to Figure 81. The results show a clear
correlation between the initial pores in particle and resultant pores in part. All analysis shows
an R-sq value above 60%. An interesting connection between the various bin ranges for each
individual powder is observed. If the plots of Powder A, B, D and E are analyzed, it is seen
that larger pore sizes in the powder gives a significantly higher number of large pores in the
deposited part. A small frequency increase in the range of 25 and 50 µm in the powder results
in a larger increase in part. This is reflected by a generally higher regression constant and
slope coefficient if compared to bin range 10 µm. This leads to the conclusion that a powder
with large pores statistically results in a part with a significantly higher number of large pores
than a powder with smaller pores. Since Powder C did not measure any pores within bin
range 10 µm, can this connection not be fully made.
92
Powder A
2,01,81,61,41,21,0
2,0
1,5
1,0
0,5
0,0
P-10
D-1
0
S 0,471405
R-Sq 76,2%
R-Sq(adj) 68,3%
A - Particle & Deposit Pores: Bin Range 10 umD - 10 = - 0,6667 + 1,333 P - 10
3,02,52,01,51,0
11
10
9
8
7
6
5
4
P-25
D-2
5
S 1,28452
R-Sq 85,9%
R-Sq(adj) 81,3%
A - Particle & Deposit Pores: Bin Range 25 umD - 25 = 1,900 + 2,750 P - 25
2,01,51,00,50,0
12
10
8
6
4
2
0
P-50
D-5
0
S 0,632456
R-Sq 97,7%
R-Sq(adj) 96,9%
A - Particle & Deposit Pores: Bin Range 50 umD - 50 = 0,4000 + 5,000 P - 50
Figure 77. Fitted line plot of Powder A and the pore size frequency for each bin range.
93
Powder B
2015105
8
7
6
5
4
3
2
P-10
D-1
0
S 0,628113
R-Sq 93,7%
R-Sq(adj) 91,6%
B - Particle & Deposit Pores: Bin Range 10 umD - 10 = 1,436 + 0,2851 P - 10
14131211109
45
40
35
30
25
20
15
P-25
D-2
5
S 5,22529
R-Sq 83,4%
R-Sq(adj) 77,8%
B - Particle & Deposit Pores: Bin Range 25 umD - 25 = - 23,36 + 4,778 P - 25
543210
17,5
15,0
12,5
10,0
7,5
5,0
P-50
D-5
0
S 0,609994
R-Sq 98,5%
R-Sq(adj) 98,0%
B - Particle & Deposit Pores: Bin Range 50 umD - 50 = 6,488 + 2,070 P - 50
Figure 78. Fitted line plot of Powder B and the pore size frequency for each bin range.
94
Powder C
54321
7
6
5
4
3
2
1
P-25
D-2
5
S 0,948683
R-Sq 81,8%
R-Sq(adj) 75,7%
C - Particle & Deposit Pores: Bin Range 25 umD - 25 = 0,5000 + 1,100 P - 25
3,02,52,01,51,00,50,0
4,0
3,5
3,0
2,5
2,0
P-50
D-5
0
S 0,320256
R-Sq 89,0%
R-Sq(adj) 85,3%
C - Particle & Deposit Pores: Bin Range 50 umD - 50 = 1,692 + 0,6923 P - 50
Figure 79. Fitted line plot of Powder C and the pore size frequency for each bin range.
95
Powder D
14121086420
3,0
2,5
2,0
1,5
1,0
0,5
0,0
P-10
D-1
0
S 0,346028
R-Sq 94,0%
R-Sq(adj) 92,0%
D - Particle & Deposit Pores: Bin Range 10 umD - 10 = - 0,1733 + 0,2256 P - 10
111098765432
30
25
20
15
10
P-25
D-2
5
S 2,05818
R-Sq 94,5%
R-Sq(adj) 92,7%
D - Particle & Deposit Pores: Bin Range 25 umD - 25 = 4,917 + 2,042 P - 25
43210
9
8
7
6
5
4
P-50
D-5
0
S 1,29099
R-Sq 66,2%
R-Sq(adj) 55,0%
D - Particle & Deposit Pores: Bin Range 50 umD - 50 = 3,875 + 0,8750 P - 50
Figure 80. Fitted line plot of Powder D and the pore size frequency for each bin range.
96
Powder E
654321
6
5
4
3
2
1
0
P-10
D-1
0
S 1,83738
R-Sq 52,2%
R-Sq(adj) 36,3%
E - Particle & Deposit Pores: Bin Range 10 umD - 10 = - 1,291 + 0,8023 P - 10
141312111098765
40,0
37,5
35,0
32,5
30,0
27,5
25,0
P-25
D-2
5
S 2,29362
R-Sq 91,7%
R-Sq(adj) 89,0%
E - Particle & Deposit Pores: Bin Range 25 umD - 25 = 15,06 + 1,768 P - 25
7654321
22
20
18
16
14
12
10
8
P-50
D-5
0
S 3,37553
R-Sq 65,8%
R-Sq(adj) 54,4%
E - Particle & Deposit Pores: Bin Range 50 umD - 50 = 9,308 + 1,779 P - 50
97
1,00,80,60,40,20,0
3,0
2,5
2,0
1,5
1,0
0,5
0,0
P-more
D-m
ore
S 0,471405
R-Sq 90,7%
R-Sq(adj) 87,7%
E - Particle & Deposit Pores: Bin Range 'more'D - more = 0,6667 + 2,333 P - more
Figure 81. Fitted line plot of Powder E and the pore size frequency for each bin range.
98
5 Conclusions
The conclusions of this project are presented below:
The various methods of powder characterization show:
□ Powder A shows the highest rank of qualitative powder morphology.
□ Powder A shows the highest rank of quantitative powder morphology.
□ Powder A shows the highest rank of powder density.
□ All five powders meet the specifications of particle size distribution.
□ Image analysis is a limited method of measuring particle size distribution at
GKN. The resulting median particle size can be viewed as an approximate.
□ Hall Flowmeter is a method not sensitive enough to detect large differences
in flowability. Various rheological measurements are needed to explain the
flowability of a powder.
□ Powder C shows the lowest resistance to confined and unconfined flow.
Powder A shows the second best result.
□ Powder C shows the best ability to be aerated. Powder D shows the second
best result.
□ Powder A is least sensitive to vibrations and the ability to pack.
□ Powder E shows the highest powder efficiency but also the lowest process
repeatability. Powder C shows the lowest efficiency.
□ Powder C shows the lowest counts of pores and thus highest part quality.
Powder A shows the second highest quality.
A statistical correlation is found between:
□ Median particle size & basic flow energy for Powder A, B, D and E.
□ Median particle size & powder efficiency for all five powders.
□ ShapeFactor & basic flow energy for Powder B, D and E.
□ ShapeFactor & powder efficiency for Powder A, D and E.
□ Basic flow energy & powder efficiency for all five powders.
□ Basic flow energy & multi-bead height for Powder A, B, D and E
□ Particle pore frequency & deposit pore frequency for all five powders.
□ Particle pore sizes & deposit pore sizes for all five powders. The frequency
of large pores, between 25-50 µm, in powder most drastically increases the
amount of large pores in a deposited part.
99
Final conclusions for the resulted powder characteristics and final part quality are:
The current powder characteristics defined in GKN’s material specification, i.e.
composition, manufacturing method, Hall Flow rate and particle size distribution,
are insufficient to alone specify and predict a powders behavior and impact on
part quality. The powder variances detected with rheological measurements and
the correlation between basic flow energy to the powder efficiency, indicates that
not only Hall Flow rate should be specified. The calculated ShapeFactor for each
powder was also reflected by the quantitative ranking of the powders. The
correlation between ShapeFactor, flow energy and efficiency for some powders
indicates that a specification towards morphology could be needed. We can also
see a clear correlation between powder pore frequency, pore size and final part
quality. This means that the powder density should be considered to be somewhat
specified. How these characteristics should be specified and by what limits cannot
be concluded in this project.
With Powder C as a reference in this project, it is Powder A that shows the highest
powder quality and highest quality of deposited material. If a secondary source
should be considered from the four powders in this project, it ought to be Powder
A.
100
6 Future Work
There are a few points that can be considered for future research to this master thesis project:
Individual process optimization (design of experiments) of second source – If
GKN sees the potential to use Powder A.
□ The question to answer is: Can we, by optimizing the process for Powder A,
lower the amount of particles in built material to match the quality of
Powder C that is in production today?
Mechanical testing of laser metal deposited parts – Assess the impact of large
and/or small pores.
□ The question to answer is: Can we find a more distinct criterion for the
material specification?
Investigation of powder efficiency – High-speed camera at the outlet of nozzle.
□ The question to answer is: Can we detect the reason for loss of powder by
implementing a high-speed camera at the outlet of nozzle? Are there
powder losses by the outlet and if not, is clogging the reason?
101
7 Bibliography
Ahsan, M. N., Bradley, R. & Pinkerton, A. J. (2011). Microcomputed tomography analysis of
intralayer porosity generation in laser direct metal deposition and its causes. Journal of Laser
Applications, 23(2), 022009-1- 022009-11.
Ahsan, M. N., Pinkerton, A. J., Moat, R. J. & Shackleton, J. (2011). A comparative study of
laser direct metal deposition characteristics using gas and plasma-atomized Ti–6Al–4V
powders. Materials Science and Engineering: A, 528(25-26), 7648–7657.
ALD Vacuum Technologies. (n.d.). Metal Powder Technology [pamphlet]. Retrieved from:
http://web.ald-vt.de/cms/fileadmin/pdf/prospekte/Metal_powder.pdf
Allen, J. (2006). An Investigation into the Comparative Costs of Additive Manufacture vs.
Machine from Solid for Aero Engine Parts. In Cost Effective Manufacture via Net-Shape
Processing, 17-1 – 17-10. Meeting Proceedings RTO-MP-AVT-139. Paper 17.
AP&C. (2015). Leading the Way in the Production of Plasma Atomized Spherical Metal
Powders [pamphlet]. Retrieved from: http://advancedpowders.com/wp-content/uploads/APC-
Corporate-brochure-fall-2015.pdf
ASTM International. (2013a). Standard Test Methods for Flow Rate of Metal Powders
Using the Hall Flowmeter Funnel (ASTM B213 – 13). Retrieved from: ASTM Compass
Online
ASTM International. (2013b). Standard Test Method for Apparent Density of Free-Flowing
Metal Powders Using the Hall Flowmeter Funnel (ASTM B212-13). Retrieved from: ASTM
Compass Online
ASTM International. (2014). Standard Guide for Characterizing Properties of Metal
Powders Used for Additive Manufacturing Processes (ASTM F3049 – 14). Retrieved from:
ASTM Compass Online
ASTM International. (2015). Standard Terminology for Additive Manufacturing – General
Principles – Terminology (ISO/ASTM 52900:2015(E)). Retrieved from: ASTM Compass
Online
Aveson, J. (2011, April 6). The Superalloys. The Naked Scientists. Retrieved from:
http://www.thenakedscientists.com/HTML/articles/article/the-superalloys/
Baumers, M. (2012). Economic aspects of additive manufacturing: benefits, costs and energy
consumption. (Doctoral thesis, Loughborough University, Leicestershire) Retrieved from:
https://dspace.lboro.ac.uk/dspace-jspui/bitstream/2134/10768/3/Thesis-2012-Baumers.pdf
102
Baumers, M., Dickens, P., Tuck, C. & Hague, R. (2016). The cost of additive manufacturing:
machine productivity, economies of scale and technology-push. Technological Forecasting &
Social Change, 102, 193-201.
Beiker Kair, K. & Sofos, K. (2014). Additive Manufacturing and Production of Metallic
Parts in Automotive Industry: A Case Study on Technical, Economic and Environmental
Sustainability Aspects. (Master thesis, Royal Institute of Technology, Stockholm) Retrieved
from: http://www.diva-portal.org/smash/get/diva2:740682/FULLTEXT01.pdf
Chen, L., He, Y., Yang, Y., Niu, S. & Ren, H. (2016). The research status and development
trend of additive manufacturing technology. The International Journal of Advanced
Manufacturing Technology. DOI 10.1007/s00170-016-9335-4.
Clayton, J., Millington-Smith, D. & Armstrong, B. (2015). The Application of Powder
Rheology in Additive Manufacturing. JOM, 67, 544-548. DOI: 10.1007/s11837-015-1293-z
Dawes, J., Bowerman, R. & Trepleton, R. (2015). Introduction to the Additive Manufacturing
Powder Metallurgy Supply Chain - Exploring the production and supply of metal powders for
AM processes. Johnson Matthey Technology Review, 59(3), 243-256.
D’Aveni, R. (2015). The 3-D Printing Revolution. Harvard Business Review, 93(5), 40-48.
Ding, D., Pan, Z., Cuiuri, D. & Li, H. (2015). A multi-bead overlapping model for robotic
wire and arc additive manufacturing (WAAM). Robotics and Computer-Integrated
Manufacturing, 31, 101-110.
Donachie, M. J. & Donachie, S.J. (2002). Superalloys: A Technical Guide, 2nd
Edition. ASM
International.
EPMA. (2015). Introduction to additive manufacturing technology [Pamphlet]. Retrieved
from: http://www.epma.com/additive-manufacturing
Ford, S. & Despeisse, M. (2016). Additive manufacturing and sustainability: an exploratory
study of the advantages and challenges. Journal of Cleaner Production, 137, 1573-1587.
Frank, G. (2016, July 26). Additive Manufacturing: Making Sense of Laser Metal Deposition
and 3D Printing [Web log post]. Retrieved from:
http://www.fabricatingandmetalworking.com/2016/07/additive-manufacturing-making-sense-
laser-metal-deposition-3d-printing/
103
Fraunhofer ILT. (2012). Additive Manufacturing by Laser Metal Deposition (LMD)
[Pamphlet]. Retrieved from:
http://www.ilt.fraunhofer.de/content/dam/ilt/en/documents/product_and_services/laser_materi
al_processing/HO_Additive_Manufacturing_by_Laser_Metal_Deposition_2012.pdf
Fraunhofer ILT. (2014a). Coaxial powder nozzles [Pamphlet]. Retrieved from:
http://www.ilt.fraunhofer.de/content/dam/ilt/en/documents/product_and_services/laser_materi
al_processing/HO_Coaxial_Powder_Nozzles_2014.pdf
Fraunhofer ILT. (2014b). Three-jet powder nozzles [Pamphlet]. Retrieved from:
http://www.ilt.fraunhofer.de/content/dam/ilt/en/documents/product_and_services/laser_materi
al_processing/HO_Three-Jet_Powder_Nozzles_2014.pdf
Freeman, R. (2007). Measuring the flow properties of consolidated, conditioned and aerated
powders — A comparative study using a powder rheometer and a rotational shear cell.
Powder Technology, 174, 25-33.
Freeman Technology. (n.d.). About the FT4 Powder Rheometer. Retrieved from:
http://www.freemantech.co.uk/_powders/ft4-powder-rheometer-universal-powder-tester
Frost, J. (2013a, May 30). Regression Analysis: How Do I Interpret R-squared and Assess
the Goodness-of-Fit? [Web log post]. Retrieved from:
http://blog.minitab.com/blog/adventures-in-statistics-2/regression-analysis-how-do-i-
interpret-r-squared-and-assess-the-goodness-of-fit
Frost, J. (2013b, July 1). How to Interpret Regression Analysis Results: P-values and
Coefficients [Web log post]. Retrieved from: http://blog.minitab.com/blog/adventures-in-
statistics-2/how-to-interpret-regression-analysis-results-p-values-and-coefficients
Frost, J. (2014, April 17). How to correctly interpret P values [Web log post]. Retrieved
from: http://blog.minitab.com/blog/adventures-in-statistics-2/how-to-correctly-interpret-p-
values
Frost, J. (2016, May 18). Understanding Analysis of Variance (ANOVA) and the F-test
[Web log post]. Retrieved from: http://blog.minitab.com/blog/adventures-in-statistics-
2/understanding-analysis-of-variance-anova-and-the-f-test
GKN Aerospace. (n.d.). Retrieved from:
http://www.gkngroup.com/aerospace/aboutus/Pages/default.aspx
Gibson, I., Rosen, D.W. & Stucker, B. (2010). Additive Manufacturing Technologies: Rapid
Prototyping to Direct Digital Manufacturing. New York, USA: Springer.
104
GTV. (n.d.). The PF Series of Powder Feeders [Pamphlet]. Retrieved from: http://www.gtv-
mbh.com/cms/upload/downloads/en/GTV_Pulverfoerderer-en.pdf
Hauser, C. (2014). Case Study: Laser Powder Metal Deposition Manufacturing of Complex
Real Parts [PowerPoint-presentation]. Retrieved from: http://www.lortek.es/files/merlin/05-
C_Hauser-TWI-Case-Study.pdf
Horiba Instruments. (2016). A Guidebook to Particle Size Analysis [Pamphlet]. Retrieved
from:https://www.horiba.com/fileadmin/uploads/Scientific/eMag/PSA/Guidebook/pdf/PSA_
Guidebook.pdf
Khademzadeh, S., Carmignato, S., Parvin, N., Zanini, F. & Bariani, P.F. (2015). Micro
porosity analysis in additive manufactured NiTi parts using micro computed tomography and
electron microscopy. Materials and Design, 90, 745-752.
Kong, C. Y., Carroll, P. A., Brown, P. & Scudamore, R. J. (2007). The effect of average
powder particle size on deposition efficiency, deposit height and surface roughness in the
direct metal laser deposition process. 14th International Conference on Joining of Materials.
Helsingør: Denmark. Retrieved from: http://www.twi-global.com/technical-
knowledge/published-papers/the-effect-of-average-powder-particle-size-on-deposition-
efficiency-deposit-height-and-surface-roughness-in-the-direct-metal-las/
Lee, S.-C., Chang, S.-H., Tang, T.-P., Ho H.-H. & Chen, K.-J. (2006). Improvement in the
Microstructure and Tensile Properties of Inconel 718 Superalloy by HIP Treatment. Materials
Transactions, 47(11), 2877-2881.
Mahamood, R. M., Akinlabi, E. T., Shukla, M. & Pitvana, S. (2013). Material Efficiency of
Laser Metal Deposited Ti6Al4V: Effect of Laser Power. Engineering Letters, 21(1), 18.
Mouritz, A. P. (2012). Introduction to Aerospace materials. Reston, Vancouver: Woodhead
Publishing.
Oerlikon Metco. (2016). Product Data Sheet: Metco Twin 150 Powder Feeder [Pamphlet].
Retrieved from: https://www.oerlikon.com/metco/en/products-services/coating-
equipment/thermal-spray/feeders/multipurpose-materials-feeder/twin-150-multipurpose-
powder-feeder/
Parimi, L. L., AswathanarayanaSwamy, R.G., Clark, D. & Attallah, M.M. (2014).
Microstructural and texture development in direct laser fabricated IN718. Materials
Characterization, 89, 102-111.
Poprawe, R. (Ed.) (2011). Tailored Light 2: Laser Application Technology. Berlin, Germany:
Springer.
105
Radhakrishna, CH. & Rao Prasad, K. (1997). The formation and control of Laves phase in
superalloy 718 welds. Journal of Materials Science, 32, 1977-1984.
Royal Academy of Engineering. (2013). Additive manufacturing: opportunities and
constraints [Pamphlet]. Retrieved from:
http://www.raeng.org.uk/publications/reports/additive-manufacturing
Sames, W. J., List, F. A., Pannala, S., Dehoff, R. R. & Babu, S. S. (2016). The metallurgy
and processing science of metal additive manufacturing. International Materials Review,
61(5). DOI:10.1080/09506608.2015.1116649
Slotwinski, J. A. & Garboczi, E. J. (2015). Metrology Needs for Metal Additive
Manufacturing Powders. Journal of the minerals, metals and materials society, 67(3), 538-
543.
SmarTech Market Publishing. (2014). Additive Manufacturing in Aerospace: Strategic
Implications [Pamphlet]. Retrieved from:
https://www.smartechpublishing.com/images/uploads/general/AerospaceWP.pdf
Special Metals Corporation. (2007). INCONEL ® Alloy 718 [Pamphlet]. Retrieved from:
http://www.specialmetals.com/assets/documents/alloys/inconel/inconel-alloy-718.pdf
Wei, H. L., Mazumder, J. & DebRoy, T. (2015). Evolution of solidification texture during
additive manufacturing. Scientific Reports, 5. DOI: 10.1038/srep16446
Zhang, C., Li, L. & Deceuster, A. (2011). Thermomechanical analysis of multi-bead pulsed
laser powder deposition of a nickel-based superalloy. Journal of Materials Processing
Technology, 211, 1478–1487.
Zhong, C., Biermann, T., Gasser, A, & Poprawe, G. (2015). Experimental study of effects of
main process parameters on porosity, track geometry, deposition rate, and powder efficiency
for high deposition rate laser metal deposition. Journal of Laser Applications, 27(4), 042003-
1-042003-8.
Zhong, C., Chen, J., Linnenbrink, S., Gasser, A., Sui, S. & Poprawe, R. (2016). A
comparative study of Inconel 718 formed by High Deposition Rate Laser Metal Deposition
with GA powder and PREP powder. Materials and Design, 107, 386-392.
106
8 Appendix
In this section various graphs and data are presented to add additional information and
understanding to previously presented methods and results.
8.1 Particle Pixel Area Fraction
These graphs are presented to view that a stable average value is reached for each powder.
Mean value: 0,060%
Mean value: 0,207%
Mean value: 0,064%
Mean value: 0,078%
Mean value: 0,174%
0,00
0,10
0,20
0,30
0 10 20 30 Pix
el a
rea
frac
tio
n (
%)
Image no.
A - Porosity
Porosity Mean porosity
0,00 0,10 0,20 0,30 0,40 0,50
0 10 20 30 40 Pix
el a
rea
frac
tio
n
(%)
Image no.
B - Porosity
Porosity Mean porosity
0,00
0,10
0,20
0 10 20 30 Pix
el a
rea
frac
tio
n (
%)
Image no.
C - Porosity
Porosity Mean porosity
0,00
0,10
0,20
0,30
0 10 20 30
Pix
el a
rea
frac
tio
n (
%)
Image no.
D - Porosity
Porosity Mean porosity
0,00
0,20
0,40
0,60
0 10 20 30
Pix
el a
rea
frac
tio
n (
%)
Image no.
E - Porosity
Porosity Mean porosity
107
8.2 Particle Pore Diameter
These graphs are presented to view that a stable average diameter is reached for each powder.
All graphs are the results from 10 analyzed images taken by optical microscope with a 50x
magnification.
0
20
40
0 10 20
Dia
met
er (
µm
)
Pore no.
A - Pore diameter
Pore dia. Average dia.
0
20
40
60
0 50 100
Dia
met
er (
µm
)
Pore no.
B - Pore diameter
Pore dia. Average dia.
0
20
40
0 10 20 30
Dia
met
er (
µm
)
Pore no.
C - Pore diameter
Pore dia. Average dia.
0
20
40
0 20 40 60 80
Dia
met
er (
µm
)
Pore no.
D - Pore diameter
Pore dia. Average dia.
0
20
40
60
0 20 40 60 80
Dia
met
er (
µm
)
Pore no.
E - Pore diameter
Pore dia. Average dia.
108
8.3 Summary of Pore Data in Powder & Deposited Part
These tables are presented to view which bin ranges that measured 0 pores. These are the
reason for a missing fitted line plot during statistical correlation. A reminder to the measured
and counted pores in the powder; investigation is made on ten images of 50x magnification
which are not of equal area as for part analysis.
Powder A – Pores
Powder (µm) Part (µm)
10 25 50 More 10 25 50 More
1 1 0 0 0 4 1 1
1 1 1 0 1 6 5 1
1 2 1 0 1 6 5 2
2 3 1 0 2 10 5 2
2 3 2 0 2 11 11 2
Powder B – Pores
Powder (µm) Part (µm)
10 25 50 More 10 25 50 More
3 9 0 0 3 15 6 0
8 9 0 0 3 20 7 0
10 11 1 0 4 33 9 1
16 12 2 0 6 39 10 1
22 14 5 0 8 39 17 1
Powder C – Pores
Powder (µm) Part (µm)
10 25 50 More 10 25 50 More
0 1 0 0 0 2 2 0
0 2 1 0 0 3 2 0
0 3 2 0 0 3 3 0
1 4 2 0 0 4 3 0
2 5 3 0 0 7 4 1
Powder D – Pores
Powder (µm) Part (µm)
10 25 50 More 10 25 50 More
1 3 0 0 0 10 4 0
2 3 1 0 0 13 5 0
3 8 2 0 1 19 5 0
6 9 4 0 1 23 6 0
14 11 4 0 3 29 9 1
Powder E – Pores
Powder (µm) Part (µm)
10 25 50 More 10 25 50 More
1 5 1 0 0 24 8 0
5 5 2 0 1 24 12 1
5 7 3 0 2 25 16 1
6 9 3 1 3 34 19 3
6 14 7 1 6 39 20 3
109
8.4 Morphology
All five ShapeFactor values obtained from image analysis for all powders are presented in the
table below.
Particle ShapeFactor
A B C D E
Mean StDev Mean StDev Mean StDev Mean StDev Mean StDev
1,0547 0,0151 1,0713 0,0225 1,0600 0,0211 1,0631 0,0204 1,0682 0,0218
1,0547 0,0202 1,0728 0,0226 1,0593 0,0207 1,0637 0,0213 1,0698 0,0224
1,0526 0,0148 1,0726 0,0228 1,0569 0,0203 1,0645 0,0215 1,0693 0,0225
1,0554 0,0154 1,0734 0,0242 1,0608 0,0224 1,0644 0,0211 1,0700 0,0228
1,0554 0,0153 1,0674 0,0203 1,0593 0,0219 1,0654 0,0232 1,0713 0,0224
8.5 Rheometer
All the values obtained from the stability and variability test are viewed below. From the BFE
results it is seen that the second test for all powders show a higher value which is believed to
be influenced by external surroundings. For that reason it is the first test results that are
plotted in Figure 46 and further analyzed.
All values from the aeration test are also presented below. The plotted values for the aeration
test are seen previously in Figure 47.
Stability & Variable Flow Energy
Basic Flowability Energy (mJ)
Test no. A1 A2 B1 B2 C1 C2 D1 D2 E1 E2 Blade tip speed (mm/s)
1 560 562 652 610 451 445 620 636 810 832 100
2 557 565 645 618 445 451 609 629 807 849 100
3 558 567 632 618 440 449 615 641 824 854 100
4 560 582 632 638 438 455 619 653 843 858 100
5 562 585 614 637 436 454 622 657 857 863 100
6 564 578 634 655 437 452 631 665 881 872 100
7 575 588 609 665 435 465 644 677 891 896 100
8 579 582 622 653 439 459 653 691 902 899 100
9 589 602 634 685 445 470 671 729 921 943 70
10 600 615 634 679 444 471 663 725 915 918 40
11 622 637 630 677 445 468 659 717 869 887 10
110
Aeration Test Data
Basic flow energy (mJ)
Air velocity (mm/s) A B C D E
0 553 615 443 505 597
1 545 561 394 499 609
2 513 509 355 452 594
3 474 465 316 427 591
4 434 425 279 397 565
5 391 390 244 365 548
6 339 351 205 324 517
8 221 249 104 186 384
10 137 185 53 95 280
8.6 Part Geometry
The measured height, width and penetration for each area section of multi and single-beads
are given in the tables below. The plotted values for all five powders are seen previously in
Figure 49 to Figure 54.
A – Average Geometry for Each Area Section
Multi-bead Single-bead
(mm) A StDev B StDev C StDev A StDev B StDev C StDev
h 1,76 0,02 1,76 0,02 1,80 0,02 0,24 0,01 0,24 0,01 0,24 0,01
w 11,52 0,05 11,54 0,05 11,53 0,04 1,54 0,01 1,53 0,02 1,52 0,01
Pmin 0,20 0,02 0,20 0,02 0,19 0,01 - - - - - -
Pmax 0,30 0,01 0,33 0,02 0,30 0,01 0,25 0,01 0,25 0,01 0,24 0,01
B – Average Geometry for Each Area Section
Multi-bead Single-bead
(mm) A StDev B StDev C StDev A StDev B StDev C StDev
h 1,77 0,03 1,75 0,02 1,81 0,02 0,25 0,01 0,24 0,01 0,26 0,01
w 11,55 0,04 11,50 0,04 11,51 0,06 1,55 0,01 1,53 0,03 1,54 0,01
Pmin 0,19 0,02 0,19 0,02 0,19 0,02 - - - - - -
Pmax 0,31 0,02 0,31 0,02 0,31 0,03 0,25 0,01 0,25 0,01 0,25 0,01
C – Average Geometry for Each Area Section Multi-bead Single-bead
(mm) A StDev B StDev C StDev A StDev B StDev C StDev
h 1,58 0,03 1,57 0,03 1,62 0,02 0,21 0,01 0,22 0,01 0,21 0,01
w 11,48 0,07 11,48 0,05 11,48 0,06 1,55 0,01 1,55 0,00 1,56 0,01
Pmin 0,17 0,01 0,18 0,01 0,18 0,01 - - - - - -
Pmax 0,27 0,02 0,28 0,02 0,28 0,01 0,23 0,01 0,23 0,01 0,23 0,01
111
D – Average Geometry for Each Area Section Multi-bead Single-bead
(mm) A StDev B StDev C StDev A StDev B StDev C StDev
h 1,82 0,03 1,80 0,03 1,83 0,03 0,25 0,01 0,24 0,01 0,24 0,02
w 11,51 0,04 11,53 0,07 11,51 0,07 1,56 0,02 1,56 0,02 1,55 0,01
Pmin 0,17 0,01 0,17 0,02 0,18 0,01 - - - - - -
Pmax 0,28 0,01 0,29 0,02 0,29 0,01 0,24 0,01 0,24 0,01 0,24 0,02
E – Average Geometry for Each Area Section Multi-bead Single-bead
(mm) A StDev B StDev C StDev A StDev B StDev C StDev
h 1,83 0,03 1,82 0,04 1,84 0,03 0,25 0,01 0,25 0,01 0,25 0,02
w 11,49 0,05 11,49 0,02 11,48 0,06 1,55 0,02 1,54 0,01 1,55 0,02
Pmin 0,15 0,02 0,16 0,02 0,16 0,02 -
- - Pmax 0,26 0,02 0,28 0,02 0,28 0,03 0,24 0,01 0,23 0,01 0,23 0,01
8.7 Statistical Evaluation
The following boxplots visually show the response data of particle size distribution from
image analysis results and various geometrical measurements. These are presented in
appendix since they were not in subject for a subsequent statistical correlation in this project
but still of interest for additional statistical observations.
EDCBA
16
15
14
13
12
Hall Flo
w r
ate
(s/
50g)
Hall Flow rate
112
EDCBA
67
66
65
64
63
62
61
60
59
58
Dista
nce
(m
m)
IA - D10
EDCBA
78
76
74
72
70
Dista
nce
(m
m)
IA - D50
EDCBA
105
100
95
90
85
80
Dista
nce
(m
m)
IA - D90
113
E-AD-AC-AB-AA-A
11,65
11,60
11,55
11,50
11,45
11,40
11,35
Dista
nce
(m
m)
MB width - Section A
E-BD-BC-BB-BA-B
11,65
11,60
11,55
11,50
11,45
11,40
11,35
Dis
tance
(m
m)
MB width - Section B
E-CD-CC-CB-CA-C
11,65
11,60
11,55
11,50
11,45
11,40
11,35
Dista
nce
(m
m)
MB width - Section C
114
E-AD-AC-AB-AA-A
0,375
0,350
0,325
0,300
0,275
0,250
Dista
nce
(m
m)
MB Pmax - Section A
E-BD-BC-BB-BA-B
0,375
0,350
0,325
0,300
0,275
0,250
Dista
nce
(m
m)
MB Pmax - Section B
E-CD-CC-CB-CA-C
0,375
0,350
0,325
0,300
0,275
0,250
Dista
nce
(m
m)
MB Pmax - Section C
115
E-AD-AC-AB-AA-A
0,24
0,22
0,20
0,18
0,16
0,14
0,12
Dista
nce
(m
m)
MB Pmin - Section A
E-BD-BC-BB-BA-B
0,24
0,22
0,20
0,18
0,16
0,14
0,12
Dista
nce
(m
m)
MB Pmin - Section B
E-CD-CC-CB-CA-C
0,24
0,22
0,20
0,18
0,16
0,14
0,12
Dista
nce
(m
m)
MB Pmin - Section C
116
E-AD-AC-AB-AA-A
0,30
0,28
0,26
0,24
0,22
0,20
Dista
nce
(m
m)
SB height - Section A
E-BD-BC-BB-BA-B
0,30
0,28
0,26
0,24
0,22
0,20
Dista
nce
(m
m)
SB height - Section B
E-CD-CC-CB-CA-C
1,60
1,58
1,56
1,54
1,52
1,50
1,48
Dista
nce
(m
m)
SB width - Section C
117
E-AD-AC-AB-AA-A
0,28
0,27
0,26
0,25
0,24
0,23
0,22
0,21
0,20
Dista
nce
(m
m)
SB Pmax - Section A
E-BD-BC-BB-BA-B
0,28
0,27
0,26
0,25
0,24
0,23
0,22
0,21
0,20
Dista
nce
(m
m)
SB Pmax - Section B
E-CD-CC-CB-CA-C
0,28
0,27
0,26
0,25
0,24
0,23
0,22
0,21
0,20
Dista
nce
(m
m)
SB Pmax - Section C