In-process monitoring for Electron Beam Additive Manufacturing using an infrared camera system

(1)

IN

DEGREE PROJECT MECHANICAL ENGINEERING, SECOND CYCLE, 30 CREDITS

STOCKHOLM SWEDEN 2018,

In-process monitoring for Electron Beam Additive Manufacturing

using an infrared camera system

VLAD ANTONIU BUGA

ROYSTEN JASON DSOUZA

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF INDUSTRIAL ENGINEERING AND MANAGEMENT

(2)

Degree Project in Production Engineering and Management (MG213X)

In-process monitoring for Electron Beam Additive Manufacturing using an infrared camera system

VLAD ANTONIU BUGA

ROYSTEN JASON DSOUZA

(3)

1

ABSTRACT

Additive Manufacturing (AM) is being embraced at a rapid rate, mainly due to its advantages over conventional machining. These include the possibility to create parts with complex geometries, while minimizing waste. The exponential growth of the technology has brought about challenges in quality assurance, which has proved a key barrier to large scale adoption.

Developing in-process monitoring techniques for AM is an ongoing challenge, and is still a long way off from the more established techniques developed for conventional machining. Previous research has brought about instances, where the technology has been implemented, with the focus on titanium alloys. This study aims to contribute to the research being carried out within in-process monitoring, and focusses on the Electron Beam Melting (EBM) process. The material being monitored is Inconel 625, to increase the scope of research to higher temperature ranges. The most suitable monitoring technology and vendor for the equipment, is narrowed down through a review of previous literature and market research. Experimental trials to analyze the performance of the monitoring technique with Inconel 625 are carried out.

The extracted data is then analyzed using image processing, which gives interesting results with regards to temperature fluctuations over successive layers of the build. The events within the build process for a layer, show interesting deviations in temperature, which are mapped and presented as graphs. The after-rake event, shows a particularly large deviation, which is then attributed to differential heating of the metal powder during the rake phase. This observation is supported by noticing ‘cold-spots’ in extracted images of the build. The results are discussed, and future scope for the study is conveyed. The intention of this study is to provide a base for further research into in- process monitoring for higher temperature ranges and contribute to the development of real-time process monitoring for AM.

(4)

2

SAMMANFATTNING

“Additive manufacturing” (AM) eller “friformsframställning” har snabbt ökat i omfattning, främst tack vare dess fördelar jämfört med konventionell bearbetning. Fördelarna inkluderar möjligheten att tillverka delar med komplexa geometrier medan slöseri minimeras. Den exponentiella tillväxten av tekniken har medfört utmaningar inom kvalitetssäkring, vilket har visat sig vara ett hinder för storskalig anpassning.

Utveckling av processövervakningstekniker för AM är en pågående utmaning, och ligger efter i utveckling jämfört med de mer etablerade teknikerna som utvecklats för konventionell bearbetning.

Tidigare forskning har visat fall där tekniken har implementerats med fokus på titanlegeringar. Denna studie syftar till att bidra till den forskning som genomförs inom processövervakning och fokuserar på EBM-processen (Electronic Beam Melting). Materialet som övervakas är Inconel 625, för att expandera forskningsområdet till högre temperaturområden. Den mest lämpliga övervakningstekniken och leverantör av utrustning väljs ut genom en gransking av tidigare litteratur och en marknadsundersökning. Experimentella försök för att analysera övervakningsteknikens prestanda med Inconel 625 utförs.

De extraherade data analyseras sedan med bildbehandling, vilket ger intressanta resultat med avseende på temperaturfluktuationer över successiva lager av byggobjektet. Händelserna inom byggprocessen för ett lager visar intressanta avvikelser i temperatur, vilka kartläggs och presenteras som grafer.

Tillståndet efter räfsning visar en särskilt stor avvikelse, som sedan tillskrivs differentialvärme av metallpulvret under räfsningsfasen. Denna observation stöds genom att notera "cold-spots" i extraherade bilder av byggobjektet. Resultaten diskuteras och vidare omfång för studien framförs.

Avsikten med denna studie är att ta fram en grund för vidare forskning i processövervakning för högre temperaturområden och bidra till utvecklingen av realtidsprocessövervakning för AM

(5)

3

ACKNOWLEDGEMENTS

We would like to thank KTH, especially Amir Rashid and Bita Daemi, for their exceptional support and guidance in carrying out this project, under their supervision.

Furthermore, we would also like to thank Xiaoyu Zhao, for her willingness to help us with testing on the ARCAM A2X machine, and being a constant support.

A special mention to Nihat Palanci, DIAS Infrared, John Renebo and Disa Fredriksson, for their continued support without whom this project would not have been possible.

(6)

4

1. INTRODUCTION

1.1 Background

As the industry grows and production processes evolve rapidly, the need to ensure quality for the products being produced, has become of paramount importance. A defective product is a lost product and either needs to be repaired or recycled. This in turn implies, additional costs and time are incurred for the production of a specific part. The traditional methods of ensuring quality, most of which were post process, such as destructive testing, are time-consuming and require specialized equipment and analysis techniques. The need to find and eradicate erroneous or defective products or procedures during the build cycle, required an approach that could be incorporated in the build cycle.

Therefore, non-destructive techniques (NDT) have been developed, mainly for inspection and assurance of product quality. This encompasses the assurance of the product’s quality and review of in-service parts at regular intervals (Boyes, 2009).

Additive manufacturing (AM) is the process of fabricating objects from a three-dimensional (3D) computer-aided design (CAD) data where materials are laid layer-by-layer (Chua, et al., 2017). This is in contrast to conventional methods such as milling and grinding, which are subtractive in nature. In recent years, AM has progressively transitioned from being regarded as a tool for rapid prototyping, to the efficient production of individually customized, and highly complex functional parts (Lu & Wong, 2018). But, this has brought about challenges such as defects and geometric inaccuracies in finished components. Further, lack of fusion leading to porosity and disjointed inter layer boundaries can be transferred throughout the whole build process. The use of NDT methods for the verification of quality and structural integrity of additively manufactured parts is necessary for the inspection of discontinuities and possible failures without destructing and damaging the part (Lu & Wong, 2018).

In-process monitoring technique for additive manufacturing is a step towards cost-effective quality assurance. The added advantage is the ability to send real-time corrective feedback to the control system, thus saving cost and time.

This report focusses on developing an in-process monitoring system, for Powder Bed Fusion (PBF) process, within the field of additive manufacturing. The focus of our research, is Electron Beam Melting (EBM). The process begins with spreading a layer of metal powder over the build platform.

A thermal source of energy fuses each layer to complete the model structure, while a new layer of powder is added over successive layers. The entire build chamber is contained in vacuum, and no external material can interfere during the build process. Previous attempts at in-process monitoring for the EBM process, have focused on titanium alloys, which will be discussed in section 2. The report will put forward research, tailored to Inconel 625 super-alloy, which is a wrought nickel-based super-

(8)

6 alloy strengthened mainly by additions of carbon, chromium, molybdenum and niobium (Shankar, et al., 2001). It has increasingly widespread applications in aeronautic, aerospace, marine and chemical industries due its high strength and excellent corrosion resistance.

The report will explain the development of in-process monitoring for EBM additive manufacturing, with a focus on infrared (IR) technology. Observations made within the experimental framework, can serve as an important base for developing a closed-loop control for additive manufacturing, which will aim to optimize machine parameters in the future.

1.2 Motivation and Problem Statement

In 2018, the AM industry was valued at $7.3 billion, which shows a 21 percent increase from the previous year (Wohlers, 2018). Investors are seeing the potential of the industry, even with the relative uncertainty surrounding the application, and how scalable it could be. For example, General Electric has tested 3D printed fuel nozzles in their leading edge aviation propulsion (LEAP) engines, which already had record orders before it was commercially released (Kellner, 2014). With such giants in the industry adopting AM techniques, the importance is being placed on rapidly developing and understanding material with high strength-to-weight ratios.

Like other additive manufacturing techniques, Powder-bed Electron Beam Melting technology is a growing field and many research and developments have been done within academia and industry sections. ARCAM AB, an originally Swedish company which is now acquired by GE, is one of the pioneers in manufacturing of powder-bed EBM, additive manufacturing systems some of which are equipped with inbuilt camera systems for in-process monitoring. Research conducted into previous attempts at in-process monitoring, brought about three major studies, by Schwerdtfeger et al., Price et al. and Rodriguez et al. All three had similarities with regards to working temperature ranges for the infrared camera and focused on titanium alloys. Breakthroughs were made with regards to understanding material discontinuities in the build structure, while a control system has also been developed which can adjust temperature across the powder bed for layers (Everton, et al., 2016).

Inconel 625, a solution strengthened nickel-based super-alloy, has been widely used to make engine components, heat exchangers, pressure valves, and etc. In the past decade, laser based additive manufacturing such as selective laser melting (SLM) has widely made use of it (Li, et al., 2017). The use of Inconel 625 in EBM is not widespread, and little is known of its behavior within the build chamber. The material has a high melting point close to 1000 degree Celsius, and previously known IR monitoring techniques were unsuitable due to their working temperature ranges being below the melting point of Inconel 625.

(9)

7 This presented an opportunity, to take elements from previous research conducted with IR imaging techniques used for titanium alloys, and apply this to Inconel-625. Previous findings with regards to suitability of IR cameras in monitoring titanium alloy, would serve as a starting point, but there was still a lot to understand, and to investigate these aspects for Inconel-625. Further scope would involve analyzing the collected data and providing a base, for further research into the material, and its suitability for EBM.

Keeping the above factors in mind, the following problem statement has been formulated:

There is a need for development of a suitable in-process monitoring solution for EBM additive manufacturing in higher temperature ranges, such as using Inconel 625.

Based on the above problem statement, two hypotheses were defined as follows:

1. Is it possible to monitor and record the temperature fluctuation of different events in an EBM additive manufacturing process (eg. pre-heating, sintering, and base-heating) for a single printed layer of Inconel 625 with an IR camera system?

2. Is it possible to monitor and record the temperature fluctuation of a single event in EBM additive manufacturing process, e.g. sintering, for several printed layers of Inconel 625 with an IR camera system?

1.3 Methodology and thesis outline

The initial phase includes conducting research on existing AM technologies to understand previous attempts that have been made at in-process monitoring. Further research is conducted through literature, on the properties of Inconel 625. The theory and literature review outlined in section 2, details these attempts and the most suitable NDT method for our application is inferred from this research. The provider for thermal imaging IR cameras, that work at our desired temperature range, is chosen after doing market research. The acquired camera is retrofitted on the EBM machine, and experimental trials are conducted, with data being recorded through the provider’s proprietary software. Section 3 and 4, detail the experimental setup and subsequent trials conducted. The results are reported and analysis of data is discussed in section 5, with image processing through MATLAB as the technique being used. The learning from our experiment, and basis for future research in the area is outlined in section 6. Conclusions are drawn based on our problem statement and hypothesis in section 7.

(10)

8

2. THEORY

2.1 AM technology and EBM

Additive manufacturing (AM), also known as 3D printing or rapid prototyping, is gaining increasing attention due to its ability to produce parts with added functionality and increased complexities in geometrical design, on top of the fact that it is theoretically possible to produce any shape without limitations (Gokuldoss, et al., 2017). Additive manufacturing is defined by ASTM (2012) as “joining materials to make objects from 3D model data, usually layer upon layer, as opposed to subtractive manufacturing methodologies”.

AM is defined as a transformative technology, that will revolutionize the conventional manufacturing sector, which has always been material and cost intensive. It is a development that is complimentary to the emerging global trend of being environmentally-friendly. AM will impact component design, cost, and product delivery; it will affect global business models and logistics; it should enable increased energy efficiency and lower environmental impact (Frazier, 2014).

The basic principle of this technology is that a part design is initially created in a CAD system, which can be directly built or manufactured without the need for complex process planning. AM avoids the needs for the careful analysis of part geometry to determine things such as the order of fabrication of different features, the type of tools and processes, and the need for fixtures. This helps AM to be more cost-effective technique than conventional manufacturing processes (Gibson, et al., 2010).

Powder bed fusion (PBF) is an additive manufacturing technology which creates a 3D part one layer at a time using a fine powder as the print medium. This powder is sintered or melted with either a laser or an electron beam as the heat source. Sintering and melting result in different outcomes but both are types of powder bed fusion metal printing (Rapid, 2018). Electron beam melting (EBM), is a PBF technique that melts metal powder through the use of a beam of electrons gathered and focused by electromagnetic coils (Figure 1). Given the inherent nature of the 3000-watt electron gun, the process must be carried out in a vacuum chamber (GE, 2018).

Prior to building an object using EBM, it is necessary to convert CAD data into object cross-sections in an STL format file. To build each new layer, the build platform is lowered incrementally.

Simultaneously, a hopper containing material powder is raised, and a roller or vibrating blade spreads the next layer across the build platform. An electron beam, guided by digital data from the STL file, selectively melts the metal powder, which fuses to the preceding layer as it cools. The layering process continues until the object is fully printed (GE, 2018).

(11)

9 Figure 1: Schematic view of a powder bed EBM additive manufacturing system (ARCAM, 2018)

2.1.1 Challenges to AM

AM technologies, like any other, is not perfect. The envisioned scenario, of complex designs being machined with ease, is hindered by the challenges in multiple aspects involved either directly, or indirectly in the manufacturing process. The most relevant challenge to this thesis topic, is the lack of robust In-process measurement and monitoring of materials’ performance (Lu & Wong, 2018) to monitor and identify defects during the printing process. Moreover, the thermal post-processing behaviors such as hot isostatic pressing and heat treatment are not well specified for AM materials.

In this scenario, understanding material properties and their behavior in-process is especially relevant.

However, inadequacy in the number of certified measurement methods for powder chemistry and size distribution illustrate a lack of material standards, making real time adjustments to process parameters in AM difficult (Sharatt, 2015; Lu & Wong, 2018). Standardization of AM process and equipment is

(12)

10 still a long way off from being realized. This is partly due to the broad classification in AM technology, and the pace of expansion. The lack of standardized guidelines, and lack of certifications for quality make it impossible to introduce design allowable standard for different AM processes.

The need of the hour is to introduce a collaborative effort to test materials, and undertake future work after thoroughly analyzing previous literature, which can then be stored in a database accessible by future researchers. This is a major challenge to overcome currently, due to the proprietary nature of some data and the inconsistency in data format (Energetics Incorporated, 2013).

Nondestructive evaluation (NDE) challenges are crosscutting and span materials, fabrication, quality assurance, testing, and modeling disciplines. Accordingly, NDE represents a pervasive need for AM and impacts all aspects from design and materials, through part build, and on to inspection and certification (Hodges, et al., 2014). Promotion of NDE methods, such as in-process monitoring, will aid in overcoming challenges in AM.

2.2 In-process monitoring in EBM

The complex nature of EBM process, and the requirement to perform it in vacuum, narrows down the method of NDE that can be used for evaluation. The electron magnetic coils used to deflect the electron beam during electron beam-PBF processing prohibit a co-axial arrangement being implemented (Schwerdtfeger, et al., 2012; Everton, et al., 2016) and evaporation and condensation of metal from the melt pool can lead to metallization of the machine viewing window (Dinwiddie, et al., 2013; Everton, et al., 2016). For the above reasons IR devices and IR thermography have been the favored methods for monitoring EBM processes.

There have been previous attempts at In-process monitoring in EBM, which yielded important observations for the scope of this thesis. Schwerdtfeger et al. equipped an ARCAM A2 electron beam- PBF system with a FLIR Systems A320 IR camera with a processing resolution of 320 × 240 pixels, while positioning the camera at a 15 degree angle and introducing a Zinc-selenide window to protect the camera lens (Schwerdtfeger, et al., 2012). The experiment was limited to titanium alloys, and visual imaging set-up allowed an understanding of how flaws are transferred from layer to layer as the build progresses to be developed (Everton, et al., 2016). Price et al. used a similar system to determine the repeatability of temperature measurements, build height effect on temperature profiles, transmission losses due to metallization of sacrificial glass, molten pool emissivity, molten pool dimensions and overhanging structure thermal effects (Price, et al., 2012). The article highlighted the phenomenon of emissivity and how it affects the quality of image obtained. A part that undergoes a sudden change in emissivity or exhibits multiple different emittances may introduce significant levels of experimental error in quantitative temperature measurements. This challenge is of particular

(13)

11 importance in powder bed additive manufacturing in which the powder shows a significantly higher emittance than the smooth printed surface (Dinwiddie, et al., 2013).

Metallization of viewing window is another phenomenon that severely degrades the quality of monitoring with IR imaging. Vacuum metallization is a form of physical vapor deposition, a process of combining metal with a non-metallic substrate through evaporation. (Dunmore, 2018). Rodriguez et al. incorporated an IR camera into an Arcam A2 electron beam-PBF machine, in order to analyze surface temperature profiles for each build layer (Rodriguez, et al., 2012). Machine modifications were required to install the IR camera, which included replacing the systems previous camera with ZnSe glass, installing a protective flap, which acts as a shutter, protecting the ZnSe window from metallization, and installing a pneumatic actuator to activate the shutter flap. Material discontinuities caused by “over-melting” during processing could be identified from the generated IR image (Image b) (Everton, et al., 2016).

Figure 2: a) Setup for In-process monitoring b)an image taken using IR camera with material discontinuities highlighted by Rodriguez, et al. (Everton, et al., 2016)

More recently, an attempt which was closest in terms of relevancy to our thesis topic, Raplee, et al, performed monitoring on a build for Inconel 718. A FLIR SC7600 thermographic camera was integrated into an ARCAM S12 electron beam melting system. To protect the camera lens from x-ray radiation and metal vapor deposition, a specialized viewport that utilized leaded glass and an integrated device that continuously scrolls Kapton film in front of the viewport was mounted to the EBM machine (Raplee, et al., 2017). The experiment explained the limitations of IR camera technology for higher temperature ranges, such as image resolution and frame rate.

(14)

12

Figure 3: Diagram of the Kapton Film canister (left) Image of the raw intensity and calibrated data collected from the IR camera (right), (Raplee, et al., 2017)

2.3 Material properties of Inconel-625

The material properties become especially important when you consider that no previous attempt has been made to monitor the build process for Inconel-625. From previous sections, it is understood that certain properties are especially important when it comes to in-process monitoring for EBM, and are summarized in the table below:

Sources: (Li, et al., 2017), (Special Metals, n.d.), (Kieruj, et al., 2016)

Property Researched Value

Composition by % weight Al-0.1, C-0.01, Co-0.1, Cr-21.6, Nb+Ta-3.89, Fe-4.1, Mn-0.23, Mo- 8.54, Ni- Bal, P- 0.015, S-0.015, Si-0.28, Ti-0.2

Melting point 982°C

Emissivity 0.16-0.20 (between 900°C -1000°C) Table 1: Properties of Inconel 625

(15)

13

3. METHOD

As a first step, the inbuilt monitoring equipment of the ARCAM A2X machine was observed. It consisted of a regular VGA camera connected to a computer, mounted above the vacuum chamber, used purely for observation purposes. The camera lens was mounted at an angle of 15 degrees, to facilitate viewing of the entire build platform. The whole equipment was enclosed in a stainless-steel casing and around its lens, rubber rings allowed the vacuum condition to be maintained by sealing the orifice. This inbuilt monitoring setup was used with Titanium alloy builds, within a temperature range of 600°C -900°C. Due to the higher melting point of Inconel-625, this camera system could not be used.

3.1 Experimental setup

A challenge to choose a new experimental setup which allowed monitoring of Inconel 625 build process, was to find a way to seal the orifice and isolate the outside environment from the high temperatures. The monitoring setup consisted of a DIAS Infrared camera system, PYROVIEW 320N, installed using a support structure, which could be adjusted for best possible viewing angles. The camera lens was protected with a sapphire glass window (shown in figure 5), selected due to its thermal resistance properties of up to 2030 °C. The window was positioned on the orifice, and sealed with rubber rings around it to maintain the vacuum. The camera was then connected to a computer through a fast ethernet gigabit cable to allow the data to be transferred and saved. Figure 4. shows the simple model of the setup within the scope of the EBM machine.

(16)

14 Figure 4: Model of complete experimental setup

(17)

15

Figure 5: Sapphire window with rubber sealing (left) and Camera Setup (right)

The model to be built would be four simple cubes (as shown in figure 6), but size and dimensions would vary over experimental trials, based on reliability of images obtained.

3.2 Data collection

The camera’s proprietary software, Pyrosoft, was used to monitor the build process. It allowed for real-time adjustments to be made to frequency of recording images, frame rate, emissivity, and also the scale to map temperature gradients.

Preliminary understanding of the importance of emissivity value, was understood by noticing the adjustability of the camera to emissivity values which were investigated to be closest to that of Inconel 625. The camera was able to give clear and accurate visuals for emissivity values in the range of 0.16- 0.20, as shown in the figures below. In this aspect, the solution was optimal for monitoring in high temperature ranges close to the melting point of Inconel 625.

(18)

16 Figure 6: Comparison of visuals from Pyrosoft 320N for optimally adjusted emissivity from 1 (above) to

0.16 (below)

The recording of the complete build process, which consisted of a certain amount of frames, was then divided into frame intervals for each individual layer. The sequence of events in the build process for each individual layer is shown in the figure below:

(19)

17 Figure 7: Sequence of events in build process for a layer

The frame intervals encompassing the above sequence of events, for each individual layer, were to be determined from exported bitmap images. The investigation was carried out on trial 4, due to reliability of recorded data. This trial contained the sharpest recorded images, until the point where recording was beginning to get affected by metallization phenomenon.

3.3 Data sorting

Initially a hands-on investigation on the recorded frames was done to determine if the captured data is suitable. The recording consisted of 56290 frames, out of which 6931 frames were reliable and free from metallization. These frames corresponded to 18 layers of the build. The next step was to identify the intervals for each of the 18 layers. Furthermore, the sequence of events (e.g. After rake, preheating, sintering, etc.) within the first three layers were identified. Since no exact pattern of repeatability over successive layers was determined, the event intervals for all layers were defined by continuing the hands-on approach. The frame intervals were exported as excel-readable files from Pyrosoft, to be used in data-processing. The exported data was organized in folders where each layer folder had files, ordered according to the sequence of events, shown in Figure 8:

(20)

18 Figure 8: Organization of recorded data

As a naming convention, each folder was named after the number of the layer it represented in the build. Inside the folders, the naming convention for the excel files was decided, to represent the name of each event (figure 8) of the build process for a layer. Each file represented a collection of 320 x 256 matrixes (this is due to the 320 x 256 pixel-resolution which the camera is able to provide), where each matrix represented one image capture of the process and contained temperature values.

3.4 Data processing

The exported excel files were processed using MATLAB. Each file contained a series of 2-dimension matrixes, where each matrix represented one image capture of a specific event. Each matrix represented a thermographic picture of an event in a certain given moment. If one after rake process was identified to contain 11 frames (images), the excel exported file contained 11 matrixes of 320 rows by 256 columns, with temperatures values.

To better understand the temperature fluctuation between different events in a layer, the average matrix for each event was calculated. Furthermore, out of the average matrix, the average temperature

(21)

19 value for each event was determined. This allowed graphical plotting of both an event’s evolution over the course of 18 layers as well as the evolution of all five events over the course of one layer.

The image processing code created in MATLAB as well as more detailed explanations for the used functions in it, are written in the Appendix section.

(22)

20

4. EXPERIMENTAL TRIALS AND DATA COMMENTARY

This section shows the setup used during the experiment trials and observations made. The trials were conducted over a 3-week span, and trial 4 provided the most reliable data. Over some trials, a table fan which was not part of the initial experimental setup, was added to maintain the temperature of the camera below its working limit which was 60°C.

4.1 Trial 1

The initial CAD model for the printing object is shown in figure 9. The printing object consisted of 4 cubes with the same lateral dimension and two different heights. The dimensions and the process parameters are listed in table 2. As shown in table 2, the images were recorded with a frequency of one image per second and an external fan was mounted outside of the chamber next to the camera to cool down the camera. Because of the fan, the camera’s temperature, during trial 1, was kept constant at 39.6°C. The camera’s distance to the printing area was 610 mm. Following the standard procedure in AM, first a support structure was printed on the base plate and then the cubes were printed on top of this support structure.

Parameter Value

Frequency of recording 1 image/second

Fan On

Camera temperature 39.6 °C constant

Print area and lateral shape dimensions

4 cubes, each with dimension of 75 x 75 mm, printed at the center of the plate

Build table area 250 x 250 mm

Height of shape 2 cubes of 20 mm, 2 cubes of 10 mm

Distance between cubes 5 mm

Camera distance to the print area 610 mm Duration of recording 64 minutes Support structure for build Present

Table 2: Trial 1 parameters

(23)

21 Figure 9: CAD model for Trial 1 build

Key observations:

After running the first trial, the observations showed that due to the high temperature in the vacuum chamber, a thin layer of metal particles would cover the sapphire window over time. This phenomenon is called metallization. The observations also showed that the metallization effect on the sapphire window influenced the temperature readings on the thermographic images recorded by the IR camera.

The readings were affected even before the actual printing started, i.e. during the support structure preparation. The metallization effect can be observed in figure 10 below. As shown in figure 10 left, the temperature of the lower right side of the image was recorded colder than the upper left corner.

These temperature differences were due to the metallization. Also, due to the large dimensions of print object, initially the field of view of the lens was smaller than the whole printing area.

As mentioned above, due to the metallization effect, there was metallic pigmentation observed on the sapphire window as shown in figure 11. At this point in our experiment the method to clean the window was not clear. Therefore, to prevent any damage to the sapphire glass, it was decided to use the window for the next trial without any attempt to clean the pigments, other than a simple wipe.

(24)

22

Figure 10: Effect of metallization for Trial 1 build at the end of support structure (left) and midpoint of trial (right)

Figure 11: Pigmentation on window after wipe

Adjustments for subsequent trial:

Smaller cube dimensions were chosen since printing the larger cubes in trial 1, caused the main plate to heat up and bend. Moreover, the smaller size of the cubes makes it possible to capture the entire printing area in camera’s field of view.

(25)

23

4.2 Trial 2

The CAD model for the printing object is shown in figure 12. The design and the process parameters are listed in table 3. As shown in table 3, an external fan was again used to cool down the camera.

Because of the fan, the camera’s temperature, during trial 1, was kept constant at 36.0°C.

Frequency of recording 1 image/second

Fan On

4 cubes, each with dimension of 20 x 20 mm, printed at the center of work table

Camera distance to the print area 610 mm Duration of recording 54 minutes Support structure for build Present

Figure 12: CAD model for Trial 2

Key observations

Metallization effect on the window, in the form of pigmentation was noticed over a greater area than in trial 1, as shown in figure 13 below, and even thought a lens cleaning solution was applied this metal build up could not be removed. At this point a fusion between the metal vapors and the sapphire glass was taken into consideration as a phenomenon that might have occurred. Metallization effect during trial 2 got critical after 25 minutes.

(26)

24 It was also understood that the frequency of recording set did not utilize the full potential of the camera and its frame rate of recording.

Figure 13: Increased pigmentation on window after trial 2

Figure 14: Critical metallization effect

Adjustments for subsequent trial:

For the upcoming trials it was decided that a new print area on the build table should be chosen for the cubes, as the inability to clean the glass affected the visibility of the camera pre-metallization and thus was capturing faulty data. Due to the computer crashes, an external hard disk drive was chosen as a storage unit to avoid further incidents. The image capturing frequency inside the software was changed to “Every image” mode, to maximize the amount of frames being captured.

(27)

25

4.3 Trial 3

Larger distance between the cubes was selected as noted in the table below. The frequency of recording was changed after trial 2 observations.

Frequency of recording Every Image

Fan On

4 cubes, each with dimension of 20 x 20 mm, printed at the lower half of work table

Camera distance to the print area 610 mm

Duration of recording 2 hours 08 minutes Support structure for build Present

During trial 3 metallization was unavoidable and affected the window at all points. Due to this, various methods of changing software parameters such as transmissivity and emissivity were attempted to compensate for bad data captured, which was unsatisfactory. It was also understood that the cooling effect of the fan use to keep the camera within normal conditions, increased the rate of metallization of the window, due to condensation. Images for the trial are present in the appendix.

(28)

26 Adjustments for subsequent trial

It was decided that before any other trials could be performed the sapphire window had to be cleaned. After testing various ways of removing the metallization marks, a combination of acetone and an abrasive pad allowed for a complete clean, although extreme care was take to not scratch the surface of the window. Regarding the printing process, it was decided that for future builds no support structure should be used so as to allow more meaningful data to be captured before metallization effects becomes prevalent. Fan was decided to be left out in trial 4 to further slow metallization buildup which was hastened by condensation effect. Images for the trial are compiled in the appendix.

4.4 Trial 4

Fan Off

Camera temperature 56.0 °C Max

Duration of recording 2 hours 30 minutes Support structure for build Absent

(29)

27 Key observations

Metallization was slowed down and the image was clearer than before when the build process started, due to omitting support structure and fan removal. However, it still became critical after 35 minutes.

Because of lack of ventilation and contact with the metallic fixture camera temperature rose to dangerous levels and the fan was introduced after 1 hour.

Due to quick rise in temperature the camera was shifted after 20 minutes and raised to a greater height from the window to slow down the heating process which could damage the equipment. However, in terms of clarity of image and process visualization, this trial provided the most reliable data. Images for the trial are compiled in the appendix.

Adjustments for subsequent trial

Another way to dissipate the heat from the metallic support, by introducing a second glass over window was thought of.

4.5 Trial 5

Fan Off

Duration of recording 1 hour 09 minutes Support structure for build Absent

The CAD file for the trial was the same as for trial 4.

Metallization seemed to increase in rate after introduction of the second glass. The continued rapid increase in camera temperature did not show any indication that it was slowing down with this modified approach.

(30)

28 Metallization effect was critical after 8 minutes recording was stopped after camera temperature was still rising beyond 50°C. Images for the trial are compiled in the appendix.

Adjustments for subsequent trial

A different way to dissipate the heat from the fixture support, by replacing the metallic parts with a heat resistant material like wood to stop the rapid transfer of heat. Further hastening the build process by printing only two blocks instead of four and removal of the additional glass added on top of the sapphire window.

4.6 Trial 6

Fan Off

Height of shape 2 cubes of 10 mm

Duration of recording 1 hour 53 minutes Support structure for build Absent

(31)

29 Key observations

Metallization was slowed down by 5 minutes, and the heat dissipation was successful. The experiment allowed us to finalize the limitations we faced with respect to camera cooling. Images for the trial are compiled in the appendix.

From the experimental trials, trial 4 was most reliable. However, the effects of metallization limited the number of layers we could use for analysis to 18.

(32)

30

5. RESULTS AND DISCUSSION

Figure 18: Trial 4 completed build

As mentioned in the previous chapter, 6 experimental trials with different designs, configurations and process parameters were carried out to collect data. The results presented in this chapter are based the date gathered from trial 4.

The data collected was evaluated using image processing in MATLAB. We decided on two methods of analyzing data The first method of analyzing data, in which all five sequential events in a build process and their evolution, were observed over each layer and plotted, with the comparison made between all layers, as shown in figure 19 and 20. The data shed light on the overall build conditions observed within the chamber. It can be observed for the first seven layers, that a “saw tooth” pattern of temperature fluctuation is maintained for all the events. Layer 1-7, is presented in figure 19.

(33)

31

Figure 19: Trend for events over layer 1

This is an indication that chamber conditions, such as temperature drop and rise are repetitive. It is believed that the amount of material until this point is negligible and has little impact on the cooling of the chamber.

However, the same thing cannot be said starting with layer eight where the discrepancy in the temperature gradient deviates from the above-mentioned pattern. One possible explanation would be, that along with the material build up, the cooling of the chamber is increased in intensity. This in turn, will lead to a larger preheat of the material to keep inner chamber conditions constant.

(34)

32

(35)

33 Figure 20: Trend for all events compared over layer 8-18

The second method, in which, the events of the build process for each layer would be individually plotted to see their evolution over the 18 layers, as seen in figure 21.

(36)

34

Figure 21: Trend for events in a build process compared for layers 1-18

It is interesting to note that temperature fluctuations were exceptionally prevalent in the ‘After Rake’

images. This was an observation, which brought to light, uneven heating of the powder metal during the rake stage. This differential heating was not clearly noticeable in the other events. However, the

‘Preheating’ stage showed an upward trend, which would merit further investigation.

Furthermore, possible explanations for the behavior in temperature variation for “After rake” and “Pre heat” events are as follow:

 After rake: As we keep adding layers on top of each other over the course of the build, the bottom layers cool down. This effect increases with size of the object, or in other words, with the amount of material added. As observed, in After rake graph (figure 19), the temperature drop is more drastic between layer 1 and layer 18, rather than layer 1 and layer 3, or layer 16 and layer 18.

 Preheating: To compensate for more material being added and its cooling effects over a layer, the temperature is increased during preheating so the material fuses better. Again we see that this temperature rises with the number of layers, or in other words with the amount of material added.

To summarize, as more material builds up, the temperature progressively decreases in the after rake event, which has to be compensated in the preheating for better material fusion.

(37)

35 The sharp drops in figure 21, indicate that as the build process proceeds, and surface conditions of the previous layers change, there is a possibility that the rake is spreading less powder over certain regions of the subsequent layer. It was noticed, that ‘cold spots’ were observed right after the rake process, which could be validated by the thermal scale, which showed a particular range of colors for different temperature ranges. The figures below which were taken right after the rake or powder spreading process, for layer 4 and layer 8 of the build process, clearly indicate these spots developing.

Figure 22: Cold-spots observed in layer 4 and layer 8 of the build

The observation of ‘cold-spots’ is particularly interesting to further the use of IR imaging, when monitoring Inconel 625. The spots represent an area where the powder may have undergone improper melting, and will leads to defects in the final build of the part.

(38)

36 Our present research showed that it is possible to monitor high temperature builds using IR thermography, however there is still a long way to go before the reliability of the solution can be presented within a robust, fail-proof method. Time constraints did not allow for development of a protection mechanism for the sapphire window to eliminate the effect of metallization, which severely hampered the amount of reliable data that could be analyzed over the build. The limitation of our IR solution with regards to frame rate and field of view, also indicate that there is a restriction on the ability to detect other possible features that may be critical to understanding how monitoring is a viable NDE method for EBM. The simple nature of the printed model also makes it uncertain whether the method will prove successful for more complex builds. Large amounts of data captured, which encompassed hundreds of gigabytes of data indicate that high-performance computing solutions will have to be developed to be able to utilize the solution for real time monitoring and process control.

(39)

37

6. FUTURE WORK

The research is a suitable base to further the use of IR thermography for in-process process monitoring with regards to high temperature ranges, specifically for Inconel 625. A mechanism for protection of the sapphire window from metallization is paramount, and a technical solution will have to be developed that will maintain the vacuum environment required for EBM. The most suitable, would be to use the methods understood from previous literature, such as the roll of film over the window, or a hydraulic shutter. The use of MATLAB in analyzing temperature fluctuations over all layers of a build should be conducted, to further understand the behavior of Inconel 625 within the build chamber.

(40)

38

7. CONCLUSION

The research was conducted with the aim to develop an In-process monitoring solution for high temperature ranges, with Inconel 625. Although, the monitoring solution was not applied to the entire build process, the research clearly showed that the data captured was reliable and also showed some important characteristics of how improper spread of powder during the rake process leads to ‘cold- spots’ developing over successive layers. Using image processing algorithms, the thesis was also able to show how temperature fluctuations over successive layers for specific events of the build process can be mapped, which will eventually contribute to an entire 3-D temperature mapping of a part. With constant developments in the area of IR thermography, combined with a better solution to overcome the metallization problem and robust algorithms for data processing and research into high-speed data handlers, real time monitoring for additive manufacturing is a reality that has the potential to be realized in the near future.

(41)

39

8. REFERENCES

1. ARCAM, 2018. http://www.arcam.com/technology/electron-beam-melting/hardware/.

s.l.:ARCAM AB.

2. Boyes, W., 2009. Instrumentation Reference Book (4th ed.). s.l.:Elsevier Science.

3. Chua, C., Wong, C. & Yeong, W., 2017. Standards, Quality Control, and Measurement Sciences in 3D Printing and Additive Manufacturing. s.l.:Elsevier Science.

4. Dinwiddie, R. B. et al., 2013. Thermographic In-process process monitoring of the electron- beam melting technology used in additive manufacturing. SPIE Proceedings, Volume 8705.

5. Dunmore, 2018. Vacuum Metallization Process. [Online]

Available at: https://www.dunmore.com/technical/vacuum-metallizing.html [Accessed 20 October 2018].

6. Energetics Incorporated, 2013. Measurement science roadmap for metal-based additive manufacturing, Columbia: National Institute of Standards and Technology.

7. Everton, S. et al., 2016. Review of In-process process monitoring and In-process metrology for metal additive manufacturing. Materials & Design, Volume 95, pp. 431-445.

8. Frazier, W., 2014. Metal Additive Manufacturing: A Review. Journal of Materials Engineering and Performance, Volume 23(6), pp. 1917-1928.

9. GE, 2018. How powder bed fusion works. [Online]

Available at: https://www.ge.com/additive/additive-manufacturing/information/powder-bed- fusion

[Accessed 14 October 2018].

10. Gibson, I., Rosen, D. W. & Stucker, B., 2010. Additive Manufacturing Technologies: Rapid Prototyping to Direct Digital Manufacturing. Boston: Springer US.

11. Gokuldoss, P. K., Kolla, S. & Eckert, J., 2017. Additive Manufacturing Processes: Selective Laser Melting, Electron Beam Melting and Binder Jetting—Selection Guidelines. Materials, Volume 10(6).

12. Hodges, K. L. et al., 2014. Nondestructive Evaluation of Additive Manufacturing State-of-the- Discipline Report, s.l.: NASA.

13. Kellner, T., 2014. Fit to Print: New Plant Will Assemble World’s First Passenger Jet Engine With 3D Printed Fuel Nozzles, Next-Gen Materials, s.l.: GE Reports.

(42)

40 14. Kieruj, P., Przestack, D. & Chwalczuk, T., 2016. Determination of emissivity coefficient of heat-resistant super alloys and cemented carbide. Archives of Mechanical Technology and Materials , Volume Vol. 36, pp. 30-34.

15. Li, C. et al., 2017. Microstructure evolution characteristics of Inconel 625 alloy from selective laser melting to heat treatment. Materials Science & Engineering A, Volume 705, pp. 20-31.

16. Lu, Q. & Wong, C., 2018. “Additive manufacturing process monitoring and control by non- destructive testing techniques: challenges and in-process monitoring,”. Virtual and Physical Prototyping, 13(2), pp. 39-48.

17. Price, S., Lydon, J., Cooper, K. & Chou, K., 2012. EXPERIMENTAL TEMPERATURE ANALYSIS OF POWDER-BASED ELECTRON BEAM ADDITIVE MANUFACTURING.

24th International Solid Freeform Fabrication Symposium.

18. Rapid, S., 2018. What is Powder Bed Fusion for Metal 3D Printing?. [Online]

Available at: https://www.starrapid.com/blog/what-is-powder-bed-fusion-for-metal-3d- printing/

19. Raplee, J. et al., 2017. Understanding the thermal sciences in the electron beam melting process through In-process process monitoring. Nondestructive Characterization and Monitoring of Advanced Materials, Aerospace, and Civil Infrastructure, Volume 10169.

20. Rodriguez, E. et al., 2012. Integration of a thermal imaging feedback control system in electron beam melting. Austin, TX, 23rd International Solid Freeform Fabrication Symposium.

21. Schwerdtfeger, J., Singer, R. F. & Körner, C., 2012. In process flaw detection by IR‐ imaging during electron beam melting. Rapid Prototyping Journal, 18(4), pp. 259-263.

22. Shankar, V., Bhanu Sankara Rao, K. & Mannan, S., 2001. Microstructure and mechanical properties of Inconel 625 superalloy. Journal of Nuclear Materials, Volume 288(2), pp. 222- 232.

23. Sharatt, B., 2015. Non-destructive techniques and technologies for qualification of additive manufactured parts and processes, Victoria, BC: Atlantic Research Centre.

24. Special Metals, n.d. Inconel 625 alloy. [Online]

Available at: http://www.specialmetals.com/assets/smc/documents/alloys/inconel/inconel- alloy-625.pdf

(43)

41 25. Wohlers, T., 2018. Wohlers Report, s.l.: Wohlers Associates.

(44)

42

9. LIST OF FIGURES AND TABLES

Figures

Figure 1: Schematic view of a powder bed EBM additive manufacturing system (ARCAM, 2018) ... 9

Figure 2: a) Setup for In-process monitoring b) an image taken using IR camera with material discontinuities highlighted by Rodriguez, et al. (Everton, et al., 2016) ... 11

Figure 3: Diagram of the Kapton Film canister (left) Image of the raw intensity and calibrated data collected from the IR camera (right), (Raplee, et al., 2017) ... 12

Figure 4: Model of complete experimental setup ... 14

Figure 5: Sapphire window with rubber sealing (left) and Camera Setup (right) ... 15

Figure 6: Comparison of visuals from Pyrosoft 320N for optimally adjusted emissivity from 1 (above) to 0.16 (below) ... 16

Figure 7: Sequence of events in build process for a layer ... 17

Figure 8: Organization of recorded data ... 18

Figure 9: CAD model for Trial 1 build ... 21

Figure 10: Effect of metallization for Trial 1 build at the end of support structure (left) and midpoint of trial (right) ... 22

Figure 11: Pigmentation on window after wipe ... 22

Figure 12: CAD model for Trial 2... 23

Figure 13: Increased pigmentation on window after trial 2 ... 24

Figure 14: Critical metallization effect ... 24

Figure 18: Trial 4 completed build ... 30

Figure 19: Trend for events over layer 1 ... 31

Figure 20: Trend for all events compared over layer 8-18 ... 33

Figure 21: Trend for events in a build process compared for layers 1-18 ... 34

Figure 22: Cold-spots observed in layer 4 and layer 8 of the build... 35

Tables Table 1: Properties of Inconel 625 ... 12

Table 2: Trial 1 parameters ... 20

(45)

43

10. APPENDIX

IMAGE PARSING CODE

The following code represents what was used in Matlab to parse the exported files from Pyrosoft.

function GetData

% Input parameters, used to at the beginning of the function to specify the number of layers to be %analyzed and the number of steps involved in producing one layer

NL=18; % Number of Layers NS=5; % Number of Steps

n=18; % Number of the Layer which you want to plot data or print the images

%the output path needed for exporting the averages from the Excel files outputPath = 'E:\Pyrosoft_trials\average_from_excel\';

% get Data = 1 load the Excel files exported from Pyrosoft for creating the average values

% getData = 0 start plotting the data for the different layers getData=0;

%If get data is 1 if getData

% T: matrix of 18x5. Each row show 5 temperature of each layer T=zeros(NL,NS);

% Save all averages into the same place M_AR=zeros(320,256,NL);

M_PH=M_AR;

M_SI=M_AR;

M_BH=M_AR;

M_PA=M_AR;

%Start counting from one to the number of layers imputed and get the average for every step by % calling the get5Steps function which will return the average for each process inside a layer for i=1:NL

[t, I_AR, I_PH, I_SI, I_BH, I_PA]=get5Steps(i);

T(i,:)=t;

%Store everything in order to be access later M_AR(:,:,i)=I_AR;

M_PH(:,:,i)=I_PH;

M_SI(:,:,i)=I_SI;

M_BH(:,:,i)=I_BH;

M_PA(:,:,i)=I_PA;

End

%After everything is in the right place move it to “AllData”

save('Alldata','T', 'M_AR', 'M_PH', 'M_SI', 'M_BH', 'M_PA');

%else if getData is 0 else

data=load('Alldata.mat');

T=data.T;

M_AR=data.M_AR;

M_PH=data.M_PH;

M_SI=data.M_SI;

M_BH=data.M_BH;

M_PA=data.M_PA;

end

%For 1 to the number of layers for k=1:n

% Note:

% These max and min temperature values are used for Normalization

% to present the temperatures in each matrix between 0 and 1 for

(46)

44

% the purpose of showing the images in MATLAB. for data analysis we used the

% tempetrature values in the original matrices.

mn=804; % min Tempetarure of after rake images of layer 1 for normalization purpose mx=959; % max Tempetarure of after rake images of layer 1 for normalization purpose

X=1:5; % X axis: 5 Steps of AR,PH,SI,BH,PA Y=T(k,:); % Y axis the temperature values of each step

%Parameters used for producing the graphs figure(1); plot(X,Y,'-b');title(['Layer ',num2str(k)]);

xticklabels({'AfterRake','','PreHeating','','Sintering','','BaseHeating','','Pause'});

ylabel('Temperature');

ylim([860,910]);

xlim([1,5]);

set(gcf,'color','w');

%save the graph in a .jpg format

saveas(gcf,strcat('E:\Pyrosoft_trials\average_from_excel\graph_layer_', num2str(k) ,'.jpg'))

%Start writing the images to the specified output path after they have been normalized with the mind and max values imwrite(normalize(M_AR(:,:,k),mn,mx), strcat(outputPath,'After Rake - Ave.Image - Layer ',num2str(k),'.jpg'));

imwrite(normalize(M_PH(:,:,k),mn,mx), strcat(outputPath,'Pre-Heating - Ave.Image - Layer ',num2str(k),'.jpg'));

imwrite(normalize(M_SI(:,:,k),mn,mx), strcat(outputPath,'Sintering - Ave.Image - Layer ',num2str(k),'.jpg'));

imwrite(normalize(M_BH(:,:,k),mn,mx), strcat(outputPath,'Base-Heating - Ave.Image - Layer ',num2str(k),'.jpg'));

imwrite(normalize(M_PA(:,:,k),mn,mx), strcat(outputPath,'Pause - Ave.Image - Layer ',num2str(k),'.jpg'));

end

% Plot each step for 18 layers X=1:18;

Name={'AfterRake','PreHeating','Sintering','BaseHeating','Pause'};

for i=1:5 Y=T(:,i);

figure(i); plot(X,Y,'-b');title(Name{i});

xticks(X);

ylabel('Temperature');

ylim([800,910]);

xlim([1,18]);

set(gcf,'color','w');

grid on;

end end

function [t, I_AR, I_PH, I_SI, I_BH, I_PA]=get5Steps(LayerNum)

%Function used to get each process that happens for each layer defined as: After rake, Pre heating, Sintering, Base heating, and Pause

%Create the file names to be read

fileName1=['1.AR_L',num2str(LayerNum),'.xlsx'];

fileName2=['2.PH_L',num2str(LayerNum),'.xlsx'];

fileName3=['3.SI_L',num2str(LayerNum),'.xlsx'];

fileName4=['4.BH_L',num2str(LayerNum),'.xlsx'];

fileName5=['5.PA_L',num2str(LayerNum),'.xlsx'];

% 1. After Rake passes

%--- I_AR=getAveIm(fileName1); % Average Image of After-Rake frames T_AR=mean(I_AR(:)); % Average Temperature of After-Rake Ave.image

% 2. Pre-Heating

%--- I_PH=getAveIm(fileName2); % Average Image of Pre-Heating frames T_PH=mean(I_PH(:)); % Average temperature of Pre-Heating Ave.image

% 3. Sintering

%---

(47)

45 I_SI=getAveIm(fileName3); % Average Image of Sintering frames

T_SI=mean(I_SI(:)); % Average temperature of Sintering Ave.image

% 4. Base-Heating

%---

I_BH=getAveIm(fileName4); % Average Image of Base-Heating frames T_BH=mean(I_BH(:)); % Average temperature of Base-Heating Ave.image

% 5. Pause

%--- I_PA=getAveIm(fileName5); % Average Image of Pause frames T_PA=mean(I_PA(:)); % Average temperature of Pause Ave.image

% Save data

%--- Name= ['Layer_',num2str(LayerNum)];

save(Name,'T_AR','T_PH','T_SI','T_BH','T_PA','I_AR','I_PH','I_SI','I_BH','I_PA');

t=[T_AR,T_PH,T_SI,T_BH,T_PA];

end

function imAve=getAveIm(fileName)

%Function used to produced the average from an exported Excel file.

% Read Excel file data=xlsread(fileName);

% Seperate the first column c1=data(:,1);

for j = 1: length(c1)

tempStr = strtok(num2str(c1(j)), '.');

if numel(tempStr) > 4 c1(j) = 1;

else c1(j) = 0;

end end

% Find the position of the NAN (where data is seperated) in the first column.

% NAN standing for Not a number, represents an indicator where one image ends and the other begins ind = find(c1 == 1);

IND=ind;

L=length(IND);

M=[]; % M is an empty tensor for i=1:L-1

r_start=IND(i)+1; % First row of the matrix r_end=IND(i+1)-1; % Second row of the matrix

im=data(r_start:r_end,:); % Seperate each matrix from the Excel data M(:,:,i)=im; % accumulate all matrices in a tensor

end

imAve=mean(M,3);

end

function In=normalize(I,mn,mx)

%function used to normalize everything based on the min and max values inputted In=(I-mn)/(mx-mn);

End

(48)

46 IMAGES OF TRIALS (Trial 3 onward)

Trial 3

Trial 4

(49)

47 Trial 5

Trial 6

(50)

TRITA -ITM-EX 2018:759

www.kth.se

In-process monitoring for Electron Beam Additive Manufacturing using an infrared camera system

In-process monitoring for Electron Beam Additive Manufacturing

using an infrared camera system

VLAD ANTONIU BUGA

ROYSTEN JASON DSOUZA

Degree Project in Production Engineering and Management (MG213X)

In-process monitoring for Electron Beam Additive Manufacturing using an infrared camera system

VLAD ANTONIU BUGA

ROYSTEN JASON DSOUZA

ABSTRACT

SAMMANFATTNING

ACKNOWLEDGEMENTS

TABLE OF CONTENTS

1. INTRODUCTION

1.1 Background

1.2 Motivation and Problem Statement

1.3 Methodology and thesis outline

2. THEORY

2.1 AM technology and EBM

2.2 In-process monitoring in EBM

2.3 Material properties of Inconel-625

3. METHOD

3.1 Experimental setup

3.2 Data collection

3.3 Data sorting

3.4 Data processing

4. EXPERIMENTAL TRIALS AND DATA COMMENTARY

4.1 Trial 1

4.2 Trial 2

4.3 Trial 3

4.4 Trial 4

4.5 Trial 5

4.6 Trial 6

5. RESULTS AND DISCUSSION

6. FUTURE WORK

7. CONCLUSION

8. REFERENCES

9. LIST OF FIGURES AND TABLES

10. APPENDIX