Mechanical Property Variance Amongst Vertical Fused Filament Fabricated Specimens via Four Different Printing Methods

Amongst additive manufacturing processes, fused filament fabrication (FFF) is one of the most affordable and cost efficient technologies that can produce complex shaped components with an increasing number of printable polymers such as the polyaryletherketone (PAEK) family, polyetherimide (PEI), and polyphenylene sulfide (PPS). Despite the gain in popularity, there is a lack of standardization in specimen’s preparation and mechanical testing of FFF samples. This study investigates the effect of different methods of printing vertical tensile specimens on the mechanical properties whilst the material and the printing parameters are fixed. A slow crystallising polyetheretherketone (PEKK) grade was selected as the printing material to exclude the effect of crystallisation on the interlayer bonding strength, leaving the temperature-dependent amorphous molecular diffusion across the layers as the governing mechanism. Vertical tensile specimens made by four printing methods: individually printed, machined, and connected (based on ISO 527-2-1A and ISO 527-2-1BA) were assessed. Individually printed vertical specimens were found to have the highest mean tensile strength, owing to the high level of diffusion induced by the very short layer time. The strengths of specimens printed via the other three methods are less sensitive to the effect of layer time, due to the rate of change during cooling and its relationship with the local temperature at the interlayer surface. This study highlights the importance of the disclosure of FFF printing methods along with any reported mechanical data.


Introduction
Material extrusion of polymers, commonly known as Fused Filament Fabrication (FFF), is one of several recognised methods of Additive Manufacturing, picking up speed and reliability as well as a wider range of high-performance materials available. The ASTM International Committee F42 on Additive Manufacturing Technologies standardised the FFF technology with the term Material Extrusion and the following definition: "an additive manufacturing process in which material is selectively dispensed through a nozzle or orifice". 1 The process has been published in multiple research papers with a vast range of polymeric materials [2][3][4][5] and their compounds. 6 Among the printable polymers, the Poly Aryl Ether Ketone (PAEK) family attracts increasing attention due to their superior mechanical properties and good biocompatibility.
Depending on the maturity of the AM system, some manufacturers provide a qualified material and parameter set which they have tested in-house and provide proven mechanical data. As such, Stratasys offers proprietary materials with fixed processing conditions to ensure components meet the published mechanical values when produced on their AM systems. One example would be Stratasys Ultem 1010. 7 Alternatively, material manufacturers offer guideline mechanical values subject to processing conditions and machine quality, such as Luvocom PET-CF. 8 Much of the published research into FFF process optimisation focuses on the variability of process parameters, and how these can be optimised for specific mechanical values such as maximum tensile stress or flexural stress and typical test samples produced in the XY (Flat) and XZ (Edge) orientations. [2][3][4][5] Maximum tensile stress of specimens aligned in the ZX (Vertical) axis offers a good indication of the layer to layer bonding strength of the process and of the material and is often significantly lower than the maximum tensile stress in the XY and XZ orientation, highlighting the anisotropy of the FFF process compared to other polymer processes such as injection moulding. Figure 1 displays the three notations of build orientation. The established test standards only provide limited guidance on the characterisation of AM specimens. The European Standard EN ISO 527-2:2012 "Plastics-Determination of tensile properties-Part 2: Test conditions for moulding and extrusion plastics" sets out the conditions for sample types and manufacturing conditions, 9 although not specific for AM. The standard states that "wherever possible, the test specimens shall be dumb-bell-shaped types 1A and 1B as shown in Figure 1 and Table 1. Type 1A shall be used for directly injection-moulded multipurpose test specimens, type 1B for machined specimens." It also states that "test specimens with machined surfaces will not give results comparable to specimens having nonmachined surfaces." The final important part of the standard mentions that "results obtained from small specimens are not comparable with those obtained from type 1 specimens." Whilst some technical datasheets list a testing standard or number of samples tested, the authors have found limited information specifying the method of manufacturing the specimens from which the data is recorded. The only study specifying the preparation method was defined for ULTEM 7 within the printer's proprietary control software from which the specimens are built and tested. This includes reference to number of specimens, support material methodology and the inclusion of a sacrificial tower built to the height of the model. 10 Other manufacturers have published just the test method identification and the value with no information of number of samples, construction methods or printing strategy. 11 Moreover, some manufacturers provide only material datasheet but not the printed component datasheet. 12 The lack of information is one of the incentives to conduct this study.
Gebisa et al. 13  However, what is missing from the current published research is a standardisation of the sample manufacture which should be considered before investigation into the process parameter variables. For example, Rinaldi et al. 16 printed ASTM D638-Type V vertical (Z) tensile specimens individually with a circular support structure at the bottom. Meanwhile, Arif et al. 17 printed four ISO 527-2-1BA type vertical (Z) tensile specimens in a single fabrication sequence. Although both studies printed Victrex PEEK450 via Indmatec printers, the former reported a Z tensile strength of 19.6 MPa whilst the latter reported a Z tensile strength of 9.99 MPa. The results of these incommensurable different printing strategies weakens the arguments made by comparing mechanical data from various sources. 18 The authors propose a study into the effect of tensile specimen manufacture on the Z mechanical performance whereby the material and process parameters are fixed and only the methods of tensile specimen preparation are changed.

FFF process
Due to the nature of the FFF process, manufacturing Z tensile specimens is more challenging, and several methods have been employed to successfully construct valid test specimens. This study proposed four methods to construct Z Tensile ISO 527-2-1A and ISO 527-2-1BA specimens respectively, see Figure 2. 10 specimens for each design were manufactured and tested.
Method 1: Individual Z specimen, unsupported. Depending on the specific AM system and their heating environment, it is possible to build an individual tensile specimen Type 1BA in the vertical (Z) axis. Some systems will require the addition of a support structure (scaffold), which may be a secondary material.
Method 2: As a first step, a wall or tower geometry with matching thickness to the tensile specimen thickness was built, then the specimen geometry was machined from the vertical walls.
Method 3: Continuously connected specimens. Various geometries can be constructed where the tensile specimen is connected to a thin wall structure, enabling the manufacture of multiple specimens that self-support during the AM process.
Method 4: Similar in design to Method 3 but using Type 1BA specimens as size. The specimens were manufactured using a 3DGence Industry F340 FFF system in Kimya PEKK-A (Poly Ether Ketone Ketone) material with identical processing parameters except the Method 1 which required the use of the part cooling fan due the short layer time, see Table   1. Layer time is the time interval between printing consecutive layers. The cross section of specimens from four methods at the gauge-length zone are displayed in Figure 3, with the corresponding layer times. ESM-10 soluble support material was used to build a raft structure below the specimens. All filaments were stored under 45 °C for at least 24-hours before printing. Method 3 specimens were machined using an Openbuilds Benchtop Router at a cutting speed of 3 mm/s. The Method 1 specimens required the use of the cooling fans due to the extremely short layer time and the radiated heat from the nozzle. Without the cooling fans the deposited polymer remained above its glass transition temperature during construction, leaving insufficient rigidity to support the next printed layer, resulting in a deformed gauge area.

Mechanical test
Tensile tests were performed to characterise the effect of sample manufacture using a 20 KN capacity standard tensile/compression machine (Shimadzu®). According to ISO 527, moduli were measured at a constant speed of 1 mm/min whilst tensile strengths and elongations at break were measured at 5 mm/min. 10 repeats for each category were achieved.
Significant differences were identified using an ANOVA technique. Means comparisons by Tukey Kramer HSD method were made using JMP (SAS, Version 15.0.0). The criteria of showing significant difference between each pair is the p-value below 0.05.

Differential scanning calorimetry
Differential scanning calorimetry (DSC) analyses were carried out using a Mettler Toledo DSC1 STAR e system on filaments of 10 mg to identify the PEKK grade. The polymer was tested by a dynamic scanning sequence of two heating scans at 10 K/min from 25 °C to 400 °C and one cooling scan at 10 K/min from 400 °C to 25 °C. All scans were protected by a 50 ml/min Nitrogen flow.

Results
The tensile modulus, tensile strength, and elongation at break per each method are shown in

PEKK grade and TTT diagram
The crystallisation speeds of PEKK materials are known to be determined by their various para/meta phenyl isomer ratios, or more commonly known as the T/I ratios. 19 The T/I ratio of The crystallisation behaviour of PEKK (60/40) has been investigated extensively. 21,22 The timetemperature-transformation (TTT) diagram of PEKK (60/40) developed by Choupin et al. 21 is partially displayed in Figure 7 (b). The TTT diagram reveals that PEKK (60/40) has the highest crystallisation speed at 230 °C. Even under this temperature condition, it still requires 150 s to initiate the crystallisation (i.e. reaching 1 % crystallinity), which suggests that PEKK (60/40) is a slow crystallising grade when comparing with other semi-crystalline PAEK grades. 23 Therefore, it is safe to conclude that all layer times in this printing matrix are too short to start crystallisation. In addition, visual inspection on all specimens confirms that they remain amorphous. For this reason, amorphous molecular diffusion across the layers is considered as the major mechanism governing the interlayer bonding of all printed samples.

Mechanical variations
Typical stress-strain behaviours of each method are exhibited in Figure 8 (a). It is noted that all specimens fractured prematurely before yielding. In Z tensile specimens, all layers are constructed in the XY plane, making them perpendicular to the loading direction. Along the loading direction, the interlayer bonding is in general weaker than the material's bulk strength. Therefore, the reported tensile strengths and elongation at break values signify the interlayer bonding strength rather than the bulk property. The interlayer bonding strength is directly linked to the level of molecular diffusion, which in turn is influenced by the local temperature near the interlayer surface. Once a new layer is deposited on top of a previous layer, thermally driven molecular motion prompts the amorphous molecular chains to diffuse across the layers. The diffusion process is temperature dependent: a high temperature promotes the diffusion, bringing a higher chance to achieve good interlayer bonding.
Meanwhile, the local temperature near the interlayer surface is determined by the layer time.
Although the processing conditions (nozzle temperature, velocity, chamber temperature etc.) are kept constant for the fabrication of all specimens, the local conditions vary depending on the method of construction. Method 1 requires the nozzle to remain directly over the previously printed surface, further heating the polymer, with a very short layer time. In Methods 2, 3 and 4, the nozzle moves away from previously printed material due to the larger cross section and printing strategy, giving the previous layer longer time to cool without the heating effect created by the nozzle. The link between the layer time and the local temperature is evident: the shorter the layer time, the higher the local temperature.
To facilitate the discussion further, the temperature profile of printing PEKK measured by Lepoivre et al. 24 is shown in Figure 8 (b). In their study, the nozzle temperature was 356 °C, which is in a close range of the nozzle temperature (380 °C) used in the current study. Hence, the local temperature in this study is considered to follow the same trend as measured by Lepoivre et al. 24 The layer time in their study was 9.8 s, which explains the fluctuation of temperature profile observed around 10 s. To quantitively investigate the rate of change in temperature between layers during cooling, the first derivative of the temperature profile was calculated. The derivative values highlight a significant change in slope up to approximately 10 s which could represent the critical layer time where mechanical differences will become substantial. This observation implies that when layer time is longer than 10seconds, the level of diffusion becomes less sensitive to the effect of layer time than when layer time is less than 10-seconds. The insensitivity explains why the strengths of specimens from Method 2, Method 3, and Method 4 are not significantly different, and why Method 1 prints specimens with higher tensile strength and high elongation at break value than the other methods.
It is interesting to notice that tensile moduli of specimens from Method 3 and Method 4 are generally higher compared with that of specimens from Method 1 and Method 2. As the crystallinity effect has been ruled out, the cause of the change in modulus is not clear at this stage.

Fracture location
A dogbone shape tensile specimen consists of five zones: two grip zones, two transition zones and one gauge-length zone, see Figure 9. In tensile test, the dogbone shape specimen is adopted to confine the deformation to the narrow parallel centre region (i.e. the gauge-length region) and to reduce the likelihood of fracture to occur outside this region. This is favoured because failure in the transition zones or in the grip zones would cause an improper measurement of the elongation values. Figure 9. Illustration of the five zones of a dogbone tensile specimen. ∆x marks the distance starting from the onset of the transition zone towards the gauge-length zone.
The fracture locations of all FFF specimens were documented and summarised in Table 2. The distribution of fracture location varied according to the printing method. To investigate the potential cause of this distribution, layer times throughout the transition zone to the onset of gauge-length zone for all methods were collected and presented in Figure 10. Figure 10. Layer time as a function of ∆x (entire transition region). ∆x is defined in Figure 9 and marks the distance starting from the onset of the transition zone towards the gaugelength zone.
The majority of Method 2 specimens fractured in the gauge-length zone, behaving the same as conventional tensile specimens, for example made by injection moulding. Method 2 ensures the same layer time of 43.9 across the whole specimen, resulting in the same level of interlayer bonding in the transition zone and in the gauge-length zone. Upon loading, the axial stress follows (transition zone) < (gauge-length zone) due to the fact that the applied force acts in series whilst the cross-sectional area increases in the transition zone. The higher axial stress in gauge-length zone helps constrain the facture to within this region. In addition, These steps may act as defects causing fracture in the transition region. Closer inspection of the toolpath information generated by the slicing software showed no variation in the extrusion strategy between the gauge-length layers and the transition area layers, therefore advanced slicer settings influenced by overhanging geometry can be excluded.
More than half of Method 1 specimens broke in the transition zone, likely due to the aforementioned stepping effects. Surprisingly, the layer time of Method 1 in the transition zone is constant around 3.6 ± 0.1 s, possibly due to the small cross-sectional area so that the majority of the time was in the period that the nozzle is positioning itself (acceleration and deacceleration phase) and not printing. Therefore, the change in printing area for each layer is not reflected in the layer time, as it is a small proportion of it.

Conclusions
Individually printed, machined, large and small connected type vertical FFF specimens were evaluated to quantify the mechanical variance between the four printing methods. Different methods lead to different layer times, bringing various local temperature values near the interface. The strength and the elongation at break are found to be linked with the interlayer bonding strength. Slow crystallising Kimya PEKK-A was selected as the printing material to exclude the effect of crystallisation on the interlayer bonding strength, leaving the temperature-dependent amorphous molecular diffusion across the layers as the governing mechanism. A maximum, mean tensile strength of 53.8 MPa was observed for Method 1 specimens, owing to the high level of diffusion induced by the very short layer time. The strengths of specimens printed via Method 2, Method 3, and Method 4 are less sensitive to the effect of layer time, due to the decreasing spontaneous cooling rate. Moreover, the fracture locations indicate that the effect of surface finish and the changing layer time bring instability to the testing. Thus, standardising the Z tensile methodology (number of samples, layer time etc.) is desirable to the additive manufacturing community so that data sets produced are comparable, with the right method selected to make it representable to likely printing processes, desired printed component layer time, and part production. Additionally, full details should be provided in relation with the printing and testing methods to allow interested parties to make informed decisions on performance of materials and printers.